3.08.2006

If only they had a time machine...


UPDATE: March 10th, 2006 - Read the follow up.

I would like to take a moment to congratulate Intel on showing Conroe to both Anandtech and Hexus at IDF. As I suggested a few months ago, Intel derailed themselves a number of years ago by staying with their Pentium 4 architecture. There is little doubt that they will be getting back on the rails later this year.

I would like to pipe in a simple counterpoint on a few things, because I feel it’s important that people understand the entire picture before rushing to judgment. In years of being in this business I cannot recall a time where a company debuts information on a product that is six months out. I can only assume that by doing this they have completely lost faith with their current desktop products.

Intel jumped into their time machine to show off a fetus that has six months to mature before it turns into a baby. They basically gave themselves the opportunity to ride in this time warp without giving AMD a chance to get in the passenger seat. Anytime you compare a future technology with an existing one you should expect to see some performance differences, but looking closer it seems as if they are comparing future technology with older technology, rather than current, in some respects.

First of all, though it’s claimed that you’re running an FX-60 processor on the machine, I have yet to see an established motherboard and processor with an image that says "processor unknown".

You’ll notice that the image I am referring to on Anandtech's website (the bios image) states that the AMD processor is “unknown” which makes me believe that the bios they are running is outdated. So, I did a bit of digging and low and behold, the DFI bios version “D49C-32” they are running is from 10/11/05. There has been 1 major revision with major fixes that include:

Set Cool 'n' Quiet Default to Disabled

- With Cool & Quiet enabled, AMD processors will throttle in order to save power and bring their thermal load down. This means the processor could be running as low as 800MHz in certain programs – no matter what the program is. In theory Cool & Quiet is supposed to throttle up to maximum in games but this is not always the case. No enthusiast PC goes out with Cool & Quiet enabled unless it’s a fanless machine or media center.

Add Support for AMD Athlon 64 FX60 CPU

- According to DFI the FX-60 will not operate correctly without this bios update. Without official support for the FX-60 CPU I’m not sure what we’re comparing against here.

Fix Memory Timings 2-1-1-1-1 and 4-1-1 Mode Wrong & Fix Read Preamble Table Error.

- Memory latency can make a massive difference in performance. If the latency was not running at the correct latency we can see a pretty big difference in all kinds of performance. Anandtech stated “The AMD system used 1GB of DDR400 running at 2-2-2/1T timings…” Apparently this isn’t the case, but they would not be able to tell without having the platform in house.

Fix Fill 3114 SVID&SSID under Cross fire mode.
- More apparent performance issues under Crossfire mode.

Next, when you take a future Intel chipset and compare it to a chipset that no enthusiast supports (RD480) with an outdated bios it’s like taking a Ferrari and putting it on Bias-Ply tires. It’s just not a good way to show off a “new” technology.

Had Intel taken an RD580 (Crossfire Xpress 3200) and coupled with the AMD Athlon FX-60 processor they almost certainly would have seen some better numbers just based on the bios issues alone. The ATi Xpress 3200 would have improved the overclocking and decoding performance as well. You don’t need a time machine to jump over to the nearest Newegg and buy the latest parts. It’s almost like Intel took their time machine 6 months ahead while throwing AMD into a time warp set a few months back in time.

So now a few numbers caught my attention based on another email I received last night.

If we go and check out the numbers on Anandtech we’ll see the Unreal Tournament 2004 benchmark showing 160fps on the unknown AMD X2 processor while the Intel Conroe at 2.66GHz came in significantly higher at 191fps.

Though this isn’t exactly conclusive, if you go back and re-read some old FX-57 reviews on Tom’s Hardware you’ll see a benchmark for the same game set at the same resolution (and the same color depth), the FX-57 running at 2.8GHz scored 183.4fps. The thing is it’s using an Nvidia Geforce 6800 GT which seems to me that there are many variables here when it comes to benchmarking. Perhaps it's somewhere locked in the settings, but I won't know until I sit down and compare our own benchmarks with consistant settings. Note that a single core Athlon 64 4000 achieved a better score in the benchmark run by Tom (160.5fps) than the one provided by Intel (160.4) at IDF. Like I said, I don't view this as conclusive, but it shows that there are variances depending on how the benchmark is setup. Here is a link to Tom’s review. I also don't believe the FEAR benchmarks are accurate, but again we'll know when we get a chance to test it ourselves.

These are just a few of the thoughts that I have been able to gather. I have received over 100 emails in the last 18 hours about this to see what my thoughts are. Quite simply, I would say go back and read my original article on Conroe – I certainly expect Intel to be back on the rails, but I don’t see them to be wiping the floor with anyone other than their Pentium 4 team come launch time.

AMD still has some big performance gains with AM2, we are talking about a new platform with low latency DDR-2 support along with new processors. While I’d love to tell you how much performance difference this would give you on a clock to clock basis, I’m afraid you’ll have to use your imaginations for now.

We can speculate all we want, but why bother? By simply changing the bios and ensuring that Cool & Quiet is disabled, that the FX-60 is fully supported, and that there are no memory timing issues I imagine we would see a fairly significant performance gain on those merits alone.

The long and the short of it is Intel has crafted an excellent marketing strategy to show off their baby in its first trimester. They are trying to win the hearts and minds of enthusiasts half a year before they have anything to show us. They created these platforms in house, and we can only hope they unknowingly crutched the AMD system by using a chipset and motherboard - with an outdated bios - that no enthusiast supports. EDITED FOR CLARITY - If you read the article you'll see that the AMD system is clearly crutched - I am hoping that Intel had no clue that this was the case. I suspect they did this unknowingly.

I’m going to give Intel an “A” for their P.R. effort, good job guys, and thanks for the show. I can’t wait to see the real substance. Conroe looks strong no doubt, hopefully it will be available in volume by the time it launches.

Anyone reading this should also assume that AMD is not standing still. Based on these articles it's clear that this year is going to be absolutely incredible.

UPDATE: March 10th, 2006 - Read the follow up.

106 comments:

Anonymous said...

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2713&p=1

here is a interesting thing that anand said

"As far as we could tell, there was nothing fishy going on with the benchmarks or the install. Both systems were clean and used the latest versions of all of the drivers (the ATI graphics driver was modified to recognize the Conroe CPU but that driver was loaded on both AMD and Intel systems)."

a modified driver for the conroe was loaded on both machines , isn't there something wrong there ?

Anonymous said...

Rahul, I do give you a lot of credit. You do know what you are talking about. Every since I was a little kid I viewed your company and your computers as The best. While I have not always exactly viewed some of Voodoo's directions and actions. You do know what you are talking about and very nice work (just lower your prices, lol)

Anonymous said...

One thing I didn't see mentioned [edit: besides the other anonymous user I see whiel previewing :P], which I wonder about, is the graphics driver. In Anand's article, it is stated that "the ATI graphics driver was modified to recognize the Conroe CPU but that driver was loaded on both AMD and Intel systems"

What I want to know is:
1) Who compiled this new revision of the driver?
2) What compiler and compiler flags were used?

Since it is stated that the modified driver was used on BOTH the AMD and Intel systems, the potential of an Intel compiler used makes me suspicious. Considering past evidence of Intel compilers creating optimized code that ignores cpu capabilities if they don't return "GenuineIntel", I wonder at how much of a performance hit AMD would be taking if all the graphics driver codepaths were 'safe' x87 instead of the fully supported SSE/SSE2 (or better, 3D Now!)

I think Intel has made a good CPU this time, but I wouldn't put it past them to fudge the numbers as much as possible to make it look like they "leapt ahead" instead of just "finally caught up." Exciting times ahead, I will eagerly await reviews from sources with less incentive to lie or cheat.

-Frid

Sharikou, Ph. D said...

My analysis of the situation is here. It becomes very amusing to watch these things play out. BTW, my estimate of Woodcrest performance is here.

Anonymous said...

You've never seen an overclocked AMD system show an unknown processor type?

Haven't seen very many systems at all have you?

Anonymous said...

You can take a look at http://www.anandtech.com/video/showdoc.aspx?i=2540&p=9 which shows FX55 running on a ATI X800XT. For the same game, at same resolution, it only scored 67.3 while 6800 scored 66.8. Of course, 6800GT perform better than 6800, but it won't make me to believe that a FX57 and 6800GT can score 183 if the same demo is used.

Based on the fact, the demo toms used is no way the same as Intel used.

Anonymous said...

BIOS used for DFI is _deafult_ coming with board announced in January.

AMD M2 will bring DDR2800, but DDR2800 will be used by Intel as well, so don't even hope there will be perf. increase (check Tom's review).

I'd even say it will be perf. decrease since Intel tolerates DDR2 latencies right now, but not K8.

The fact that Intel shows this monster "baby" now, means that it could be released 1-2 months from now, but marketing "back to school" dictates it be released in July.

Moreover, even currently I'd by 2Mb L2 65-nm CedarMill/Presler and overclock it >50% instead of any Athlon.

Server Woodcrest 3GHz@1333FSB first time ever will be faster then desktop offerings. Not many noticed "128SSE" - it's important, it's x2 gain in FP, so no more AMD FP leadership.

Next year is transition to 45-nm ...

my advise to you guys, just stop design your HW on AMD, and buy Intel shares while they are ~20$ now.

RotoSequence said...

You make some interesting points; youre more than likely right that Intel has done some optimizations for itself (And probably a few agaisnt AMD).

Bit-tech however, did run comparative tests with their own test rig (which has very similar parts) and got comperable results as the Intel system with FEAR (Graphics turned down to minimum to reduce impact).

Ultimately, time will tell just what kind of differences there are, but Intel does have a six month gestation period left before Conroe's release. Things will be VERY interesting come September.

Anonymous said...

Hi Rahul,

Thanks for reassuring us and giving your prespective on the "early" benchmarks. I would have to assume that anytime a vendor (Intel in this case) places their top-of-the-line processor pitted against their competitor's best under their watch can't possibly be a fair benchmark.

Also, I'm actually disappointed that Anandtech didn't raise these issues (processor unknown, in Intel's house, etc) before crowning Intel the champ. I have followed Anandtech for years and found them very resourceful, but now I question "Did Intel pay Anandtech off?". Anandtech should immediately run the same benchmarks in their own labs for the Athlon FX-60 to confirm the results they found in Intel's lab was indeed correct (also to see if Tom's results are close).

Thanks,
Mike Lee

TheKhalif said...

The benchmarks on Toms' are for SLI not for a single GPU. At least he's using an SLI board. I doubt that the 4000+ and 1 6800GT would have 89 frames on DOOM III 1280 HQ. My 7800GT maybe, not a 6800GT.

Since X1900XT will TOTALLY DESTROY 6800GT and the FX57 got 183 fps, it seems like there was a problem at best.

Anonymous said...

I don't think you can just compare benchmark results across the board from different sources. The demos run for the benchmarking critically affect the performance. You can set up two different demos, and have a 100% or more variation in performance, depending on for example the number of bots used in the demo. Fact is, we either believe the results shown by Intel, or we do not, but unless you are running the same benchmark demos, you can't verify a whole lot. Had Intel used a more up-to-date platform, with better performance and a newer bios, maybe the comparison would have convinced more people. As it is now, certain questions still remain.

Rahul Sood said...

Anonymous said

a modified driver for the conroe was loaded on both machines , isn't there something wrong there ?


No, I don’t think there is anything wrong – they needed a modified driver in order to operate correctly on the Intel chipset.

Wed Mar 08, 02:25:53 PM
Anonymous said...

Rahul, I do give you a lot of credit. You do know what you are talking about. Every since I was a little kid I viewed your company and your computers as The best.


Thank you for the nice comments.

Wed Mar 08, 02:37:31 PM
Anonymous said...

I wonder at how much of a performance hit AMD would be taking if all the graphics driver codepaths were 'safe' x87 instead of the fully supported SSE/SSE2 (or better, 3D Now!)


I would say these benchmarks on the AMD side are hindered slightly overall based on the bios and the things I wrote about. I don’t think it would offer a huge difference, but certainly enough to close the gap slightly.

Exciting times ahead, I will eagerly await reviews from sources with less incentive to lie or cheat.

Agreed, exciting times ahead no doubt.

Wed Mar 08, 02:55:20 PM
Anonymous said...

You've never seen an overclocked AMD system show an unknown processor type? Haven't seen very many systems at all have you?


I guess there is such a thing as a dumb question. Any of the new processors have their unique identifiers setup regardless of overclocking. The bottom line is DFI said that the FX-60 will not operate at full potential with the installed bios.

Wed Mar 08, 03:04:42 PM
Anonymous said...

Based on the fact, the demo toms used is no way the same as Intel used.


I would agree with this – it can’t possibly be the same.

Wed Mar 08, 03:10:35 PM
Anonymous said...

BIOS used for DFI is _deafult_ coming with board announced in January.

The bios was updated in December. Check it again.

AMD M2 will bring DDR2800, but DDR2800 will be used by Intel as well, so don't even hope there will be perf. increase (check Tom's review).

Intel uses DDR2 now. 667 vs 800 will make a difference, but not as much as DDR to DDR2.

I'd even say it will be perf. decrease since Intel tolerates DDR2 latencies right now, but not K8.

I’d say you’re wrong, I’m expecting to see lower latency DDR-2 memory in a few months.

The fact that Intel shows this monster "baby" now, means that it could be released 1-2 months from now, but marketing "back to school" dictates it be released in July.

I don’t agree with this timeline, but who knows.

Moreover, even currently I'd by 2Mb L2 65-nm CedarMill/Presler and overclock it >50% instead of any Athlon.

..and currently I’d buy a 2009 Chrysler Imperial, but damn I don’t have a time machine.

my advise to you guys, just stop design your HW on AMD, and buy Intel shares while they are ~20$ now.

My advice to you, wait until AMD posts their next quarter… 

Wed Mar 08, 04:06:25 PM
RotoSequence said...

Ultimately, time will tell just what kind of differences there are, but Intel does have a six month gestation period left before Conroe's release. Things will be VERY interesting come September.


Agreed.

Wed Mar 08, 04:13:21 PM
Anonymous said...

Thanks for reassuring us and giving your prespective on the "early" benchmarks. I would have to assume that anytime a vendor (Intel in this case) places their top-of-the-line processor pitted against their competitor's best under their watch can't possibly be a fair benchmark.


Well, it’s not independent – and it’s not built by people who have the best interests of the competitor in mind. It’s still a good rough indication of what to expect from their side though.

"Did Intel pay Anandtech off?".

Anand is way too honest to get paid off by anyone. He’s writing as he sees fit, and I would never question his integrity.

Wed Mar 08, 04:17:04 PM
TheKhalif said...

Since X1900XT will TOTALLY DESTROY 6800GT and the FX57 got 183 fps, it seems like there was a problem at best.


We are running our own tests now.

Wed Mar 08, 04:19:21 PM
Anonymous said...

I don't think you can just compare benchmark results across the board from different sources.


Agreed, we need to do our own internal tests for sure.

Richmond said...

I was really amazed by the results of Conroe's performance especially in gaming.

There were discrepancies as you pointed out.

But for me this is a wake up call for AMD.

I would like to see AMD increaase the EV6 bus which is at 200Mhz now to 400Mhz. You pair it with DDR II 800 and for sure Intel's Conroe will get thrased again.

Parker said...

The benchies are very impressive but I think untill it is actually availlable you can't really compare them. The FX-60 is on the market and can be bought right now but the conroe cannot. Intel is headed the right direction but untill conroe is available I have no reason to hype about it. And what is AMD going to do for the next 4-6 months? Sit there? Stop making better processors? We all should wait to go crazy untill after we can actually buy a conroe chip.

Sharikou, Ph. D said...

There is no BUS at all for AMD64. AMD64 uses Direct Connect Architecture. EV6 bus was for the old K7 which now only exists as the Geode LX.

I challenged INTEL to loan AnandTech the conroe box for a test at home, the INTEL dudes apparently have no balls to meet that challenge.

This INTEL test is rigged period. Just like the rigged INTEL benchmarks ( http://www.vanshardware.com/reviews/2002/08/020822_AthlonXP2600/020822_AthlonXP2600.htm ) , rigged INTEL compilers ( http://www.swallowtail.org/naughty-intel.html ) and rigged Skype ( http://blogs.zdnet.com/ip-telephony/?p=947 ).

A company without basic ethics will fall hard.

Wil Harris said...

I too was suspicious of Intel's tests. To get a real comparison, I ran the Conroe in a benchmark I have a reference for: FEAR. In our previous CPU tests on bit-tech, we have run FEAR at 640x480 0x/0x as a CPU benchmark, and we have comparative numbers for a range of processors using this benchmark.

I tested Conroe in exactly the same method, to get a decent reference. You can also see that the Intel AMD rig performed worse than our reference FX-60 lab machine in my testing.

However, Conroe still rocks compared to our properly-configured FX-60 machine.

Interesting food for thought: http://www.bit-tech.net/news/2006/03/08/intel_conroe_benchmark_fear/

Anonymous said...

Why not ask the Intel guys for the timedemos they used? That would be the quickest way to verify at least the results on the athlon.

Anonymous said...

Interesting to note the way these things seem to immediately polarise the tech community. Intel fans will always be Intel fans, AMD fans will always be AMD fans. Time will reveal the reality.

Point not to miss is that a weak Intel is not good for AMD and vice versa. If AMD had not become what they are now, we may still be paying $8000 for a PC. Technological advances always leapfrog between the main players and that's something to be welcomed. So trying to prematurely "bury" AMD (as some comments around the web advocate) will not necessarily benefit the consumer, and neither will it work the other way around.

Take a step back from the hype and think about it.

Sharikou, Ph. D said...

<<" Intel fans will always be Intel fans, AMD fans will always be AMD fans. ">>

Will INTEL fans keep buying Netbust P4, Xeon and 32 bit Core Duos to keep their beloved company alive or will they wait for the super duper Conroe/Wodcrest while their beloved co runs into operating losses?

Rahul Sood said...

Point not to miss is that a weak Intel is not good for AMD and vice versa.

Agreed, but don't kid yourself - Intel is anything but weak. Even though they are losing market share and posting poor results Intel is an absolute monster. It puts things into perspective when you consider what AMD has been up against for the last number of years. That's management at its best.

Rahul Sood said...

Wil Harris said...

I too was suspicious of Intel's tests. To get a real comparison, I ran the Conroe in a benchmark I have a reference for: FEAR. In our previous CPU tests on bit-tech, we have run FEAR at 640x480 0x/0x as a CPU benchmark, and we have comparative numbers for a range of processors using this benchmark.


Will, thanks for stopping by. I frequent your site quite often actually, so thanks for the great work.

I tested Conroe in exactly the same method, to get a decent reference. You can also see that the Intel AMD rig performed worse than our reference FX-60 lab machine in my testing.

I would like to have a discussion with you on the benchmarks when you get a chance - drop me an email sometime.

However, Conroe still rocks compared to our properly-configured FX-60 machine.

No doubt Intel has a strong product, but once again people assume AMD is standing still. Give it a few months to mature and see how it all plays out.

I still believe large performance gains could be had on a properly configured (overclocked to 2.8) AMD system. This closes the gap further on technology that's actually available.

Anonymous said...

Well... FEAR results can be explained easily here. FEAR is single threaded and *very* cache sensitive (a 165 will beat a 3800+ here, where it is usually the other way around). For this app the one conroe core is going to have the full 4 megs of L2 while the FX-60 will only have 1. Of course, FEAR isn't that indicative the performance of other games, and seems to be rather poorly coded.

The rest of the game benchmarks don't really mean squat though as long as they are using an Intel supplied timedemo. Remember the old quake 3 timedemos and how AMD would win on some of them and Intel on others? And those timedemos were made by a third party.

Sharikou, Ph. D said...

I recommend everyone who is interested in knowing more about INTEL and AMD to read the book "Inside Intel" by Tim Jackson, it costs only about $2 on amazon.com . Intel is indeed a monster.

Anonymous said...

The 6.00PG BIOS shown in Anand's image is the 12/13 BIOS from DFI, not the 10/15 one.

Richmond said...

so the amd64 no longer use the ev6 bus

thanks.

so let me correct my statement

AMD should bump the FSB to 400Mhz...which is 200Mhz right now and match it with an 800MHz DDR II.

that should thrash Conroe

Rahul Sood said...

Anonymous said...

The 6.00PG BIOS shown in Anand's image is the 12/13 BIOS from DFI, not the 10/15 one


The bios they are running is the one I said "D49C-32".

Sharikou, Ph. D said...

A lot of fishy stuff is going on with FEAR (see http://www.beyond3d.com/forum/showthread.php?t=25456 ). Though the situation is differnt, the spirit it the same. So unless you have the exact same exe running on the conroe and the FX, and you can play with the thing to figure out the anomalies, you can't be sure about the results.

T2k said...

Excellent observations, congrat. :)

Anonymous said...

Anandtech never stated that they were comparing it against a FX-60. They overclocked a X2 4800+ TO FX-60 speeds.

Anonymous said...

And what price would the intel chip actually be compared to the AMD?

Not to mention by the time it actually appears - AMD could be up to an FX-64 or 66!

Anonymous said...

so let me correct my statement

AMD should bump the FSB to 400Mhz...which is 200Mhz right now and match it with an 800MHz DDR II.


No no no. Athlon64 does NOT use a FSB in the sense as you think. The CPU is NOT using an "FSB-to-chipset-to-memory" to access the memory. The Athlon 64 have a built in memory controller that directly connects to the memory banks, no FSB needed in the sense you think, and the new DDR2 controller will be running at the proper speed for DDR2 800 memory. There would be no sense to modify the memory controller to handle DDR2 if they did not do that

Anonymous said...

cool and quiet was disabled. Bios was the latest which supported FX-60. It was an FX-60 overclocked to emulate an FX-62.

Anonymous said...

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2716

Conroe Performance Preview Follow-Up

Anonymous said...

Intel Magic show is all about share price - they are sweating bullets and hanging on at 20.00 by a thread.
Come early April (quarterly reports) when its official that they have lost the desktop market big time (Its going to take more than a trick to make people forget the AMD name) no one will remember this rabbit in the hat stunt

Rahul Sood said...

Wed Mar 08, 10:32:44 PM
Anonymous said...

Anandtech never stated that they were comparing it against a FX-60. They overclocked a X2 4800+ TO FX-60 speeds.


Impossible, you can't change the multiplier on a 4800+. They were comparing an FX-60.

Wed Mar 08, 10:42:05 PM
Anonymous said...

And what price would the intel chip actually be compared to the AMD?


Who knows...You are talking about 6 months from now, and we don't know what's going to happen until then.

Not to mention by the time it actually appears - AMD could be up to an FX-64 or 66!

Maybe, maybe not. The point is it's 6 months out, so until then we can't assume anything other than Intel has woken up and Conroe looks to be bad-ass. AM2 also looks to be great, so I think 2006 is shaping up to be the most interesting year ever in the CPU space.

Thu Mar 09, 04:13:48 AM
Anonymous said...

cool and quiet was disabled. Bios was the latest which supported FX-60. It was an FX-60 overclocked to emulate an FX-62.


Yeah, I saw Anandtech's follow up article. It's good to see the bios was updated, but I would still never let a machine ship with a major bios upgrade like that unless we re-installed the OS. I imagine we could close the gap further - maybe not a huge amount, but enough to make a slight difference. This is what we do for a living, so I can't wait to get Conroe in house as well.

It's good to see Anandtech followed up changing the claim of 20-41% to 20% overall as a rough estimate. This means RD580 still has a bit of play in some of the numbers, then AM2, and evolutionary clock speed changes. At the end of the day I would say that AMD will close the gap further.

Intel Conroe Extreme Edition will certainly start the buzz on the AMD engineering side. I'm sure they are now aware of what to expect.

Anonymous said...

AMD fans are really in denial mode. Believing that BIOS update can close 20-30% gap is not wise. And to Intel's credit i have to say that we are comparing a prerelease system with mature S939 platform. Maturity that brings a lot things like video drivers optimised for processor. It's not like Intel will sit during the next several month. As assembly developer myself i'd say that 1 cycle SSE instructions require a bit different approach. ( It actually depends on many yet undisclosed things, but there is definately quite some extra performance to be extracted. It might be as easy as recompiling with new ICC for some apps, or hand tuning as is the case with graphics drivers)

I have a lot of faith in AM2 myself, but seeing 20-30% deficit i am staying sceptical about AM2 ( Yup, sceptical about AM2 and not about Conroe). DDR2 even at 800 is not the best thing for Athlon 64 architecture - if you were following fan sites low latency DDR memory was beating memory running with relaxed timing at greater speed. With integrated memory controller memory latency is a king!

Talking about memory, i think Intel has a lot of room to improve performance with 1333 FSB and 666/800 DDR2 memory. And i am 100% certain that once Conroe is out we will see the return of high FSB overclocking. While AMD can probably release 3.0 CPU on 90NM process, they can't match Intel entusiasts overclocking their systems to huge FSB ( read 1300-1600 range) and still matching their bandwith with CPU using DDR2 800-1000.

Anonymous said...

When have you seen a benchmark of CPU manufactures that will be a valid one?

NEVER!!

and... this issue is more than evident when Intel is behind of that... past years, Intel has lived of "Marchitecture" instead of Netburst processor.

On the other hand, I have seen that Anandtech article is biased to Intel, at least, he could emphasize more that benchmark is NOT accurate, it is only a marketing show...

I dont have doubts that future Conroe will be quicker on clock per clock basis due to 4-issue and optimized and revisted arquitecture, in addition...up to 4MB of L2! (some games will fly with this cache!) but it is the FUTURE, and the performance could be up to 20% or even less, but for that times, AMD will have new chips out there so... as always, I imagine that AMD will win some tests and Intel others.

I have to thanks to AMD since a few years ago, I can buy a good CPU(Intel or AMD) for a reasonable amount of money due to "sane" competition between them, and I hope this will occur in the future :)

RPP

Rahul Sood said...

Anonymous said...
AMD fans are really in denial mode. Believing that BIOS update can close 20-30% gap is not wise.


I totally agree - with a simple bios update we won't see huge gains, but with a chipset upgrade, and a better builder we'll see the gap close further. You also have to factor in the time - AMD is not sitting still, you don't know what they're releasing other than the rumours that are flying around. AMD didn't bother to showcase technology 6 months out because they aren't in the same position as Intel. Intel probably felt this would deflect their negative results I think.

And to Intel's credit i have to say that we are comparing a prerelease system with mature S939 platform. Maturity that brings a lot things like video drivers optimised for processor.

Agreed - but the same goes to AMD's credit in the sense that you are talking about a part that isn't released - we don't know what type of yields we will see, we don't know what the thermal issues are dealing with a high density part like this, and we don't know exactly when it's coming out.

It's not like Intel will sit during the next several month. As assembly developer myself i'd say that 1 cycle SSE instructions require a bit different approach. ( It actually depends on many yet undisclosed things, but there is definately quite some extra performance to be extracted. It might be as easy as recompiling with new ICC for some apps, or hand tuning as is the case with graphics drivers)

Yes, there are room for optimizations on both sides (next generation AM2).

I have a lot of faith in AM2 myself, but seeing 20-30% deficit i am staying sceptical about AM2 ( Yup, sceptical about AM2 and not about Conroe). DDR2 even at 800 is not the best thing for Athlon 64 architecture - if you were following fan sites low latency DDR memory was beating memory running with relaxed timing at greater speed. With integrated memory controller memory latency is a king!

You are correct, memory latency is a huge focus at this point, and although I have faith in AM2 I think AMD realizes that Conroe looks to be a strong part.

Talking about memory, i think Intel has a lot of room to improve performance with 1333 FSB and 666/800 DDR2 memory. And i am 100% certain that once Conroe is out we will see the return of high FSB overclocking. While AMD can probably release 3.0 CPU on 90NM process, they can't match Intel entusiasts overclocking their systems to huge FSB ( read 1300-1600 range) and still matching their bandwith with CPU using DDR2 800-1000.

Do keep in mind there will be complex cooling issues with a high density 65 nm component. Although I'm sure we'll be able to overcome it instantly with liquid cooling other OEMS may have challenges. In any case, you are correct the higher FSB could be interesting, we are still talking months away and speculating at best.

Anonymous said...

AMD should bump the FSB to 400Mhz...which is 200Mhz right now and match it with an 800MHz DDR II.

that should thrash Conroe


Then you haven't seen any benchmarks at all, then.

The Athlon64 is far from a bandwidth starved processor as it is. There are plenty of single-channel vs. dual-channel memory benchmarks out there. The only ones that show over a 10% performance increase are synthetics (worthless) and very memory intensive computational benchmarks. Most (real) benchmarks show less than 5%increase. Even DDR-333 vs. DDR-400 in single-channel mode shows only marginal gains (tested on my own single-channel S754 Athlon64) for 'real' programs (performance testing on my work).

So, we have a processor that isn't bandwidth limited on current memory technology (DDR) going to DDR2. The DDR2-800 memory has more bandwidth and probably equal or worse latency. After all, the entire AMD fanboi community laughed at Intel for going to DDR2 because of the increased latency. I predict at most a 10% increase in performance in general usage (not synthetics or linpack types) given by DDR2-800.

Other than that, the only technologies that AMD has announced have been F, G, H, and L revisions due out spread out over the next year. We know that DDR2 support is in those, and virtualization (no effect unless you are running virtual machines) and an increased HT speed. That increased HT speed is probably a part of the DDR2 support so is inclusive to those gains. The only other real information that we have is that the L stepping will have 2x the number of FPUs for an expected 50% increase in performance in FPU heavy code. This will help games but not nearly by 50% and these cores will probably be Opteron only anyway.

The fact is that AMD hasn't announced anything, even in combinations, that will add up to 30%+ performance increases for the desktop. I think we'll see 10%. I'll even give an outside 15% for the desktop parts.

Now, Conroe was tested using DDR2-667 and at 2.67GHz. All Conroe parts have a faster FSB than that so expect production systems to have faster memory (although I doubt this will give a big performance increase). The other thing to remember is that Conroe will release at 3GHz top speeds and there are strong rumors of a 3.33GHz part but not officially announced. Speculation that if things are as they seem, AMD had better be ready to release 3.6+GHz parts in July (when Conroe will be out) to stay ahead. But, the only official AMD announcement of speed bumps are to 2.8GHz.

Given that by the time AMD is releasing 65nm parts (and probably not many), Intel will be releasing 45nm parts which will give another speed bump.

I buy what makes sense. I have seven AMD machines right now (three Athlon64, one Athlon64x2, the rest Athlon XPs - and only two Intel machines and they are both Dell laptops) but I wouldn't hesitate to upgrade to Intel if they were better at the time I plan to buy. AMD had better have a rabbit in the hat that would have to be a very well kept secret (with all the leaks we have these days) and have a plan to compete or they will be a year behind very quickly.

JohnnyBeGood said...

Some valid questions brought up rahul, and apparently anandtech had the exact same questions because they've put up an update that answers some of the exact things you were asking.

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2716

The interesting thing I find is that, even with the updated bios/etc, the performance gains were almost negligable. Include in the fact that apparently intels own memory timing was off and they got a slight boost, well....

And from Tom's own preview article about the AM2, the performance increase seems to be pretty negligable to say the least.

While we all still have to wait till the systems finally get into unbias journalists, I'd say that intel really has a shot at the big leagues!

Anonymous said...

I used to work in the procurement side of a bank. You won't believe the extent intel will push to get their kit on your floor. On one occassion they came with two server rigs to show us how it superior cpu's, 2 amd opteron 2xx series and their own rig had 2 Xeons with 16MB of cache each. (This was early last year by the way)

Anonymous said...

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2716&p=1\


I have one question for you. Do you feel like an idiot at all?

Rahul Sood said...

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2716&p=1\

I have one question for you. Do you feel like an idiot at all?


Why should I feel like an idiot? I was right - the bios was wrong, and I still believe there are tweaks that could be made to close the gap. I am also happy to see that Anand realized the mistake on FEAR because that benchmark was way off. Read my comments before you make such idiotic posts.

Anonymous said...

We can't give any credit to this test for the following:

- Testing a chip without being sure this is the piece of hardware that will go on sale

- You can't make it that way. Supply the rigs and give no choice of using own's hardware

- Testing months ahead Conroe against current processor is a waste of time

- No official audit was done on the hardware like AMD uses to do

Now tell me why Intel did not answer AMD challenge if the company was so sure of its coming supremacy?

Sounds all like another laùe attempt to keep people waiting for some kind of paperware just like the company is doing since 3 years already

Anonymous said...

"I was right..."

Uhum.

"I certainly expect Intel to be back on the rails, but I don’t see them to be wiping the floor with anyone other than their Pentium 4 team come launch time."

Based on what? A64 is not bandwidth-starved.

"We can speculate all we want, but why bother?..."

Indeed... but let's do it anyway, right?


"...By simply changing the bios and ensuring that Cool & Quiet is disabled, that the FX-60 is fully supported, and that there are no memory timing issues I imagine we would see a fairly significant performance gain on those merits alone.

Wrong.

"If you read the article you'll see that the AMD system is clearly crutched.."

Really?

Rahul Sood said...

Anonymous said...

Uhum.

"I certainly expect Intel to be back on the rails, but I don’t see them to be wiping the floor with anyone other than their Pentium 4 team come launch time."


I still don't. Wiping the floor means they will absolutely kill AMD on the performance side and I don't see this happening. I see them catching up fast - and there is certainly a chance that they might get ahead, but not like AMD has been for the last few years. You have to remember that Intel is the won getting killed at the moment on the desktop, server, and workstation side. They have huge ground to gain and none to lose at the moment.

Even if it turns out that AMD is slightly behind in performance in 6 months it won't matter. They will have already taken much of the market that they intended to take - and that will give them enough momentum to carry them to their next core.

"We can speculate all we want, but why bother?..."
Indeed... but let's do it anyway, right?


I see your point, and while there will be some speculation I'm still not willing to make a final assessment until I have the product in my lab. It would be wrong of me to assume a performance lead either way on product released 6 months from now. I do know that AMD has a few tricks up their sleeve and whether or not that will lead to big gains is anyones guess.

"...By simply changing the bios and ensuring that Cool & Quiet is disabled, that the FX-60 is fully supported, and that there are no memory timing issues I imagine we would see a fairly significant performance gain on those merits alone.

Wrong.


Well, that's technically not true. If you bought a machine from me and it had that bios on it you would know that it wasn't properly tweaked to begin with. My technicians would reinstall Windows after doing that type of bios update.

As well when I said the AMD system was crutched, I believe using an RD480 is a crutch - ATi would agree with me compared to RD580. Yeah it wouldn't make massive gains, but it would close the gap from 20% further.

"If you read the article you'll see that the AMD system is clearly crutched.."

Really?


Yes, really. :) I would love to build a machine to go up against the one Intel build for AMD, and watch the difference. We would obviously do our own tweaking as I'm sure they did with theirs, we would close the gap though we still wouldn't win. We are talking about unreleased technology afterall.

It was also good that Anand spotted the FEAR error because we had replicated a lesser system yesterday and beat the scores he was getting so that would have come into question as well.

Sharikou, Ph. D said...

As I wrote here, INTEL has the burden of proof. And I have solid proof there that Woodcrest (same as Conroe) will be 10-20% slower than Opteron 280.

Anonymous said...

Hi Rahul,

Thanks for all your answers and comments.

What people forget are a couple of things:

1. 65nm Intel process vs. 90nm AMD process. This to me is the most amazing part of all of this...Intel had to rush to their 65nm process to meet (or slightly exceed) AMD's current 90nm process offerings. I would imagine there will be no question who will own the performance crown once AMD transitions to their 65nm process (assuming they can do this cleanly)???
2. Intel's quad-core is really two dual-core processors...not a true quad-core processor!!! Again, Intel is taking a short-cut of "reacting to AMD" than providing a clean solution. I would imagine that AMD will be releasing a true quad-core processor, although there has been no word on this.
3. From what I've read below, so what if Intel's Conroe is approximately 20% better than AMD's current processor today (according to Anand's benchmark)...If Intel couldn't produce a better chip in the future (6 months out), the company would continue to lose ground. Intel has shown us their cards, now the only question is what does AMD have up their sleeves???

I'm Looking forward to more of your blogs. :)

Regards,
Mike Lee

Chandler said...

Great stuff, all of you. The ammount of information held in this discussion is mind boggling, you all should be very proud of your accomplishments and the knowledge you hold. Keep it coming!

Rahul Sood said...

At Thu Mar 09, 11:23:21 AM, Anonymous said...

Hi Rahul,

Thanks for all your answers and comments.


No problem.

1. 65nm Intel process vs. 90nm AMD process. This to me is the most amazing part of all of this...Intel had to rush to their 65nm process to meet (or slightly exceed) AMD's current 90nm process offerings. I would imagine there will be no question who will own the performance crown once AMD transitions to their 65nm process (assuming they can do this cleanly)???

Well, the next core is probably the key for AMD. If we look at this seriously and we assume that Conroe will get a performance lead - we still have over 3 years+ of AMD being on top of the performance foodchain and this year they have the mainstream momentum to carry them beyond the next 6 months.

That means AMD may lose an edge on performance but in no way will they lose their momentum, it's impossible. Their momentum should carry them through to 65 nm process - but then we hope they will be able to keep up with demand.

2. Intel's quad-core is really two dual-core processors...not a true quad-core processor!!! Again, Intel is taking a short-cut of "reacting to AMD" than providing a clean solution. I would imagine that AMD will be releasing a true quad-core processor, although there has been no word on this.

Doesn't matter how you slice it - Intel can duct tape two dual core processors on a processor and call it quad core - because it is quad core. At the end of the day this is a big deal no matter what. I don't think we can compare it to their EMT 64 implementation, it's a threat that AMD needs to acknowledge.

3. From what I've read below, so what if Intel's Conroe is approximately 20% better than AMD's current processor today (according to Anand's benchmark)...If Intel couldn't produce a better chip in the future (6 months out), the company would continue to lose ground. Intel has shown us their cards, now the only question is what does AMD have up their sleeves???

AMD has quite bit going for them. They have a new architecture on the way that's easy to adopt, they have huge OEM momentum, they have retail momentum, and they have their server business. AMD doesn't need to show their cards right now to be perfectly honest, their cards are already there for the people to see. Wait until their next quarterly results - they will bust a cap in Wall Street Analysts *yet again*.

I'm Looking forward to more of your blogs. :)

Thanks.

At Thu Mar 09, 11:52:17 AM, Chandler said...

Great stuff, all of you. The ammount of information held in this discussion is mind boggling, you all should be very proud of your accomplishments and the knowledge you hold. Keep it coming!


Thanks Chandler, there are certainly a ton of great comments being posted here. Thanks for your comments!

Anonymous said...

Don't ever trust benchmarks from AMD, INTEL, ATI or NVIDIA.

Simple as that.

A_Pickle said...

As a self admitted Intel fan, before I compose this I'd like to say to AMD, ya done good. Competition is always a welcome thing, and frankly Intel deserved to be kicked around for a little bit for pulling that whole "Prescott" thing.

As noted, Anandtech has released a revised Conroe review (with Intel's participation) taking into account your criticisms. It's much, much more rewarding to see Conroe doing so well with "foul play" effectively ruled out of the question. That would not have been made possible without your insight, and so, as usual, excellent job.

Your laptops look wonderful, I might add :).

-Pikl

PS: Sharikou... is a moron. I'm sorry, I just had to get that out.

Anonymous said...

In reading the various articles on the Conroe vs. Athlon FX duel, It is interesting to note that they (intel) didn't try and compare Conroe to the Opteron 180/185 processors.

Opterons would be a closer comparison since they have a bigger cache (1 mb X 2) and apparently a better memory controller.

As everybody knows, the opterons have been scaling up very easily with 3 ghz not uncommon.

I was alway curious why the opteron 1xx series, 939 was launched by AMD in the first place. Originally I thought that it was a bonehead move, ... why compete with your own best and most expensive chips (ie FX series)??. In hindsight, I suspect that this was an AMD "beta test market" to see how well the enthusiast could push the chips.

AMD has always had cards up their sleeves before (ie - dual core opterons) and nobody seems to have realized that the opteron 1xxx would probably meet or exceed the current conroe specs....maybe.

Just sharing my thoughts... you have an interesting blog

David

Rahul Sood said...

A_Pickle said...

As a self admitted Intel fan, before I compose this I'd like to say to AMD, ya done good. Competition is always a welcome thing, and frankly Intel deserved to be kicked around for a little bit for pulling that whole "Prescott" thing.


Here here...

As noted, Anandtech has released a revised Conroe review (with Intel's participation) taking into account your criticisms. It's much, much more rewarding to see Conroe doing so well with "foul play" effectively ruled out of the question. That would not have been made possible without your insight, and so, as usual, excellent job.

Agreed, and I'm happy to see the FEAR benchmark was misquoted, so at least we all have an idea of what to expect.

Your laptops look wonderful, I might add :).

Thank you!

Thu Mar 09, 12:15:09 PM
Anonymous said...

In reading the various articles on the Conroe vs. Athlon FX duel, It is interesting to note that they (intel) didn't try and compare Conroe to the Opteron 180/185 processors.

Opterons would be a closer comparison since they have a bigger cache (1 mb X 2) and apparently a better memory controller.


Intel is comparing their future desktop part with the best they can get from AMD on the desktop side. I don't blame them for this.

As everybody knows, the opterons have been scaling up very easily with 3 ghz not uncommon.

I was alway curious why the opteron 1xx series, 939 was launched by AMD in the first place. Originally I thought that it was a bonehead move, ... why compete with your own best and most expensive chips (ie FX series)??.


Hmm, interesting point to note no doubt.

AMD has always had cards up their sleeves before

...yes they have.

Just sharing my thoughts... you have an interesting blog

Thank you for the comments.

Anonymous said...

Rahul, will you never "sell" an Intel system? I mean AMD is great with their processors, but wouldnt you agree that Conroe is a sheer great piece of engineering, and not the sheer piece of marketing dead-end that was Netburst.

Whether or not the 20% advantage now erodes to 5-10% advantage at launch, I think you should be a little more commending on Intel for working out a great piece of silicon, as much as you root your heart out for AMD.

Brand loyalty (Intel/AMD) and fanboyism (Intel/AMD) are two things that prevent true engineering from being recognised.

My personal thoughts: Conroe looks like a (near) perfect Computer-Architecture textbook approach to designing a high-performance CPU. AMD hasnt gone to 65nm yet, but then Intel hasnt yet used a IMC either.

ProperMethodz said...

Thanks Raul,

Your information is always thought provoking.

I'm not all that interested in the next gen processors yet. I feel that right now the processors that are out do the job to handle programs which are currently released. They do the job very well and are able to play games and perform tasks at alarming rates with awesome multi tasking capabilities. Furthermore, I can have Photoshop CS2 open with 100 4 Mega pixel pictures opened, FireFox with roughly 10 tabs (we all know this is a memory hog) and still pull 60 FPS in Unreal Tournament 2004 on my current setup (AMD Athlon 64 3500+, 2GB PQI 2-2-2-5 TCCD, Gigabyte NF3, and a 6800U).

The system I want is an AMD X2 4400+ on the DFI Lan Party With 2 Gigs of Corsair XMS. Throw a 7800 GT in that thing and you've got yourself a machine that will stand up to Vista or any other program out there currently, or for the next two years for that matter.

The only reason I want the new processors to come to fruition is to see the current technology's price to drop.

Now some of the latest posts on this here blog have mentioned that intel is doing the 65nm process, where AMD is still on the 90nm...

I'd like to point out that a lot of people have already forgotten AMD's new relationship with IBM which "wipes the floor" with Intel in terms of capital. With Intel just dropping the bricks on their 45nm processes... AMD is working on the 32/22nm processes and is backed by IBM.

http://www.theregister.co.uk/2005/11/01/amd_ibm_chip_deal/

Rahul Sood said...

Rahul, will you never "sell" an Intel system? I mean AMD is great with their processors, but wouldnt you agree that Conroe is a sheer great piece of engineering, and not the sheer piece of marketing dead-end that was Netburst.

We currently sell Intel systems. We love their notebook architecture, so we support what we feel is best. I would absolutely sell an Intel based desktop if it was a smart buying decision. I believe the Intel Centrino is a killer platform and that's why we support it.

Our loyalty with AMD is strong, we believe in them as a company and we've had a relationship with them for years (since around 1993). In good times or in bad AMD has always understood the enthusiast market better than their competition. Believe me, even if Conroe is faster, AMD will still have big momentum in so many markets - that momentum will carry them to their next core.

Whether or not the 20% advantage now erodes to 5-10% advantage at launch, I think you should be a little more commending on Intel for working out a great piece of silicon, as much as you root your heart out for AMD.

I hear what you're saying, but try and understand we haven't seen a good desktop part from Intel in a very very long time. I am looking forward to Conroe and I will gladly give them the props if it's deserving.

Brand loyalty (Intel/AMD) and fanboyism (Intel/AMD) are two things that prevent true engineering from being recognised.

I agree, don't let my positive comments on anyone make you think I'm a fanboy. People thought that of me with Nvidia - and when R580 came out I was all over ATi Crossfire. I will actively give props when I see a good technology. I put ATi through the wringer last year - and you know what? They listened to everything I said, and they didn't cry about the stuff I wrote. They agreed with all of it - and they made changes for the better.

My personal thoughts: Conroe looks like a (near) perfect Computer-Architecture textbook approach to designing a high-performance CPU. AMD hasnt gone to 65nm yet, but then Intel hasnt yet used a IMC either.

Yep, it certainly looks good from a distance - no doubt.

Anonymous said...

I am curious why an overclocked FX60 cannot out perform a stock one?
http://www.bit-tech.net/news/2006/03/08/intel_conroe_benchmark_fear/

Anonymous said...

"AMD had better have a rabbit in the hat that would have to be a very well kept secret (with all the leaks we have these days) and have a plan to compete or they will be a year behind very quickly."

There's one reason why I don't believe AMD has a short term counter to Conroe: Financial disclosure laws.

The SEC demands tha any financially relevent data be disclosed to the public. Considering how long it takes to spin silicon, AMD would have to have been working on a new core/release/architecture for about since months now. Since none of AMD's publically available roadmaps contain this information, one must assume that there cannot be a major response this year to Conroe.

Assuming of course that AMD doesn't want to get sued by every state and every shareholder that sold stock in the last six months.

Anonymous said...

People what you are not getting with your comments of "closing the gap" is, that Conroe E6700 is only midrange, and running 200mhz slower than the FX60 2.8 with a slower 1066 FSB. Just wait till the 3.33 EE comes out with the 1333 FSB LOL. AMD is in BIG trouble. Just take it on the chin and admit it :-), give credit where credit is due, instead of making excuses.
Anand has followed up with the "complaints" of the AMD fanboys with the Intel testing, time to get over it and wait a couple of years till AMD catch up again.

Anonymous said...

The second those benchmarks were posted, my first concern was: "are those FX-60 numbers consistent with real world performance?" I am fairly certain that if Intel knew how to tweak software performance to benefit Conroe, they'd find a way to put it on the chip. However, it's fairly clear that the answer to my first question is a big, fat, "NO!"

Heck, my concern wasn't even the software side, it was the fact that Intel tinkered with the AMD processor at all. Since it's physically impossible for somebody from Intel to replicate the performance of the FX-60, not only because nobody at Intel actually knows what that is but because there are some very large differences between the FX-60 and FX-62. Because of that the test is invalid by any objective standard.

Also, I am concerned that nobody from the press got to see what was under the hood, so to speak. For all we know, Intel resorted to physical means to handicap the FX-60, beyond the might-as-well-be-trash RD480 chipset. Stock cooling fan for an overclocked part, inferior thermal paste, low-performance RAM, maybe even no CPU fan at all. While I'm no system builder, I'm sure any one of those could adversely affect the performance of the AMD part used in the test.

Anonymous said...

My question is, Why didn't they just compare it to one of their own existing processors? Why even try to compare to an AMD chip with questions about the setup for the competition? Just my two cents as for a reason is why would you show your own hardware as incompetant and why give the competition a exact performance comparision.
Can you imagine comparing a Conroe chip to say their top Pentium Dual chip? In my opinion they would either fudge the numbers for one chip or the other, hard for them to say their old processors suck since they have to sell them for 6 more months or show marginal improvements over the old chips so that people would see that the new chips weren't worth getting.

RotoSequence said...

Well Rahul, the question is this; will Voodoo be making a badass machine to cater to the ultra high end, if Conroe turns into the current generation Athlon 64 (and potentially next gen) killer that it appears to be shaping up to be, that contains this processor?

Voodoo shoots for the best of the best, and strong ties aside, Im betting you'd go with the winner ;)

Anonymous said...

This didn't even help their stock price today granted it hurt AMD a little. One thing is for sure it sure is not going to help Intels 1st quarter report. When you think of the people that now know about AMD that didn't a year ago (the desk & lap top market). It seems to me that AMD has already won the race - It is for all intensive purposes a equal competitor now (with alot less overhead).
Oh yea think about all the extra bucks you have paid through the years because of intels business practices.

Those customers are going to be real hard to get back.

Sharikou, Ph. D said...

It's amusing to see the IDF turned into a truth searching event for INTEL's questionable integrity or lack thereof. Even INQ has spoken on this. http://www.theinquirer.net/?article=30177

I have clear and convincing proof that Woodcrest will be 10-20% slower than Opteron 280. No one, not even Intel folks can challenge that. INTEL failed to meet the minimum requirement to establish the minium credibility for Conroe(which has the same core as Wood): allow Anand to take the Conroe out for a drill.

INTEL fan boys can say whatever they like. The fact INTEL is pouring $10 billion into Itanium is further indication that NGMA won't do any wonders.

I projected 15-20% 1Q06 revenue fall for INTEL, that's been validated. I projected 3Q06 operating loss for INTEL. Mark my word.

Game over for INTEL.

sreeram said...

Hi Rahul,
I read a post in intel forum at hardforum.com where the poster is commenting on reasons behind ur response to conroe. Have you any comments to counter this?

" Rahul has everything to lose. He's a little shop who has built his entire profit margin model across the line around AMD. Over 75% of his gross sales on a monthly basis are built around AMD-equipped machines. He's been loving the healthy margins he's been allowed to cultivate due to Dell's decision not to carry AMD processors. Good for him, I applaud him for cornering a market, as did all the other "performance" builders. And he knows what the rest of us in the industry know: A strong Dell means a weaker everyone else...including VoodooPC. Hence why he also posted about the potential Dell/Alienware deal a week ago with similarly negative comments. Oh and as well as his blog entry about how "offended" he was not be offerend Quad-SLi initially by nVidia...as Dell was.

Rahul is very worried. And I would be too in his shoes.

Hopefully he can make his model work at much lower profit margins, as that's what he'll be forced to do as the entire white-box world sells Conroe in large quantities. If not, he'll have to move to another space, perhaps home theater media PCs which he has an interest in.

Not a conspiracy theory in the least, and you labelling my comments as such is naive. More a man trying to protect a niche he carved out in a very combative sector over the past three years. Again, I'd do the same. I'd spin more than a top...AMD will respond, AMD won't let this go unchecked, don't worry, AMD will be back on top for sure, just wait kids, I know things that you don't, etc, etc, etc, all of which he stated in that blog.

Desperation.

I wish Rahul the best, I'm sure he'll be fine. He will just have to transition the bulk of his sales to a supplier he hasn't supported in a large way previously via Intel. And he'll have to rework his base profit margin percentage to be something more realistic than the egregious levels he charges currently. He's not alone. They all will. From Alienware, to Falcon, to GamePC, to all the others. It's going to be a Conroe world soon, and we're all just livin' in it."

Sharikou, Ph. D said...

Why didn't Anand report his, but instead wrote "Intel regains..."?

http://techreport.com/etc/2006q1/conroe/index.x?pg=1

"We used test systems pre-configured by Intel before
the show, and we had very limited time to conduct
testing or inspect the systems. We were not allowed to
look inside of the case of either PC, and the scope of
the benchmarks we were allowed to run was defined by
Intel. We weren't given the leeway to record our own
custom timedemos for the games, and we didn't have
enough time to run each test three times or even
reboot between the tests....

our role really was confined largely to clicking a few
icons and menu items to kick off a test and then
writing down the results."

Rahul Sood said...

Anonymous said...

I am curious why an overclocked FX60 cannot out perform a stock one?
http://www.bit-tech.net/news/2006/03/08/intel_conroe_benchmark_fear/


Simple, because the stock one wasn't configured with the best possible settings. I assure you a properly overclocked FX-60 would do much better. Anyone who tries to argue that with me should read my credentials first. I am not - in any way - suggesting that it will be 20% faster, but it will certainly be faster than what was shown.


Thu Mar 09, 01:52:40 PM
Anonymous said...

"AMD had better have a rabbit in the hat that would have to be a very well kept secret (with all the leaks we have these days) and have a plan to compete or they will be a year behind very quickly."


...even if AMD is behind on the desktop side they still have big momentum. You cannot stop a freight train. I believe their momentum will carry them to the next core.

There's one reason why I don't believe AMD has a short term counter to Conroe: Financial disclosure laws.

The SEC demands tha any financially relevent data be disclosed to the public. Considering how long it takes to spin silicon, AMD would have to have been working on a new core/release/architecture for about since months now. Since none of AMD's publically available roadmaps contain this information, one must assume that there cannot be a major response this year to Conroe.


I can't comment on this.

Assuming of course that AMD doesn't want to get sued by every state and every shareholder that sold stock in the last six months.

Anyone who sold AMD stock in the last 6 months should probably sue themselves before they think about suing anyone else.

Thu Mar 09, 02:50:33 PM
Anonymous said...

People what you are not getting with your comments of "closing the gap" is, that Conroe E6700 is only midrange, and running 200mhz slower than the FX60 2.8 with a slower 1066 FSB. Just wait till the 3.33 EE comes out with the 1333 FSB LOL. AMD is in BIG trouble. Just take it on the chin and admit it :-), give credit where credit is due, instead of making excuses.


Who is making excuses? Closing the gap means bringing the performance lead closer - not closing it to zero. I am not suggesting a current FX-60 can beat a Conroe that's 6 months out! Where are you guys reading this stuff? As far as the "1333MHz FSB LOL!" goes you won't see a large difference in these particular benchmarks - this will effect graphic intensive benchmarks at higher resolutions.

Anand has followed up with the "complaints" of the AMD fanboys with the Intel testing, time to get over it and wait a couple of years till AMD catch up again.

Sir/Ma'am, you really need to do your research. AMD has been way ahead for years - anyone expecting Intel to stay behind forever needs to give their head a shake. Intel is WAY behind at the moment on desktops, servers, and workstations - their only recourse is to show what they have 6 months from now - at that point AMD won't be left in the dust - they will still have huge momentum from the mainstream. You have to understand this concept.

Thu Mar 09, 03:21:25 PM
Anonymous said...

The second those benchmarks were posted, my first concern was: "are those FX-60 numbers consistent with real world performance?" I am fairly certain that if Intel knew how to tweak software performance to benefit Conroe, they'd find a way to put it on the chip. However, it's fairly clear that the answer to my first question is a big, fat, "NO!"


Perhaps, but do realize that Conroe is still fast, those numbers aren't made up - they're everywhere. AMD needs to prepare for it and I'm sure they know this.

Heck, my concern wasn't even the software side, it was the fact that Intel tinkered with the AMD processor at all. Since it's physically impossible for somebody from Intel to replicate the performance of the FX-60, not only because nobody at Intel actually knows what that is but because there are some very large differences between the FX-60 and FX-62. Because of that the test is invalid by any objective standard.

They didn't do a good job tweaking it - I can get much better performance from an FX-60 than what Intel got on their FX-60 - I totally stand by that claim.

Also, I am concerned that nobody from the press got to see what was under the hood, so to speak. For all we know, Intel resorted to physical means to handicap the FX-60, beyond the might-as-well-be-trash RD480 chipset. Stock cooling fan for an overclocked part, inferior thermal paste, low-performance RAM, maybe even no CPU fan at all. While I'm no system builder,

Yes, you aren't a system builder I can tell :) - but conspiracy theories are out here. Intel didn't do this - they just didn't spend the time to make the PC run optimally. I can't blame them for this, I mean why would they bother? They are trying to showcase their technology in the best light possible. Believe me Intel wouldn't risk doing something so stupid, we're all going to have these products in our labs soon.

Thu Mar 09, 03:38:22 PM
Anonymous said...

My question is, Why didn't they just compare it to one of their own existing processors?


Now that's a great question - the answer is simple though - they need to do what they can to stunt AMD's momentum. I believe AMD is going to have a blockbuster quarter this time around - mark my words.

Can you imagine comparing a Conroe chip to say their top Pentium Dual chip?

There's no comparison - why does Intel want to rub salt in their own wound? The P4 engineers have heard it enough already I'm sure.

Thu Mar 09, 04:17:13 PM
RotoSequence said...

Well Rahul, the question is this; will Voodoo be making a badass machine to cater to the ultra high end, if Conroe turns into the current generation Athlon 64 (and potentially next gen) killer that it appears to be shaping up to be, that contains this processor?


The answer is this: We will continue to sell both AMD and Intel as we do now. If Conroe rocks, we will certainly sell it. At the same time our flagship systems will still have AMD and Intel no doubt.

Thu Mar 09, 04:28:07 PM
Anonymous said...

This didn't even help their stock price today granted it hurt AMD a little. One thing is for sure it sure is not going to help Intels 1st quarter report. When you think of the people that now know about AMD that didn't a year ago (the desk & lap top market). It seems to me that AMD has already won the race - It is for all intensive purposes a equal competitor now (with alot less overhead).


You cannot stop a freight train - no doubt. AMD has big momentum, people selling the stock just don't understand this business at all. Speculating based on technology that's 6 months out - technology that isn't going to be available in volume (20% of Intel's high end will be Conroe by the end of the year in my estimation) - not to mention the fact that AMD has momentum. Investors are too emotional.

Oh yea think about all the extra bucks you have paid through the years because of intels business practices.

Well, regardless - we pay bucks for all our high end stuff :)

Those customers are going to be real hard to get back.

Very true, on the mainstream momentum front it's going to take time for Intel to rebuild.

Thu Mar 09, 05:23:34 PM
sreeram said...

Hi Rahul,
I read a post in intel forum at hardforum.com where the poster is commenting on reasons behind ur response to conroe. Have you any comments to counter this?


I'll see :)

" Rahul has everything to lose. He's a little shop who has built his entire profit margin model across the line around AMD. Over 75% of his gross sales on a monthly basis are built around AMD-equipped machines.

On the desktop and workstation side it's more like 95%. On the notebook side it's around 22% now.

He's been loving the healthy margins he's been allowed to cultivate due to Dell's decision not to carry AMD processors.

Our healthy business has nothing to do with Dell - it has to do with building powerful personal authentic PCs, offering great service, and fantastic technjology.

I applaud him for cornering a market, as did all the other "performance" builders.

Thank (him).

And he knows what the rest of us in the industry know: A strong Dell means a weaker everyone else...

A strong Dell means a strong industry. Dell will create more interest in our space as they have in the past.

Hence why he also posted about the potential Dell/Alienware deal a week ago with similarly negative comments.

How were my comments negative? I feel there is a good chance Dell will buy Alienware. How is that negative? If it comes true is it a positive comment?

Oh and as well as his blog entry about how "offended" he was not be offerend Quad-SLi initially by nVidia...as Dell was.

We're actually a launch partner with Quad SLI - we have stock of the 7900 GTX QUAD SLI now. Nvidia obviously appreciates the business we do with them.

Rahul is very worried. And I would be too in his shoes.

Worried about Intel Conroe being better than AMD's next generation? No I'm not worried even if this is the case.

AMD will respond, AMD won't let this go unchecked, don't worry, AMD will be back on top for sure, just wait kids, I know things that you don't, etc, etc, etc, all of which he stated in that blog.

AMD will respond, no doubt. ..and I truly don't believe this is a "dire" position as many people are trying to make it out to be. They seem to be forgetting the dire position their competition is in for the next few quarters.

It's going to be a Conroe world soon, and we're all just livin' in it.

???

Anonymous said...

Hi, Rahul.

According to this link, AMD's current list price for an FX-60 is $1,031 (in 1000-unit quantities).
http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_609,00.html

And according to this link (which repeats other rumors), the Conroe 2.67GHz is rumored to sell in July for $529, and the Conroe 2.40GHz for $315.
http://www.mikeshardware.co.uk/RoadmapQ306.htm

Based on this, if a Conroe 2.67GHz were just equal in performance to an FX-60, the FX-60 would need a price cut of some 49% just to match Conroe price-performance.

And if a Conroe 2.40GHx were equal in performance to an FX-60, then the FX-60 would need a price cut of some 69% just to match Conroe price-performance.

Now I know that you don't like to speculate, but please make an exception here. Do you agree that FX-60 price cuts of such magnitudes are coming in a matter of a few months? If so, what do you venture the FX-60 will be selling for when, say, Conroe 2.40GHz becomes available in volume?

Will AMD be able to command a price premium for FX-60, or will it need to discount it, on a price-performance basis? If so, what will be the price-performance, relative to Intel, of AMD processors in the Conroe 2.40GHz range?

Also, do you expect Conroe 2.40GHz to become available in volume in the July timeframe? If not, what is your estimate?

Anonymous said...

I'll tell you why their is a difference between IDF system score and the reviewer's system.

Look here: http://www.anandtech.com/showdoc.aspx?i=1810&p=9

My friend has a 2.4C overclocked to 3.2C using P4P800 with Samsung PC3200 memory at default timings and Radeon 9800XT, Segate Barracuda 7200.7 with SATA 150 without NCQ.

We tested out with the setting in Quake 3 called "fastest". The results?? Around 280 frames per second. Could it be that "fastest" settings with everything low and 512x384 settings be getting lower frames per second then 1024x768 results the review sites got?? I highly doubt it.

Are the review sites flawed?? How do they achieve such high fps with better graphics and same system?? BIOS and driver make that much of a difference?? RAM make that much of a difference??? Probably, but it shows that such a huge difference is made that comparing different sites are not favorable.

If you want X2 4800+ with Crossfire X1900XT benchmarks that seems close to IDF system plus or minus clock speed difference, take a look, here: http://www.tbreak.com/reviews/printpage.php?id=430

Anonymous said...

I think these parts will bring much needed relief to Intel's deteriorating quarterly results. Q1 looks bad, Q2 will probably be worse. Core Duo would be of some help here, but not much. Q3 would probably be okay, while Q4 would be best for them.

On AMD's side, I dont think they are going to have bad quarters at all this year. But they will have less bone-headed analysts and less rave reviews on their side, which means stock may become stagnant later on. I dont know how much would revenue slow down because they have their second fab ramping up, and should sell lots of products.

On 65nm, probably just Opteron and Turion would move to that node by mid-to-end 2006. I think Turion DC would be 65nm.

But really their next desktop 65nm part is Brisbane, which is due only in 2007. So you will have 90nm AM2 vs 65nm Conroe for some time in Q3/Q4.

I read that PC desktop sales are slowing down, but for many of us I think Conroe has inspired us somewhat to remain with the PC.

Anonymous said...

Some questions:

1) Is Conroe a true 64-bit chip like the AMD chips are? If so, how do you think it will perform with 64-bit Vista and the new 64-bit capable games coming out next year?

2) Does Intel have a tradition of having yield and volume issues when they move to a new manufacturing process like 65nm with Conroe?

3) Rahul, are you worried about your AMD/Dell bet now that the Conroe benchmarks are out?

4) What has been the average cycle for an AMD new core to come out (4 years)? And how long as the current AMD core been out?

5) If the AMD chips had a huge 4MB cache like Conroe, would that improve performance?

Rahul Sood said...

More comments;

Anonymous said...

...Based on this, if a Conroe 2.67GHz were just equal in performance to an FX-60, the FX-60 would need a price cut of some 49% just to match Conroe price-performance.


The AMD FX program usually means they pull the previous versions and bring on the new FX at the same price. I suspect AMD will take FX to entirely new level eventually, and they will still position it as a premium product. The FX brand is all about performance. The FX-60 will be history by the time Conroe comes out.

And if a Conroe 2.40GHx were equal in performance to an FX-60, then the FX-60 would need a price cut of some 69% just to match Conroe price-performance.

FX-60 will be history when Conroe comes out.

Now I know that you don't like to speculate, but please make an exception here. Do you agree that FX-60 price cuts of such magnitudes are coming in a matter of a few months? If so, what do you venture the FX-60 will be selling for when, say, Conroe 2.40GHz becomes available in volume?

I see you are trying to hedge your bets on this - clearly you aren't an enthusiast and probably an investor. As an investor you need to understand the concept of "momentum" in the PC space. It's different from other industries - it takes time to build momentum, and it's hard to stop a freight train. Read some of my other comments above.

Also, do you expect Conroe 2.40GHz to become available in volume in the July timeframe? If not, what is your estimate?

As I am not under NDA at the moment with Intel, I don't expect large volumes of Conroe to show up in July.

Thu Mar 09, 06:51:08 PM
Anonymous said...

If you want X2 4800+ with Crossfire X1900XT benchmarks that seems close to IDF system plus or minus clock speed difference, take a ...


Interesting review - but with all due respect, in the broad scheme of things it means nothing. You don't know what time demos were run, you don't have the same hardware, etc.

Thu Mar 09, 07:08:21 PM
Anonymous said...

I think these parts will bring much needed relief to Intel's deteriorating quarterly results. Q1 looks bad, Q2 will probably be worse. Core Duo would be of some help here, but not much. Q3 would probably be okay, while Q4 would be best for them.


Hmm, I would guess that AMD Q2, Q3, and Q4 will all be amazing.

On AMD's side, I dont think they are going to have bad quarters at all this year. But they will have less bone-headed analysts and less rave reviews on their side, which means stock may become stagnant later on. I dont know how much would revenue slow down because they have their second fab ramping up, and should sell lots of products.

I don't think there will be a revenue slow down. They only have upside on servers, notebooks, desktop, and workstation. When Conroe comes out they will already have 95% of the market covered.

I read that PC desktop sales are slowing down, but for many of us I think Conroe has inspired us somewhat to remain with the PC.

Conroe will boost interest in the market again - as usual that time of year tends to be a busy time.

Thu Mar 09, 08:48:14 PM
Anonymous said...

1) Is Conroe a true 64-bit chip like the AMD chips are? If so, how do you think it will perform with 64-bit Vista and the new 64-bit capable games coming out next year?


Apparently yes, but I haven't seen it so I can't comment yet.

2) Does Intel have a tradition of having yield and volume issues when they move to a new manufacturing process like 65nm with Conroe?

Most people have problems with new processes - but Intel will have plenty of time to work out their issues. I believe they will have yield issues regardless of the process.

3) Rahul, are you worried about your AMD/Dell bet now that the Conroe benchmarks are out?

Are you joking? You won't see me in a bunny fabbit suit.

4) What has been the average cycle for an AMD new core to come out (4 years)? And how long as the current AMD core been out?

They keep making evolutionary changes to their core.

5) If the AMD chips had a huge 4MB cache like Conroe, would that improve performance?

Yes most likely, but it would also increase the amount of silicon required to produce. It also increases complexity pertaining to thermal management, voltage, etc.

Anonymous said...

There was a 2nd round of benchmarks, and this time with a proper BIOS which recognizes the FX-60. Check out http://www.anandtech.com/tradeshows/showdoc.aspx?i=2716 As with previous tests, the Conroe still blazes ahead. As for the DDR2 issue, using the AM2 sample.. it seems that AMD does have some issues and in fact did not perform faster than the existing DDR as expected. Check out http://www.tomshardware.com/2006/02/21/a_look_at_amds_socket_am2_platform/index.html and http://www.anandtech.com/weblog/default.aspx#279

Anonymous said...

There are a lot of valuable arguments all over the internet which comment to these new benchmarks, but I can't shake of the feeling that AMD fans are all like "yes, ok, BUT if then else, AMD would win".

Mark my words, in 6 months AMD will not have the performance crown anymore. And this is just the beginning.

Anonymous said...

good reading here, my 2 cents would be that I would expect that a 4 instruction / clock vers a 3 instruction / clock would be about 20% faster..maybe AMD could do something here as well and some more z-ram cache to up the ante in the CPU stakes.?

Anonymous said...

Commenting on the prices from
http://www.mikeshardware.co.uk/RoadmapQ306.htm:
the prices in pounds? Would this mean that a $529 Conroe will actually be $1060 USD?

Anonymous said...

i have found an interesting link
http://www.tgdaily.com/2006/02/11/intel_conroe_to_launch_in_q3/
where you con see that Conroe 2.66 ghz
can have 4 MB of cache !!! (the top model)
Is this the contender of AMD FX60 ?

Anonymous said...

Hi everyone
I found an interesting link
http://www.tgdaily.com/2006/02/11/intel_conroe_to_launch_in_q3/
where you can see that the Conroe top model is a 2.66 ghz 4 MB cache !!!! beast.

Is this the contender of AMDFX60 ?

Bye

Anonymous said...

Hi Rahul,

I love your commentary: well written and expressed.

I have two questions that coincide. Do you have any inside information on upcoming AMD processors that could elevate their performance? If so, is it possible to share this insight?

Thanks.

Anonymous said...

It looks like Anandtech went back and redid some of the benchmarks to address the BIOS and F.E.A.R. issues and there was really not much difference in the result. Conroe still kicked the crap out of the OVERCLOCKED FX-60. I can't wait to get my hands on a Conroe based system.

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2716&p=1

Tartan said...

I don't understand why some people are placing this new chip in competition with the FX-60. In 6 months time, I am sure AMD will have advanced so far, and AMD have no incentive to showcase any ground-breaking hardware at this moment, as they have all the momentum in the market.

AMD need not come out in retiliation at this time; I am sure in 6 months time they will deal a competitor to the Conroe. If AMD leak anything, it gives Intel 6 months to catch up. It's all game theory...

Anonymous said...

Hi, Rahul,

You wrote: "FX-60 will be history when Conroe comes out."

What does this mean? I certainly don't think you are saying that AMD will no longer offer any consumer-oriented dual-core processor in the FX-60 performance range when Conroe comes out.

I simply was trying to elicit your guess on the price of such a processor, even if AMD decided to kill "FX-60" by that name, but offer one of similar performance under a different name.

So, what is your guess of the price of the consumer-oriented performance equivalent of the FX-60 when Conroe 2.40GHz becomes available in volume.

If it is any help, my guess is that AMD will have to discount them at least 10% from whatever Intel will be charging for the Conroe 2.40GHz at that time. If that is $315 (in 1000-unit quantities), as per the cited rumor, then the FX-60 equivalent would be at some $275 in quantity.

Volume shipment of Conroe 2.40GHz to me appears likely somewhere in the July-September timeframe.

Does this seem about right to you? If not, what do you expect the price of AMD's dual-core performance equivalent of the FX-60 (whatever its name or socket) will be when Conroe 2.40GHz becomes available in volume?

P.S. It's a simple question, and has nothing to do with "momentum."

Anonymous said...

Sure, the FX-60 will no longer be AMD's top dog, but the FX-62 (Socket AM2 based) will be. The whole idea of using an OC'd FX-60 at 2.8ghz was to simulate the FX-62, which will under those exact same conditions, except with DDR2 (and obviously on AM2, but that's displayed no significant performance boost, say 10% at the most?).

I doubt AMD will have a more powerful AM2 FX CPU out by July, seeing as they will just have released the AM2 FX-62. If they do, It will still come in at least 10-20% less in performance than the Conroe 2.66ghz. The FX's always cost around $1000+, and the Conroe is to cost around $500.

Pick your deal? Pay half as much and get a 20% performance boost or pay twice as much and get a worse performing processor.

Rahul Sood said...

Anonymous said...

There was a 2nd round of benchmarks,


You're a bit late, read the comments. Then go here: follow up

Anonymous said...

There are a lot of valuable arguments all over the internet which comment to these new benchmarks, but I can't shake of the feeling that AMD fans are all like "yes, ok, BUT if then else, AMD would win".


AMD fans are missing the point of momentum

Mark my words, in 6 months AMD will not have the performance crown anymore. And this is just the beginning.

...indeed it is just the beginning for Intel, but AMD will have already established themselves where they need to be for the time being.

Fri Mar 10, 02:07:02 AM

Anonymous said...

good reading here, my 2 cents would be that I would expect that a 4 instruction / clock vers a 3 instruction / clock would be about 20% faster..maybe AMD could do something here as well and some more z-ram cache to up the ante in the CPU stakes.?


There are many things AMD can and should do to improve performance.

Fri Mar 10, 02:29:20 AM

Anonymous said...

Commenting on the prices from
http://www.mikeshardware.co.uk/RoadmapQ306.htm:
the prices in pounds? Would this mean that a $529 Conroe will actually be $1060 USD?


Go to www.xe.com for the exchange rate. Intel won't drop their flagship much below $1000.

Fri Mar 10, 07:34:18 AM

Anonymous

Is this the contender of AMDFX60 ?


Um, no.

Fri Mar 10, 07:41:40 AM
Anonymous said...

Hi Rahul,

love your commentary: well written and expressed.


Thank you.

I have two questions that coincide. Do you have any inside information on upcoming AMD processors that could elevate their performance? If so, is it possible to share this insight?

Yes, and no, sorry. Check this out though.

Fri Mar 10, 08:47:18 AM
Anonymous said...

It looks like Anandtech went back and redid some of the benchmarks to address the BIOS and F.E.A.R. issues and there was really not much difference in the result. Conroe still kicked the crap out of the OVERCLOCKED FX-60. I can't wait to get my hands on a Conroe based system.


Conroe looks good no doubt. You should certainly read the follow up I did in any case.

Fri Mar 10, 09:54:27 AM
Tartan said...

I don't understand why some people are placing this new chip in competition with the FX-60. In 6 months time, I am sure AMD will have advanced so far, and AMD have no incentive to showcase any ground-breaking hardware at this moment, as they have all the momentum in the market.


Great comment.

AMD need not come out in retiliation at this time; I am sure in 6 months time they will deal a competitor to the Conroe. If AMD leak anything, it gives Intel 6 months to catch up. It's all game theory...

Interesting point, thank you for the comments.

Fri Mar 10, 09:54:48 AM
Anonymous said...

Hi, Rahul,

You wrote: "FX-60 will be history when Conroe comes out."

What does this mean?


It means the FX-60 will be discontinued - old news - it's 6 months away man! When was the last time you have heard of a high end processor staying around for 6 months?

I simply was trying to elicit your guess on the price of such a processor, even if AMD decided to kill "FX-60" by that name, but offer one of similar performance under a different name.

I think you need to re-research your estimated prices from Intel.

So, what is your guess of the price of the consumer-oriented performance equivalent of the FX-60 when Conroe 2.40GHz becomes available in volume.

There won't be an FX-60 replacement, there will be a new high end in its place.

Volume shipment of Conroe 2.40GHz to me appears likely somewhere in the July-September timeframe.

That's quite the timeframe :)

Does this seem about right to you? If not, what do you expect the price of AMD's dual-core performance equivalent of the FX-60 (whatever its name or socket) will be when Conroe 2.40GHz becomes available in volume?

Wow, I think you asked the same question 4 times in the same post :)

P.S. It's a simple question, and has nothing to do with "momentum."

Yes, but clearly there's no simple answer - because the question seems flawed.

Fri Mar 10, 12:31:56 PM
Anonymous said...

Sure, the FX-60 will no longer be AMD's top dog, but the FX-62 (Socket AM2 based) will be. The whole idea of using an OC'd FX-60 at 2.8ghz was to simulate the FX-62, which will under those exact same conditions, except with DDR2 (and obviously on AM2, but that's displayed no significant performance boost, say 10% at the most?).


I can't really disclose anything on the future roadmap for AMD.

I doubt AMD will have a more powerful AM2 FX CPU out by July, seeing as they will just have released the AM2 FX-62. If they do, It will still come in at least 10-20% less in performance than the Conroe 2.66ghz. The FX's always cost aroundhttp://www.blogger.com/comment.g?blogID=17375431&postID=114184911431790820 $1000+, and the Conroe is to cost around $500.

I think you need to research your pricing and comparitives.

Pick your deal? Pay half as much and get a 20% performance boost or pay twice as much and get a worse performing processor.

Sounds like quite a deal ;)

Anonymous said...

Hi, Rahul,

Gosh, I thought it might be interesting to speculate about how much an FX-60 (or its performance equivalent) might cost in the third quarter, perhaps as early as July. But I just don't seem to be able to put it in the right words.

I guess I tried asking the same thing in different ways in my second post when you failed to answer the substance of that question in my first post.

Perhaps you will not be interested in consumer-oriented dual processors of the power of an FX-60 in 4-6 months, but I think many other people may be. So why not help them out and give your best guess on how much AMD will charge for such processors in that timeframe.

(I can't believe that you truly contend that AMD will kill the FX-60 and not offer a comparable processor in that timeframe.)

P.S. This question is directed to Rahul Sood and not Bill Clinton. ;)

Rahul Sood said...

Anonymous said...

Hi, Rahul,

Gosh, I thought it might be interesting to speculate about how much an FX-60 (or its performance equivalent) might cost in the third quarter, perhaps as early as July. But I just don't seem to be able to put it in the right words.


:)

I guess I tried asking the same thing in different ways in my second post when you failed to answer the substance of that question in my first post.

Alright I'll bite. I imagine that however AMD decides to kit the latest FX package in July/August/September/whenever will remain at the current FX pricing.

Perhaps you will not be interested in consumer-oriented dual processors of the power of an FX-60 in 4-6 months, but I think many other people may be. So why not help them out and give your best guess on how much AMD will charge for such processors in that timeframe.

I am absolutely interested in this though I'm certain the kit will be interesting.

(I can't believe that you truly contend that AMD will kill the FX-60 and not offer a comparable processor in that timeframe.)

I didn't contend this, but I took the bait anyways ;)

P.S. This question is directed to Rahul Sood and not Bill Clinton. ;)

Damn, Bill was just here too.

Anonymous said...

Here are the benchmark results from another site, hothardware.com. Aside from the strange F.E.A.R. result, they show an advantage of the Conroe 2.67GHz over the 2.80GHz overclocked FX-60 of between some 13.6% and 27.9%, lending some additional support to AnandTech's 20% SWAG.

The first column is my calculation of the percentage Conroe performance advantage, using the raw figures from the following link:
http://hothardware.com/viewarticle.aspx?articleid=794&cid=1


25.2% DivX Encode: 1080p Source File
18.4% iTunes MP3 Encode
13.6% Windows Media Encoder
21.6% Futuremark PCMark05: CPU Performance
20.7% Unreal Tournament 2004 - 1024x768 - Max Quality
26.3% Half Life 2: Medium Quality - 1024x768
23.3% Quake 4: HQ 1024x768 - SMP
27.9% Quake 4: HQ 1024x768 - no SMP
46.2% F.E.A.R.: LQ 640x480
Bill, I don't what you have done with Rahul, but please let him go: I would like him to comment on these figures. (If you promise not to kidnap him again, I promise I won't ask him about future CPU prices.)

Duane Bolt aka BlaqMale said...

I'VE FIGURED IT OUT!!!!...(i think)

Like everyone else, i too was puzzled as to how Intel managed to pull off that on average 25% (24.8% based off hothardware's figures, yeah i like math :p ). I also thought, yeah, something must be up with the hardware to cripple the AMD machine. I was here leisurely researching the benefits of dual core vs single core and it's effect on SLI when it hit me.

I don't know from practical use but isn't ATI's implementation of dual graphics cards (crossfire) beneficial in all senarios unlike nvidia's (SLI) which works for some games, not saying Xfire is better than SLI since it is not. However,I came up with this, like Xfire (and RAID) Intel came up with a way to increase dual core performance with Conroe (stating the obvious). What is not obvious is the method in which they did it, right?
Looking at Intel's test system, the only thing different from hardware currently available is the processor (stating the obvious again). However. note that the OS is the same old, well updated!!! Windows Xp Pro, updated in that intel would obviously need new drivers for Xp support, right??? (Hint: modification to the ATI catalyst driver mentioned by Anand).

There lies the key to this whole fiasco, Intel was not demoing the performance of the conroe architecture (hardware)being faster than AMD's but VANDERPOOL (software). Vanderpool (VT) being Intel's version of "Virtualisation" (Amd's being hardware based - codenamed Pacifica, please forgive me if the term virtualisation refers to AMD technolgy alone and not computer technolgy in general) that VT isn't software but the tech heads should know what i mean. Yes, Vanderpool has to be supported by harwdare, that being Conroe atm. I'm not the most technically versed but i believe virtualisation is supposed to advance the way software talks to the OS and OS to hardware or something to that effect.

Could it be then that software based virtualization (working in XP and not Vista) coupled with Intel's change of processor architecture (based off the Pentium M) was what gave Intel the performance boost??? I say Yeah.


Random specularisation (i know it's not a word) from here on.

So then, software virtualisation (i like the sound of 'isation' want to use z but it's american to use s right? and i'm Jamaican :p )
Anyways, back to specularis(z)ing, software virtualisation gives the underperforming Intel a nice speed boost. What then should we expect from AMD's hardware based version codenamed Pacifica???
If i remember correctly, it was Henri Richard, AMD's chief sales and marketing officer that said they (AMD) are certain they can maintain their technology leadership until 2007/08. But wouldn't that depend on coding support for AMD's new hardware???
and isn't that what AMD's CEO Hector Ruiz is advising The US Congress about, innovation and competition, could that be why AMD is making available their IO virtualisation technology royalty free??? for industry wide support.
I mean, AMD is trying to take down a monster here. (monster meaning anything that helps my points)

Hmmm, i need a new computer like last year, this under 1ghz not cutting it, so do i wait three more months for AM2??? screw NDAs, somebody tell me something. Is it wise to go Athlon 64/X2 now???(not like i can afford anything higher than a sempron anyways). AMD will keep intel in check by one upping them like they have started doing since athlon 64.

well that's about it i guess, make sense?


I think i visit the I'nQ too much. :p

Hey Rahul and all the anonymous ones , if you ever feel like a vacation, please consider Jamaica, we need your foreign exchange to better the economy here.

Lastly (for Rahul and his friend Fuad), AMD pwns the retail at the moment, servers doing good, technically they don't need Dell but you'ld be suprised to know how many people buy intel cause of the name, AMD needs the OEM market, can you get (or try to influence) AMD to ADVERTISE!!! please. And my bet is on Rahul since Dell isn't stupid, AMD is here to stay, Dell will use AMD chips in their line.




P.S. Any chance someone can hook me up with a subscription to the I'nQ. :p

Duane Bolt aka BlaqMale said...

oh and can you beeing credible tell me if dual core makes a difference over single in respect to SLI and any resolution but particularly high ones, i'm only finding info in no sli setups??

Anonymous said...

Hi there Raul ,

Well , I think Conroe is really a great chip, but I do believe too that AMD will have some kinda jazz jackrabbit in the hat.

Utilizing DDR2 wont save amd from conroe, but who knows maybe amd have some improvement in algorithm just like for conroe?

I have my faith in AMD but currently , for dual core i encourage my customer to buy pentiumD 2.66 for the best bang for their buck. Same as u , we buy and sell what best for the money we paid for. for single core I think Amd sempron or amd64 will be better, not bcos of the speed, but the technology and the power consumption of it.

hmmm another thing to ask u Raul, why dont someone come out with a quad channel memory controller for AMD? is it really that hard than adapting DDR2 ?

nice blog indeed which prevent me from going to bed early :P

seeyah around
Homura

Rahul Sood said...

Anonymous said...

Bill, I don't what you have done with Rahul, but please let him go: I would like him to comment on these figures. (If you promise not to kidnap him again, I promise I won't ask him about future CPU prices.)


Well, the only comment I can give you on these figures is - based on the numbers I have seen and the people I have spoken with, Conroe looks to be a strong part when it comes out.

Read this for comment;

Fri Mar 10, 09:19:08 PM
Duane Bolt aka BlaqMale said...

...well that's about it i guess, make sense?


Sort of. No matter how you slice it if you take one platform and it's faster than another through whatever optimization are built-in - who cares...

Hey Rahul and all the anonymous ones , if you ever feel like a vacation, please consider Jamaica, we need your foreign exchange to better the economy here.

Ya mon.

can you get (or try to influence) AMD to ADVERTISE!!! please. And my bet is on Rahul since Dell isn't stupid, AMD is here to stay, Dell will use AMD chips in their line.

I believe AMD wants to draw consumer demand based on innovation. They want their customers to build with AMD, and their customers can advertise to their customers.

Sat Mar 11, 01:02:09 AM
Duane Bolt aka BlaqMale said...

oh and can you beeing credible tell me if dual core makes a difference over single in respect to SLI


Depends on the game.

Sat Mar 11, 01:05:34 AM
Anonymous said...

Hi there Raul ,

Well , I think Conroe is really a great chip, but I do believe too that AMD will have some kinda jazz jackrabbit in the hat.


...

Utilizing DDR2 wont save amd from conroe, but who knows maybe amd have some improvement in algorithm just like for conroe?

...

hmmm another thing to ask u Raul, why dont someone come out with a quad channel memory controller for AMD? is it really that hard than adapting DDR2 ?

As you may know AMD uses a built-in memory controller. They do read this blog often, so I'm sure they'll see your ideas ;)

nice blog indeed which prevent me from going to bed early :P

Yeah, same here, thanks.

Anonymous said...

History repeats itself. Intel does the same in every IDF. Intel wants to convince consumers to stop buying processors until Conroe is out

Intel’s R&D budget is equal to 2xAMD. So you guys should give AMD credit for the competition it brought to the table. Who knows what would have happened if AMD vanished like the rest. We would be using Pentium Pro still.

I would have been more interesting if Intel used also its Extreme Edition in that benchmark. Intel always compares its future product line with its current one but this time it would have looked so bad I think.

Competition is healthy for consumers and consumers should understand that and support it otherwise we’re ……. As always by Intel

Anonymous said...

No one talk about the 90nm to 65 nm when AMD makes the change you will see Intel shaking.

Anonymous said...

If you read the tests about chipsets with dual setups at tomshardware you will figure out that Intel has choosen the worst scenario for AMD.

http://www2.tomshardware.de/motherboard/20060303/ati_crossfire_xpress
_3200_chipsatz-18.html

In this we see that for Fear and Quake we have the following SLI scaling:

ATI R480 (used old): 138 %
ATI R580 (current): 164 %
NVidia NForce (default): 192 %

If you recheck the numbers knowing this you come to the point where Intel still lags behind in these benchmarks.

For the encoding benchmarks it is clear that Intel made a good job with their SSE units (also an old weak point of AMD).

A_Pickle said...

Hardly. ATI's RD480 chipset works great in games, it just suffers from anything that uses the southbridge a lot. Essentially, none of the benchmarks in IDF were using southbridge functions, with the possible exception of media encoding.

As you can see, here, 3D rendering is not in any way hindered significantly by the dismal southbridge, nor does encoding. AMD didn't take a hit other than by the likely outcome that Conroe is just a damn good processor.

-Pikl

Rahul Sood said...

Anonymous said...

History repeats itself. Intel does the same in every IDF. Intel wants to convince consumers to stop buying processors until Conroe is out


Yep, very true.

Intel’s R&D budget is equal to 2xAMD. So you guys should give AMD credit for the competition it brought to the table. Who knows what would have happened if AMD vanished like the rest. We would be using Pentium Pro still.

Great point.

I would have been more interesting if Intel used also its Extreme Edition in that benchmark. Intel always compares its future product line with its current one but this time it would have looked so bad I think.

Same here, but they don't need to prove Conroe is way better than the Pentium 4, that should go without saying.

Competition is healthy for consumers and consumers should understand that and support it otherwise we’re ……. As always by Intel

Correct.

Sat Mar 11, 10:42:14 PM
Anonymous said...

No one talk about the 90nm to 65 nm when AMD makes the change you will see Intel shaking.


I'm not so sure Intel will be shaking - I think at this point there is respect for AMD from Intel - whereas before Intel didn't even acknowledge AMD. It took them awhile but they seem to have humbled slightly.

Sun Mar 12, 07:17:45 AM
Anonymous said...

If you read the tests about chipsets with dual setups at tomshardware you will figure out that Intel has choosen the worst scenario for AMD.


...Yes, correct.

In this we see that for Fear and Quake we have the following SLI scaling:

ATI R480 (used old): 138 %
ATI R580 (current): 164 %
NVidia NForce (default): 192 %


RD580 kills RD480.

Sun Mar 12, 11:11:13 AM
A_Pickle said...

Hardly. ATI's RD480 chipset works great in games, it just suffers from anything that uses the southbridge a lot.


Not quite, RD480 does not work well with software crossfire. Thus the 1300 and 1600 run much faster on RD580.

Essentially, none of the benchmarks in IDF were using southbridge functions, with the possible exception of media encoding.

Agreed on that point.

As you can see, here, 3D rendering is not in any way hindered significantly by the dismal southbridge, nor does encoding. AMD didn't take a hit other than by the likely outcome that Conroe is just a damn good processor.

Yes, Conroe will be a damn good processor, It's pretty safe to say.

Anonymous said...

Anonymous wrote: "Intel’s R&D budget is equal to 2xAMD. So you guys should give AMD credit for the competition it brought to the table. Who knows what would have happened if AMD vanished like the rest. We would be using Pentium Pro still."

R&D is usually a reasonable proportion of revenue: lots of revenue = lots of R&D.

I give Intel most credit for spreading some of its R&D budget to Israel (where it did the most good).

I give AMD most credit for using some of its R&D budget to buy the Alpha design team (where it did the most good).

The right key designers has far, far more to do with success than the size of the R&D budget (if the latter is even correlated with success at all). ;)

Anonymous said...

Hi! First of all..Great Blog!
Don't really think switching from DDR to DDR2 will make that big difference. But doesn't AMD have signed a 5 year licensing deal with Rambus? Doesn't AMD have also a deal for builind LZRAM into their new processors? Just think about an AM2 FX62 with 4MB+ of LZRAM Cache and an enhanced memory controller build on Rambus Technology!!

Anonymous said...

I don't care if the Conroe is 20% faster than an OC'ed FX60. I will not pay extra $$$$ just to gain 20% in performance. Intel chips has always been more expensive. So what if you can get 190 FPS and I get 160 FPS...big deal...

From my point of view, having better stuff is only for bragging rights. In reality, you can hardly tell the difference. So what if you can encode a video 10 secs faster? There isn't a time bomb waiting to go off if. I will not all of a sudden throw away my FX60 and grab a Conroe...


BTW, fire time poster in here. Very interesting site...*bookmarked it **

Anonymous said...

oops... i meant first.. not 'fire' lol!

Anonymous said...

I tell you all this.... Let AMD build there own machine with the latest chips and mother board and INTEL theirs and let the drag races begin! I 'll place my $$$$ on INTEL.

Rahul Sood said...

Anonymous said...

R&D is usually a reasonable proportion of revenue: lots of revenue = lots of R&D.


Sure, but then why does Apple spend the same on R&D as Dell does?

I give Intel most credit for spreading some of its R&D budget to Israel (where it did the most good).

Agreed on that.

I give AMD most credit for using some of its R&D budget to buy the Alpha design team (where it did the most good).

Yep.

The right key designers has far, far more to do with success than the size of the R&D budget (if the latter is even correlated with success at all). ;)

True enough. You can't make a baby in one month with nine wives.

Sun Mar 12, 07:18:10 PM
Anonymous said...

Hi! First of all..Great Blog!
Don't really think switching from DDR to DDR2 will make that big difference.


No, but combined with the chipset/platform optimizations we should see a reasonable difference.

But doesn't AMD have signed a 5 year licensing deal with Rambus? Doesn't AMD have also a deal for builind LZRAM into their new processors? Just think about an AM2 FX62 with 4MB+ of LZRAM Cache and an enhanced memory controller build on Rambus Technology!!

Yeah sounds great on paper no doubt.

Tue Mar 14, 01:59:19 PM
Anonymous said...

I don't care if the Conroe is 20% faster than an OC'ed FX60. I will not pay extra $$$$ just to gain 20% in performance. Intel chips has always been more expensive. So what if you can get 190 FPS and I get 160 FPS...big deal...


The FX processors are usually the most expensive from AMD's line, just like the EE are the most expensive from Intel.

From my point of view, having better stuff is only for bragging rights. In reality, you can hardly tell the difference. So what if you can encode a video 10 secs faster? There isn't a time bomb waiting to go off if. I will not all of a sudden throw away my FX60 and grab a Conroe...

My customers want the best performance we can give them regardless of price. They also want to ensure that they're making smart buying decisions - so performance isn't always king.

BTW, fire time poster in here. Very interesting site...*bookmarked it **

Thank you.

Wed Mar 15, 03:24:31 AM
Anonymous said...

oops... i meant first.. not 'fire' lol!


We know what you meant.

Wed Mar 15, 03:25:23 AM
Anonymous said...

I tell you all this.... Let AMD build there own machine with the latest chips and mother board and INTEL theirs and let the drag races begin! I 'll place my $$$$ on INTEL.


When?

Darkskypoet said...

As I am sure many others have done, i wasted (sic) too much time (perhaps) reading through this entire blog. What has become evident, however, is that many people posting don't understand all that much about what they are posting about.

For instance, many believe that conroe's success or failure is going to either suddenly wreak the demise of AMD, or if it is a bunch of FUD, wreck intel. This of course, is not the case.

Secondly, Momentum is key. Although you can stop a freight train, it requires more then a thousandth of a percent of all of those purchasing cpu's to jump infront of it. For all of you holding fast to the notion that conroe will kill AMD, walk into a room of tech neophytes, and ask them if they love conroe.

Point being, at one point in time, AMD could make a processor that walked through walls while making you an espresso, and 99% of people would not buy it. Fact or fiction intel was the best. Now, AMD doesn't have to, and people understand that AMD is as good, or better then a comperably priced intel processor. In fact, the old axiom intel good, amd cheap, has been replaced with intel hot, amd faster and (still because of momentum) cheap.

Conroe could come out tommorow, make you an espresso, and walk through walls. But more then say 80% of the market wouldn't know about it, wouldn't care about it, and would be told by an undereducated salesperson at a big box store, that AMD is cooler, quieter, faster, etc. Why?

Because the majority of people don't spend 3 hours researching "conroe benchmark", "IBM AMD DSL 65nm", "SOI 90nm", etc every night.

The majority of sales people couldn't tell you that an integrated memory controller, and a finely tuned (to avail bandwidth) core will not benefit past a small margin with 5000x the memory bandwidth it actually needs.

I, like the author of this blog, and other builders that have chimed in, could more then likely exceed the performance given by intel's AMD test system. And I could probably do it with a nice sample of an x2:4400(never mind an FX:60), better motherboard, etc.

As well, realize that for $175 CDN I can buy an A64 that runs at 2.7ghz +.
Well short of the $1000 Fx 60, or conroe chip somewhere out yonder window.

Single core you retort? Well Yeah, cause given the same budget, I'll stomp your conroe, if I can afford 2 x x1900xtx's for the same price as your conroe, and one x1900xt (in 6 months). You will kill me in super PI, but my clients want to game, not calculate meaningless extraneous PI digits all day. (As threading comes into its own this will change substantially, however, i have taken so many x2:3800's past 2.5ghz, and a64 3200's past the same at stock, or near stock voltage and cooling so as to make them defacto uberchips price per performance metric)


All in all, the mainstream, and momentum, etcetera serve to illustrate a point:
When AMD first released Athlon, I fought with people to adopt the tech for all sorts of applications, and they fought back with "intel is better". I argued with them relentlessly that that Athlon 500 creams the then recalled P3-600 in their apps. They did not listen. Momentum, was on chipzilla's side. Even though on paper, in benchmarks, in the cheap assed duron that spanked the more expensive P3's on their desk in a demo: "It wasn't a pentium."

Well at this point, AMD is not yet that dominant, by any stretch. Yet, regardless of intel's dominance on paper,and in benchmarks, I have no conroe system to shove onto someones desk to overcome the prevailing feeling: "My buddies p4 sucks in games, AMD's been beating intel for years."

And there in lies the rub; it's not what is true today that most people know, and rely on, but what has been true in the past. If we lived in a world of perfect information, then yes, intel would have sold far less P4's, AMD would own the market in all but mobile, and the instant conroe was availible, then AMD would sell only those chips that were needed to back up lack of supply on the conroe side.

Thanks to miscues, and momentum, there will be a lot of people still pissed at intel for the 423 -> 478 slap, the fact that their 1.3ghz P4 was slower then their P3 1ghz, and that you can indeed fry an egg on a Prescott 3e, but not really go any faster then then the northwood it replaced.

P.S. Gratz to intel on conroe, you athlon'd the athlon.

Anonymous said...

INTeL's NGMA is a P6 and that basic arch dates back to 1995 with the pentium pro!! also INTeL is known to play dirty, Skype Anyone, BAPCo Anyone, Intel C Compiler Anyone. at INTeL marketing guys run the Co. but at AMD the R&D guys run the Co.

Anonymous said...

YOU GOT PWNT WITH YOUR NUBNESS and TROLLNESS