NVIDIA just recently released a new stable 270 driver, version 270.41.06. I just loaded it up on my Linux system and the nv270 CUDA app is running well. This new driver supports all the latest cards as well.
Some days ago i've statrted getting computation errors on CUDA apps. I've got 270.41.06 driver, my card is recognized by BOINC: Чт. 12 Ð¼Ð°Ñ 2011 22:32:15 NVIDIA GPU 0: GeForce GTX 460 (driver version unknown, CUDA version 4000, compute capability 2.1, 1024MB, 717 GFLOPS peak)
But i'm getting immediate computation error for every CUDA app: Пт. 13 Ð¼Ð°Ñ 2011 08:13:30 Einstein@Home Output file PM0108_02171.dm_128_0_2 for task PM0108_02171.dm_128_0 absent
The E@H's CUDA app is the only one i'm running, so i can't tell if its BOINC's or E@H's problem., since driver functions fine.
../../projects/einstein.phys.uwm.edu/einsteinbinary_BRP3_1.08_i686-pc-linux-gnu__BRP3cuda32nv270: error while loading shared libraries: libstdc++.so.6: cannot open shared object file: No such file or directory
Did the new driver add this as a dependency? Or is this different between the apps ("fullCPU" and "nv270")?
What are the GPU and driver requirements for the new Windows CUDA applications?
I'm getting BRP3 CUDA workunits for one of my computers:
267.24 driver
GTS 450
Computing capability 2.1
993 MB graphics memory
but not the other:
258.96 driver
G 105M
Computing capability 1.1
485 MB graphics memory
Both have 8 GB main memory and 64-bit Windows Vista SP2.
I'm interested in finding a way to run two workunits in parallel on the GTS 450 but only one at a time on the G 105M.
I've now found that a 275.33 driver is available for the G 105M, and have downloaded it, but plan to drain all my current GPU workunits on that computer before deciding whether to install it.
What are the GPU and driver requirements for the new Windows CUDA applications?
I'm getting BRP3 CUDA workunits for one of my computers:
267.24 driver
GTS 450
Computing capability 2.1
993 MB graphics memory
but not the other:
258.96 driver
G 105M
Computing capability 1.1
485 MB graphics memory
Both have 8 GB main memory and 64-bit Windows Vista SP2.
I'm interested in finding a way to run two workunits in parallel on the GTS 450 but only one at a time on the G 105M.
I've now found that a 275.33 driver is available for the G 105M, and have downloaded it, but plan to drain all my current GPU workunits on that computer before deciding whether to install it.
The ansewer is to use a app_info.xml file on the machine with the GTS450 and set the parameter
CUDA
0.5000
Then you run two BRP in parallel
Fpr the other machine leave the app_info file away, or set the count-value to 1.0000
And, I think, the minimum graphics driver version is 266.??
NVIDIA just recently released
)
NVIDIA just recently released a new stable 270 driver, version 270.41.06. I just loaded it up on my Linux system and the nv270 CUDA app is running well. This new driver supports all the latest cards as well.
http://www.nvidia.com/object/linux-display-amd64-270.41.06-driver.html
nVidia's up to v.4.0.14 on
)
nVidia's up to v.4.0.14 on their Mac drivers, did this address any of those issues?
RE: nVidia's up to v.4.0.14
)
Nope, the issue persists with 4.0.14.
BM
BM
Some days ago i've statrted
)
Some days ago i've statrted getting computation errors on CUDA apps. I've got 270.41.06 driver, my card is recognized by BOINC:
Чт. 12 Ð¼Ð°Ñ 2011 22:32:15 NVIDIA GPU 0: GeForce GTX 460 (driver version unknown, CUDA version 4000, compute capability 2.1, 1024MB, 717 GFLOPS peak)But i'm getting immediate computation error for every CUDA app:
Пт. 13 Ð¼Ð°Ñ 2011 08:13:30 Einstein@Home Output file PM0108_02171.dm_128_0_2 for task PM0108_02171.dm_128_0 absentThe E@H's CUDA app is the only one i'm running, so i can't tell if its BOINC's or E@H's problem., since driver functions fine.
The actual error is visible
)
The actual error is visible here:
Did the new driver add this as a dependency? Or is this different between the apps ("fullCPU" and "nv270")?
BM
BM
This thread might be of
)
This thread might be of help:
http://einsteinathome.org/node/195779&nowrap=true#112033
Thanks, that post is exactly
)
Thanks, that post is exactly my case.
What are the GPU and driver
)
What are the GPU and driver requirements for the new Windows CUDA applications?
I'm getting BRP3 CUDA workunits for one of my computers:
267.24 driver
GTS 450
Computing capability 2.1
993 MB graphics memory
but not the other:
258.96 driver
G 105M
Computing capability 1.1
485 MB graphics memory
Both have 8 GB main memory and 64-bit Windows Vista SP2.
I'm interested in finding a way to run two workunits in parallel on the GTS 450 but only one at a time on the G 105M.
I've now found that a 275.33 driver is available for the G 105M, and have downloaded it, but plan to drain all my current GPU workunits on that computer before deciding whether to install it.
See BRP CUDA requirements (
)
See BRP CUDA requirements ( READ ONLY thread ) ;-)
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
RE: What are the GPU and
)
The ansewer is to use a app_info.xml file on the machine with the GTS450 and set the parameter
CUDA
0.5000
Then you run two BRP in parallel
Fpr the other machine leave the app_info file away, or set the count-value to 1.0000
And, I think, the minimum graphics driver version is 266.??