Original URL: http://www.theregister.co.uk/2004/11/19/amd_top500_loss/
AMD's Opteron loses ground where it kind of counts
Not in my HPC cluster
The last couple of weeks have been full of AMD buzz. Analysts have upped their price targets for the company's share price. Executives have bragged about better than expected future performance. The rumors about Dell picking up Opteron have kicked into full gear once again.
Despite all this glory, AMD recently fell well short of expectations in one of the more public battles in the processor kingdom. It gained hardly any ground as the processor supplier of choice among high performance computing customers, as measured by the most recent Top 500 supercomputer list. By gained hardly any ground, we actually mean AMD and its Opteron processor lost ground. This is a dismal state of affairs for a young chip meant to be gaining steady market share.
The number of supercomputer installations using AMD's chips fell from 34 sites in June of 2004 to 31 sites in November. Meanwhile, Intel, which barely registered on this list three years ago, moved from 285 sites to 318.
Before the letters start rushing in, let's make a couple of things clear. First off, Intel has two chips to play with on this list - Xeon and Itanium versus Opteron (Athlon is negligible) for AMD. Secondly, Intel has far more resources than AMD to put toward "encouraging" labs and researchers to pay attention to its products. Intel's willingness to sweeten deals that block the purchase of Opterons has become legendary over the past year. In addition, much of the Top500 list is dominated by fairly well-established high-end processors that have loads of software already prepared for them and large server vendors helping out the cause of their respective chips.
It, however, must be noted that AMD, simply put, failed its backers over the past year in the high performance computing realm.
AMD last year used major wins at the Shanghai Supercomputer Center and Los Alamos National Laboratory to drum up massive amounts of press around Opteron. It claimed that Opteron would storm through the Top500 ranks and put the lowly 32-bit Xeon to shame. This clearly never happened.
The Top500 list is a pretty arbitrary measure of much of anything. For all its flaws though, it does give an indication of the direction the labs are taking with their systems. And high performance computing customers are often ahead of the curve, picking up gear that will later be used by enterprise customers. With that in mind, the lack of interest in Opteron clusters must be disheartening for AMD.
Opteron is a class chip, but if AMD is to gain serious ground against Intel it needs as much marketing and customer win help as it can get. Losing the supercomputing wars just as Intel begins to fight on the x86-64-bit front, doesn't bode well for the little processor maker that could.