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.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: Elmer who wrote (72711)	2/28/2002 12:37:03 AM
From: milo_morai	Respond to of 275872

Elmer, Kap is correct. Just on this one test alone shows more users the gains go down rapidly. Most FTP servers have over 400 users on them. Web sites can even be more. infoworld.com

To: Elmer who wrote (72711)	2/28/2002 3:35:30 AM
From: Petz	Read Replies (2) \| Respond to of 275872

Elmer, but in the Hyperthreading and Web apps benchmark, the advantage of HTh drops to 20% as the server becomes fully loaded (100 users). Fully loaded is the only benchmark that counts. (I see Milo said the same thing.) Petz

To: Elmer who wrote (72711)	2/28/2002 8:49:14 AM
From: minnow68	Read Replies (1) \| Respond to of 275872

Elmer, And for many applications, Kap was more correct than I would have ever guessed. From the gamepc.com review, there are some _shocking_ benchmark results. For example, the 50 MB windows media encoding on a dual Xeon took 2:34 without HT. It took 3:03 with HT turned on. Turning HT on slowed things down by 18%!!! Looking at all the gamepc.com benchmarks and the infoworld benchmark, I would not turn on HT on a machine under my control unless it was running a limited number of applications and I knew that HT would actually help those apps in the way that I used them. I've only seen about 20 benchmarks using HT so far, and the results as a whole have not been impressive. There seem to be just as many pathological cases where HT seriously harms performance as there are good cases where the HT seriously helps. Obviously, this picture may very well change as more benchmarks come in. Mike

To: Elmer who wrote (72711)	2/28/2002 10:52:01 AM
From: Gopher Broke	Read Replies (5) \| Respond to of 275872

Our tests of a hyperthreading-enabled Intel Xeon DP server showed, on average, a 45.71 percent increase in SQL transaction performance and a 31.13 percent increase in three-tier Web application performance, versus the same system with hyperthreading disabled. It doesn't seem right for them to just disable hyperthreading for the comparison system. Who knows what performance penalty that gives to a core that is designed to be hyperthreading. Also, it does look like these people have the objective of making the Xeon look good, which devalues their results. For example, their "averages" are taken over a range of number of connected users, yet the benefit of hyperthreading decreases with more users. Why on earth do this averaging? It is the benefits to a more heavily loaded system that are more relevant, so why include the 60% benefit for 15 clients, when it goes down to 34% for 50 clients? If the server only has 15 clients connected then you are not exactly worried about its performance are you? And why do they stop at 50 clients for the database test? I would expect the numbers for 75 and 100 clients to be equally interesting/significant.