To get NVIDIA Settings to work I installed the NV Repository driver 440 first, and ran the Coolbits tweak to create an xorg.conf. Then installed the New 20.10 driver using ./amdgpu-pro-install --opencl=legacy --headless . I used an app_config.xml file to run two AMD tasks since it's a little easier than using an app_info.xml.
It works very nicely on the BIOSTAR Mining board since it only has an 8 core CPU, but can run 12 GPUs. With the one AMD card running two tasks it still uses less CPU than it was running 8 NV tasks. So, even the 7 NV tasks are running a little faster. Hmmm, I need to replace these other GTX1060s with a RX570 ;-) It's similar to replacing a GTX1060 with a RTX2070, only much cheaper, and much better CPU usage.
I just got it going a little while ago, seems to be working well.
But can I get more total production with 1, 2 or 3 tasks?
You'll have to try and find out for yourself. run 2x for a while, and keep track of WU run times to get an average production rate. then change to 3x, and repeat. don't think anyone will have the "this is best" answer.
But can I get more total production with 1, 2 or 3 tasks?
You'll have to try and find out for yourself. run 2x for a while, and keep track of WU run times to get an average production rate. then change to 3x, and repeat. don't think anyone will have the "this is best" answer.
I have some baselines on 1:1 or 0.1:1so I have switched to 2 tasks on each gpu. The NV is running 1 thread per gpu task. The Amd is running 0.1 thread per gpu task.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor)
the CPU values really don't mean much as far as actual CPU use, ie. you can't "assign" or "limit" CPU resources used by the GPU app, the app will use whatever it needs. you could set the nvidia to 0.1 CPU and it'll still use a whole core. that number is only used for BOINC's resource management, so it knows how much free resources it has for assigning work. I would just observe what the app is using, and set the value close to it's actual use so that BOINC can properly account for it.
for example, if you tell it 0.1 CPU on nvidia tasks, but its really using 1.0, BOINC thinks more resources are free than there really are, and will assign more jobs to run and overcommit the system. just set the value to whatever percentage of a CPU thread you see being used by the AMD processes.
sounds like something to do with Secure Boot. set that to "other OS" or turn it off, if possible.
Any Windows setting was toggled back and forth numerous times without any change in the AMD Error. But there was a change with the NV cards, change the settings from what they are and then the NV cards also don't work. That's a pretty good indication the settings are correct, change the settings and nothing works, use them as are and at least the NV cards work. The only thing that makes sense is there is some error in the BIOS Microcode for this particular ASUS Strix 570. The Card is fine on every other board, only the MSI board gives this error. Next time I have a reason I'm going to flash the last BIOS to the board and see if that works. That may be some time away.
It's a done deal, All sorted.
)
It's a done deal, All sorted. I even have both the NV & AMD GPU temps in GKrell. This machine, https://einsteinathome.org/host/12820885
To get NVIDIA Settings to work I installed the NV Repository driver 440 first, and ran the Coolbits tweak to create an xorg.conf. Then installed the New 20.10 driver using ./amdgpu-pro-install --opencl=legacy --headless . I used an app_config.xml file to run two AMD tasks since it's a little easier than using an app_info.xml.
It works very nicely on the BIOSTAR Mining board since it only has an 8 core CPU, but can run 12 GPUs. With the one AMD card running two tasks it still uses less CPU than it was running 8 NV tasks. So, even the 7 NV tasks are running a little faster. Hmmm, I need to replace these other GTX1060s with a RX570 ;-) It's similar to replacing a GTX1060 with a RTX2070, only much cheaper, and much better CPU usage.
I just got it going a little while ago, seems to be working well.
TBar wrote:and kept getting
)
sounds like something to do with Secure Boot. set that to "other OS" or turn it off, if possible.
told ya ;)
_________________________________________________________________________
Tom M wrote: But can I get
)
You'll have to try and find out for yourself. run 2x for a while, and keep track of WU run times to get an average production rate. then change to 3x, and repeat. don't think anyone will have the "this is best" answer.
_________________________________________________________________________
Ian&Steve C. wrote: Tom M
)
I have some baselines on 1:1 or 0.1:1so I have switched to 2 tasks on each gpu. The NV is running 1 thread per gpu task. The Amd is running 0.1 thread per gpu task.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor)
the CPU values numbers really
)
the CPU values really don't mean much as far as actual CPU use, ie. you can't "assign" or "limit" CPU resources used by the GPU app, the app will use whatever it needs. you could set the nvidia to 0.1 CPU and it'll still use a whole core. that number is only used for BOINC's resource management, so it knows how much free resources it has for assigning work. I would just observe what the app is using, and set the value close to it's actual use so that BOINC can properly account for it.
for example, if you tell it 0.1 CPU on nvidia tasks, but its really using 1.0, BOINC thinks more resources are free than there really are, and will assign more jobs to run and overcommit the system. just set the value to whatever percentage of a CPU thread you see being used by the AMD processes.
_________________________________________________________________________
Ian&Steve C. wrote: sounds
)
Any Windows setting was toggled back and forth numerous times without any change in the AMD Error. But there was a change with the NV cards, change the settings from what they are and then the NV cards also don't work. That's a pretty good indication the settings are correct, change the settings and nothing works, use them as are and at least the NV cards work. The only thing that makes sense is there is some error in the BIOS Microcode for this particular ASUS Strix 570. The Card is fine on every other board, only the MSI board gives this error. Next time I have a reason I'm going to flash the last BIOS to the board and see if that works. That may be some time away.
Meanwhile the NV machine with 10 GPUs has run down another Windows machine, now only One Windows machine left, https://einsteinathome.org/community/stats/hosts