I promoted the 1.14 out of Beta status and re-issued 1.08 as 1.15 (Beta). You can still decide what version you get (1.14 with RecalcGPU or 1.15 with RecalcCPU) by means of the "Beta work" switch, but now reversed as before. The logic behind this is that people that just run BOINC with more or less default settings (including no Beta work) and don't actively manage their configuration will only run one such task per GPU, and for those the 1.14 should work better.
Not sure that change quite worked as you intended. After receiving a 1.15 I decided I'd prefer to run 1.14 over 1.08/1.15 so I set beta to NO.
I am now getting 1.07 OpenCL instead of 1.14 Cuda on my Linux box.
Is there a plan to build a CUDA O3AS app for Windows?
I tried getting the Linux one to run via WSL2 Ubuntu but BOINC doesn't recognize the GPU in that setup. Not sure if it's a WSL2 issue or me not knowing how to set things up correctly.
OK, so if I "baby sit" the BOINC Manager and suspend all CPU tasks when the GPU one needs the CPU during the start/middle/end, I can drastically reduce the overall time for the GPU task, albeit while stopping work on the CPU tasks for several minutes.
in Linux while ALSO maxing out CPU with CPU tasks, 1.15 cuda:
~2179
Linux while suspending CPU tasks when the GPU one needs the CPU, 1.15 cuda:
1031
Unfortunately all the suspending and resuming of the CPU-only tasks MAY have caused issues, so use with caution!
The CPU tasks all eventually ended up getting stuck just under 90%, even though running and with the elapsed time going up. While that COULD be normal for them, my current GPU task then wouldn't get above 50% GPU usage. After (sadly) aborting all the CPU tasks, the GPU usage immediately increased (although this could just be it needing CPU to manage it, and whatever the CPU tasks had been doing the GPU wasn't getting enough?).
Anyway, that's enough fiddling for me. Back to just running 1.15 (GW-cuda) tasks again now, with run time again what it was originally without any CPU tasks.
David, maybe if you were to un-hide your computer(s) so we could see what you are actually working with, you would get more responses to your 5 previous posts.
I'm just saying that because many new members don't realize that their computer(s) are hidden and we can be of more help if you did un-hide them. You may not be aware but this is quite a secure system and many of us do have our computers open for others to see, and be of help if need be.
JohnDK wrote: So there's no
)
No. It's the exact same binary.
BM
Bernd Machenschalk wrote:I
)
Not sure that change quite worked as you intended. After receiving a 1.15 I decided I'd prefer to run 1.14 over 1.08/1.15 so I set beta to NO.
I am now getting 1.07 OpenCL instead of 1.14 Cuda on my Linux box.
Is there a plan to build a
)
Is there a plan to build a CUDA O3AS app for Windows?
I tried getting the Linux one to run via WSL2 Ubuntu but BOINC doesn't recognize the GPU in that setup. Not sure if it's a WSL2 issue or me not knowing how to set things up correctly.
I would also love to see a
)
I would also love to see a Win build of this.
Hi, Are 1.14 with
)
Hi,
Are 1.14 with RecalcGPU ("1.14 (GW-cuda)") still being issued?
With Beta apps off I receive 1.07 (GW-opencl-nvidia-2) tasks.
With Beta apps on I receive 1.15 (GW-cuda)
Many thanks.
Some stats (while I wait on a
)
Some stats (while I wait on a reply about 1.14 tasks above).
GW run times on my 2080 Super in different configurations (all with just one GPU task running):
in Windows while ALSO maxing out CPU with CPU tasks:
~4200
in Windows with NO CPU tasks:
~1100
in Linux, NO CPU tasks, 1.07 opencl:
~897
in Linux, NO CPU tasks, 1.15 cuda:
877 (one finished so far)
Hence my interest in 1.14 tasks to see if those figures can be reduced any further.
out of curiosity: in
)
out of curiosity:
in Linux while ALSO maxing out CPU with CPU tasks, 1.15 cuda:
~2179
So obviously slower than with no CPU tasks, but nowhere near the slowdown when doing the same in Windows.
OK, so if I "baby sit" the
)
OK, so if I "baby sit" the BOINC Manager and suspend all CPU tasks when the GPU one needs the CPU during the start/middle/end, I can drastically reduce the overall time for the GPU task, albeit while stopping work on the CPU tasks for several minutes.
in Linux while ALSO maxing out CPU with CPU tasks, 1.15 cuda:
~2179
Linux while suspending CPU tasks when the GPU one needs the CPU, 1.15 cuda:
1031
in Linux, NO CPU tasks, 1.15 cuda:
~876
Unfortunately all the
)
Unfortunately all the suspending and resuming of the CPU-only tasks MAY have caused issues, so use with caution!
The CPU tasks all eventually ended up getting stuck just under 90%, even though running and with the elapsed time going up. While that COULD be normal for them, my current GPU task then wouldn't get above 50% GPU usage. After (sadly) aborting all the CPU tasks, the GPU usage immediately increased (although this could just be it needing CPU to manage it, and whatever the CPU tasks had been doing the GPU wasn't getting enough?).
Anyway, that's enough fiddling for me. Back to just running 1.15 (GW-cuda) tasks again now, with run time again what it was originally without any CPU tasks.
David, maybe if you were to
)
David, maybe if you were to un-hide your computer(s) so we could see what you are actually working with, you would get more responses to your 5 previous posts.
I'm just saying that because many new members don't realize that their computer(s) are hidden and we can be of more help if you did un-hide them. You may not be aware but this is quite a secure system and many of us do have our computers open for others to see, and be of help if need be.
Proud member of the Old Farts Association