Het is een tijdje geleden, maar Thomas Pabst heeft weer eens een blurb in elkaar gedraaid. Wederom haalt hij in deze editie weer uit naar de kloterige Athon mobo support van Taiwanese bedrijven en Intel's i820 fiasco. Lees zijn complete verhaaltje hier.
This is funny enough, since Intel doesn't really have much to laugh about. The disaster with i820 and RAMBUS is the biggest blunder Intel made since the Pentium floating-point flaw. Now we're on the eve of the Coppermine-release (a detailed architectural overview of Coppermine will follow this week), but this processor that is supposed to make Intel look better again doesn't have a platform to run at. Intel's own i820-chipset might not be out until next year, so that Coppermine will either have to reside in a highly pathetic i810e-platform with integrated slow-motion 3D-graphics, 133 Mhz FSB, but only PC100 memory support or in a non-Intel VIA 640-platform, which at least supports AGP4x and PC133 memory. This is almost as if Athlon would have been released without a chipset. It seems incredible to me that the motherboard makers put up with this and still support Intel. Thousands of i820-motherboards that were ready to ship on the planned release date of i820 had to be called back and taken apart. This must have cost hundreds of thousands. It would have been the perfect timing to push sales of Athlon-platforms, but something kept the Taiwanese motherboard makers from doing that. It seems rather obvious where this 'something' comes from and it has to be possible that the FTC finds some proof for that. Intel is so scared by Athlon's performance and popularity, that they even made up a rumor of releasing mysterious 'Willamette' at 1100 MHz 6 months early in December 1999, just to get some positive headlines. May I ask on which platform Willamette is supposed to run at? Maybe Intel should try and use the good old 430FX-chipset for it. Intel looks highly stupid, but they decided to make sure that AMD looks at least as stupid as well. [break]Verder vertelt hij ook nog een klein stukje over de lagere L2 cache snelheden die hoog geklokte Athlon's mee gaan krijgen:[/break] AMD plays its own part to worsen the situation. Dana Krelle admitted to me that the motherboard situation is currently the worst thing for them. However, the next Athlon processors, starting with the Athlon 750, will clock the second level cache at only 2/5 of the core clock, not at 1/2 as Athlon 500-650. This will worsen the performance-scaling, especially once the multiplier shrinks down to 1/3 in Athlons beyond 800 MHz. Coppermine will continue to run it's L2-cache at 1/1, regardless how high the clock speed. It's a matter of fact that AMD needs to get their .18 micron Athlons with on-die L2-cache ready as soon as possible, because once i820 does run, Intel only needs to increase Coppermine's clock and it will overtake Athlon. This could be the end of AMD.