A few questions from a new folder.

Geckovich

Geckovich

Geckovich

Associate
Joined
27 Nov 2008
Posts
2,273
Location
Ayia Napa
My 'rig is in the sig'.

I'm currently running the GPU client.

Basically which would be most productive to run, GPU or CPU?

Can I run both?

Cheers.
 
Not got an ATI card, but i would imagine you would be able to get ~3000-3500 points from the 4870 with the latest client and drivers. Vista is a good starting point - GPU clients under vista use very little cpu time, so you should have almost all of your quad to fold cpu units too if you are interested. If your going to leave the PC on for a long time, you should be looking at the SMP client (MPICH client, not the deino client). It works the PC hard, and is pretty sensitive to 'slightly' dodgy overclocks, but gets good points. You can run two side by side with the gfx client, and each should get about 1500-2000 points per day (i would imagine). Stanford looks down on this a bit - they prefer the faster turnaround of just running one client at once, but they certianly dont mind people running 2.

The best way to maximise your points output from the cpu is to either run linux (and use a 'bodge' to get the gpu to fold), or run the linux client in a virtual machine. This isnt so easy, but in my experience it is more stable. Theres a few people round here who have done it (including myself) and im sure they wouldnt mind lending a hand if you get stuck.

If you dont really leave the pc on that much, the standard clients have long deadlines, and will get you a trickle of points per day (nothing on the gpu and SMP clients), but its all good for the project (and OCuK of course!).

EDIT: To clarify, which would be more productive could be difficult to say - try both and see. You can certainly run both at once for even more points!
 
Cheers, I think for now I'll stick with GPU2 and one iteration of the CPU non-graphical client. I think I'll pass on Linux as it doesn't seem to play well with my wireless card.

See you on the team stats page!
 
hey im pretty new to this too :) ive just installed the gpu2 client 6.23 i think (also on vist x64 ult) but just curious, when looking at everyone elses vids of their folding at home they seem to go really fast and smoothly and have a performance of a few thousand... where as mine is a bit jaggerdy and only getting 160ish performance.... is it just because my GPU is older? or is something not right? i remember i tried it once before about 2 months back and i remember the performance being a lot better...(unless it depends on the workunit as its a pretty big looking one).. any input?
 
Last edited:
160? That's awful for any client!

You might want to check the log and see if it is running Standard Loops or something.
 
yea i was thnking that, says about 3 hours to do the work unit, i remeber it taking about 25 mins before on the gpu client... only thing that has changed is ive installed the newest nvidia drivers..also noticed when running yes the GPU is maxed out as its goign to its peak temps... but also its using up quite a lot of cpu power.. that normal or something wrong :p ..any ideas?

oh it only uses loads when i have the viewer open so guess thats just beacuse it has to draw it :p but yea... any ideas why my performance is so low and how i can get back up on form again?
 
Last edited:
I have no idea tbh. Better wait for someone who has experience with the GPU2 - instead of relying on me who is just winging it :p
 
hehe but you were doing a great winging job :D yea ill check some time tomorrow to see if anyone has got any ideas :)
 
Sasso - can you post a gpu2 log file (FAHlog.txt - should be in doc's and setting/username/app data/folding@home-gpu) the end of my current file looks like this

Code: 
[08:19:12] Completed 96%
[08:20:33] Completed 97%
[08:21:55] Completed 98%
[08:23:16] Completed 99%
[08:24:37] Completed 100%
[08:24:37] Successful run
[08:24:37] DynamicWrapper: Finished Work Unit: sleep=10000
[08:24:47] Reserved 1127244 bytes for xtc file; Cosm status=0
[08:24:47] Allocated 1127244 bytes for xtc file
[08:24:47] - Reading up to 1127244 from "work/wudata_07.xtc": Read 1127244
[08:24:47] Read 1127244 bytes from xtc file; available packet space=785303220
[08:24:47] xtc file hash check passed.
[08:24:47] Reserved 34800 34800 785303220 bytes for arc file=<work/wudata_07.trr> Cosm status=0
[08:24:47] Allocated 34800 bytes for arc file
[08:24:47] - Reading up to 34800 from "work/wudata_07.trr": Read 34800
[08:24:47] Read 34800 bytes from arc file; available packet space=785268420
[08:24:47] trr file hash check passed.
[08:24:47] Allocated 560 bytes for edr file
[08:24:47] Read bedfile
[08:24:47] edr file hash check passed.
[08:24:47] Allocated 130374 bytes for logfile
[08:24:47] Read logfile
[08:24:47] GuardedRun: success in DynamicWrapper
[08:24:47] GuardedRun: done
[08:24:47] Run: GuardedRun completed.
[08:24:50] - Writing 1293490 bytes of core data to disk...
[08:24:50] Done: 1292978 -> 1155537 (compressed to 89.3 percent)
[08:24:50]   ... Done.
[08:24:50] - Shutting down core 
[08:24:50] 
[08:24:50] Folding@home Core Shutdown: FINISHED_UNIT
[08:24:52] CoreStatus = 64 (100)
[08:24:52] Unit 7 finished with 97 percent of time to deadline remaining.
[08:24:52] Updated performance fraction: 0.963376
[08:24:52] Sending work to server
[08:24:52] Project: 5016 (Run 2, Clone 477, Gen 22)
[08:24:52] - Read packet limit of 540015616... Set to 524286976.


[08:24:52] + Attempting to send results [December 4 08:24:52 UTC]
[08:24:52] - Reading file work/wuresults_07.dat from core
[08:24:52]   (Read 1156049 bytes from disk)
[08:24:52] Connecting to [url]http://171.64.65.20:8080/[/url]
[08:29:33] Posted data.
[08:29:34] Initial: 0000; - Uploaded at ~4 kB/s
[08:29:34] - Averaged speed for that direction ~7 kB/s
[08:29:34] + Results successfully sent
[08:29:34] Thank you for your contribution to Folding@Home.
[08:29:34] + Number of Units Completed: 187

[08:29:38] Trying to send all finished work units
[08:29:38] + No unsent completed units remaining.
[08:29:38] - Preparing to get new work unit...
[08:29:38] + Attempting to get work packet
[08:29:38] - Will indicate memory of 2047 MB
[08:29:38] - Connecting to assignment server
[08:29:38] Connecting to [url]http://assign-GPU.stanford.edu:8080/[/url]
[08:29:40] Posted data.
[08:29:40] Initial: 40AB; - Successful: assigned to (171.64.65.106).
[08:29:40] + News From Folding@Home: GPU folding beta
[08:29:40] Loaded queue successfully.
[08:29:40] Connecting to [url]http://171.64.65.106:8080/[/url]
[08:29:41] Posted data.
[08:29:41] Initial: 0000; - Receiving payload (expected size: 70754)
[08:29:42] - Downloaded at ~69 kB/s
[08:29:42] - Averaged speed for that direction ~57 kB/s
[08:29:42] + Received work.
[08:29:42] Trying to send all finished work units
[08:29:42] + No unsent completed units remaining.
[08:29:42] + Closed connections
[08:29:42] 
[08:29:42] + Processing work unit
[08:29:42] Core required: FahCore_11.exe
[08:29:42] Core found.
[08:29:42] Working on queue slot 08 [December 4 08:29:42 UTC]
[08:29:42] + Working ...
[08:29:42] - Calling '.\FahCore_11.exe -dir work/ -suffix 08 -checkpoint 15 -verbose -lifeline 2276 -version 620'

[08:29:43] 
[08:29:43] *------------------------------*
[08:29:43] Folding@Home GPU Core - Beta
[08:29:43] Version 1.19 (Mon Nov 3 09:34:13 PST 2008)
[08:29:43] 
[08:29:43] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86 
[08:29:43] Build host: amoeba
[08:29:43] Board Type: Nvidia
[08:29:43] Core      : 
[08:29:43] Preparing to commence simulation
[08:29:43] - Looking at optimizations...
[08:29:43] - Created dyn
[08:29:43] - Files status OK
[08:29:43] - Expanded 70242 -> 360060 (decompressed 512.5 percent)
[08:29:43] Called DecompressByteArray: compressed_data_size=70242 data_size=360060, decompressed_data_size=360060 diff=0
[08:29:43] - Digital signature verified
[08:29:43] 
[08:29:43] Project: 5758 (Run 8, Clone 29, Gen 17)
[08:29:43] 
[08:29:43] Assembly optimizations on if available.
[08:29:43] Entering M.D.
[08:29:49] Working on Protein
[08:29:51] Client config found, loading data.
[08:29:51] Starting GUI Server
[08:31:11] Completed 1%
[08:32:32] Completed 2%
[08:33:57] Completed 3%
[08:35:17] Completed 4%
[08:36:38] Completed 5%
[08:37:59] Completed 6%
[08:39:19] Completed 7%
[08:40:40] Completed 8%
[08:42:01] Completed 9%
[08:43:21] Completed 10%
 
hey here, i think this is what ya wanted... not sure how to put it in the scroll box, if ya let me know ill edit it so its doesnt take so much room..... and well how does it look? did you want the whole thing cos ive taken a snippet from the end. noticed that its the 32bit version, is there a 64 bit one? hope this gets sorted :)
Code: 
[02:33:36] *------------------------------*
[02:33:36] Folding@Home GPU Core - Beta
[02:33:36] Version 1.19 (Mon Nov 3 09:34:13 PST 2008)
[02:33:36] 
[02:33:36] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86 
[02:33:36] Build host: amoeba
[02:33:36] Board Type: Nvidia
[02:33:36] Core      : 
[02:33:36] Preparing to commence simulation
[02:33:36] - Ensuring status. Please wait.
[02:33:46] - Looking at optimizations...
[02:33:46] - Not checking prior termination.
[02:33:46] - Expanded 98797 -> 492276 (decompressed 498.2 percent)
[02:33:46] Called DecompressByteArray: compressed_data_size=98797 data_size=492276, decompressed_data_size=492276 diff=0
[02:33:46] - Digital signature verified
[02:33:46] 
[02:33:46] Project: 5751 (Run 13, Clone 11, Gen 4)
[02:33:46] 
[02:33:46] Assembly optimizations on if available.
[02:33:46] Entering M.D.
[02:33:52] Will resume from checkpoint file
[02:33:52] Working on Protein
[02:33:54] Client config found, loading data.
[02:33:54] Starting GUI Server
[02:33:54] Resuming from checkpoint
[02:33:54] Verified work/wudata_05.log
[02:33:54] Verified work/wudata_05.edr
[02:33:54] Verified work/wudata_05.xtc
[02:33:54] Completed 16%
[02:35:54] Completed 17%
[02:37:31] + Paused
[02:37:33] + Working ...
[02:37:34] Suspending work thread...
[02:37:34] Resuming work thread...
[02:37:54] CoreStatus = 62 (98)
[02:37:54] + Restarting core (settings changed) 
[02:37:54] 
[02:37:54] + Processing work unit
[02:37:54] Core required: FahCore_11.exe
[02:37:54] Core found.
[02:37:54] Working on queue slot 05 [December 4 02:37:54 UTC]
[02:37:54] + Working ...
[02:37:54] 
[02:37:54] *------------------------------*
[02:37:54] Folding@Home GPU Core - Beta
[02:37:54] Version 1.19 (Mon Nov 3 09:34:13 PST 2008)
[02:37:54] 
[02:37:54] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86 
[02:37:54] Build host: amoeba
[02:37:54] Board Type: Nvidia
[02:37:54] Core      : 
[02:37:54] Preparing to commence simulation
[02:37:54] - Ensuring status. Please wait.
[02:38:04] - Looking at optimizations...
[02:38:04] - Not checking prior termination.
[02:38:04] - Expanded 98797 -> 492276 (decompressed 498.2 percent)
[02:38:04] Called DecompressByteArray: compressed_data_size=98797 data_size=492276, decompressed_data_size=492276 diff=0
[02:38:04] - Digital signature verified
[02:38:04] 
[02:38:04] Project: 5751 (Run 13, Clone 11, Gen 4)
[02:38:04] 
[02:38:04] Assembly optimizations on if available.
[02:38:04] Entering M.D.
[02:38:10] Will resume from checkpoint file
[02:38:10] Working on Protein
[02:38:12] Client config found, loading data.
[02:38:12] Starting GUI Server
[02:38:13] Resuming from checkpoint
[02:38:13] Verified work/wudata_05.log
[02:38:13] Verified work/wudata_05.edr
[02:38:13] Verified work/wudata_05.xtc
[02:38:13] Completed 17%
[02:40:05] Completed 18%
[02:41:53] Completed 19%
[02:43:38] Completed 20%
[02:45:24] Completed 21%
[02:47:16] Completed 22%
[02:49:03] Completed 23%
[02:50:50] Completed 24%
[02:52:36] Completed 25%
[02:54:21] Completed 26%
[02:56:05] Completed 27%
[02:57:49] Completed 28%
[02:59:32] Completed 29%

Folding@Home Client Shutdown.
 
Last edited:
Everything's fine there - that works out at about 4118 points per day. :cool:

When you mention the 160, I guess you're talking about the iterations/sec figure that's shown in the graphical mode? The graphical mode is still pretty buggy and eats up a lot of precious GPU power, so it's best not to use it or believe anything it says. If you want to monitor your client, it's good to install FahMon which will tell you directly how many PPD you're doing. Once you've installed it, add a new client and point it to the folder that contains your log file.

As for CPU usage, I'm guessing you're on XP? The GPU client tends to use quite a lot of CPU on XP - it's better on Vista.

Hope this helps :D
 
ahhh really? just the performance was liek 150 ns/day... what ever that is :P but still 3 hours for a work unit? and im on vista x64 ;) i just liked looking at the crazy strands wriggling but now its jittering :( left the pc on at home to get rid of that one and see if it was just that unit being big :p
 
When the client came out, the work units were really short (half an hour-ish). Now it's been tested a bit more the WUs are longer. They also give you a lot more points, so all is good :)
 
ahh rocktastic :D does make sense because the graphical visualisation on the mesh does look a lot bigger and complex tbh...

just want some clarification for what performance "ns/day" people seem to be getting on the same card or what ever :p im still not completely convinced its all fine :p
 
if you click the 'Go Advance' you get more options to the normal quick reply.
But all you need to do is rap code tags round the text. [ code ] before and [ /code ] at the end. I've added spaces so that it shows btw :)
 
ah cheers :) i've edited it now :) so anyone.... your performance in ns/day? is your graphics really smooth of does it jitter? and how long do ** WU's take to complete now?

8800gts512 150ns/day 2h50min :p
 
You must log in or register to reply here.
Back
Top Bottom