The CUDA application for 32 bit Windows has been deployed on BOINC. In terms of hardware a NVIDIA GPU supporting Compute Capability 1.3 is required. The following GPUs: GeForce GTX 295, 285, 280, 260, Tesla S1070, C1060, Quadro Plex 2200 D2, Quadro FX 5800, 4800 are known to have CUDA 1.3 support. The GPUs also need to have 256 MB of Video RAM and NVIDIA Driver 190.xx or higher needs to be installed. If these prerequisites are not met the CUDA application will not be downloaded through BOINC.

Hello,

That compute capability, where can I find if my cards have that? I have searched the nVidia site but could not find it with the specs of the cards.
My cards however are listed in the list with CUDA-enabled producsts. (8800GTS and 8600GT).

---

Greetings from,
TJ

That compute capability, where can I find if my cards have that? I have searched the nVidia site but could not find it with the specs of the cards.
My cards however are listed in the list with CUDA-enabled producsts. (8800GTS and 8600GT).

In the messages tab in the first 100 or so messages where it lists the detected CUDA devices ...

YOu need a 200+ GTX card for MW though yours may work at GPU Grid (they are on the ragged edge and may be always on the edge of not getting back in the deadline), and would be Ok at SaH and Collatz

Thanks for the answer Paul,

For GPUGRID they are indeed to "slow", I tried.
I have one GTX285 and is taking about 3 seconds for MW, so I hoped that the older cards could do something as well, but unfortunately they are 1.1 (reading in the message list indeed).

---

Greetings from,
TJ

Thanks for the answer Paul,

For GPUGRID they are indeed to "slow", I tried.
I have one GTX285 and is taking about 3 seconds for MW, so I hoped that the older cards could do something as well, but unfortunately they are 1.1 (reading in the message list indeed).

Put the other cards on Collatz. The pay is about the same as MW (for the same time invested) and with the latest optimized versions seems to run about as fast as a MW task batch ... or, shudder, shudder, you could run SaH ...

I have one GTX285 and is taking about 3 seconds for MW,

Wait, 3 seconds? Are you sure that isn't CPU time? The GPU should be taking about 250 seconds per WU if I recall correctly.

Wait, 3 seconds? Are you sure that isn't CPU time? The GPU should be taking about 250 seconds per WU if I recall correctly.

That could be the case. I look at the web site under my tasks and then the at Run Time (sec). There I see around 3 seconds.

---

Greetings from,
TJ

Hi,

i'm running MW@H on my ATI-computer, but i do have a pc equipped with a GT260 as well. Unfortunately it runs vista64. I just heard from that cuda app and want to run this app. How do i get it, will it run on vista64?

Regards,
Alexander

Hi,

i'm running MW@H on my ATI-computer, but i do have a pc equipped with a GT260 as well. Unfortunately it runs vista64. I just heard from that cuda app and want to run this app. How do i get it, will it run on vista64?

Regards,
Alexander

Hi Alexander,
A GTX 260 (compute ability 1.3) running Vista 64 with BOINC 6.10.13 up will run M@H. A GTS 260 will NOT as it only single precision math (compute ability 1.1).

---

Hi David,

it's a GTX260. Wherefrom can i get the files?

All the versions of Boinc are here: http://boinc.berkeley.edu/dl/?C=M;O=D.

Download Cuda 2.3 from here: http://www.nvidia.com/object/cuda_get.html

When Boinc is running you should, in the start up messages, see something like the following:

NVIDA GPU 0: GeForce GTX 260 (driver version 19062, CUDA version 2030, compute ability 1.3, 896M, 633GFLOPS peak)

This is what I get running 6.10.17 After I convert this box to Linux I will change to 6.10.18.

Edit:
If you have any problems, check out Ice's Brilliant Apps thread http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=631

---

Hi David,

the message from startup is there, CUDA Version 2030, ability 1.3, 896M, 477GFLOPS
I do have some cuda dll's, named cudart.dll in the gpugrid project directory, versions 2.2 and 2.3. BOINC is 6.10.18.
I copied the cudart.dll version 2.3 to my mw project directory. Is that all i have to do?
The brilliant-site has apps for the ATI-Cards, i do use them, but i did not find any cuda application. And in Ice's thread they talk about ATI-problems.
What else have you done to get this app?

Hi David,

the message from startup is there, CUDA Version 2030, ability 1.3, 896M, 477GFLOPS
I do have some cuda dll's, named cudart.dll in the gpugrid project directory, versions 2.2 and 2.3. BOINC is 6.10.18.

That's good, Boinc recognizes your card.

I copied the cudart.dll version 2.3 to my mw project directory. Is that all i have to do?

I also have cutil32.dll for what it's worth. When the site comes back up you should be a go. Afaik the dll's are no longer needed, and when the site comes back up M@H should automatically see the cuda and download a file called milkyway_0.21_windows_intelx86_cuda23. The WU's follow.

Important: Make sure you have Use GPUs for Computation? set to yes under M@H preferences.

The brilliant-site has apps for the ATI-Cards, i do use them, but i did not find any cuda application. And in Ice's thread they talk about ATI-problems.
What else have you done to get this app?

My bad, sorry.

---

Hi David,

thanks a lot for your help. I hope, that m@h comes up soon, if it still causes problems i will post that again.

Best regards

Alexander

The CUDA application for 32 bit Windows has been deployed on BOINC. In terms of hardware a NVIDIA GPU supporting Compute Capability 1.3 is required. The following GPUs: GeForce GTX 295, 285, 280, 260, Tesla S1070, C1060, Quadro Plex 2200 D2, Quadro FX 5800, 4800 are known to have CUDA 1.3 support. The GPUs also need to have 256 MB of Video RAM and NVIDIA Driver 190.xx or higher needs to be installed. If these prerequisites are not met the CUDA application will not be downloaded through BOINC.

A new preference has been added to the Milkyway@Home preferences page that specifies whether the project is allowed to use your GPU for computation, this has to be enabled in order to receive work from the CUDA application. This preference does not affect the anonymous plan class used by the ATI application.

While the CUDA application is running there will be a slowdown in the responsiveness of the user interface of the operating system, more specifically the explorer.exe process will use a lot of CPU. This is because the GPU is performing computations therefore it does not have a lot of free time to update the user interface, therefore the CPU has to do more work. There is a preference in BOINC to only use the GPU when the computer is idle.

If computation errors are being generated by the CUDA application please reference the results and the error message so the problem can be diagnosed. With the tremendous amount of variety in products, drivers, and operating systems it is high unlikely for everything to go smoothly these first few weeks, the Milkyway@Home team sincerely appreciates your patience in reaching this milestone and we look forward to your feedback.

This has been successfully tested on Windows XP x64 Professional Edition, BOINC 6.6.36 (32 bit), NVIDIA Drivers 190.62 (64 bit), and with a NVIDIA GeForce GTX 285 that was generously donated from NVIDIA to the Milkway@Home research team.

Just to let you know, the NV GT240 also works :)

How about my 5870, ccc9.11, win7-64? :)

---

Cheers,

PeterV

.