why low score for so many wu completed

[Ancient_Dragon]

When I view my team's status I have completed 89 WU and have 8670 score. Yet another team member has completed only 65 wu but has 13651 score. Can you tell me why such a large difference ? My computer is running 24/7, with 240 gig hard drive, 2 gig ram, and 1.5 gig cpu. Yet I don't seem to be getting very high score for all the work it is doing.

[Flathead74]

Your score to WU ratio would very much depend upon which client you are running,
and also whether you are getting a lot of E_U_Es or not.

Although you have included some system specs, without client/configuration information,
it is hard to speculate with any kind of reliability.

[Ancient_Dragon]

My computer is running Vista Home Premium, if that helps. Here are the first few lines of the log file

--- Opening Log file [May 9 16:41:37]


# Windows Graphical Edition ###################################################
###############################################################################

Folding@Home Client Version 5.03

http://folding.stanford.edu

###############################################################################
###############################################################################

Launch directory: D:\Program Files\Folding@Home


[16:41:38] - Ask before connecting: No
[16:41:38] - User name: Ancient_Dragon (Team 45978)
[16:41:38] - User ID: 512BAAEC491F3D9E
[16:41:38] - Machine ID: 1
[16:41:38]
[16:41:40] Loaded queue successfully.
[16:41:40] Initialization complete
[16:41:40] + Benchmarking ...
[16:41:46]
[16:41:46] + Processing work unit
[16:41:46] Core required: FahCore_82.exe
[16:41:46] Core found.
[16:41:46] Working on Unit 02 [May 9 16:41:46]
[16:41:46] + Working ...
[16:41:56]
[16:41:56] *------------------------------*
[16:41:56] Folding@Home PMD Core
[16:41:56] Version 1.03 (September 7, 2005)
[16:41:56]
[16:41:56] Preparing to commence simulation
[16:41:56] - Looking at optimizations...
[16:41:56] - Files status OK
[16:41:58] - Expanded 77037 -> 553787 (decompressed 718.8 percent)
[16:41:58]
[16:41:58] Project: 2170 (Run 27, Clone 109, Gen 0)
[16:41:58]
[16:41:58] Assembly optimizations on if available.
[16:41:58] Entering M.D.
[16:47:45] Protein: p2170_lambda_obc_300K
[16:47:45]
[16:47:45] Completed 446175 out of 500000 steps (89)
[16:57:05] Writing checkpoint files
[17:07:34] Writing local files

[Hyperlife]

You're running the single-CPU client, which is benchmarked to produced 110 points per day on a P4 2.8GHz system with SSE disabled. Actual points per day vary based on your system's hardware when compared to the benchmark machine.

There are a couple of high-performance beta clients available for people with multi-core CPUs and/or certain ATI video cards. These clients produce more points per day -- the SMP multi-processor client produces 1760 points per day on the benchmark machine (a 2-CPU dual-core Xenon at 2.33GHz with 4MB cache on each CPU), and the new GPU2 client for high-end ATI cards does 1000 points per day on a Radeon 3850 with 512MB. PlayStation 3s generate 900 points per day.

If your fellow team member is running any of the high-performance clients, then it's likely he/she is earning more points per work unit. In general, beta work units are worth more points (though this is not always true). The higher point rewards reflect the increased science produced by these clients and the instability of the clients during the beta stage of development.

[uncle_fungus]

You're running the single-CPU client, which is benchmarked to produced 110 points per day on a P4 2.8GHz system with SSE disabled.

Correction, SSE is enabled, but SSE2 is disabled to prevent CPUs without SSE2 performing badly on WUs requiring double precision.

[Hyperlife]

Correction, SSE is enabled, but SSE2 is disabled to prevent CPUs without SSE2 performing badly on WUs requiring double precision.

Caught before I could edit it myself. Thanks, uncle_fungus.

[uncle_fungus]

Quick fingers

I thought you'd know, but I remember seeing someone else saying the same thing fairly recently so I'll leave the clarification.

[sneakers55]

Correction, SSE is enabled, but SSE2 is disabled to prevent CPUs without SSE2 performing badly on WUs requiring double precision.

The upshot of that is the Double GROMACS positively SCREAM on a machine with SSE2, particularly if it has Intel Core2 Inside.

[v00d00]

Are DGROMACS workunits still limited by that damned intel compiler?

As for performance, as i remember on the original run, that certain people were getting pretty decent numbers on AMD cpus, although you had to mod the core to get them (which is naughty and not allowed (read the EULA)).

[Ancient_Dragon]

Thanks for all your replies.