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
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: Saturn V who wrote (67753)8/5/1999 8:11:00 PM
From: kapkan4u  Read Replies (3) of 1576951
 
<PB: I believe the balance of the information contained in the article is extremely positive for Intel, and the box makers.>

<Saturn V: Hm, I am still pondering over the possible reasoning behind your statement.>

PB's statement is so bizarre that it prompted many questions on this thread but resulted in no answers. Since PB is not giving an explanation, let me try to help him out. Please note that what I say is a total speculation. I have no inside information, and just put together bits of information that were posted on SI or elsewhere to make the scenario I am about do present plausible (but not likely).

The desktop CuMine was delayed one week after the release off Athlon benchmarks showing 40% performance advantage in specFP among other things. When the strength of the Athlon FP implementation became known to Intel they faced two alternatives:
1. Release CuMine on schedule and rely solely on process technology to try to keep up with Athlon.
2. Make a quick architectural fix by adding some new instructions. A good candidate would be adding double-precision scalar SSE instructions to help with specFP numbers. The latest Intel compiler already supports single-precision FP SSE instructions as an alternative to x87. It would be easy to extend this to double precision in the compiler. I don't know how hard would it be to implement this on the hardware side.

If the second alternative was chosen, then the CuMine process/speed problems could be a smoke screen.

What made me think that the second alternative is possible?

First, Elmer posted that CuMine would be competitive with Athlon on specFP, within few % points (in his words) due to compiler use of SSE. The problem with Elmer's statement is that only one specFP95 benchmark is using single-precision FP. I hope this doesn't get Elmer in trouble. Second, the mobile CuMine is also delayed (it has to be if new instructions are added). Third, PB posted that the latest slip is “extremely positive for Intel”. I hope this doesn't get PB in trouble. Fourth, the Register is quoted someone at Intel saying the original CuMine delay was a knee-jerk reaction to Athlon.

Far fetched? I think so, but amusing nevertheless.

Kap
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  

HomeMailHotSubjectMarksPeopleMarksKeepersSettings
Terms Of UseContact UsCopyright/IP PolicyPrivacy PolicyAbout UsFAQAdvertise on SI
© 2025 Knight Sac Media.  Data provided by Twelve Data, Alpha Vantage, and CityFALCON News