V10/11 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

[svenie25]

Okay, the problem is this: I only want to use the optimized apps at my GPU. On the CPU should run Einstein@home. I am using BOINC-Manager 6.6.20. Down loaded and installed the latest package. But if I get an WU for SETI the Manager isn´t run it on the GPU, I have only 2 active tasks. Is there something that I can do?

[M_M]

Hi,

This is what Boinc sees:
"31.03.2009 21:18:16      CUDA device: GeForce GTX 295 (driver version 18208, CUDA version 1.3, 896MB, est. 106GFLOPS)", which is correct, as there is only one physical card in system - with two GPU cores.

I use Raistmer's_opt_pack_V10a_x64_SSSE3_MB_CPU_GPU_team, as as said earlier - I've set the value to 2 in "number_of_GPUs" file.

Morten

you have to disable SLI for CUDA to see both GPUs as "independent" GPU's...

[svenie25]

I think, I said wrong. I have a C2D and 1 GPU for crunching. The GPU should do SETI with the optimized app and the 2 CPUs Einstein.

[M_M]

I think, I said wrong. I have a C2D and 1 GPU for crunching. The GPU should do SETI with the optimized app and the 2 CPUs Einstein.

I don't think boinc is smart enough for this... if it sees a unused core (cpu or gpu) and is has a seti task, it can give it to CPU as well. And then your einstain comes, and there is no free cpu core, and it waits for it....

Have you tried latest development 6.6.20? Maybe this one is smart enough?

[Claggy]

Okay, the problem is this: I only want to use the optimized apps at my GPU. On the CPU should run Einstein@home. I am using BOINC-Manager 6.6.20. Down loaded and installed the latest package. But if I get an WU for SETI the Manager isn´t run it on the GPU, I have only 2 active tasks. Is there something that I can do?

You just need a Cuda only app_info, Raistmer's V10/V11 app can be run standalone without the team part.

Here's a Cuda only app info using Raistmer's V11 app, you'll need to put in your own GPU flops figure, but Multiply it by 0.2 first, get it from the Boinc messages log and Don't forget all the extra zeroes seeing as its in gigaflops.

Claggy

Code: [Select]

<app_info>
<app>
<name>setiathome_enhanced</name>
</app>
<file_info>
<name>MB_6.08_mod_CUDA_V10.exe</name>
<executable/>
</file_info>
<file_info>
<name>cudart.dll</name>
<executable/>
</file_info>
<file_info>
<name>cufft.dll</name>
<executable/>
</file_info>
<file_info>
<name>libfftw3f-3-1-1a_upx.dll</name>
<executable/>
</file_info>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>608</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>0.127970</avg_ncpus>
<max_ncpus>0.127970</max_ncpus>
<flops>12000000000</flops>
<plan_class>cuda</plan_class>
<file_ref>
<file_name>MB_6.08_mod_CUDA_V10.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>cudart.dll</file_name>
</file_ref>
<file_ref>
<file_name>cufft.dll</file_name>
</file_ref>
<file_ref>
<file_name>libfftw3f-3-1-1a_upx.dll</file_name>
</file_ref>
<coproc>
<type>CUDA</type>
<count>1</count>
</coproc>
</app_version>
</app_info>



[svenie25]

Thanks a lot. Looks like it works. If BOINC mangaer says CUDA has 85GFlops, I have only to change "<flops>12000000000</flops>" to "<flops>85000000000</flops>"?

[Claggy]

Thanks a lot. Looks like it works. If BOINC mangaer says CUDA has 85GFlops, I have only to change "<flops>12000000000</flops>" to "<flops>85000000000</flops>"?

Sorry, you've got to Multiply that figure by 0.2

My 9800GTX+ was 87 Est.Gflops, ie 87 000 000 000 * 0.2 = 17 400 000 000

see MarkJ's app_info thread

Claggy

[svenie25]

Tahnks a lot again. Now I think, I have it.

[Torix]

Sorry for beeing a noob first
im using Bonic Client 6.4.7
Intel Core2Duo 8400 45nm with support ut to SSE4.1
3Gb Ram PC1023
XP 32bit Pro SP3
Gainward 260 55nm version 216 shaders

i see 2 versions one called x86_SSSE3 and x86_SSE3 and im not shure whitch version i sould use.
Second Question is where is this cc_config.xml you are talking about, i dont have this file on my pc (checkt manuell and via search funktion includet hidden/system ... files.

[sunu]

You can use the SSSE3 one.
You say that you use boinc 6.4.7. I suggest you download the latest beta 6.6.20 from berkeley. It's beta but it hasn't any big problems and it has much much better handling of cuda.
The cc_config.xml file is found in the boinc directory together with the boinc executables, client_state.xml etc. I think it doesn't exist by default, you have to write one by yourself.

[Torix]

thx sunu have found the place and createt the file 

[sunu]

Ok Torix. Forgot to tell you, here you can find everything you can put in cc_config.xml.

[Torix]

thx Raistmer for your aps.
with the original astropulse_v5_503 my computer needet 54h to complete the unit with your moddet ap_5.03r112_SSE3.exe it went down to 7h 25min now. Also changed the Bonic client now to the Beta 6.6.20 Version (computet first all downloadet WU´s and update it without any active WU´s) seems running fine on my pc.

[fusion2905]

Hi everyone. I have a quick question. If I try the package with 6.6.20 will I need to edit app_info to include flops (or edit it in any other way for that matter), or will it just be enough to do what's said in the first post?

thanks


--
Orin

my PC

[Raistmer]

thx Raistmer for your aps.
with the original astropulse_v5_503 my computer needet 54h to complete the unit with your moddet ap_5.03r112_SSE3.exe it went down to 7h 25min now. Also changed the Bonic client now to the Beta 6.6.20 Version (computet first all downloadet WU´s and update it without any active WU´s) seems running fine on my pc.

Yes, opt AP is a LOT faster than 5.03 stock app but, please, don't forget that current state of opt AP is a result of whole Lunatics team work. All entered some speeding up mods or gave some speeding up ideas.
BTW, some of these modifications already merged in stock too so we can hope new 5.05 and up AP will be much faster than its precursors.