SI
SI
discoversearch

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: Charles R who wrote (73905)10/5/1999 2:18:00 AM
From: Paul Engel  Respond to of 1572369
 
Chuckles - Re: "I forgot a biggie here. Software! Today the x86 world of software is optimized for PII/PIII. Optimizations for Athlons (including the new instructions) should yield significant additional benefits"

AMD didn't forget.

Didn't they specifically re-code a graphics benchmark - by Futuremark - to SPECIFICALLY MAKE USE OF ATHLON OPTIMIZED CODE ?

"A Canadian benchmarking operation claimed Advanced Micro Devices Inc. modified tests to give its Athlon microprocessor an unfair advantage over Intel Corp.'s Pentium III. "

Here's the PROOF !

{===========================================}
August 31, 1999

COMPUTER RESELLER NEWS : Special from Electronic Buyers' News -

A Canadian benchmarking operation claimed Advanced Micro Devices Inc. modified tests to give its Athlon microprocessor an unfair advantage over Intel Corp.'s Pentium III.

FutureMark Software Corp. Ltd. said AMD is violating the trademark of the Toronto company's 3DMark 99 MAX test suite and terms of a related licensing agreement. The company said AMD's actions were "inappropriate."

According to FutureMark, AMD modified the 3DMark 99 MAX test by optimizing the DLL test code for its recently released Athlon. AMD's own test results demonstrated that, in one of the two 3DMark tests, the 600MHz Athlon outperformed the 600MHz Pentium III by more than 30 percent. Of the 13 test results accompanying AMD's Athlon performance assertions, the 3DMark figures were among the highest in favor of the chip.

AMD erred by comparing the optimized Athlon results to the generic test suite used with the Pentium III, said FutureMark.

Mention of the customized 3DMark test suite was fully disclosed on page 39 of a 42-page software performance guide that accompanied the test results, an AMD spokesman said.

"The reference should have been included in some other places," he said. "But in the excitement of the [Athlon] launch, that reference was overlooked."

AMD and FutureMark were in talks to resolve the matter, the spokesman said.

{=================================}

Paul



To: Charles R who wrote (73905)10/5/1999 12:51:00 PM
From: Steve Porter  Respond to of 1572369
 
Charles,

Allow me to jump in here.. since we are talking about software optimizations for the Athlon.

Athlon is much more Pentium II optimization friendly than the K6-x is/was. (This is a dead product line in my mind.. sure they may still 3 million of them, but I doubt anyone is really focusing on a chip that probably has a gross margin accross the entire product line of 5-10%).

Athlon has a 3 cycle load, just like the P6 core. This allows means that code generated by compiler will be more optimally scheduled on Athlon than the K6-2. However, there are some significant differences as well. For example, there are only 2 execution units (for int) in the P6. Athlon has 3. This means that you need a maximum of 8 instructions between a load and the use of the data loaded. However, luckily Athlon has a large scheduling buffer, which allows it to absorb these differences in all but the most memory intensive code.

Additionally with Athlon's 3 load/store units, the scheduler can actually advance load quite far up a code chain, thereby internally creating the necessary 3 cycle (or 8 instruction) padding. I'm just glad I don't have to design today's microprocessors. The logic of the scheduler must be nightmarish. Add to that the register renaming, and the fact teh FPU is re-schedulable too.. ouch..

Steve