Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?

To: Cirruslvr who wrote (43964)	12/24/1998 12:24:00 AM
From: Cirruslvr	Read Replies (1) \| Respond to of 1574485

L3 cache improves performance up to 9.5% in Winstone 98 Anand posted more benchmarks at his site. Although this review was of the Kyrotech K6-2 500, he has a few benchmarks of a K6-3 running a 350MHz with 0 L3 cache, 1MB L3 cache, and 2MB L3 cache. Here are the numbers when running Winstone 98: K6-3 350 (2MB L3) - 28.9 K6-3 350 (1MB L3) - 28.0 K6-3 350 (0MB L3) - 26.4 Difference btwn size of L3 cache: 2MB and 1MB - 3% 1MB and 0MB - 6% 2MB and 0MB - 9.5% There won't be a desktop version of the 350 K6-3 so I don't know why he tested it at that speed. Tomorrow, he will post an article that deals specifically with the K6-3 and L3 cache. He will probably test it at more speeds at that time. Here are the numbers when running Winstone 99: K6-3 350 - (2MB L3) - 20.2 K6-3 350 - (1MB L3) - 19.2 K6-3 350 - (0MB L3) - 18.2 Difference btwn size of L3 cache: 2MB and 1MB - 5% 1MB and 0MB - 5.5% 2MB and 0MB - 11% Go here to see everything: 206.132.42.114

To: Cirruslvr who wrote (43964)	12/24/1998 12:26:00 AM
From: Elmer	Read Replies (1) \| Respond to of 1574485

Re: "I can understand you thinking it was run on NT after you saw the K7's score. ;) Speaking of the K7, what do you think about its score?" Assuming these benchmarks are meaningful, the K7 performs well but certainly not as astonishingly as we have been lead to believe. If the K7 numbers don't improve under NT, and the Katmai's numbers surely will, then the K7 will have a hard time beating Katmai, clock for clock, and I think the Katmai will run away from the K7 in frequency, at least on .18u. Keep in mind that both the K7 and Katmai aren't available yet so these samples may not reflect the true performance potential. The early versions of the K7 will be handicapped by a slow, narrow L2 bus. Later versions may bring the L2 onboard and improve performance. The Katmai will do the same with it's derivatives. Right now, I think a Katmai running at 500mhz under NT will post higher Winstone98 numbers than those posted for the K7 at the link I previously posted. However I don't think those numbers posted are indicative of the K7's true potential. I expect it to be a bit faster when introduced. Re: "Also, a real Katmai, run at 5x100, will score lower than what that guy got. This is because he has the bus speed at 112 (4.5x112). This makes the PCI bus, AGP card, and memory run 12% faster (37 1/3 vs 33 1/3) than at "normal" clock speed. When the 133MHz bus hits the Katmai, the PCI bus and AGP card will still run at 33MHz (133/4) while the memory will of course run at 133MHz. The PCI bus and AGP of the K7 will run at 33MHz also (200/6)." You're wrong about the AGP port (not a bus). The current version runs at 133mhz and once the 133mhz FSB camino chipset is here the AGP will run at 266mhz. EP