Can AMD survive another Core 2 Duo?

Posted by Wolfgang Gruener

OpinionIDF is just around the corner and Intel will provide a flood of new information about its upcoming Nehalem (Core i7) processor as well as its 32 nm and 22 nm successors, new architectures such as Larrabee, the ready-to-launch WiMax mobile platform, CE processors such as Tolapai and new partners such as Dreamworks. IDF’s marketing machine typically buries anything from AMD, but this year, AMD is reverting to a strategy from the past: AMD is setting up its camp in a hotel nearby in an effort to balance Intel’s messages. A first briefing discounting Intel’s current product line was given to journalists earlier today. And if the tone of this briefing is any indication, then Nehalem feels like Core 2 Duo all over again.

If you are somewhat interested in what Intel is up to these days, next week’s IDF should bring very few surprises. The new Core i7 processor, the first chip based on Nehalem architecture, will debut in desktop flavors in the fourth quarter, there will be more Nehalem mobile and server chips in 2009; there will be more details on the company’s much discussed 2009/2010; Larrabee graphics card; we will see USB 3.0 and WiMax demonstrations; Intel will announce that it is now producing more 45 nm than 65 nm processors and it will tell attendees that its mobile CPUs are overtaking desktop CPUs in terms of production volume. There will be plenty of information of future technologies that we can expect to see within the next four to five years. In fact we hear that the information coming out of IDF Fall will exceed any other previous IDF and we expect Intel to take full advantage to make its pitch to the 5000-or-so developers and a few hundred journalists from around the world.

There is not much AMD can do during this time, but I understand that it is a bit awkward for AMD executives and the marketing department to keep quiet and simply watch as Intel conquers the front pages of tech publications for a full week. Rather than wait for journalists to check back with AMD and ask the green team’s take on Intel’s claims, the company this year decided to take a pro-active approach again and relay its message before the feeding frenzy begins.    

So, what is AMD’s message?

1. It’s all about graphics

Note to AMD: We got it. A great processor alone does not make you happy anymore. I believe most journalists out there are absolutely with AMD on this one, but repeating this story over and over does not convince us even more. There is nothing new to this story for IDF, other than I noticed a new marketing term for visual quality – “Eye Definition.” I leave that up to you how well this one works.

AMD’s balanced platform approach consisting of a good-enough processor and strong graphics are probably the best weapon AMD currently has against Intel. As long as Intel graphics chipsets remain weak, this is AMD’s most effective pitch. However, the Radeon 4800 was just announced, so there is really nothing new from AMD and the company did not provide anything about what may come after the 4800 series. The company needs to follow up on this one quickly to keep its traction in the market.  

 

 

 

2. Larrabee: Discussing a paper product

Nvidia already zeroed in on Larrabee and sent a document questioning Larrabee’s product claims to journalists. AMD now chimes in and asks questions about Larrabee’s scalability, power consumption and a possible failing software model for the new architecture. AMD’s executive said that “Larrabee feels like the next Itanium” to them, which may be a bit early to say, since we really haven’t seen a product yet. Also, AMD failed to mention that it does not have the best track record with choosing a software model for its stream processors either – the software model was changed twice within the past 12 months.

 

 

Attacks on Larrabee are to be expected and Nvidia in fact does this in a very efficient way while providing a very convincing platform, which is the market leader at this time. But these shots are little more than scatter shooting right now as neither AMD nor Nvidia know what they are actually shooting at. Intel needs to provide a lot more details about Larrabee before anyone can make any reasonable claims what this product will be able to do and what not.

Larrabee is without doubt on AMD’s radar, but the company says it is “delighted” with its own direction, which involved a switch to an open programming model of its graphics cards. Beyond that, we don’t know much. Criticizing a high power consumption of Larrabee does not work (yet), as we don’t know how much power Larrabee will consume and AMD just rolled out its own 270 watt graphics card .

3. Opteron is fast, Xeon is slow and expensive

We also heard (again) that the company’s Opteron chip is faster and more power efficient than Intel’s Xeon processors. We don’t want to go too deep into this one, but anyone who is researching this topic knows that at least the power claims are justified as far as certain benchmarks are concerned. However, Intel has been cutting deep into Opteron revenues and shipments and, ultimately, AMD margins. There is an obvious problem. AMD needs to come up with more convincing messages and products that are even more compelling than the B3 Opterons it is offering today.

Nehalem-based Xeons are just around the corner and it does not take much to see that these processors will continue the march that was started with Intel’s Woodcrest Xeons back in Q3 2006. So, what is AMD’s plan? First, AMD hopes that the dual-socket and four-socket Nehalems won’t be available anytime soon and then there is the  45 nm Shanghai that will be shipping in Q4. AMD said that it believes that actual Shanghai server will be available in Q4 as well. Desktop processors will follow “quickly”, which, if AMD will keep its traditional product strategy in place, means that 45 nm desktop processors will roll out within 60 – 90 days after the Shanghai debut.

There was no additional information about Shanghai available.

 

Read on the next page: 45 nm is late, Conclusion

 


4. 45 nm is late

AMD will not agree with this statement. But let’s cut through the chase and let’s be honest: Dirk Meyer, recently promoted to the role of CEO, stated during AMD’s Q2 2006 conference call that AMD would accelerate its 45 nm development and ship 45 nm processors by mid-2008 and close the gap to Intel by 6 months. It is unclear what Meyer exactly meant with “shipping” at the time (Samples? Volume?), and you can always twist and stretch the meaning of words. What we know, however, is AMD certainly has not closed the gap to Intel. AMD was one year behind Intel with 65 nm and AMD will be one year behind Intel with 45 nm, senior vice president Randy Allen confirmed today. He told us that AMD aims to catch up with future generations.

However, Allen now claims that 45 nm “is ahead of schedule”, which just does not make much sense to us, even if he explained this remark by stating that 45 nm is ahead of schedule in terms of “clock speed and performance”. In some way, it appears Allen and Meyer may not have been on the same page what mid-2008 actually means and the inconsistency of the message originates in miscommunication.

No one of us journalists really cares if AMD is behind or ahead by a few months. If AMD can squeeze more months out of 65 nm and remain competitive, then this is actually a good thing and just shows how strong the manufacturing process really is. I just don’t like being told that AMD is ahead of schedule, when they are actually behind.

 

 

      

Conclusion: Is AMD ignoring a threat?

I was left scratching my head over the briefing AMD gave today. There was no news, just the mantra we have heard multiple times this year already and lots of doubts and concerns about Intel’s technology. My first impression was that in targeting Intel’s products with today’s technology, AMD really had not much to say about how it will be answering the obvious threats Nehalem and Larrabee are – with the exception of the balanced platform message. No news about Shanghai, no news about the 45 desktop part, no news about a future graphics part. Isn’t IDF a great opportunity to provide future product plans that would enable us to see what the future will bring?

This is a completely subjective impression, but at least to me, Nehalem feels like Core 2 Duo all over again. Back in spring of 2006, AMD had a similar marketing and PR strategy in place, discounting Core 2 Duo as a paper product that won’t be able to touch the Opteron and Athlon X2 in terms of performance and power efficiency. We all know that Core 2 Duo and its derivates almost bankrupted AMD, stripping AMD of virtually all advantages and market gains it had made in the 18 months before. Back then, I was concerned that AMD would underestimate Intel and I have the same feeling again.

So, is ignoring AMD an obvious threat? I doubt that AMD is underestimating Intel internally, but AMD’s public presentations leave the impression that the company isn’t taking Intel’s products serious enough. AMD may already have been at the mercy with Intel and its Core 2 Duo processor and Nehalem may feel like a reboot to Q3 2006 for AMD. Only now, AMD has a substantial weaker position from a financial point of view, but is prepared much better to deal with another hard hitting technology: Intel’s bases are loaded and AMD really needs to come up with a really good pitcher.

Following the conference call, I actually got in touch with AMD to discuss much of the content that was discussed - content I did not agree with and content that should have been discussed, but was not discussed during the call. And there are certain aspects that shed light on teh reasons why AMD gave the presentation in this way. Perhaps most importantly, there were journalists with different technology knowledge levels on the call and AMD chose to emphasize its graphics message, to provide a counter weight against Larrabee. There is a tremendous amount of pressure on AMD to limit the possible damage that can result out of IDF and questioning certain claims about Nehalem & Co. was the preferred way to build up a defense.

Also, we should not forget that AMD constantly lives with the dilemma that it is out-resourced and out-spent in comparison with Intel. If it reveals too much about its future strategy and Intel likes that strategy there is at least the theoretical chance that Intel could take this idea and deliver a product even before AMD. This was probably a key reason for the company to switch its entire communications strategy and remain completely quiet about a new product until it exactly knows its specs and capabilities – and is convinced that Nvidia or Intel can’t beat it to market anymore. In turn, if AMD does not provide any information, which really was the case today, there will be doubt about the competitiveness of the company’s products. 

Undoubtedly, next week there will be questions whether AMD will be able to withstand the force of Nehalem and Larrabee. But will AMD go away? Well, it may split, but it won’t go away. Even if Intel could strip AMD of its remaining life lines, a functioning AMD is in the best interest of us all: Intel can control its anti-trust concerns even if AMD turned out to be a major pain in the you know what for them – but keeping AMD small and at a manageable level is actually much better for Intel than no AMD at all. System vendors and consumers need AMD as a balance in the market and they need two strong players as suppliers. And strong AMD is the reason why we saw the stunning innovation in the microprocessor industry this decade.

I don’t want to paint a completely dark scenario. In fact, it is obvious that the company enjoys a great deal of sympathy in the market right now. I also agree that AMD has impressive products in its lineup today and there are reasons its PR strategy has to be different than what Intel does. But the current approach just does not work and will create more problems for AMD.