So it is a wash as far as Windows/Drivers vs. Linux/Drivers are concerned.
It's a little more complicated than that. You haven't said which driver version or versions you used for testing with both OS's and I don't believe you can directly compare the performance of the Linux driver xyz v's the Windows version of the same driver.
As I no longer run any Linux machines all I can say is that the latest drivers for the Radeon vii give terrible performance in a Windows environment...
Always used the latest Linux driver from Amd. And tried several different windows drivers upto the latest.
Given the number of windows top ranked Radeon VIi gamma ray gpu processing systems I doubt the windows drivers have a problem with Radeon VIi hardware.
But I have limited experience and will finish selling the Radeon VIi tonight since it can't co-exist with legacy OCL1.2 under either windows or linux using the Boinc manager.
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
But I have limited experience and will finish selling the Radeon VIi tonight since it can't co-exist with legacy OCL1.2 under either windows or linux using the Boinc manager.
Selling?
Ideas are not fixed, nor should they be; we live in model-dependent reality.
I just installed a Radeon RX 580 video card with 8Gb VRAM.
I am running two GPU units at a time right now (utilization factor= 0.5), and I see that the VRAM is only about 12% utilized, though the GPU utilization reaches up to 100%.
I have 8 CPUs, thus 16 cores, so the CPU could handle more units simultaneously.
So how many work units should that card be able to run at one time? 4? 8? 16??
I was running 2 per gpu. Note a 1 task baseline. Then bump it to two. If the processing time increases to less than twice the baseline you have a winner. Ditto for 3x etc.
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Tom M wrote:So it is a wash
)
It's a little more complicated than that. You haven't said which driver version or versions you used for testing with both OS's and I don't believe you can directly compare the performance of the Linux driver xyz v's the Windows version of the same driver.
As I no longer run any Linux machines all I can say is that the latest drivers for the Radeon vii give terrible performance in a Windows environment...
Always used the latest Linux
)
Always used the latest Linux driver from Amd. And tried several different windows drivers upto the latest.
Given the number of windows top ranked Radeon VIi gamma ray gpu processing systems I doubt the windows drivers have a problem with Radeon VIi hardware.
But I have limited experience and will finish selling the Radeon VIi tonight since it can't co-exist with legacy OCL1.2 under either windows or linux using the Boinc manager.
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Tom M wrote:But I have
)
Selling?
Ideas are not fixed, nor should they be; we live in model-dependent reality.
I just installed a Radeon RX
)
I just installed a Radeon RX 580 video card with 8Gb VRAM.
I am running two GPU units at a time right now (utilization factor= 0.5), and I see that the VRAM is only about 12% utilized, though the GPU utilization reaches up to 100%.
I have 8 CPUs, thus 16 cores, so the CPU could handle more units simultaneously.
So how many work units should that card be able to run at one time? 4? 8? 16??
I was running 2 per gpu.
)
I was running 2 per gpu. Note a 1 task baseline. Then bump it to two. If the processing time increases to less than twice the baseline you have a winner. Ditto for 3x etc.
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Glenn Hawley, RASC Calgary
)
I am confused.
I am a beginnder and don't unserstand a lot of things.
Maybe you mean you have 1 PC with 1 CPU that has 8 Cores and (having HT on) thus 16 threads?
Or do you have perhaps 4 Workstations each with 2 CPUs, where each CPU has 2 Cores? With HT off and thus giving an overall of 16 Cores?
Or 8 PCs each with one CPU where each CPU has 2 Cores and HT (hyperthreading) off?
Life is very complicated, so relax and have a beer on me ...