Intel Core i7 920

[Pappa]

i can get a Sparkle GeForce GTS 250 Video Card - 512MB DDR3, PCI Express 2.0,
(2) Dual Link DVI, SLI Ready  for $129.95 now

My BFG OC 250GTS has finally crossed 6000 RAC using a test hybrid GPU app in Seti Beta. How I got it was Frys.com had it listed for $xxx.xx Bestbuy.com had it listed for $xxx.xx with a price match. I carried in the print screen for the price match and the 250GTS 1 gig cost me $125 and pennies plus tax. The other day I stopped at Frys for a MB and could not help myself went back and found an open box EVGA 250GTS for 123.00 the only caveate is that you should have a "Good" 475 watt power supply for the G92 chips

So other than cores, the 250 has better timings than the 260 or 275

I would be curious to see what the 220 would do. But that would take a month of doing Seti Cuda only on the GPU.

So as mentioned RAC is a fickle thing, the best is seeing how much credit it can produce in a month and then divided by 30 days.

http://boincstats.com/stats/host_graph.php?pr=sah_beta&id=16588
and pendings for the machine is over 70000
To make matter worse it is running Aqua on the CPUS which is fairly intenisve but give up time to the GPU fairly nicely. However I am losing about two minutes per Cuda WU (0.40 AR)  which adds up.

That is objective as I can be. The machine is producing over 12000 cobblestones/day

Regards

[TouchuvGrey]

hmmm, BFG GeForce GTS 250 Video Card - 1024MB GDDR3, PCI Express
 2.0, (2) Dual Link DVI, SLI Ready $129.95 at Circuit City. Perhaps put that
in main machine and put GT220 in one of my other machines, if any can use it.

[Pappa]

hmmm, BFG GeForce GTS 250 Video Card - 1024MB GDDR3, PCI Express
 2.0, (2) Dual Link DVI, SLI Ready $129.95 at Circuit City. Perhaps put that
in main machine and put GT220 in one of my other machines, if any can use it.

There are options, you just have to find them.

[TouchuvGrey]

i just got my GTS 250 video card ( 1024 MB GDDR3 ). i will be installing it on my
ASUS P6TD Deluxe motherboard using the cc_confic.xml  file claggy
suggested. Is there any preference as to which slot the 220 and
250 should go in and is there any additional configuration needed ?

[Pappa]

i just got my GTS 250 video card ( 1024 MB GDDR3 ). i will be installing it on my
ASUS P6TD Deluxe motherboard using the cc_confic.xml  file claggy
suggested. Is there any preference as to which slot the 220 and
250 should go in and is there any additional configuration needed ?

I would put the 250 into the slot closest the CPU. There it will be most effective. As it is installed, reinstall the nvidia drivers as some features/functions will not take effect until that happens.

If you have a machine that will support one of the other video cards that would be better.

Regards

[TouchuvGrey]

If i add a 2nd video card does it need to be the same as the 1st ?.

Nope, it doesn't have to be the same, but Boinc 6.10.x will only run the most capable by default,
you'll need to make a cc_config.xml (you don't have one by default) just copy the following, and paste it using Notepad,
then save it as cc_config.xml making sure it doesn't have a .txt extension, drop the file in your Boinc Data folder and restart Boinc.

Code: [Select]

<cc_config>
<log_flags>
<sched_op_debug>1</sched_op_debug>
</log_flags>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>


What percentage of the number crunching comes from the GPU
as opposed to from the CPU ?

Depends on the GPU / which project the GPU/CPU are running, expect some where between 50% and 90% total RAC coming from the GPU.

What sort of gain could i expect from adding a 2nd video card ?

Just downloaded and installed Multibeam optimized app for
64 bit SSE 4.1

Claggy

OK, i'm obviously a congenital idiot or have not had enough coffee.
Where does the Boinc Data folder hide in Windows 7 64 bit ?

[Claggy]

If i add a 2nd video card does it need to be the same as the 1st ?.

Nope, it doesn't have to be the same, but Boinc 6.10.x will only run the most capable by default,
you'll need to make a cc_config.xml (you don't have one by default) just copy the following, and paste it using Notepad,
then save it as cc_config.xml making sure it doesn't have a .txt extension, drop the file in your Boinc Data folder and restart Boinc.

Code: [Select]

<cc_config>
<log_flags>
<sched_op_debug>1</sched_op_debug>
</log_flags>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>


What percentage of the number crunching comes from the GPU
as opposed to from the CPU ?

Depends on the GPU / which project the GPU/CPU are running, expect some where between 50% and 90% total RAC coming from the GPU.

What sort of gain could i expect from adding a 2nd video card ?

Just downloaded and installed Multibeam optimized app for
64 bit SSE 4.1

Claggy

OK, i'm obviously a congenital idiot or have not had enough coffee.
Where does the Boinc Data folder hide in Windows 7 64 bit ?

Windows Vista/Windows 7: C:\ProgramData\BOINC\
and it'll be hidden, so you'll need to unhide it.
See: The Big BOINC 6 Answer Thread.

Claggy

[TouchuvGrey]

Thank you Claggy. Done,  and this is what i'm getting now:

2/10/2010 6:55:52 PM   SETI@home   [sched_op_debug] Starting scheduler request
2/10/2010 6:55:52 PM   SETI@home   Sending scheduler request: To fetch work.
2/10/2010 6:55:52 PM   SETI@home   Reporting 2 completed tasks, requesting new tasks for GPU
2/10/2010 6:55:52 PM   SETI@home   [sched_op_debug] CPU work request: 0.00 seconds; 0 idle CPUs
2/10/2010 6:55:52 PM   SETI@home   [sched_op_debug] NVIDIA GPU work request: 777600.86 seconds; 1 idle GPUs
2/10/2010 6:56:02 PM   SETI@home   Scheduler request completed: got 0 new tasks
2/10/2010 6:56:02 PM   SETI@home   [sched_op_debug] Server version 611
2/10/2010 6:56:02 PM   SETI@home   Message from server: No work sent
2/10/2010 6:56:02 PM   SETI@home   Project requested delay of 11 seconds
2/10/2010 6:56:02 PM   SETI@home   [sched_op_debug] handle_scheduler_reply(): got ack for result 01dc06ac.27997.198061.11.10.28_0
2/10/2010 6:56:02 PM   SETI@home   [sched_op_debug] handle_scheduler_reply(): got ack for result 01dc06ac.27890.198061.10.10.174_1
2/10/2010 6:56:02 PM   SETI@home   [sched_op_debug] Deferring communication for 11 sec
2/10/2010 6:56:02 PM   SETI@home   [sched_op_debug] Reason: requested by project

      can you or anyone else enlighten me as to what is going on ?

[TouchuvGrey]

i modified the code Claggy supplied to read:
<cc_config>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

moved the 250 to Slot 1
the 220 to slot 2 restarted the client and now have:

2/11/2010 9:56:31 PM      NVIDIA GPU 0: GeForce GTS 250 (driver version 19621, CUDA version 3000, compute capability 1.1, 1024MB, 470 GFLOPS peak)
2/11/2010 9:56:31 PM      NVIDIA GPU 1: GeForce GT 220 (driver version 19621, CUDA version 3000, compute capability 1.2, 1024MB, 131 GFLOPS peak)
2/11/2010 9:56:31 PM   SETI@home   Found app_info.xml; using anonymous platform


2/13/2010 6:19:35 AM   SETI@home   Sending scheduler request: To fetch work.
2/13/2010 6:19:35 AM   SETI@home   Reporting 16 completed tasks, requesting new tasks for GPU
2/13/2010 6:19:45 AM   SETI@home   Scheduler request completed: got 0 new tasks

    so if i'm reading this correctly i have both GPU's crunching now.

[Claggy]

Thank you Claggy. Done,  and this is what i'm getting now:

2/10/2010 6:55:52 PM   SETI@home   [sched_op_debug] Starting scheduler request
2/10/2010 6:55:52 PM   SETI@home   Sending scheduler request: To fetch work.
2/10/2010 6:55:52 PM   SETI@home   Reporting 2 completed tasks, requesting new tasks for GPU
2/10/2010 6:55:52 PM   SETI@home   [sched_op_debug] CPU work request: 0.00 seconds; 0 idle CPUs
2/10/2010 6:55:52 PM   SETI@home   [sched_op_debug] NVIDIA GPU work request: 777600.86 seconds; 1 idle GPUs
2/10/2010 6:56:02 PM   SETI@home   Scheduler request completed: got 0 new tasks
2/10/2010 6:56:02 PM   SETI@home   [sched_op_debug] Server version 611
2/10/2010 6:56:02 PM   SETI@home   Message from server: No work sent
2/10/2010 6:56:02 PM   SETI@home   Project requested delay of 11 seconds
2/10/2010 6:56:02 PM   SETI@home   [sched_op_debug] handle_scheduler_reply(): got ack for result 01dc06ac.27997.198061.11.10.28_0
2/10/2010 6:56:02 PM   SETI@home   [sched_op_debug] handle_scheduler_reply(): got ack for result 01dc06ac.27890.198061.10.10.174_1
2/10/2010 6:56:02 PM   SETI@home   [sched_op_debug] Deferring communication for 11 sec
2/10/2010 6:56:02 PM   SETI@home   [sched_op_debug] Reason: requested by project

      can you or anyone else enlighten me as to what is going on ?

the sched_op_debug flag is showing that you're asking for GPU work, and how much work you're asking for, that you're got an idle GPU without any work,
that the server didn't send you any work, and that two tasks got reported correctly,

Claggy

[Claggy]

i modified the code Claggy supplied to read:
<cc_config>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

moved the 250 to Slot 1
the 220 to slot 2 restarted the client and now have:

2/11/2010 9:56:31 PM      NVIDIA GPU 0: GeForce GTS 250 (driver version 19621, CUDA version 3000, compute capability 1.1, 1024MB, 470 GFLOPS peak)
2/11/2010 9:56:31 PM      NVIDIA GPU 1: GeForce GT 220 (driver version 19621, CUDA version 3000, compute capability 1.2, 1024MB, 131 GFLOPS peak)
2/11/2010 9:56:31 PM   SETI@home   Found app_info.xml; using anonymous platform


2/13/2010 6:19:35 AM   SETI@home   Sending scheduler request: To fetch work.
2/13/2010 6:19:35 AM   SETI@home   Reporting 16 completed tasks, requesting new tasks for GPU
2/13/2010 6:19:45 AM   SETI@home   Scheduler request completed: got 0 new tasks

    so if i'm reading this correctly i have both GPU's crunching now.

What messages did Boinc show before you moved the GPU's about?, Did it show 'Config: use all coprocessors' on start up?, did it say one of the GPU's was not used?
What does it say now?
The cc_config needs a 'Start' Options tag, as well as and 'End' Options tag, so:

<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

Claggy

[TouchuvGrey]

i don't recall and did not record what it said before moving the cards.
it now says:

2/11/2010 9:56:30 PM      Starting BOINC client version 6.10.18 for windows_x86_64
2/11/2010 9:56:30 PM      Config: use all coprocessors
2/11/2010 9:56:30 PM      log flags: file_xfer, sched_ops, task
2/11/2010 9:56:30 PM      Libraries: libcurl/7.19.4 OpenSSL/0.9.8l zlib/1.2.3

    i did not look carefully at what i pasted before. My cc_config.xml file reads:

<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

so i do have both the "start" and "end" tags

   i removed the red sections as i understand the "use all GPU's" part,
but not the other parts. i'm learning about this stuff now.

<cc_config>
<log_flags>
<sched_op_debug>1</sched_op_debug>
</log_flags>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

[TouchuvGrey]

restored cc_config.xml to your original. Now getting:

2/14/2010 8:10:47 AM      Starting BOINC client version 6.10.18 for windows_x86_64
2/14/2010 8:10:47 AM      Config: use all coprocessors
2/14/2010 8:10:47 AM      log flags: file_xfer, sched_ops, task, sched_op_debug
2/14/2010 8:10:47 AM      Libraries: libcurl/7.19.4 OpenSSL/0.9.8l zlib/1.2.3
2/14/2010 8:10:47 AM      Data directory: C:\ProgramData\BOINC


2/14/2010 8:10:47 AM      Memory: 11.99 GB physical, 23.98 GB virtual

2/14/2010 8:10:47 AM      NVIDIA GPU 0: GeForce GTS 250 (driver version 19621, CUDA version 3000, compute capability 1.1, 1024MB, 470 GFLOPS peak)
2/14/2010 8:10:47 AM      NVIDIA GPU 1: GeForce GT 220 (driver version 19621, CUDA version 3000, compute capability 1.2, 1024MB, 131 GFLOPS peak)
2/14/2010 8:10:47 AM   SETI@home   Found app_info.xml; using anonymous platform
2/14/2010 8:10:47 AM      Not using a proxy


2/14/2010 8:31:10 AM   SETI@home   [sched_op_debug] Starting scheduler request
2/14/2010 8:31:10 AM   SETI@home   Sending scheduler request: To fetch work.
2/14/2010 8:31:10 AM   SETI@home   Requesting new tasks for GPU
2/14/2010 8:31:10 AM   SETI@home   [sched_op_debug] CPU work request: 0.00 seconds; 0 idle CPUs
2/14/2010 8:31:10 AM   SETI@home   [sched_op_debug] NVIDIA GPU work request: 777600.86 seconds; 2 idle GPUs
2/14/2010 8:31:20 AM   SETI@home   Scheduler request completed: got 0 new tasks
2/14/2010 8:31:20 AM   SETI@home   [sched_op_debug] Server version 611
2/14/2010 8:31:20 AM   SETI@home   Message from server: No work sent
2/14/2010 8:31:20 AM   SETI@home   Project requested delay of 11 seconds
2/14/2010 8:31:20 AM   SETI@home   [sched_op_debug] Deferring communication for 11 sec
2/14/2010 8:31:20 AM   SETI@home   [sched_op_debug] Reason: requested by project

    are there no GPU work units available ?

[Claggy]

restored cc_config.xml to your original. Now getting:

2/14/2010 8:10:47 AM      Starting BOINC client version 6.10.18 for windows_x86_64
2/14/2010 8:10:47 AM      Config: use all coprocessors
2/14/2010 8:10:47 AM      log flags: file_xfer, sched_ops, task, sched_op_debug
2/14/2010 8:10:47 AM      Libraries: libcurl/7.19.4 OpenSSL/0.9.8l zlib/1.2.3
2/14/2010 8:10:47 AM      Data directory: C:\ProgramData\BOINC


2/14/2010 8:10:47 AM      Memory: 11.99 GB physical, 23.98 GB virtual

2/14/2010 8:10:47 AM      NVIDIA GPU 0: GeForce GTS 250 (driver version 19621, CUDA version 3000, compute capability 1.1, 1024MB, 470 GFLOPS peak)
2/14/2010 8:10:47 AM      NVIDIA GPU 1: GeForce GT 220 (driver version 19621, CUDA version 3000, compute capability 1.2, 1024MB, 131 GFLOPS peak)
2/14/2010 8:10:47 AM   SETI@home   Found app_info.xml; using anonymous platform
2/14/2010 8:10:47 AM      Not using a proxy


2/14/2010 8:31:10 AM   SETI@home   [sched_op_debug] Starting scheduler request
2/14/2010 8:31:10 AM   SETI@home   Sending scheduler request: To fetch work.
2/14/2010 8:31:10 AM   SETI@home   Requesting new tasks for GPU
2/14/2010 8:31:10 AM   SETI@home   [sched_op_debug] CPU work request: 0.00 seconds; 0 idle CPUs
2/14/2010 8:31:10 AM   SETI@home   [sched_op_debug] NVIDIA GPU work request: 777600.86 seconds; 2 idle GPUs
2/14/2010 8:31:20 AM   SETI@home   Scheduler request completed: got 0 new tasks
2/14/2010 8:31:20 AM   SETI@home   [sched_op_debug] Server version 611
2/14/2010 8:31:20 AM   SETI@home   Message from server: No work sent
2/14/2010 8:31:20 AM   SETI@home   Project requested delay of 11 seconds
2/14/2010 8:31:20 AM   SETI@home   [sched_op_debug] Deferring communication for 11 sec
2/14/2010 8:31:20 AM   SETI@home   [sched_op_debug] Reason: requested by project

    are there no GPU work units available ?

Can we have a link to your host please?, snipping bits out your startup messages doesn't help us find out why you're not getting any Cuda work,
and if you're ever managed to do any, there's Plenty of Multibeam work available, so that's not the problem.

Claggy

[TouchuvGrey]

http://stats.free-dc.org/stats.php?page=host&proj=sah&hostid=5241862

http://setiathome.berkeley.edu/show_host_detail.php?hostid=5241862

if that's not what you need, let me know please