I've set GPU to no, but it it's trying for it anyway, and I'm not getting any work at all:

Use GPUs for Computation? no
Use CPU for Computation? yes

27/08/2009 01:03:07 Milkyway@home Requesting new tasks for GPU
27/08/2009 01:03:13 Milkyway@home Scheduler request completed: got 0 new tasks
27/08/2009 01:03:13 Milkyway@home Message from server: No work sent

I'm using BOINC 6.10.1

---

I've set GPU to no, but it it's trying for it anyway, and I'm not getting any work at all:

Use GPUs for Computation? no
Use CPU for Computation? yes

27/08/2009 01:03:07 Milkyway@home Requesting new tasks for GPU
27/08/2009 01:03:13 Milkyway@home Scheduler request completed: got 0 new tasks
27/08/2009 01:03:13 Milkyway@home Message from server: No work sent

I'm using BOINC 6.10.1

Does this still happen with subsequent scheduler updates? When I had the Use GPUs for Computation set to no it would return a message saying something like "Preferences set to not accept GPU work, but work is available". Sometimes I would receive the same message "Scheduler request completed: got 0 new tasks" but once the next scheduler update came it disappeared and switched back to saying "Requesting new tasks for CPU".

Edit:
With BOINC 6.6.36 I see something like so

(no to GPU computation, no to CPU computation)
8/26/2009 8:22:12 PM Milkyway@home Resetting project
8/26/2009 8:22:17 PM Milkyway@home Sending scheduler request: To fetch work.
8/26/2009 8:22:17 PM Milkyway@home Requesting new tasks for GPU
8/26/2009 8:22:22 PM Milkyway@home Scheduler request completed: got 0 new tasks
8/26/2009 8:22:22 PM Milkyway@home Message from server: Not sending work - last request too recent: 15 sec
8/26/2009 8:23:27 PM Milkyway@home Sending scheduler request: To fetch work.
8/26/2009 8:23:27 PM Milkyway@home Requesting new tasks for GPU
8/26/2009 8:23:32 PM Milkyway@home Scheduler request completed: got 0 new tasks
8/26/2009 8:23:32 PM Milkyway@home Message from server: No work sent
8/26/2009 8:23:32 PM Milkyway@home Message from server: GPU jobs are available, but your preferences are set to not accept them
8/26/2009 8:24:37 PM Milkyway@home Sending scheduler request: To fetch work.
8/26/2009 8:24:37 PM Milkyway@home Requesting new tasks for CPU
8/26/2009 8:24:42 PM Milkyway@home Scheduler request completed: got 0 new tasks
8/26/2009 8:24:42 PM Milkyway@home Message from server: No work sent
8/26/2009 8:24:42 PM Milkyway@home Message from server: CPU jobs are available, but your preferences are set to not accept them
(no to gpu computation, yes to cpu computation)
8/26/2009 8:25:47 PM Milkyway@home Sending scheduler request: To fetch work.
8/26/2009 8:25:47 PM Milkyway@home Requesting new tasks for GPU
8/26/2009 8:25:52 PM Milkyway@home Scheduler request completed: got 0 new tasks
8/26/2009 8:25:52 PM Milkyway@home Message from server: No work sent
8/26/2009 8:25:52 PM Milkyway@home Message from server: GPU jobs are available, but your preferences are set to not accept them
8/26/2009 8:26:57 PM Milkyway@home Sending scheduler request: To fetch work.
8/26/2009 8:26:57 PM Milkyway@home Requesting new tasks for GPU
8/26/2009 8:27:02 PM Milkyway@home Scheduler request completed: got 0 new tasks
8/26/2009 8:27:02 PM Milkyway@home Message from server: No work sent
8/26/2009 8:27:02 PM Milkyway@home Message from server: GPU jobs are available, but your preferences are set to not accept them
8/26/2009 8:28:07 PM Milkyway@home Sending scheduler request: To fetch work.
8/26/2009 8:28:07 PM Milkyway@home Requesting new tasks for CPU
8/26/2009 8:28:12 PM Milkyway@home Scheduler request completed: got 2 new tasks
8/26/2009 8:28:14 PM Milkyway@home Started download of milkyway_0.19_windows_x86_64.exe
8/26/2009 8:28:14 PM Milkyway@home Started download of ps_s222_3s_02w_03r_21_search_parameters_3880771_1251332799
8/26/2009 8:28:15 PM Milkyway@home Finished download of ps_s222_3s_02w_03r_21_search_parameters_3880771_1251332799
8/26/2009 8:28:15 PM Milkyway@home Started download of ps_s222_3s_02w_03r_21_search_parameters_3880772_1251332799
8/26/2009 8:28:17 PM Milkyway@home Finished download of milkyway_0.19_windows_x86_64.exe
8/26/2009 8:28:17 PM Milkyway@home Finished download of ps_s222_3s_02w_03r_21_search_parameters_3880772_1251332799
8/26/2009 8:28:18 PM Milkyway@home Starting ps_s222_3s_02w_03r_21_3880771_1251332799_0
8/26/2009 8:28:18 PM Milkyway@home Starting task ps_s222_3s_02w_03r_21_3880771_1251332799_0 using milkyway version 19
8/26/2009 8:28:18 PM Milkyway@home Starting ps_s222_3s_02w_03r_21_3880772_1251332799_0
8/26/2009 8:28:18 PM Milkyway@home Starting task ps_s222_3s_02w_03r_21_3880772_1251332799_0 using milkyway version 19
8/26/2009 8:28:26 PM Suspending computation - user request

To be honest the CUDA application is not running at its full potential, the computation is sliced into pieces so that it does not occupy the whole GPU in order to decrease the slowdown of the user interface. There are two ways the application can be executed

1. Using all of the GPU's power, work completes faster but the OS is pretty much unusable because it takes ~2s for the user interface to update, meaning it can only run when the computer is idle
2. Using a partial amount of the GPU's power, work takes longer, updates to the OS user interface are still slow, but hardly noticeable in most situations, meaning it can be run when the computer is in use

I chose to go with #2, but which one makes more sense? I'll look into the feasibility of having this as a project specific preference if there is a 50/50 divide between the two choices.

What actually limits the responsiveness of the user interface is not the full use of the GPU resources, but the runtime of one kernel call. The GPU is blocked for the duration of a call, even if it would be designed to use only half of the available resources.

What helps is to implement some mechanism, which adapts the execution domain size (i.e. the number of "threads" fired off by a kernel) to the speed of the used GPU in such a way that the runtime of one kernel call remains more or less constant (let's say 40ms). That is what the parameter f of the ATI app controls (default is 30Hz, i.e. 33ms per kernel call).
Actually, one of the reasons I rotated the orientation of the integration plane (I use nu_steps calls of kernels integrating in the mu-r plane as I said already several times) was to have enough threads in such a domain (1.12 million for current WUs) to play with. From the used GPU and its clockspeed I estimate how long it is going take to integrate a complete mu-r plane and slice it dynamically so each kernel call shouldn't take longer than a specified time (some milliseconds are enough to get a very high GPU usage).

In the moment all GT200 based cards are quite uniform in speed (opposed to compatible ATIs) so it could be done statically for now. But that will change with the GT300. So it may be worth thinking about such a scheme.

I've set GPU to no, but it it's trying for it anyway, and I'm not getting any work at all:

Use GPUs for Computation? no
Use CPU for Computation? yes

27/08/2009 01:03:07 Milkyway@home Requesting new tasks for GPU
27/08/2009 01:03:13 Milkyway@home Scheduler request completed: got 0 new tasks
27/08/2009 01:03:13 Milkyway@home Message from server: No work sent

I'm using BOINC 6.10.1

Does this still happen with subsequent scheduler updates? When I had the Use GPUs for Computation set to no it would return a message saying something like "Preferences set to not accept GPU work, but work is available". Sometimes I would receive the same message "Scheduler request completed: got 0 new tasks" but once the next scheduler update came it disappeared and switched back to saying "Requesting new tasks for CPU".

I'm still getting the same messages in my BOINC manager, but couldn't say when the scheduler updates.

[edit]This is what I am still getting (Use GPUs for Computation? is still set to no) ;

27/08/2009 01:33:02 Milkyway@home update requested by user
27/08/2009 01:33:05 Milkyway@home Sending scheduler request: Requested by user.
27/08/2009 01:33:05 Milkyway@home Requesting new tasks for GPU
27/08/2009 01:33:10 Milkyway@home Scheduler request completed: got 0 new tasks
27/08/2009 01:33:10 Milkyway@home Message from server: No work sent

further edit;

I have a NVIDIA GPU, but not compute capability 1.3;

27/08/2009 01:00:15 NVIDIA GPU 0: GeForce 9600M GT (driver version 19038, CUDA version 2030, compute capability 1.1, 512MB, est. 14GFLOPS)
27/08/2009 01:00:15 No CAL Runtime Libraries installed.

---

I'm not quite sure how BOINC decides which platform to request tasks for (GPU or CPU) but I will try BOINC 6.10 over the next day to see if that is the source of the issue.

Edit:
Just saw your update with respect to your CUDA card, the server is set to reject sending the GPU application to users without a compatible card. It is possible that I messed up that part of the source code. Since I do have an 8800 GTX as well I'll put that in tomorrow afternoon to see if I can reproduce the situation.

I'm not quite sure how BOINC decides which platform to request tasks for (GPU or CPU) but I will try BOINC 6.10 over the next day to see if that is the source of the issue.

I tried to downgrade to 6.6.36, but got an error and the BOINC manager wouldn't run, so had to go back to 6.10.1

---

I'm not quite sure how BOINC decides which platform to request tasks for (GPU or CPU) but I will try BOINC 6.10 over the next day to see if that is the source of the issue.

I tried to downgrade to 6.6.36, but got an error and the BOINC manager wouldn't run, so had to go back to 6.10.1

I sent you a PM based on my analysis, if other users start reporting similar issues I'll post a public message without any private information.

Not sure where to put this. Not cuda related per say, I have only been getting 5 wu at a time today. Did this change from 6, or just my luck?

---

Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.

Not sure where to put this. Not cuda related per say, I have only been getting 5 wu at a time today. Did this change from 6, or just my luck?

The server side software was upgraded recently so it might have been that, I'm not well versed in the workings of BOINC to be able to answer that question correctly.

:-( My new gpu is a gts 250 . it looked good in comparison with the gtx 260 in gpu grid's chart, close in bang for half the cost. almost bought a 260 today (cheaper now)but ended up getting a hub, some mice, keyboards, cables, kvm etc. woulda put it in a 64 bit linux box though (too cheap to buy windoze, can't stand it nywayz) as ThierryH says tho, #1 is best 4 Me cuz i don't need to use the pc.

why the 32 bit BOINC ?
my server2003 64 bit 'doze wont run that. (disingenuous (ungenuine) of course ;-)
is there a 64bit 'nix app in the pipeline ?

I'm not quite sure how BOINC decides which platform to request tasks for (GPU or CPU) but I will try BOINC 6.10 over the next day to see if that is the source of the issue.

I tried to downgrade to 6.6.36, but got an error and the BOINC manager wouldn't run, so had to go back to 6.10.1

I sent you a PM based on my analysis, if other users start reporting similar issues I'll post a public message without any private information.

I've replied to that, thanks Anthony. Very impressive analysis - you're very CUDA capable methinks ;)

---

I've set GPU to no, but it it's trying for it anyway, and I'm not getting any work at all:

Maybe any local settings in the manager?

On Win7 64b BM 6.10.1 64b, 190.38 64b and GTX260SP216 all WU crushed immidetly after run.

I detached project, attached again, restart, stop GPUGRID but it does not help.

Now I will try on xp.....

---

A proud member of the Polish National Team

COME VISIT US at Polish National Team FORUM

finally got it going on vista 64 after downloading the new driver
Does it need cuda 2.3?
aver crunch time is 6.5 mins 2 apps running a a time
at the moment very poor perforance compared to ATI cards but early days I suppose.
My 2 4870x2 are turning in 300k credits or more a day
Ross

---

I've set GPU to no, but it it's trying for it anyway, and I'm not getting any work at all:

Maybe any local settings in the manager?

I don't think so aad. I've tried suspending all other projects except for MW and it still seems to be trying to get GPU WUs, even though I have set GPU = no, and CPU = yes. It could be that it detects I have a CUDA capable GPU, ignores my request for GPU = no, and then decides I am not compute capable 1.3, but does not recognize I am asking for CPU WUs. Anthony is looking into this I expect. Right now I am unable to get any work for this host since, it seems, I have a CUDA GPU, but not capable enough, which also stops me getting CPU WUs.

[edit] I've had another look at the local settings in my manager and can't see anything that could be causing the problem. I have 'Use GPU while computer is in use' checked. I tried unchecking it, but my manager is still requesting new tasks for GPU and I'm still getting no MW work.

---

Ok, after few restarts it works!!!!

But I install driver 190.62, BOINC 6.10.1 and detached + atached to projest and set only GPU...

It works witch AQUA@HOME CPU.

Win7 64 bit.

On GTX260 (625Mhz GPU 1348 Shader) = 11 m. 25 s. = 685s. 74.24 credits so 1 credit in 9.23s.

GPUGRID pays better..... (1 credit at 4-7sec.)

But MW app is "young" and I hope soon it will be improved to have better performance :)

---

A proud member of the Polish National Team

COME VISIT US at Polish National Team FORUM

:-( My new gpu is a gts 250 . it looked good in comparison with the gtx 260 in gpu grid's chart, close in bang for half the cost. almost bought a 260 today (cheaper now)but ended up getting a hub, some mice, keyboards, cables, kvm etc. woulda put it in a 64 bit linux box though (too cheap to buy windoze, can't stand it nywayz) as ThierryH says tho, #1 is best 4 Me cuz i don't need to use the pc.

why the 32 bit BOINC ?
my server2003 64 bit 'doze wont run that. (disingenuous (ungenuine) of course ;-)
is there a 64bit 'nix app in the pipeline ?

Server 2003 64 bit should be able to execute the application, my personal setup is using Windows XP x64 Professional Edition (which is based off of Server 2003) with a 32 bit BOINC install and it receives the 32 bit GPU application and the 64 bit CPU applications.

The application keeps segfaulting when executed through BOINC on Linux/GNU. It successfully runs as a standalone application, but once BOINC tries to execute it, it fails. Please see here [1] for more information.

[1] http://forums.nvidia.com/index.php?showtopic=105178

I have a Quad Intel processor, NVIDIA GPU running under 190.62 and is CUBA 2.3, I use BOINC 6.6.36. I upgraded to the BOINC 6.6.36 yesterday and since then I have not been able to get new work units.
I receive a message stating my card is not compatible (CUBA 2.3 and not 1.3).

How do I disable the GPU / CUBA function? The only option I can find is USE GPU WHILE COMPUTER IS IN USE. I have not selected this option.

Do I just have to Detach this quit this project? This is dumb if this is the case. I hope I am just having trouble finding the option.

Thanks for any help.

Yep, less credit than gpugrid more than collatz or almost the same.
But works on 260 gpu's that don't crunch gpugid anymore, happy!
Credit/Wu could be higer or wu faster.......

Ok, after few restarts it works!!!!

But I install driver 190.62, BOINC 6.10.1 and detached + atached to projest and set only GPU...

It works witch AQUA@HOME CPU.

Win7 64 bit.

On GTX260 (625Mhz GPU 1348 Shader) = 11 m. 25 s. = 685s. 74.24 credits so 1 credit in 9.23s.

GPUGRID pays better..... (1 credit at 4-7sec.)

But MW app is "young" and I hope soon it will be improved to have better performance :)

---

I have a Quad Intel processor, NVIDIA GPU running under 190.62 and is CUBA 2.3, I use BOINC 6.6.36. I upgraded to the BOINC 6.6.36 yesterday and since then I have not been able to get new work units.
I receive a message stating my card is not compatible (CUBA 2.3 and not 1.3).

How do I disable the GPU / CUBA function? The only option I can find is USE GPU WHILE COMPUTER IS IN USE. I have not selected this option.

Do I just have to Detach this quit this project? This is dumb if this is the case. I hope I am just having trouble finding the option.

Thanks for any help.

Jim, as I can see on your computers page, you have only series 8000 cards which are not compatible with Milkyway/CUDA application (see fisrt post of this thread).