Project: 6509 (Run 16, Clone 43, Gen 21)

[AgrFan]

I'm getting 44PPD on this unit. It's running on a P4 2.6 HT Northwood with 512KB cache. I''ve had two units produce 50% PPD recently on this box. I normally get 100-125PPD with GROMACS and Protomol units. I copied the files to a box with a P4 2.8 HT Prescott with 1MB cache and got 110PPD as expected.

Someone really needs to look into this stuff. It's ridiculous to have to deal with large differences in PPD between similiarly clocked P4 chips (Northwood vs. Prescott). If there's a reasonable answer to this, I'd be interested to hear it. For now, this box is getting turned off. The low PPD is not worth the electricity.

And before anyone suggests it, these boxes will not run the GPU client. None of them have PCI-E slots.

Code: Select all

[07:22:52] *------------------------------*
[07:22:52] Folding@Home Gromacs Core
[07:22:52] Version 1.90 (March 8, 2006)
[07:22:52] 
[07:22:52] Preparing to commence simulation
[07:22:52] - Looking at optimizations...
[07:22:52] - Created dyn
[07:22:52] - Files status OK
[07:22:53] - Expanded 993612 -> 5047657 (decompressed 508.0 percent)
[07:22:53] - Starting from initial work packet
[07:22:53] 
[07:22:53] Project: 6509 (Run 16, Clone 43, Gen 21)
[07:22:53] 
[07:22:54] Assembly optimizations on if available.
[07:22:54] Entering M.D.
[07:23:00] Protein: TR574_17 in water
[07:23:00] 
[07:23:00] Writing local files
[07:23:01] Extra SSE boost OK.
[07:23:02] Writing local files
[07:23:02] Completed 0 out of 250000 steps  (0%)
[08:24:54] Writing local files
[08:24:55] Completed 2500 out of 250000 steps  (1%)
[09:26:44] Writing local files
[09:26:44] Completed 5000 out of 250000 steps  (2%)
[10:28:35] Writing local files
[10:28:36] Completed 7500 out of 250000 steps  (3%)
[11:30:36] Writing local files
[11:30:36] Completed 10000 out of 250000 steps  (4%)
[12:32:30] Writing local files
[12:32:30] Completed 12500 out of 250000 steps  (5%)
[13:34:21] Writing local files
[13:34:21] Completed 15000 out of 250000 steps  (6%)
[14:36:15] Writing local files
[14:36:15] Completed 17500 out of 250000 steps  (7%)
[15:38:08] Writing local files
[15:38:08] Completed 20000 out of 250000 steps  (8%)
[16:40:04] Writing local files
[16:40:04] Completed 22500 out of 250000 steps  (9%)
[17:42:02] Writing local files
[17:42:02] Completed 25000 out of 250000 steps  (10%)
[18:43:56] Writing local files
[18:43:56] Completed 27500 out of 250000 steps  (11%)
[19:45:51] Writing local files
[19:45:51] Completed 30000 out of 250000 steps  (12%)
[20:47:48] Writing local files
[20:47:48] Completed 32500 out of 250000 steps  (13%)
[21:49:46] Writing local files
[21:49:46] Completed 35000 out of 250000 steps  (14%)
[22:51:44] Writing local files
[22:51:44] Completed 37500 out of 250000 steps  (15%)

[bruce]

FAH does not design any WU for any particular chip. Intel (or others) design chips with certain features that improve performance and FAH uses whatever they find.

I can't answer your question directly, but I'd guess that the differences in CPU speed and FSB speed make a small difference but the lack of a 1MB cache may be particularly important for this particular project.

I'd probably turn that machine off, too. The Northwood is over 8.5 years old and in processor generations, that's a long, long time. The savings in your cost of power, alone, would fund an upgrade in a relatively short period of time. In any case, that's your choice to make.