There will be a new version of the OpenCL AMD app version on Monday fixing this problem. Till then (as the Windows versions also have problems), NVIDIA under Linux is the only remaining OpenCl app version in the beta test. Stay tuned.
Any chance the memory requirements will be lowered? I have a 512MB NVIDIA ready to do a few tests.
Sorry, I have very little hope that we can squeeze the current Cas A workunit data to fit onto 512MB cards in a safe way (so the card can also perform its main task of displaying graphics).
Any chance the memory requirements will be lowered? I have a 512MB NVIDIA ready to do a few tests.
Running double precision (as required) your GT 430 would not be faster than a single CPU core.. which applies to almost any GPU with 512 MB of VRAM or less.
Thanks for your replies. I didn't expect much throughput from this GPU anyway, just an opportunity to make some good use before retirement. I'll have to face the fact that there's no use anymore.
Frankly, I'm a bit worried by the tendency towards big work units. Sure, that fits the high performance GPUs better, but even my main GPU "only" has 1GB of RAM, which seems to be the minimum for all new applications. Just one more step in that direction and I'll be cut off.
Running double precision (as required) your GT 430 would not be faster than a single CPU core.. which applies to almost any GPU with 512 MB of VRAM or less.
I have to correct myself here: in the mean time I learned that the bulk of computations is still single precision, so it would be too early to rule cards out due to low double precision performance.
Thanks for reporting. The next build of the Beta app should have this fixed, I hope that this will be issued on Monday, together with a fixed Windows build.
BM
Thanks for the version 1.07 release. The tasks are running without error now. Run time is averaged at 574 seconds via a 7970. GPU load is in the range of 63-78% when the GPU is being used. So far the tasks that have completed are pending validation. My system is currently running driver 14.3.
I started up the new application on my GTX 780 Ti in Windows 64-bit. So far, the tasks are completing successfully. Five tasks have validated and the rest of the tasks are pending. The average run time of completed tasks so far is 664 seconds. GPU load ranges from 84-99%.
Thanks for
)
Thanks for reporting.
There will be a new version of the OpenCL AMD app version on Monday fixing this problem. Till then (as the Windows versions also have problems), NVIDIA under Linux is the only remaining OpenCl app version in the beta test. Stay tuned.
HB
Any chance the memory
)
Any chance the memory requirements will be lowered? I have a 512MB NVIDIA ready to do a few tests.
I have the same issue as
)
I have the same issue as Nigel Garvey with my PowerMac G5.
The actual error is:
process exited with code 193 (0xc1, -63)
and
SIGBUS: bus error
Thanks for testing. For the
)
Thanks for testing. For the time being I disabled the ALTIVEC-Beta app.
BM
BM
RE: Any chance the memory
)
Sorry, I have very little hope that we can squeeze the current Cas A workunit data to fit onto 512MB cards in a safe way (so the card can also perform its main task of displaying graphics).
HBE
RE: Any chance the memory
)
Running double precision (as required) your GT 430 would not be faster than a single CPU core.. which applies to almost any GPU with 512 MB of VRAM or less.
MrS
Scanning for our furry friends since Jan 2002
Thanks for your replies. I
)
Thanks for your replies. I didn't expect much throughput from this GPU anyway, just an opportunity to make some good use before retirement. I'll have to face the fact that there's no use anymore.
Frankly, I'm a bit worried by the tendency towards big work units. Sure, that fits the high performance GPUs better, but even my main GPU "only" has 1GB of RAM, which seems to be the minimum for all new applications. Just one more step in that direction and I'll be cut off.
RE: Running double
)
I have to correct myself here: in the mean time I learned that the bulk of computations is still single precision, so it would be too early to rule cards out due to low double precision performance.
MrS
Scanning for our furry friends since Jan 2002
RE: Thanks for reporting.
)
Thanks for the version 1.07 release. The tasks are running without error now. Run time is averaged at 574 seconds via a 7970. GPU load is in the range of 63-78% when the GPU is being used. So far the tasks that have completed are pending validation. My system is currently running driver 14.3.
I started up the new
)
I started up the new application on my GTX 780 Ti in Windows 64-bit. So far, the tasks are completing successfully. Five tasks have validated and the rest of the tasks are pending. The average run time of completed tasks so far is 664 seconds. GPU load ranges from 84-99%.