You, or others on this board, may recall when I was active on the Intel board (until mid-2005) and Intel was pretty good to me. I accepted a three year assignment FAR from the USA in summer of that year, and moved almost my entire position into real estate, mostly land in high quality areas (beachfront for the most part).
Now that I'm back in the states with a more relaxed schedule, I'm wading cautiously into what looks like a potentially interesting buyer's market. The announcement this week induced me to take a position in AMD.
While I think that AMD made some questionable investments over the last few years, I don't think that their current situation qualifies as their most desperate condition - after all, AMD made an art out of desperate conditions in the 90s.
On the plus side, AMD took advantage of Intel's single minded pursuit of IA64 in the high end and Netburst on the desktop to make a game changing move with X64 that put them in the server space for the first time ever, and pretty much ended both Itanium as anything but a niche product, and consigned the whole tejas line to oblivion. On the negative side, they used that momentum to open up an as yet unproven move to integrate graphics and traditional processing, rather than consolidate.
Despite the negative comments here and elsewhere, AMD still has a 20% share in servers - something that would have been unthinkable 6 or 7 years ago. They also have share about equal to Intel in the North American retail space. Leaving aside the birth pains of Barcelona, the product problems they have are not in the 'flag on the mountain' space, where Intel has had a lead since Conroe, but in power usage in mobile, which is the real money battleground in client.
In the server space, for real workloads as opposed to benchmarks, AMD even today has some key advantages - in particular in driving large numbers of virtual machines, which is a hot area, and in other loads that need frequent context switching. This is because Intel, up until now, has not implemented either page table nesting or integrated memory control, which are the keys to that higher performance.
Intel has addressed both issues with nehalem, with the power benefits of their core architecture included, but performance in the server workloads they had stumbled with in the past is not yet proven. My guess is that nehalem will be a strong part and will challenge AMD in those areas where AMD has an edge today, but it is not likely to be a walk away, and the story has yet to play out.
From what I have seen publicly, the AMD news this week ought to make them more able to maintain the pace, and keeps them from being hamstrung by a cash crunch. While there are of course many questions as to how well they will perform with this new business model, I think that the opportunity is interesting enough to justify a modest investment... |
