AMD’s server and workstation comeback

[shahed26]

Analyst Opinion - This week I have been spending some time with AMD listening to an update of their workstation and server roadmap. AMD’s message: We are healthy and we are executing once again. However, they admitted that their misses in 2007 hurt the company a great deal in revenue, profitability and - even more importantly - credibility.

These events are similar to a pep talk before a big game. They are long on promise and hope, but may not reflect what will happen when the teams hit the field. For much of what they are talking about, the game is still months off so take that into consideration as you read this. In effect, I am in the AMD locker room and only telling their side of the story today. Someone standing in the Intel locker room at IDF (Intel Developer’s Conference) next week will undoubtedly get a different picture and both will probably not reflect what happens in the real game.

AMD’s OEM fan club

For AMD, one of the things that has always worked for them, at least while they are much smaller than Intel, is that the OEMs (and this applies to PCs, servers, and workstations) want them in the market, because they don’t like the idea of not being able to bid companies against each other. In addition, when you have a single vendor and that vendor has a problem executing, it takes out the entire eco-system and these OEMs don’t like that kind of risk.

Typically, the dominant vendor has the advantage, because the industry standardizes on them, but because of this single vendor fear, AMD gets a significant boost, if they can either meet or beat Intel values. Currently, the industry is struggling with a change in how you measure the server side of this market shifting from pure performance metrics to those that factor in energy efficiency. AMD believes they have a significant advantage once you factor in energy.

Shanghai

Shanghai replaces Barcelona later this year and, on paper, it looks impressive. AMD is promising better performance, lower idle power consumption, improved/pricing for performance and minimal changes over Barcelona. These are significant improvements and AMD argues that this new Shanghai part will significantly exceed Intel’s Harpertown in performance. We will have to wait for independent benchmarks to confirm whether this is the case and Intel clearly isn’t going to be standing still. Their roadmap will be presented at the Intel Developer Forum next week.

Shanghai won’t really be aggressively moving into markets until the first quarter of next year, so the real competitive results, in terms of sales, probably won’t be known until AMD and Intel report sales numbers at the end of the first quarter in 2009.

One of the reasons for the delay between launch and market ramp is the validation process that has to occur when any new part is brought to market. While Shanghai is similar to Barcelona, it has enough differences to still require a validation test. A big problem with Shanghai now is that it is both substantially better than Barcelona and coming relatively soon, which could cause buyers to defer server and workstation purchases until the CPU becomes available. However, one good part of a bad market (and economic conditions in general are bad), is that funds available this year may not be available next year and funds for purchases may be in the “use it or lose it” category, which should offset this tendency to defer somewhat.

Microsoft’s role in AMD’s future

One of the advantages that AMD has over Intel is they don’t seem to bump heads with Microsoft very much. Intel and Microsoft are both dominant in their respective segments and neither likes to give up leadership to the other. As a result, Intel tends to be very aggressive on Linux and Microsoft tends to support AMD more aggressively in response. Microsoft apparently is tuning its virtualization offerings, which have actually started to sell impressively well and really stunned VMware, the firm that really put virtualization on the map.

What is often kind of funny to watch is how Microsoft and Intel do this dance. It’s almost like watching a bad marriage where everyone knows the couple is living separately, fights a lot and have new partners. But the couple publically acts like they are still close. In the case of Linux for Intel and AMD for Microsoft I often wonder, how this would pan out if these companies were people.

This means that while Microsoft does favor AMD, their public posture doesn’t fully represent just how close the two companies actually are. And for servers in particular, this should show some strong additional benefits for AMD’s offerings as the next generation of products from both AMD and Microsoft roll to market.

Industry problems

One of the things that both AMD and Intel have to deal with is that the future is uncertain. Currently both Intel and AMD are chasing, particularly in the server space, the idea of massive multi-core roadmaps, which assume the future will be able to use these new systems. While Microsoft has shifted to a per socket pricing model, other firms like Oracle still price their software on a per-core basis, which works against this multi-core trend as the software cost negates much of the hardware advantage. In addition, programming for massive multi-core systems has been incredibly difficult and most existing applications have trouble scaling to more than two or three cores let alone the 16 to 32 that are clearly coming.

Consolidation - combining many servers into fewer servers with more cores - and virtualization (which allows a single server to act as if it were a large number of servers) clearly help, but these shifts are still in their infancy and the market often shifts between models suggesting the possibility that the current massively multi-core roadmap may not accurately reflect the real future. And this idea of cloud computing, which shifts the market from a focus on selling hardware to selling services, has massive implications for this segment.

Wrapping up

AMD feels like they have their act together now and their recent competitive successes in the graphics segment appears to confirm this.

Barcelona is in the market, finally, and Shanghai, if it arrives as promised, should help confirm that AMD is on track again. Granted, until it actually ships, this is more promise than reality, which adds a cautionary element. Shanghai, everything else aside, will be the most valuable product to AMD in terms of restoring the company’s credibility if it ships in Q4. Intel clearly isn’t giving any ground and we’ll likely see next week that Intel has come to the court to play as well. This should be an interesting second half in the processor wars, at least for servers and workstations.

http://www.tgdaily.com/content/view/38883/128/