I must apologize for not writing any updates since E3. The bottom line is I have been travelling more now than ever before, and I am currently sitting in the San Francisco airport and I thought I would grab a few minutes to write a blog.In the meantime I thought I'd write a quick note to Charlie at the Inquirer. Many of you recall the bet we had where I said Dell would eventually come around and carry AMD... It looks like this is indeed the case and AMD's stock is going crazy in after hours trading.
I don't think this is the last surprise, this is simply the beginning...
Over the next week or so I plan to write a few articles on PC Gaming, what I saw at E3, and the future of our industry. I will also soon talk about the consolodation of graphics and CPU, I believe there is convergence in these technologies in the future. In the meantime I can't talk too much about Conroe other than to say I am very impressed with the performance - it's $%*#ing fast man, seriously. I still think yields may be low to start , but I think Voodoo will be alright with allocation...
Last note (back on topic): I want to get some ideas for the bunny suit decoration for Charlie at the Inquirer. Please feel free to grab his picture from the blog and do your best in Photoshop... We are allowed to hang a sign around his neck, so the best photo wins.
The winner will receive a very cool signed Voodoo EDGE Mousepad...
35 blogger comments:
Should I consider to submit, I would like to know whose signature goes on the mouse pad to make it all worth it?
But actually, I really do want to see Charlie in that bunny suit badly, so it may not matter.
I may even end up throwing a Dell Quad core Opteron server for the winner myself to accompany that mouse pad as a gaming pc.
Just a reference design:
http://hollywoodcostumesandparty.com/easter/1091.jpg
http://sharikou.blogspot.com/2006/05/dell-goes-amd.html
Congratulation Rahul! Has Groo called to plead for amnesty. (Just kidding). Your bet is the second most talked about thing on the tech message board tonight.
Congratulations on winning the bunny bet, Mr. Sood. Just to inform you that your comment about "Conroe" being [expletitive] fast made me cease being dismissive about Core2 processors as a threat to AMD; and that I quote you in my blog, here
I have shown that Conroe's advantages were solely due to shared 4MB cache in single threaded applications. In fact, even Core Duo demonstrated an advantage over Athlon 64 in some gaming tests. In general heavy multitasking loads, Conroe and Athlon 64 are close clock/clock. Moreover, Athlon 64 enjoys a 20-40% performance boost under 64 bit mode. I am not seeing similar improvements for Conroe. By the time Intel has enough units for Conroe, it will be 2007 already and AMD's next round of frags will out for Intel to bear.
Congrats!
But why did you remove the sentence on AM2 from your comment? I was surely neither saying too much nor negative in any way, be it towards Intel or AMD. Just wondering.
You keep harping on conroe's yields. Woodcrest is going to me 70% of server cpu sales for intel. Infact "Charlie" at inquirer wrote yesterday that TDP of woodcrest is going to fall to 65w TDP. Still u think intel would not have enough conroes. Is it bcos of process issues or bcos it would focus on woodcrest.
w00t a mousepad!
I was hoping for a Voodoo SLI notebook. :(
You might arrange for one or two nice female bunnies to accompany Charlie
so he won't look so alone and out of style. I don't know if IDF is the ideal
location. They might have "dress-code objections" (Since he would remind
people of the Dell-AMD deal :^)
Regards, Hans
You WILL post pictures of Charlie in the bunny suit, won't you?
Hey,
I enjoy yuor blog. I kinda figured that after nVidia's business platform Dell would bend. I mean, they are getting killed by the DL585. MS uses them for ALL of their 2003 X64 servers. Alienware gets them X2, this gets them Opteron. And now with inroads into film rendering(Episode III), Dell can make a real killing. I mean 8xx can be used in a dual config with scale up.
With Conroe they can use X2 as the low end - Intel would be fools to leave the prices as low as they were announced. Their margins will be crap for 9 months or until the next gen. That means margining the Woodcrest high-end which won't work above 4 way.
This will be a tough year for Intel financially.
"Infact "Charlie" at inquirer wrote yesterday that TDP of woodcrest is going to fall to 65w TDP."
65w TDP using Intel's TDP definition, which Intel defines as "power dissipation under load".
as a result, it means its maximum TDP is significantly higher.
On the other hand, AMD's TDP definition is "the absolute maximum power consumption designed". so, under normal circumstances, AMD's processor actually consume a lot less power than its TDP.
so what's Woodcrest's real TDP? we may know the real truth after some 3rd party sites actually got a hand on it and test it.
http://www.silentpcreview.com/article169-page3.html
correction:
Intel's TDP is actually measured under normal usage, not under load.
Just to let you know, the current mobile Core Duo (Yonah) and unreleased Core 2 Duos (Conroe and Meroms) have already broken ALL benchmark (SuperPi, PC Mark, 3DMark, Aquamark, etc..) records by a BIG MARGIN. Imagine SuperPi 1M in 12.9 seconds (with overclocked Conroe @3.9GHz)! An extemely overclocked FX (@ 3.8GHz) can only managed a measly 21 seconds (when comparing clock to clock wise). Conroe is @#$%ing fast indeed. Even X-bit labs guys are testing it out with a mobile Core Duo at http://www.xbitlabs.com/articles/cpu/display/core-duo-overclocking.html (AMD FX-60 got PWN in most benchmarks, by an Intel mobile processor!).
And what does AMD responded with? #1 - AM2.. sorry this one isn't gonna push AMD any forward, basically still stuck with current K8 performance. #2 - Low power.. Trying to take attention from performance issues, and performance (and gaming) freaks will overlook this. For laptops maybe, but currently Core Duos rule the roost.. #3 - Turion.. Lower power maybe, but not performance, still at K8-like levels. Even the Core Duos are still faster! Only advantage is 64-bit (until Merom arrives) #4 - K8L... A "paper launch". That die pix doesn't look like proper silicon at all.. Photoshop patch up? #5. 65nm... When, oh when? Intel already has 65nm on the market and I've yet to see one from AMD. #6 - Prices.. Intel is pruning all prices (check latest news updates) while AMD still keeping (almost the same) level? AM2 isn't gonna be cheaper than FX either. AMD needs to reduce prices further soon otherwise the performance per price issues will start again (like when AMD pwn Intel, are we in reverse here?). #7 - Megahertz war.. Rumors only, faster clock speeds with their current K8 architecture, they are not gonna do any damage here either.
Yes, bring on the K8L (which is more likely targeted for server and workstation) but WHEN is the first silicon? New architectures (like Intel's Core) takes time... AMD's been sleepin' (growing fat and lazy) on their K8 thingy for past few years (with the attitude "Intel can't touch us"), and now they got a rude wakeup call (see how AMD panicked with their press releases and keeping mum about competition). That's good because now AMD will have get back on track again. And Intel isn't sleepin' either. Competition is good. Cheaper processors soon.. especially "older generation" ones. I do hope AMD bring down their prices more as well (and Let more peeps afford AMD 64 machines). Till then and for now, I'm for one is going Conroe's way for my next upgrade.
Regards.
"#1 - AM2.. sorry this one isn't gonna push AMD any forward, basically still stuck with current K8 performance. "
AM2 was never a response to Conroe, it is an platform that supports DDR2. It is just an upgrade to support next year's chips.
"#7 - Megahertz war.. Rumors only, faster clock speeds with their current K8 architecture, they are not gonna do any damage here either."
You must be kidding that it wouldn't going to do any damage. Most people I met, the first thing they ask about a computer is "What gigahertz is it?".
"#6 - Prices.. Intel is pruning all prices (check latest news updates) while AMD still keeping (almost the same) level? AM2 isn't gonna be cheaper than FX either. AMD needs to reduce prices further soon otherwise the performance per price issues will start again (like when AMD pwn Intel, are we in reverse here?)."
Dropping prices on all P4 chips? P4 is consider a budget cpu now. Conroe is cheap! I gotta agree it is cheap, but don't talk about it now. Talk about it when it is in retail and you can actually buy one.
I think the fact is, you are bias against AMD. Conroe is next generation chips, but why are you comparing it to a three years old AMD chips.
Please read what Goldman has to say on the matter
It tends to make sense of the ups and downs. But undoubtedly good news for AMD
Just hope Intel doesn't let its cash do the talking
now, that will be funny
good luck :D
"Core 2 Duos (Conroe and Meroms) have already broken ALL benchmark (SuperPi, PC Mark, 3DMark, Aquamark, etc..) records by a BIG MARGIN."
And with real apps ?
Hey Rahul,
You always seem to have an idea of what's going...Does AMD have a secret weapon against Conroe in terms of performance?
"#4 - K8L... A "paper launch". That die pix doesn't look like proper silicon at all.. Photoshop patch up?"
that is just a total fanboy statement. this statement inevitably forced me to question your intelligence.
"#5. 65nm... When, oh when? Intel already has 65nm on the market and I've yet to see one from AMD."
65nm means nothing but a shrink in die size. proof? take a look at P-D lineup. is 900 series significantly better than 800 series? of course not. so 65nm doesn't matter, the architecture matters. using the same logic, 45nm is not significantly better than 65nm, and so is 32nm. so cut this "oh my processor's nm is smaller so it must perform a lot better" idiotic statements.
Can I get a decent picture of charlie to use for the photochop?
haha, charlie is in big trouble right now.
Charlie's face must show while wearing the bunny suit, or else he can chicken out and make some other dude wear it for him without anybody noticing.
Rahul,
It seems I knew more than you when it cane to Intel pricing strategy for the upcoming trio. I wrote one time telling you that Intel intends to take market share back and hurt AMD by aggressively pricing the next generation chips. You responded by saying that would not be an Intel strategy.. I guess it is.. Think about it, it is the only way to recover market share and investor confidence in a very short time. With this move, Intel can really hurt AMD's (and Charter's) bottom line.
About the Dell move to AMD, I predict it is going to be either temporary or insignificant move. Why? For one, I do not think Dell being in their tough position now will resist Intel's pricing for the trio. As a matter of fact, I think this may have been a rescue move done freely by Dell to rescue Intel in their defense against AMD lawsuit. Do you also think that Dell will pay primium for AMD chips and pass more powerful (or equally competitive) from Intel for less. If you think deeper about this matter, you will see that Dell is playing damage control and trying to please its investors.
What are your thoughts?
Anonymous said...
"#4 - K8L... A "paper launch". That die pix doesn't look like proper silicon at all.. Photoshop patch up?"
that is just a total fanboy statement. this statement inevitably forced me to question your intelligence.
Compare the "paper launch" die pix at http://www.chip-architect.com/news/Quad_vs_Dual_.jpg with the current AMD's dual core one.. (and also compare it to Conroe's at http://images.dailytech.com/nimage/711_conroedie.jpg ). I know how a die suppose to look like. That does NOT look like one at all. Lots of missing "interconnects" on the outer edges and central areas (between the cores, and where's the "northbridge"?). Only the four cores look legit at the moment.
"#5. 65nm... When, oh when? Intel already has 65nm on the market and I've yet to see one from AMD."
65nm means nothing but a shrink in die size. proof? take a look at P-D lineup. is 900 series significantly better than 800 series? of course not. so 65nm doesn't matter, the architecture matters. using the same logic, 45nm is not significantly better than 65nm, and so is 32nm. so cut this "oh my processor's nm is smaller so it must perform a lot better" idiotic statements.
Squeezing a quad core into 90nm? Are you kiddin'? Do you know how BIG will the die size become? Also at 65nm, less propogation delays as distances are shorter. This enables the chip to run faster (internally) as well (without resorting to higher clock speeds).
Anonymous said...
I think the fact is, you are bias against AMD. Conroe is next generation chips, but why are you comparing it to a three years old AMD chips.
Conroe's gonna launch soon, while AMD has a 3-years old chips (with DDR2 controller) as competition. And we may not be seein' K8L until 2008, unlike quoted by AMD as early 2007. Check this out (translated) http://www.google.ca/translate?u=http%3A%2F%2Fwww.hkepc.com%2Fbbs%2Fnews.php%3Ftid%3D600827&langpair=zh-CN%7Cen&hl=en&ie=UTF8
"However, according to Taiwan's first motherboard manufacturers to our enquiries, AMD plans to launch the second half of 2007 the four core mainframe processors Deerhound will not adopt a new framework, and will be the first half of 2008 will be the emergence of a K8L of products including desktop processors Greyhound, workstations and server processors processors Cadiz Zamora."
Seems like a long wait. Regards.
"Squeezing a quad core into 90nm? Are you kiddin'? Do you know how BIG will the die size become? Also at 65nm, less propogation delays as distances are shorter. This enables the chip to run faster (internally) as well (without resorting to higher clock speeds). "
AMD/Intel needs to go to 65nm because of the die size. That does not mean transition to 65nm with the same architecture significantly improves its performance. Intel's transition to 65nm was not a very wise move, as their 65nm product yields are terrible.
"Seems like a long wait. Regards."
Intel's quad core will likely to come sometimes in late 07, where "true" quad core will come a lot later. Intel still has a lot to work on.
Anonymous said..."Intel's transition to 65nm was not a very wise move, as their 65nm product yields are terrible."
At least they have released 65nm products, AND also engineering samples (and that's some real effort materialize). I've yet to see one from AMD so far (not even any "Brisbane" core engineering samples), or is AMD having problems moving or transition to 65nm? One wonders...
Anonymous said..."Intel's quad core will likely to come sometimes in late 07, where "true" quad core will come a lot later. Intel still has a lot to work on.".
Intel already has a working/running quad core silicon. Check out http://www.xbitlabs.com/news/cpu/display/20060415072338.html and url=http://www.hkepc.com/bbs/news.php?tid=583149 It may not be "true" quad core but certainly they have silicon!
Also (new!), Woodcrest previewed at http://www.techreport.com/etc/2006q2/woodcrest/index.x?pg=1 Looks like it beat Optys in EVERY TEST, including power usage and dissipation.
Heat is on for AMD this time around.
Anonymous said..."Also (new!), Woodcrest previewed at http://www.techreport.com/etc/2006q2/woodcrest/index.x?pg=1 Looks like it beat Optys in EVERY TEST, including power usage and dissipation."
That is good, Intel finally has a superior product that I can buy in the near future. Now, when can I buy a 4 socket box from from HP or Dell? Whoops, it is only 2 socket. We're going to have to wait until next year for Tigerton. Are you going to make a fanboi comment about how Tigerton is going to kill Optys?
As I pointed out previously, CONROE/WoodCrest will be 10% slower than K8 at the same clock. The TechReport benchmarks validated this analysis again: in those 64 bit multi-threaded benchmarks, Intel's CORE2 is clearly underperforming AMD. Intel may have an advantage in 32 bit single threaded load due to the 4MB cache. Although Intel's advantage will vanish as we move to 64 bit computing, let's wait and see AMD's responses on June 1.
"Sharikou, Ph. D said...
As I pointed out previously, CONROE/WoodCrest will be 10% slower than K8 at the same clock. The TechReport benchmarks validated this analysis again: in those 64 bit multi-threaded benchmarks, Intel's CORE2 is clearly underperforming AMD. Intel may have an advantage in 32 bit single threaded load due to the 4MB cache. Although Intel's advantage will vanish as we move to 64 bit computing, let's wait and see AMD's responses on June 1."
On your debunking efforts, based on one SINGLE benchmark, the POVRay.. Well, according to the Tech Report folks - "Once more, Woodcrest comes out on top of the Opteron. The margin of victory here appears to narrow as we add threads, but I should be clear: that is no fault of Woodcrest. The newest version of SMPOV takes more time than usual spinning off new render threads, and I'm certain that's why neither system scaled as well as it could have. The single-threaded result is the most notable here, in many ways. Woodcrest finishes rendering the scene 42 seconds before Opteron, a substantial margin of victory.". You are debunked.
As your previous debunking (often based on a single benchmark), it still holds no ground because in the end the benchmark results is what counts (or how people evaluate). The 4MB theory may have some grounding but still does account for the OTHER results. Remember that the cache will be occupied by parts of the operating system kernel, hardware drivers, background services and DLLs.. and not forgetting the (multiple) stacks as well!
And yes, that higher speed RAMs may contribute to better performance, BUT it should be NOTED that this performance (no thanks to Intel's old design) "FSB limited". BTW, I've yet to see these Woodcrest in 4P config, since (possibly) only 2P config may give its some advantage.
New (old generation-based) 65nm processors and more Intel price cuts ahead at http://www.dailytech.com/article.aspx?newsid=2455 At this rate, Intel may be churning out budget processors.. LOL.
"I've yet to see one from AMD so far (not even any "Brisbane" core engineering samples), or is AMD having problems moving or transition to 65nm? One wonders..."
as i said before, AMD does NOT need to move to 65nm for single/dual core processors, because AMD's 90nm is simply far superior than intel's 65nm.
"Intel already has a working/running quad core silicon. Check out...It may not be "true" quad core but certainly they have silicon!"
LOL. what's the point of having a dual die quad core, when it clearly is inferior to any "true" quad core? P-D may beat AMD to be the first on the dual core market, but that doesn't mean it is any good.
The original e-mail I sent to Charlie suggested wearing an Intel bunny suit, and doing a 2 minute intrepretive jazz dance expressing Intel's joy or sorrow (depending on the outcome of the bet).
Charlie said he couldn't get a hold of an Intel bunny suit, so it became a "regular" bunny suit, but I'd still love to see a video of the jazz dance part of the bet :-)
I found out that Sharikou has been busted for fraud. He get his PhD from Dr. Dre school of computing!
Well, check out the benchmark results here: http://www.pconline.com.cn/diy/evalue/evalue/cpu/0605/791941.html which pits MID/LOW END Conroe E6300 and E6400 VERSUS AMD FX-60 (stock) and overclocked AMD FX-60 (2.8GHz). These Conroes have only 2MB L2 cache! So much for the 4MB theory.. debunked again.
Conroe's performance should be derived more from its instruction/decode efficiency and parallelism (as well as faster FPU and SSE execution). Large caches doesn't always make processors faster (which is proven in the Pentium EE). Regards.
Back on topic - I would ask the AMD marketing department to supply one.
Post a Comment