Getting your GTX680 to Fold

Biffa · 11 Apr 2012 at 06:13

You have to use the v7 Client and you need to add the command:
"client-type:beta" to the gpu slot

And don't expect any help or sympathy from the folding peeps unless you are a signed up beta member

Biffa · 11 Apr 2012 at 13:19

They are no better than the 580 at the moment, but work is ongoing on a more optimised client so I have heard.

Biffa · 13 Apr 2012 at 18:56

I think you need to be on the beta team, just checking that now. I had the same prob.

Biffa · 14 Apr 2012 at 10:22

FahCore_15.exe is ok, it just has to be 2.24 instead of 2.22

Biffa · 16 Apr 2012 at 01:00

Thats the version of FahCore_15.exe

Biffa · 16 Apr 2012 at 10:32

Ok so I am not sure if it is a coincidence, and I'd been wanting to do this for a while anyway given the range of hardware I have to fold on, but I joined the beta team and managed to get the correct client.

But even then I had to delete the client files and work files I had already downloaded an been unable to fold. Getting around 16.5K PPD when left on its own, but drops down if the system is in use.

Code:

06:43:52:WU01:FS00:Running FahCore: D:\fah\FAHClient/FAHCoreWrapper.exe D:/FAH/FAHClientData/cores/www.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/beta/Core_15.fah/FahCore_15.exe -dir 01 -suffix 01 -version 701 -lifeline 5068 -checkpoint 15 -gpu 0
06:43:52:WU01:FS00:Started FahCore on PID 4832
06:43:52:Started thread 10 on PID 5068
06:43:52:WU01:FS00:Core PID:3320
06:43:52:WU01:FS00:FahCore 0x15 started
06:43:52:WU01:FS00:0x15:
06:43:52:WU01:FS00:0x15:*------------------------------*
06:43:52:WU01:FS00:0x15:Folding@Home GPU Core
06:43:52:WU01:FS00:0x15:Version                2.24 (Thu Feb 23 13:21:02 PST 2012)
06:43:52:WU01:FS00:0x15:Build host             ASUS_Laptop
06:43:52:WU01:FS00:0x15:Board Type             NVIDIA/CUDA
06:43:52:WU01:FS00:0x15:Core                   15
06:43:52:WU01:FS00:0x15:
06:43:52:WU01:FS00:0x15:Window's signal control handler registered.
06:43:52:WU01:FS00:0x15:Preparing to commence simulation
06:43:52:WU01:FS00:0x15:- Looking at optimizations...
06:43:52:WU01:FS00:0x15:DeleteFrameFiles: successfully deleted file=01/wudata_01.ckp
06:43:52:WU01:FS00:0x15:- Created dyn
06:43:52:WU01:FS00:0x15:- Files status OK
06:43:52:WU01:FS00:0x15:sizeof(CORE_PACKET_HDR) = 512 file=<>
06:43:52:WU01:FS00:0x15:- Expanded 121843 -> 544090 (decompressed 446.5 percent)
06:43:52:WU01:FS00:0x15:Called DecompressByteArray: compressed_data_size=121843 data_size=544090, decompressed_data_size=544090 diff=0
06:43:52:WU01:FS00:0x15:- Digital signature verified
06:43:52:WU01:FS00:0x15:
06:43:52:WU01:FS00:0x15:Project: 8036 (Run 5, Clone 1, Gen 2)
06:43:52:WU01:FS00:0x15:
06:43:52:WU01:FS00:0x15:Assembly optimizations on if available.
06:43:52:WU01:FS00:0x15:Entering M.D.
06:43:54:WU01:FS00:0x15:Tpr hash 01/wudata_01.tpr:  3511027989 1414838518 1491937849 3442628229 1972443818
06:43:54:WU01:FS00:0x15:GPU device id=0
06:43:54:WU01:FS00:0x15:Working on Protein
06:43:54:WU01:FS00:0x15:Client config unavailable.
06:43:55:WU01:FS00:0x15:Starting GUI Server
06:43:59:WU00:FS00:Upload 96.95%
06:43:59:WU00:FS00:Upload complete
06:43:59:WU00:FS00:Server responded WORK_ACK (400)
06:43:59:WU00:FS00:Final credit estimate, 3843.00 points
06:43:59:WU00:FS00:Cleaning up
06:44:59:WU01:FS00:0x15:Setting checkpoint frequency: 250000
06:44:59:WU01:FS00:0x15:Completed         3 out of 25000000 steps (0%).
06:48:42:WU01:FS00:0x15:Completed    250000 out of 25000000 steps (1%).
06:52:25:WU01:FS00:0x15:Completed    500000 out of 25000000 steps (2%).

Biffa · 17 Apr 2012 at 09:44

Just to note thats around 2K slower than a GTX470

Biffa · 6 Aug 2012 at 12:51

I'd say they are probably a level below their 5 series counterparts, i.e a 680 performs more like a 570/560Ti, with equivalent power usage. So a 680 is definitely less power hungry than a 580 but then it has lower performance and costs almost twice as much. The maths is pretty straight forwards for me

Remember there are other BOINC projects that do similar science to Folding, or even other types of non-seti science, but again I think the 5 series and even the (shock, horror) AMD gpu's get much better returns than the 680's.

As per someone elses thread in the gpu forum, I said it there and I'll say it here, for the return you get better PPD from a 2600K than you do from any GPU. I'm currently pulling 40K on a 2600K running a normal SMP WU while crunching Seti on my 7970 @ 20K RAC