Folding Forum

Subject line says it all......why such a big difference?????

Maurice

Unles you fold on Stanfords benchmark machine, you will see this happen. Every piece of hardware handles each of the various projects differently, and every project will have slightly different hardware stress points (clock speed, RAM, L2 cache, etc.). With Win SMP, I do very well on the 2653 and poorly with any of the other projects on an Athlon64 X2 6000+. I just spent a week with a q6600 making more points on 4 CPU clients than I could with the SMP.

It's magic, just pure magic. Logic doesn't enter into it.

Since you didn't give any details about your setup, all I can say is it's because your setup isn't the same as the benchmark machine at Stanford.

yes,stupid of me not to give details.....sorry

2.4 gig Intel Core 2 Quad.........Asustek Basswood 3g1.05....in an HP systemk

ChelseaOilman wrote:Since you didn't give any details about your setup, all I can say is it's because your setup isn't the same as the benchmark machine at Stanford.

What is the benchmark machine for SMP ?

From: http://folding.stanford.edu/English/FAQ-SMP#ntoc13

a dedicated Macintosh Pro with 2 - 2.33 GHz Dual Core Xeon processors. (more specifically, 2 Woodcrest 5140 processors with 4 MB cache (each), 5 GB FBDIMM Memory (667 MHz DDR2), 1.33 GHz Bus)

In this post, viewtopic.php?f=12&t=1143&p=9779&hilit=dual+woodcrest#p9779, the user of a dual Woodcrest (5160) also complains about the ppd on the 306x WUs. I suppose there may be some other factor in the configuration of his machine causing his dual Woodcrest to underperform the benchmark machine or perhaps not.

This is normal ... p2653 always have a higher output than other SMP projects

Before I get jumped on for my previous post, there could, just as likely, be some factor in the dual 5160 poster's hardware configuration to cause his machine to outperform the benchmark machine on p2653. However, on a Q6600 @ 3.33 GHz running Linux SMP, the 306x series produce about 1000 ppd less than p2653 and 700 ppd less than p305x WUs.

ChasR wrote:Before I get jumped on for my previous post, there could, just as likely, be some factor in the dual 5160 poster's hardware configuration to cause his machine to outperform the benchmark machine on p2653. However, on a Q6600 @ 3.33 GHz running Linux SMP, the 306x series produce about 1000 ppd less than p2653 and 700 ppd less than p305x WUs.

From my experience 306x WUs have the largest fluctuation in PPD compared to any other SMP WU I've run across. I regularly see several hundred PPD difference between 3060s on the same machine even the same WU % to % varies as much as 10%.