7.22.2006

AMD + ATi: One + One = Three


Digg this Story

I am busting at the seams here, I have been holding my opinions on this potential ATi + AMD deal for many months now. Obviously everyone knows Intel’s latest product offering is excellent and news on Conroe/Woodcrest is humming along – but this deal with ATi + AMD has positive cataclysmic effects on the entire industry. This deal slightly overshadows any news going on in our industry at the moment.

I am going out on a limb – I, like many others, believe this deal is imminent and I am offering up my reasons why I believe this deal makes sense.

As many of you know I have always said that ATi is a target for acquisition, and I truly believe they are a diamond hidden deep in the rough – and with a little bit of polishing this acquisition will go down in history as the best move that AMD ever made. ATi + AMD = a go big strategy that will shake the very foundation of our industry.

Based on the feedback I’ve received from many analysts, I am going to assume that most people will not be able to make sense of this deal. The following obvious questions come to mind;

1) Why would AMD buy ATi, why not Nvidia? Isn’t Nvidia a stronger company?

The answer this question requires some methodical strategic thought. This is not unlike a chess game – and I believe that AMD is positioning its pieces carefully before it works its way over to the king.

- ATi is generally undervalued. They have been trading in a tight little window for what seems like forever. ATi does not have the market prowess that Nvidia has; quite simply Nvidia knows how to manage their relationships with Wall Street better than ATi does.

- ATi is very much like AMD was 5+ years ago before the new AMD management came on board and started to fix the company. Under Hector Ruiz, new management leveraged the awesome innovations that AMD had under the hood and turned the company on its head. For the last 3+ years AMD has been the undisputed leader in almost all areas of the market.

- ATi has some killer technologies under their hood – not least of which are publicly available now. ATi is the best third party notebook chipset manufacturer in the world. They understand power management like no ones business, and they have the capabilities to build killer platforms for mobile platforms. There is no denying that the Intel Israeli team is talented when it comes to designing some killer platforms - especially for notebook. ATi + AMD under one roof = mobility platforms that will rival anything that anyone else can put out.

- ATi’s handheld business is strong and AMD has some great OEM relationships to help raise this portfolio.

- No need to talk about the benefits of XBOX 360 + Nintendo Wii, I think you all know what this translates to.

- ATi’s decentralized organization could use a big tune up. They are very easy to work with, and they are incredibly flexible – this acquisition will come as a welcome challenge to their team.

- ATi has polished their skills in platform chipsets, and with a little bit of honing they could be the next big force in the market. With AMD building their own platforms “in house” via ATi I think we’ll see some pretty amazing pre-engineered products.

- It’s not the time for Nvidia yet: If anyone knows Nvidia they know that their CEO is very much in control of the organization. Nvidia is also much more expensive than ATi, and rightfully so – Nvidia has been making all the right moves for many consecutive months. Nvidia is known for their aggressive business moves, and those who know Jen-Hsun will tell you that he wouldn’t let anyone buy the company without him remaining at the helm of the larger organization. Think about the future opportunities here –- We could see the beginnings of a much larger organization.

2) Why is AMD fixated on the graphics market?

Long term thinkers already know that CPU and GPU computations may eventually end up on one piece of silicon. Back in the days of the separate math co-processors industry “pundits” didn’t believe they would ever morph into one – but those in the know realized that this bottleneck needed to be cut out as soon as possible. In the case of CPU + GPU the possibilities are endless – imagine a multi-core piece of silicon where one core handles massive CPU computations while another handles graphics, and perhaps another handles the traffic between the multiple cores.

3) Will Nvidia react negatively and work exclusively with Intel?

Hardly. There is no way that AMD will alienate Nvidia. In fact I am quite confident that Nvidia will actually work closer with AMD, and perhaps share and co-develop graphics in the long run. There are no ego’s at AMD or ATi – if they feel it’s good for their shareholders and their customers to work closer with Nvidia they will do it. AMD knows part of the reason for their success is the fact that they offer choices to their customers. Also, the worldwide demand for chipsets is higher than what ATi could produce so obviously Nvidia will still play a major role in AMD based machines.

AMD has some great platform solutions slated for next year, including server, workstation, notebook, etc. I can’t get into too many details, but in my mind AMD + ATi = one large company with the potential of becoming a massive juggernaught. Any debt that AMD incurs to make this transaction a reality will be short lived. The long term outlook on AMD is unbelievable if this deal is approved – this company will go big, and those along for the ride are going to reap huge rewards.

As usual, initial reactions to change are always negative - but long term thinkers who truly understand the technology will tell you this deal is the one way that AMD can turn the tides in their favor.

On another note, someone needs to write a book on AMD, I’m telling you it would be a best seller. The management over at AMD really understands the future of technology and the needs of their customers. I think a book about how they did it would be a very interesting read.

I welcome any and all of you to visit my blog at www.rahulsood.com for more details on this transaction as it progresses – and I welcome all comments on all sides of the deal.

The follow up to this article (with response to comments) is here, please check it and reserve your comments for that entry.

21 comments:

Richard Evans said...

I appreciate your insight into this deal Mr. Sood. It's been tough to get a good tech perspective on it these past couple of days, since the INQ broke the story. Since you're not writing as a "fanboy" of any of the companies, just as a guy who assembles the best parts available into something he can sell to a hungry marketplace, you have a more balanced perspective.

I agree with you, I think this will be a very positive thing for the computer market, should the deal go through. And I admit, I *am* a fanboy of both ATI and AMD, so I'm very excited by the possibilities.

For one, I think it will force both Intel and Nvidia to not only market themselves well (as they have been doing), but to keep pushing the boundaries of tech themselves to keep up with the potential super-company they'll be working against, and hopefully, with. Nvidia can't afford to lose the AMD market they have, they sell WAY too many motherboards to AMD owners, and thus, will have to work, as you said, maybe even more closely with AMD to make sure they aren't left out in the cold.

Graphics-wise, I think this will give ATI a very big boost, as AMD owns several patents to technologies that could assist in major improvements to the graphics industry. Nvidia will keep up I'm sure, as they're a very saavy company, tech-wise, and aren't leaders by accident.

All in all, I'd say this is one of the most exciting times to live in the computer industry, and the next 5 years should be VERY interesting indeed.

Anonymous said...

Great synopsis of this monuments event,
I believe that another leader of the industry, Microsoft would be encouraging this move by AMD+ATI for its own reasons, not because M$ doesn’t like Intel but because of the scalability of the X86-64 architecture introduced by AMD. Microsofts next Vista for us all needs a close relationship between CPU and GPU.

For me the most telling example of the leadership AMD offers to the industry was the presentation done by Cray for its next generation processing goliath using array's of AMD's K8L's and "other accelerator processors" An ATI GPU would be a great feature to visulize massive data sets processed by this supercomputer.

Anonymous said...

A good reason for AMD to acuire ATI is to keep all its new fabs in full production.. those fabs cost huge amounts of money to build and can make one bankcrupt if they have no designs to churnout chips.

Anonymous said...

Hi Rahul,

I personal have been negative on this idea of AMD/ATI merger, but reading your ideas on the matter have put me at ease some. I guess I look at it as mostly a financial play and not so much as what tech AMD would gain from such a merger. It seems very risky move for AMD to go for ATI now, I mean why rock the boat when you already have a good thing going for you? Plus the financial risk of integrating a company and their technology is always high(they could be inflicting a wound on themselves while Intel gains an advantage from it). Nvidia didn't get SLI going till what 4 years after it bought out 3dfx? Mergers between roughly equal size companies are usually not successful. One thing that bothers me is that it seems to go against AMD's strategy of allowing companies to work with the company. Where this case would be AMD taking ATI and pushing them in a particular direction.

However, I have to say I do trust the judgement of AMD management if such a merger transpires. If AMD is willing to go through with it, they obviously think ATI has tech that will make them grow. There is a financial risk, but maybe they will be able to capitalize off of it.

Matt

Anonymous said...

Hi Rahul,

One other thing, If AMD and ATI are potentially developing shared processors, what does that mean for the future of Nvidia? Will Nvidia be pushed out from the market? I know this is super long term thinking, but it's harder for me to see where Nvidia fits into the picture.

Matt

Anonymous said...

hey rahul, think you're on the mark. If this is real, once folks get over the disruptive and industry transforming nature of this deal, agree that it tees up AMD to integrate graphics (inevitable) and helps them in the mobile and commercial space.

Also per previous poster, don't believe AMD needs this to fill fabs, they're filling their fabs with CPUs and want to keep up with customer ramping. Bet they go fab and foundry.

Carson

Anonymous said...

hey rahul,

although I admit that you analysis is very thorough, I am still not convinced that this merger will work in the long term. Technological analysis aside, I fail to see any increases in the bottom line to justify this merger.

You mentioned that ATI is undervalued, xbox360 and Wii will be of significance. However I feel that all tech stocks have been overvalued, otherwise there's no need for them to plunge on weak earnings. Due to that, I firmly believe that xbox360 and wii earnings potential are already reflected in ATI's stock price. Since AMD's paying a premium of about a billion dollars, I'm not convinced that AMD+ATI is worth an extra billion over AMD and ATI separately. What I see in AMD's earnings history is that during a small window when the Athlon was superior to the Pentium IV, AMD made money, but during the time between the rise of the Athlon 64 and the Athlon XP 3200+, P4 was superior and AMD lost money. Now during the past few quarters when the Athlon 64 was superior, AMD again made profit. Now that Core 2 is on the verge of being released and another price war imminent, I not even convinced AMD won't be seeing red at least once before K8L, should it be competitive, if not more. We can attribute AMD's success as much to their great leaders as we can to Intel's bad ones, if not more.

Moving on the mobile computing, I cannot see how ATI's prowess with mobile chipsets will translate into success for AMD. Current crop of Turion based notebooks are at a disadvantage in terms of performance and battery life compared to Intel ones, even Intel ones based on ATI chipsets, suggesting the problem lies in AMD's current lineup, and not some external partner. Since AMD needs partnerships for all its platforms currently, I do not see it being a stretch that they're working with any partner committed to bringing out a solid platform; thus I do not believe a merger will be of any benefit in this arena either.

In regards to CPU+GPU, it will get to a point where thermal dissipation will become a very big issue, and Moore's law will reach a wall. While it could be done, there are many issues that would need to be solved, and I'm not sure if resource allocated to doing this would be the most optimal. The case of FPU isn't very convincing, since the interconnects back in the 80s are a fraction of what they are now. Is it any wonder that AMD is opening up its platforms for new co-processors? Hypertransport is one very effective method of communication, one seen in multiple socket configurations; there's no reason to believe that such an interconnect need be limited to AMD's in-house design, or that such a design will be necessarily superior.

Although this merger will not affect Nvidia in the short term, I'm pretty sure they are aware that a strong AMD+ATI will erode sales of their nforce chipsets down the line. ATI cannot supply AMD's chipset demand simply because they never enjoyed high demand to begin with. While Nvidia might not have a knee jerk reaction, it would be surprising if they didn't act coolly since on one hand they'll have Intel who loves to keep their platform closed, and on the hand facing new competition from AMD+ATI.

I agree that AMD+ATI can create some amazing chipsets for all of AMD's platforms. However, whether this increased cooperation will result in more profits (above and beyond what AMD and ATI would make separately) or not is still hard to determine at this point. Further, ATI's primary profit source is still their gpu division, and it'll be interesting to see how AMD integrates this into their corporate image.

The one place where I do see synergism is handheld businesses, where both ATI and AMD's joint resources will make them better off than each alone.

Negative or positive reactions don't necessarily make one a short or long term thinker. It's merely a result of one's interpretation of the status of the industry now, and where it will be tomorrow.

Jalf said...

At the risk of being called old-fashioned or ignorant, I'm not so sure GPU's are going to end up on the CPU.

FPU's were different. They're made to assist the CPU, and it's vital that you get the lowest-possible latency between the two. FPU operations become much more efficient when they're integrated into the CPU. And there's a limit to how big each of these could grow in isolation. It wouldn't make sense to quadruple the FPU size if it was still detached from the CPU. Performance wouldn't follow.

But the GPU? It's a very different beast. The CPU sends data to the GPU, and that's it. The CPU doesn't care how long latency there is, it doesn't care about getting anything back. It's a fundamentally asynchronous operation. The GPU receives data, and, when it has time, it pushes data onward to the screen.

Occasionally they exchange a bit of data, but usually not enough to really pose a serious bottleneck.

On the other hand, if they were put on a single chip, they would have to deal with:
Area constraints. GPU's are huge. And CPU's are too. Both have almost unlimited growth potential. CPU's can easily add more cores if they have a corner of a die unused. GPU's can add more shader units. So no matter how tiny transistors get, both CPU's and GPU's will always be able to fill up the die area. If they have to share one die, they both get cut down to 50% of the space they could otherwise use. Lower performance on both CPU and GPU, in return for low-latency traffic, which wasn't really important to begin with.
- Memory contestion: At the moment, the CPU uses a dualchannel DDR2 bus. GPU's use a high speed point-to-point DDR3 connection. The bandwidth difference is immense. A GPU wouldn't be able to work with the CPU's memory constraints. And neither of them would be able to work if they had to *share* that one single bus. Even with faster memory, there's still only so much you can transfer per pin, and there's only room for so many pins on a chip. Which means sooner or later, you're out of bandwidth. Again, what did you get in return? Lower latency on CPU/GPU traffic, which isn't really needed.

Now, a couple of disclaimers. Of course, the usage patterns of GPU's might change radically, and low-latency CPU/GPU traffic might become a neccessity. I just don't see it yet. (Although the physics functionality might be a small step in that direction. We'll have to wait and see where that leads)

Second, on stuff like notebooks, an integrated GPU might be a big win. It'd be efficient for low-performance systems, it'd use little power, and it'd simplify motherboards a lot and take up less space.

Wirmish said...

Mixing a CPU and a GPU ??

Why not make a Torrenza GPU ?
This will give us a way to upgrade the GPU like we change the CPU.

Why would AMD buy ATI ??

ATI make GPUs for the XBOX 360 and the Wii. But neither Microsoft nor Nintendo needed to buy ATI. If AMD wants a Torrenza GPU, they just have to sign a contract with ATI, nVidia, VIA or SIS. If AMD want a chipset, they just have to do the same thing. This way AMD will not alienate Nvidia nor everybody else.

And if AMD buy ATI they will not have the cash to build the New York FAB and rebuild their FAB 30...

One day INTEL will enter in competition with nVidia and ATI with a powerful GPU. While waiting for this day, AMD has only to maintain a good relation with ATI and nVidia.

AMD just have to be patient...

Anonymous said...

It's just going to generate bad consumer options if AMD and ATI combine forces.

Look for your next pc to be something like an Xbox or a Playstation if it happens.

This dynamic duo might benefit gaming enthusiasts, but far more PC's are sold for business and home office use. By driving up the price of the core millions of customers who don't really appreciate the extra gpu power will find it to be a waste of resources and turn to Intel CPU's and a cheap mobo gpu.

In fact, I'd wager that the more integrated a manufacturer becomes (in this case) the worse it will fare. I think to reduce your product line to work with only one GPU is just plain poor judgement, especially when the other gpu manufacturer sells far more units and has almost always had better products. Maybe I'm crazy?

Like wirmish said (I THINK he/she said anyways), AMD would be far better off just maintaining good relationships with Nvidia and Ati...

Anonymous said...

"No need to talk about the benefits of XBOX 360 + Nintendo Wii, I think you all know what this translates to."

sorry, but i have no idea what you mean by this?

love your stuff!

jackall said...

People, stop ranting about why you don't like the idea of mixing GPUs and CPUs. No one ever said it going to happen, they merely said it was a possiblility. Besides, if it does happen it will only be because we gain some benefit from it which you may not be able to see right now but if or when the time comes it will be much clearer. Personally, I like the idea of a merger between both companies because it will allow both companies to bring their talents together and make some truly innovative and amazing products which will ultimately benefit us.

From a financial standpoint based on the information we have right now it is a gamble for AMD to be doing this right now because they are presently unsure of the impact that the recent intel product launches will have on their corporation.

Anonymous said...

Why is your blog titled 1+1=3?

Anonymous said...

I think AMD just panicked.. Seriously.. And they are moving toward closing their platform rather than openning it.. AMD is realizing that growth in CPU will not be as easy as it has been in the past couple of years. Growth apparently was not in flash or the Alchemey so they got rid of the business.. It seems they believe that growth is in Platformazation.. CPU + Chipset + Graphics + ...

Putting that in mind explains, at least to me, why ATI makes sense to AMD..

Now, it they actually believe that, they are about to get into something very hard to do.. Just ask Intel..

Anonymous said...

well this is funny, Intel chipsets for Conroe are supporting ATI crossfire but not SLI, looks like humble Canadian ATI surrender their technology to Intel. Will Intel let AMD take over ATI that easy?
BTW CPU industry is sinking, the industry no longer consider growth industry, even AMD+ATI produce what wonderful products, their sources of growth are come from taking rival market share. Given competition between ATI and Nvidia, I don't see any need that AMD need to pay premium for ATI. And from my consumer sense, the wonderful last couple years of the PC industry are deal to competition between AMD and INtel, ATI and Nvidia, which speedup product cycle and innovation, and also drive down prices. If AMD+ATI really work closely with Nvidia, I would consider less competition and less effective use of resources. Given consumer now are familiar with fast price down of computer industry I don't think AMD can get much juice from high-end graphics business. SO it down to low-end integrated graphics chipset like Intel's. But I think Intel's ppl could think a way to dry up the juice from ATI and AMD syngery.
As a Canadian, I believe mighty american can find lay-low humble ATI is easy to deal with, but may not be the case of AMD. And everyone is afraid to create another monopoly like microsoft, so I think ATI's sales on HD-TV and handheld mobile chips may slow down. AMD may not be able to handel nonPC areas of ATI business. Remember Intel X-scale chips? It is possible ATI become losing business to drag down AMD profit.
And if talents from ATI don't like the deal, they can quit ATI and startup a new firm anyways. So I think AMD really making a gamble bet.

mikerotech said...

I just listened to the stream of the conference at AMD, and it looks like now we will get to see first hand if your predictions are correct Mr. Sood. I'm excited to see what AMD will have to offer now that they have aquired ATI.
I hope we will be able to watch some real innovations in 2007 from AMD with the aquasition of ATI in addition to the new K8L architecture.

May I ask, Mr. Sood, are you at liberty to speak about what ATI has "under their hood?"

Please excuse my ignorance. Perhaps I have not been following technology long enough to realize why ATI is being made out to seem like such an underdog. Last time I checked, even though Nvidia has been on an upstream, ATI has had more market share than Nvidia for quite some time - in mobile and desktop arenas. And this is not to mention ATI has arguably had the highest performing graphics chip avalible since the X1900XT's release. I'm just trying to clear things up, please enlighten me :)

thorkia said...

We all know that AMD is going to push it's 4x4 for the perfomance market. Hell, who wouldn't want 2 dual core FX-62s, but that would be a little extreme in price for me.

Here is what I think...

Currently, most software either doesn't support, or has minimal support for mutli threading. So, the extra 2 cores would be pretty much useless for most people (ie office, and most home use). The extra slot would be much more useful if we could say, drop in an AM2 compatable X1900XT chip, with software support. Having the GPU sitting right beside CPU, and using HT to communicate would eliminate the entire bandwidth to the card problem. Then, AMD could license the socket to NVidia, and have them produce drop in chips for 4x4

That's just my 2 cents. I'd like to see what Rahul thinks of that idea...

Anonymous said...

I would say that a price war would definitely slow the momentum of what this merger could do... It reminds me of HP buging out Compaq..

SmartM0F0 said...

"I would say that a price war would definitely slow the momentum of what this merger could do... It reminds me of HP bu(y)ing out Compaq.. "

I think you've got the concept all wrong... HP & Compaq basically merged to consolidate doing the same thing (building PC's), to save money, and not really leveraging off each other's technology or creating anything new. It was purely a cost merger/buyout.

The AMD + ATI merger makes a lot of sense in the fact that it's different with each company having its very STRONG areas of expertise! It's not about saving money, but taking the collective expertise of both companies and their engineers, creating synergy, and coming up with something innovative that will truly revolutionize the industry. Just by surfing around the net today alone, I feel a lot of energy and enthusiasm for what's ahead and what can be created to advance technology as a whole. I for one, as excited as Rahul thinking of the possibilities that lie ahead.

I think AMD+ATI will learn from Intel how NOT to irritate and alienate your partners. I see this as creating an eco-system of "platforms" that EVERYONE can benefit from. And I agree, that Nvidia can thrive just as much in the AMD eco-system with their open platform approach.

Rahul: What lies in your crystal ball for the future of computing 3-4 years from now if this does pan out? And how is Intel going to react to this?

Keep up the good blog! Very un-biased and informative view of the tech industry... :)

satosphere said...

Hi rahul
That was quite a nice write-up analyzing the repercussions.
I do have a question - was the OEM market (read Dell, HP) a big influence in this decision, ie because now AMD-ATI can provide a platform to OEMs like Intel, its much easier for them to bring out more AMD products?

Rahul Sood said...

I responded to all the main comments here.

For the last comment from Satosphere - yes indeed that was an important factor in the decision I'm sure - among many others.