Ten, re:<AMD, on the other hand, isn't held to the same standards. Their definition of high volume isn't the same as Intel's. That's not meant to be a slap on AMD. It's meant to show that AMD as a smaller company can release higher speed grades without having to demonstrate volumes. No one expects AMD to deliver Intel-sized volumes, and that's not even AMD's goal (or at least it shouldn't be).>
My comments on the 800 MHz CuMine and Tenchusatsu's comments about AMD above:
1. AMD is really in a tougher position here. If total production of Athlons is only 600K by 11/1 and they yield 5% at 700 MHz, thats only 30K chips. Not enough chips for IBM or CPQ to release a SKU to the public. The big guys would get mad if AMD sold them to the little guys first. If Intel yields 5% at 733 MHz and makes 2M chips by 11/1, thats 100K chips -- enough for 1 or 2 SKU's, but not the usual dozen.
2. The fact that the demo did not use "exotic" cooling does not mean it didn't use, say, a Peltier, not suitable for a high humidity environment.
3. The demo machine did not have to run 24/7.
4. The demo machine only ran its demo software -- any disclousure here? Did anyone see it run Winbench, for example?
5. According to AMD specifications, a chip has to run at its rated speed at 65-70 degC case temperature and 0.05v BELOW rated voltage. I assume Intel is equally stringent -- these requirements are usually equivalent to a least a speed grade's worth of de-rating.
6. Taking 2,3,4,5 into consideration, the CuMine speed grade I expect to see at intro is no higher than 733 MHz.
7. tomshardware.com has run several Athlon's through Winbench, high end Winbench, etc., at 750 MHz without stability problems.
8. JC reports today that an Ebay vendor is selling rewired 500 and 550MHz Athlons as 700 MHz units, with a 90 day warranty. This implies that the majority of Athlons, even the 500 MHz units run at 700 MHz.
I predict that AMD will release, and a Tier 1 will announce use of, a 750 MHz Athlon within 48 hours of Intel's CuMine 733. (BTW, 766 is not a possibility -- not a .5 multiple of 133)
Petz |
