...For me it seems more like a scheduler problem. The result is running out of memory on the gpu and than have no empty ram and so killing a wu.
Yes, and if I recall correctly, it was fixed in 6.6.38 and a 6.10.x version after 6.10.4.
Gruß,
Gundolf
Ok I will give it a try. But 6.6.38 has another bug (atleast in win7 64bit). It only recognizes 1 graphicscard. The second card is not detected. 6.5 is working fine so far.
I will try 6.6.38 in winxp later.
hotze
Not detected, or detected but not used?
If you have two graphics cards of different specifications, BOINC v6.6.xx (and v6.10.xx) is designed only to use the "most capable". You can change that by setting the 1 configuration flag documented in client configuration.
If you have two graphics cards of different specifications, BOINC v6.6.xx (and v6.10.xx) is designed only to use the "most capable". You can change that by setting the 1 configuration flag documented in client configuration.
Not detected. I have 2 9800GX2 from the same vendor with the same specifications. 6.6.38 in winxp 32 seems good so far. Running on all 4 gpus and not killing wus.
One gpu is running in a PCIe16x slot and the other is in a PCIe4x slot. There is no significant difference in runtime yet. So there is some room for performance improvement ;)
BOINC 6.6.38 (winxp32 sp3) hasn´t trashed any wu so far. But I have seen a runtime variance between the gpus in the PCIe16 slot and the PCIe4 slot (which is reasonable).
PCIe16: 14200 seconds
PCIe4: 15400 seconds
on a Q6600 @ 3.6GHz.
Just so you folks know, for 6.10.x the only "usable" versions are .3, .7, and .11 based on experiences on GPU Grid, Collatz, MW ... oh, and my own personal use. :)
Though a few people have had .6 work for them there are serious issues with that build for most people. And yes, task deaths are among the problems with several of those versions ... as is unconstrained downloads.
I am mostly running 6.10.11 and 6.10.7 -- I think I have a few other iterations out there -- probably need to address that -- though I tend not to change until properly burned on a specific workstation...
Quote:
Just so you folks know, for 6.10.x the only "usable" versions are .3, .7, and .11 based on experiences on GPU Grid, Collatz, MW ... oh, and my own personal use. :)
Though a few people have had .6 work for them there are serious issues with that build for most people. And yes, task deaths are among the problems with several of those versions ... as is unconstrained downloads.
Quite all my CUDA Wu are aborting with a "Compute error" after 4-5 elapsed hours.
Now I have 21 WUs in the tube.
I think that if also the next one will abort, I'll force the abort for the other WUs, and I'll wait for a next ABP1 version.
Win XP 64b. BM 6.10.11 64b version. CUDA 190.62. NVIDIA 9800GT
No reason to change if there is no problem or you don't need a feature. I had been moving up because of the need for ATI support and then left it on a couple machines even after I moved cards about.
I have at least two machines on 6.6.2x I forget the exact versions ... for one thing on the Mac I don't know that there is any improvement in moving up ...
RE: RE: RE: ...For me
)
Not detected, or detected but not used?
If you have two graphics cards of different specifications, BOINC v6.6.xx (and v6.10.xx) is designed only to use the "most capable". You can change that by setting the 1 configuration flag documented in client configuration.
RE: Not detected, or
)
Not detected. I have 2 9800GX2 from the same vendor with the same specifications. 6.6.38 in winxp 32 seems good so far. Running on all 4 gpus and not killing wus.
One gpu is running in a PCIe16x slot and the other is in a PCIe4x slot. There is no significant difference in runtime yet. So there is some room for performance improvement ;)
hotze
BOINC 6.6.38 (winxp32 sp3)
)
BOINC 6.6.38 (winxp32 sp3) hasn´t trashed any wu so far. But I have seen a runtime variance between the gpus in the PCIe16 slot and the PCIe4 slot (which is reasonable).
PCIe16: 14200 seconds
PCIe4: 15400 seconds
on a Q6600 @ 3.6GHz.
Just so you folks know, for
)
Just so you folks know, for 6.10.x the only "usable" versions are .3, .7, and .11 based on experiences on GPU Grid, Collatz, MW ... oh, and my own personal use. :)
Though a few people have had .6 work for them there are serious issues with that build for most people. And yes, task deaths are among the problems with several of those versions ... as is unconstrained downloads.
I am mostly running 6.10.11
)
I am mostly running 6.10.11 and 6.10.7 -- I think I have a few other iterations out there -- probably need to address that -- though I tend not to change until properly burned on a specific workstation...
Quite all my CUDA Wu are
)
Quite all my CUDA Wu are aborting with a "Compute error" after 4-5 elapsed hours.
Now I have 21 WUs in the tube.
I think that if also the next one will abort, I'll force the abort for the other WUs, and I'll wait for a next ABP1 version.
Win XP 64b. BM 6.10.11 64b version. CUDA 190.62. NVIDIA 9800GT
No reason to change if there
)
No reason to change if there is no problem or you don't need a feature. I had been moving up because of the need for ATI support and then left it on a couple machines even after I moved cards about.
I have at least two machines on 6.6.2x I forget the exact versions ... for one thing on the Mac I don't know that there is any improvement in moving up ...
RE: Quite all my CUDA Wu
)
Your tasks abort with the "Maximum elapsed time exceeded" message.
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
Quite all my CUDA Wu are
)
Quite all my CUDA Wu are aborting with a "Compute error" after 4-5 elapsed hours...
Yew I saw it. But I can't do anything.
Another WU with Compute Error.
I aborted all the remaining WU
"Maximum elapsed time
)
"Maximum elapsed time exceeded"
Yes I know about it. But I can't do anything.
In next few days I'll receive a new GTX260. I hope it will better !!