GPU enabled Seti@home Beta

Just a heads up for those of you who crunch on GPUs or would like to
http://setiweb.ssl.berkeley.edu/beta/cuda.php
Up graded to BOINC 6.4.5 today , haven’t seen any units yet, but My GPUs are fully loaded with GPUGRID units at the moment .
Edit – you need a Beta account to participate

Interesting…
The quad we do most crunching on is ATI Radeon HD 3400
But the boat anchor comp will qualify. So how can I resist testing this out.
Thanks for the info!

[QUOTE=Vortex;431068]Interesting…
The quad we do most crunching on is ATI Radeon HD 3400
But the boat anchor comp will qualify. So how can I resist testing this out.
Thanks for the info!

[/QUOTE]

Holy crap Vortex, I just checked out your connection speed. What are the details of the connection, cost, cap etc if you don’t mind divulging the details.

Nah I don’t mind, and have achieved faster.
No bandwidth limit, has speed boost on long downloads that has hit 36000kb/s
Cost is about $30.00 US

[QUOTE=Vortex;431072]Nah I don’t mind, and have achieved faster.
No bandwidth limit, has speed boost on long downloads that has hit 36000kb/s
Cost is about $30.00 US[/QUOTE]

Wow, that is awesome. In SA my connection is 4096kb/s with a 3GB cap and cost the equivalent of $130. Time to move I think;)

Its looking good so far , Crunched 17 WUs at 70 points each - a little over 4 minutes per WU. it could be a big deal when it goes public.

Report deadline	8 Jan 2009 16:34:38 UTC
CPU time	257.1988
stderr out	

<core_client_version>6.4.5</core_client_version>
<![CDATA[
<stderr_txt>
cudaAcc_initializeDevice: Found 2 CUDA device(s):
   Device 1 : GeForce 8800 GT 
   Device 2 : GeForce 8800 GT 
cudaAcc_initializeDevice is determiming what CUDA device to use...
   user specified SETI to use CUDA device 1: GeForce 8800 GT
SETI@home using CUDA accelerated device GeForce 8800 GT
setiathome_enhanced 6.02 Visual Studio/Microsoft C++
libboinc: 6.3.22

Work Unit Info:
...............
WU true angle range is :  0.378409
Optimal function choices:
-----------------------------------------------------
name                
-----------------------------------------------------
              v_BaseLineSmooth (no other)
            v_GetPowerSpectrum 0.00024 0.00000 
                   v_ChirpData 0.02199 0.00000 
                  v_Transpose4 0.00863 0.00000 
               FPU opt folding 0.00276 0.00000 

Flopcounter: 27893493293631.379000

Spike count:    1
Pulse count:    0
Triplet count:  0
Gaussian count: 0
called boinc_finish

</stderr_txt>
]]>

Validate state	Initial
Claimed credit	70.2591072865855
Granted credit	0
application version	6.05

That screams, 4 minutes for 70 credit points?

SO a $90 video card looks like it way outruns any CPU combination.

My laptop is using a 8600GT to do WU’s in 15 mins T7700 cpu!

G

ok, so am I missing a trick here or something - I’ve not been involved in the alpha for some time. By Beta do you mean the astropulse site? Are these units being dished out by the standard set@home?

DT.

At the moment they are only available as a tester on the SETI@home/AstroPulse Beta site , you need to open a Beta account .
I had a few yesterday 12 or 13 points for 24 seconds crunching - unfortunately none have been converted into hard points yet.

Edit - I see that you are already a member , but last contact was last year

cheers, I get mixed up with all the different alpha and beta I get involved with.

I’ll add it to my rig and see how things go.

DT.

[QUOTE=Mortlake;431227]At the moment they are only available as a tester on the SETI@home/AstroPulse Beta site , you need to open a Beta account .
I had a few yesterday 12 or 13 points for 24 seconds crunching - unfortunately none have been converted into hard points yet.

Edit - I see that you are already a member , but last contact was last year[/QUOTE]

Errr, I just installed the lastest enhanced client 603 and updated boinc on my laptop. My laptop is getting credit on my normal account. Credits completed yesterday have between 15 - 65 credit on them already, and the host average has risen.

G

Can this be run on a Quadro FX 370?

As long as it has more than 256MB - Yes
See http://setiweb.ssl.berkeley.edu/beta/cuda.php

[QUOTE=Gandelf;431240]Errr, I just installed the lastest enhanced client 603 and updated boinc on my laptop. My laptop is getting credit on my normal account. Credits completed yesterday have between 15 - 65 credit on them already, and the host average has risen.

G[/QUOTE]

If I looked at the right machine on your account it is only doing normal WUs on the CPU using an Alex Khan optimised app, not CUDA WUs on the GPU (currently 6.06)

19/12/2008 10:38:43||CUDA devices found
19/12/2008 10:38:43||Coprocessor: GeForce 8600M GT (1)

Does this mean it found it but it’s not using it?

G

From reading the forums it seems that 6.4.5 is the first version to enable GPU coprocessor detection for SETI . But I suppose until the CUDA app is released into the general population your GPU won’t get used unless you sign up for the Beta site or GPUGRID.net.

For CUDA , on the task details you get something like

stderr out	

<core_client_version>6.4.5</core_client_version>
<![CDATA[
<stderr_txt>
cudaAcc_initializeDevice: Found 2 CUDA device(s):
   Device 1 : GeForce 8800 GT 
   Device 2 : GeForce 8800 GT 
cudaAcc_initializeDevice is determiming what CUDA device to use...
   user specified SETI to use CUDA device 2: GeForce 8800 GT
SETI@home using CUDA accelerated device GeForce 8800 GT
setiathome_enhanced 6.02 Visual Studio/Microsoft C++
libboinc: 6.3.22

Work Unit Info:
...............
WU true angle range is :  0.378447
Optimal function choices:

in your laptop the T7700 CPU is used

stderr out	

<core_client_version>6.4.5</core_client_version>
<![CDATA[
<stderr_txt>
Unrecognized XML in parse_init_data_file: computation_deadline
Skipping: 1230207977.000000
Skipping: /computation_deadline
Unrecognized XML in parse_init_data_file: computation_deadline
Skipping: 1230207977.000000
Skipping: /computation_deadline
Windows optimized S@H Enhanced application by Alex Kan
Version info: SSSE3x (Intel, Core 2-optimized v8-nographics) V5.13 by Alex Kan
SSSE3x Win32 Build 41 , Ported by : Jason G, Raistmer, JDWhale

     CPUID: Intel(R) Core(TM)2 Duo CPU     T7700  @ 2.40GHz 
     Speed: 2 x 2408 MHz 
     Cache: L1=64K L2=4096K
  Features: MMX SSE SSE2 SSE3 SSSE3 
 
Work Unit Info:
...............
Credit multiplier is :  2.85
WU true angle range is :  2.714026

Flopcounter: 5215330804434.171900

Spike count:    6
Pulse count:    0
Triplet count:  0
Gaussian count: 0
called boinc_finish

</stderr_txt>
]]>

From what I’ve played with (under XP) there are other pitfalls too.

It’s now Windows mainstream, under 6.4.5. But the installation advice given is a little misleading. For a start, you need a utility like Techpowerup’s GPUZ (I’ll dig some addresses out in a bit). That will tell you the ‘core’ of the GPU.

It’s Nvidia only.

I eventually discovered I had a G80 core 8800 GTS 320 (64 shaders) which is CUDA 1.0 and, well, stuffed. The application is CUDA 2.0 only. And apparently, CUDA 2.1 beta is a bit buggy.

Generally, if it works with GPU Grid, (in Windows) it’ll work with Seti@Home. See here http://www.gpugrid.netforum_thread.php?id=316

As far as I can see, it needs Seti@Home 6.05. You can’t crunch with your CPU AND GPU, although the GPU app. uses a bit of a core. You can use your CPU for other projects tho.

That means, at least at the moment, if you’re used to using something like “Win32_AK_v8_SSSE3x”, Alex’s optimised cpu app, you’re going to have to disable it and its xml file.

Once that’s done, if your graphics card supports the new CUDA app, the rest will download and feed your card. Only… there’s a bit of a bit of bottleneck at Seti at the mo.

Oh, and the fact that some shorties seem to complete so fast (20 seconds anyone?) is leaving some to wonder if it’s buggy :slight_smile:

If you can’t do Seti on Graphics card, there’s always Folding :slight_smile: Feel free to correct me.

Mustangs done Today’s stats !

does everyone else get a lot of failures? I’ve got one CUDA unit to crunch using my setup, every other unit bombed out :confused:

DT.

[QUOTE=DoubleTop;431278]does everyone else get a lot of failures? I’ve got one CUDA unit to crunch using my setup, every other unit bombed out :confused:

DT.[/QUOTE]

I have found that I cannot overclock my windows GPUs at all with BOINC based crunching , although they both ran fine overclocked when I used them for folding (2x 8800GT-1GB) although my Linux hosted 8800GTS-512MB does 750 clock and 1000 memory without problems on GPUGRID
So far my GPUs are doing WUs at about 80 seconds CPU = 8mins Wall clock for a 14.56 unit and 210 seconds (25mins) for a 61.52 unit