AMD IS BETTING its shirt on the x86-64 AMD64 architecture. The Opteron has been released and has delivered some impressive results. But there are already rumours that the Athlon 64 might be pushed back to next year. But there's evidence for an even more startling conclusion. There is evidence that could mean the Athlon 64 isn't just going to be pushed into next year, there is evidence that it could be pushed altogether.
The first piece of evidence in this conspiracy theory is provided directly by AMD. Travel to any show where the firm has been flashing its AMD64 wares and there has been a distinct lack of Athlon 64 machines present. They have been seen but only in minimal quantities. Compare that with the Opteron. It's much more difficult to make than the Athlon 64 but AMD has had bucket loads of them about.
The next piece of evidence is on the benchmark front. A little while back Fudo managed to get some benchmarks done on an Athlon 64. The results were less than impressive. Even considering that some drivers were beta in the machine, they were a long way short of what everyone was hoping for. Indeed, they were a long way short of what an equivalent clockspeed Athlon XP is capable of. In comparison, the Opteron is startlingly fast only a month later. Did all that beta stuff get fixed in such a short space of time?
Then there's financial evidence. One of the reasons for the introduction of the Athlon 64 was the expense of the socket 940 used by the Opteron. But now that Opteron boards are available it's quite possible to check that. Comparing like with like was fairly easy. Digging up the MSI K8D Master on a site revealed a price of around $530. The MSI E7501 Master-LS2 carries a very similar specification but is for the Xeon. It costs about $50 more. Strike cost from the list of reasons. These boards may be expensive in themselves but they're low volume server boards, a mainstream product would cost much less.
Then there's the price of the processors themselves. The bottom end Opteron is priced at about $300. That's fine for a reasonably top end product right now.
But enough of the finances, what about the surrounding players? Take Nvidia and Via for example. They've been touting Athlon 64 chipsets for over a year. Neither of them showed any interest in the Opteron. All of a sudden these traditional home market players are showing off "workstation" motherboards. There is no doubt that these are the same chipsets as the Athlon 64 ones. After all, the processor links to the chipset by Hypertransport so there's not much to change.
Then there's the interesting little marketing war going on inside AMD. Some of the marketing droids want to push the Opteron as a games machine. No reasons were given for this other than the Opteron being really very good at it. But what if it's also because the Athlon 64 really isn't very good at it?
Let's make this absolutely clear. This is a conspiracy theory. All of the evidence is circumstantial. AMD, when asked to comment, denied everything and said that the Athlon 64 was going to be launched as planned.
But then AMD would deny it, wouldn't it? If the conspiracy is right, AMD's best move would be to keep postponing the Athlon 64 until there was no point in releasing it at all. Either that or release an Opteron with a smaller cache, say 256KB or 512KB, and call that the Athlon 64. It could claim that socket 940 was better for the customer and so socket 754 was being dropped.
The simple thing about what's written here is that, even if it's not true, the top brass at AMD should be taking a long hard look at the evidence and asking themselves whether they should make it true.
the inquirer
