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: TGPTNDR who wrote (41931)	11/20/1998 3:47:00 PM
From: Petz	Read Replies (1) \| Respond to of 1576310

TGPTNDR and Tenchusatsu, regarding K6-X integer performance. Winbench '98 is NOT an integer benchmark. In fact, of the 7 or 8 applications used in it THREE of them are spreadsheets, in which all calculations are floating point (even 'integer' calculations are done in FP in a spreadsheet). Integer performance includes data movement, test and branch and arithmetic. Of these, the K6-2 core is slower than a Pentium II in data movement, slightly faster in arithmetic and much faster in 'test and branch.' As a true integer comparison, look at the chess benchmarks someone posted a while back, or Norton SI; in both of these K6-2 beats Pentium II, clock for clock. I looked for some application-specific benchmark results for things like Word and Wordperfect but couldn't find any right now. The K6-2 core for the K6-2 CPU's above 350 MHz (referred to as the 'CXT') features some improvements to boost integer performance even more. According to AMD's website the new 'CXT' core implements "write combining" and improved "write allocation." These features alone account for at least a 10% boost in Winbench 98 performance at the same clock speed. Rumor has it that even the K6-350's will be using this new core as well. So, AMD has not ignored integer performance in the K6-3 and the K7. The weakest link in the K6-2's integer performance, data movement, has already been partially addressed. It will be further improved via the on-chip L2 in the K6-3 and the 1/2 speed L2 (>250 MHz vs 100 MHz) in the K7. The K7 also features a deeper lookahead buffer and more execution units, which should improve the test and branch aspect of integer performance. Petz