Variable, and probably depends significantly on the speed of your CUDA card too.
Your Quadro FX 580 isn't listed in the developer documentation yet (at least not the public release version - see page 110 of that 145 page PDF document) - but its predecessor the Quadro FX 570 is very much at the bottom of the range - it has only 4 multiprocessors, where other consumer-grade cards have up to 30.
Overall, I would expect that the later CPU-only v3.11 Einstein ABP1 release would probably run faster on your machine that this initial CUDA test version. But maybe the balance will shift with the next CUDA release.
BTW, Bernd, how are you getting on? Anything new for us to test in the near future?
11.11.2009 18:50:20 Einstein@Home Sending scheduler request: Requested by user.
11.11.2009 18:50:20 Einstein@Home Requesting new tasks for CPU
11.11.2009 18:50:35 Einstein@Home Scheduler request completed: got 0 new tasks
11.11.2009 18:50:35 Einstein@Home Message from server: To get more Einstein@Home work, finish current work, stop BOINC, remove app_info.xml file, and restart.
11.11.2009 18:50:35 Einstein@Home Message from server: No work sent
11.11.2009 18:50:35 Einstein@Home Message from server: Your app_info.xml file doesn't have a version of Hierarchical S5 all-sky GW search #6.
What's wrong?
11.11.2009 18:50:04 Starting BOINC client version 6.10.18 for windows_x86_64
11.11.2009 18:50:04 Processor: 2 AuthenticAMD AMD Athlon(tm) 64 X2 Dual Core Processor 5200+ [AMD64 Family 15 Model 67 Stepping 2]
11.11.2009 18:50:04 Processor: 1.00 MB cache
11.11.2009 18:50:04 OS: Microsoft Windows 7: Ultimate x64 Edition, (06.01.7106.00)
11.11.2009 18:50:04 NVIDIA GPU 0: GeForce GTX 260 (driver version 19107, CUDA version 2030, compute capability 1.3, 896MB, 497 GFLOPS peak)
11.11.2009 18:50:04 Einstein@Home Found app_info.xml; using anonymous platform
11.11.2009 18:50:20 Einstein@Home Requesting new tasks for CPU
...
11.11.2009 18:50:35 Einstein@Home Message from server: Your app_info.xml file doesn't have a version of Hierarchical S5 all-sky GW search #6.
What's wrong?
1. You'll have to wait for a request for GPU tasks.
2. Does your app_info.xml contain S5R6-related information?
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
And else, should I change an app_info on other machine, pure-CPU host?
There is currenty no CPU version in beta test anymore (they were promoted to stock app by now), so could remove that app_info.xml after fully consuming the cached work.
Hello,
I download (automaticaly) the new cuda zip archive today from Einstein.
Using the GPU slew down my computer graphics !
I installed the latest driver for my graphic card (Nvidia - 9400 GP)
intalled the latest Boinc version 6.10.18 and... my graphics still moving VERY slowly.. SoI went back to my last Boinc version 6.6.36... nothing change.
In Boinc preferences, everything return normal when I unchecked use the GPU while the computer is in use.
Can you explain me how to come back to the last-1 version of your cuda driver ?
In advance, thanks a lot.
Lionel.
In fact it takes a few seconds to slow down the graphics. Seems like the memory of the GPU is getting full and the computer spend is time to swap between the GPU memory and the rest... All the graphics a terribly slow (close window, open window...)
Yes, I forgot in my last email : I'm using a Pentium IV 3.06 GHz, 1 G Ram, and Nvidia GPU 9400 GP.
Bests regards.
Lionel.
Hello,
I download (automaticaly) the new cuda zip archive today from Einstein.
Using the GPU slew down my computer graphics !
I installed the latest driver for my graphic card (Nvidia - 9400 GP)
intalled the latest Boinc version 6.10.18 and... my graphics still moving VERY slowly.. SoI went back to my last Boinc version 6.6.36... nothing change.
In Boinc preferences, everything return normal when I unchecked use the GPU while the computer is in use.
Can you explain me how to come back to the last-1 version of your cuda driver ?
In advance, thanks a lot.
Lionel.
Hello Kty
Is it the first time you use your gpu to compute ? If yes, it will be interesting to compare your feelings on an other gpu-computing project, like seti.
It is usual to have some screening slow down when gpu-computing, moreover when your card is a little one. The fact is that you sature your GPU by computing. Your solution to stop computing when user is active is a good compromise. That's what I've done on some machines.
Variable, and probably
)
Variable, and probably depends significantly on the speed of your CUDA card too.
Your Quadro FX 580 isn't listed in the developer documentation yet (at least not the public release version - see page 110 of that 145 page PDF document) - but its predecessor the Quadro FX 570 is very much at the bottom of the range - it has only 4 multiprocessors, where other consumer-grade cards have up to 30.
Overall, I would expect that the later CPU-only v3.11 Einstein ABP1 release would probably run faster on your machine that this initial CUDA test version. But maybe the balance will shift with the next CUDA release.
BTW, Bernd, how are you getting on? Anything new for us to test in the near future?
11.11.2009
)
11.11.2009 18:50:20 Einstein@Home Sending scheduler request: Requested by user.
11.11.2009 18:50:20 Einstein@Home Requesting new tasks for CPU
11.11.2009 18:50:35 Einstein@Home Scheduler request completed: got 0 new tasks
11.11.2009 18:50:35 Einstein@Home Message from server: To get more Einstein@Home work, finish current work, stop BOINC, remove app_info.xml file, and restart.
11.11.2009 18:50:35 Einstein@Home Message from server: No work sent
11.11.2009 18:50:35 Einstein@Home Message from server: Your app_info.xml file doesn't have a version of Hierarchical S5 all-sky GW search #6.
What's wrong?
11.11.2009 18:50:04 Starting BOINC client version 6.10.18 for windows_x86_64
11.11.2009 18:50:04 Processor: 2 AuthenticAMD AMD Athlon(tm) 64 X2 Dual Core Processor 5200+ [AMD64 Family 15 Model 67 Stepping 2]
11.11.2009 18:50:04 Processor: 1.00 MB cache
11.11.2009 18:50:04 OS: Microsoft Windows 7: Ultimate x64 Edition, (06.01.7106.00)
11.11.2009 18:50:04 NVIDIA GPU 0: GeForce GTX 260 (driver version 19107, CUDA version 2030, compute capability 1.3, 896MB, 497 GFLOPS peak)
11.11.2009 18:50:04 Einstein@Home Found app_info.xml; using anonymous platform
RE: 11.11.2009
)
1. You'll have to wait for a request for GPU tasks.
2. Does your app_info.xml contain S5R6-related information?
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
RE: 2. Does your
)
Don't know. It's from newly downloaded archive with cuda apps (einsteinbinary_ABP1_3.10_windows_intelx86_cuda.zip).
What should be there else?
RE: RE: 2. Does your
)
That zip was created before S5R6 began, so it's a bit out of date. You'd have to stop BOINC, make some updates to the app_info.xml
(analogous to http://einsteinathome.org/node/194500&nowrap=true#100336, see items highlighted in bold (don't copy and paste as that example is for Linux)
and restart BOINC.
Bikeman
Thanx, it started
)
Thanx, it started work.
But why not to insert these changes into archive at "Einstein@Home Beta Testing" page?
And else, should I change an app_info on other machine, pure-CPU host?
RE: And else, should I
)
There is currenty no CPU version in beta test anymore (they were promoted to stock app by now), so could remove that app_info.xml after fully consuming the cached work.
Bikeman
RE: Hello, I download
)
In fact it takes a few
)
In fact it takes a few seconds to slow down the graphics. Seems like the memory of the GPU is getting full and the computer spend is time to swap between the GPU memory and the rest... All the graphics a terribly slow (close window, open window...)
Yes, I forgot in my last email : I'm using a Pentium IV 3.06 GHz, 1 G Ram, and Nvidia GPU 9400 GP.
Bests regards.
Lionel.
RE: Hello, I download
)
Hello Kty
Is it the first time you use your gpu to compute ? If yes, it will be interesting to compare your feelings on an other gpu-computing project, like seti.
It is usual to have some screening slow down when gpu-computing, moreover when your card is a little one. The fact is that you sature your GPU by computing. Your solution to stop computing when user is active is a good compromise. That's what I've done on some machines.
See you
GuL