Now with another card... GTX 760, underclocked, running three BRP6-Beta-cuda55's simultaneously. Again, 356.43 made a clear reduction on the CPU time on every task. These are average results I chose by eye:
No, I don't know what that is. I didn't install anything else than the display driver (and non-avoidable PhysX...) from the 356.43-vkonly_geforce_win10_64bit_international.exe .
well, this exe-file is are the API Runtime Librarys, must be importent for the regular use. I dont know if it changes anything, i installed the drivers, and dont saw anything that said me that the Vulkan devices drivers were installed, then i searched on my own, and got it.
Vulkan Driver Support
Windows driver version 356.45 and Linux driver version 355.00.29 provide beta support for Vulkan. Release Updates: March 2nd
Vulkan on NVIDIA Hardware FAQ
.....
4) Will there be an official (non-beta) NVIDIA driver with Vulkan support?
Yes. We plan to include Vulkan in both Windows and Linux drivers in an upcoming driver release. These will be available to all users to run Vulkan content as part of our regular driver releases.
.....
Can wait :-)
I know I am a part of a story that starts long before I can remember and continues long beyond when anyone will remember me [Danny Hillis, Long Now]
Yes. We plan to include Vulkan in both Windows and Linux drivers in an upcoming driver release. These will be available to all users to run Vulkan content as part of our regular driver releases.
.....
Can wait :-)
Driver has now been released by Nvidia 364.47 for Win vista/7/8 and Win10 32 & 64 WHQL
Vulkan as I understand it allows closer usage of the hardware, with less overhead. But I assume that requires that the api be addressed by programming for it with whatever application is being run..
Didn't think any e@h app did so.. unless there is some sort of tie in with Cuda I'm not aware of.. And since I'm most certainly NOT a programmer I leave that to those who are:-)
Hi Folks,
Well I tried Nvidia's latest & greatest:-( And for me it sucks!
I run an Asus sabertooth FX 990 Mobo, Amd FX 9590 cpu and 2 GTX 980ti GPU's.
1 x 980ti is dedicated to E@H
1 x 980ti is dedicated to MW@H
cpu runs 1 x A@H cpu task.
Both GPU's have NVI corrected p2 state memory timings to 3505MHz
Cpu has OC FSB of + 11 Mhz.
Under the new driver I observed a reduction in time to complete [TTC] of 5/6 secs for MW@H tasks down from 1 min 55 secs to 1 min 49/50 secs.
For E@H I had 3 min plus added to TTC for BRP6 tasks!!
For A@H no significant change.
I have reverted to a prior driver by simply installing it over the new one.
And saw an immediate improvement in BRP6 times. Back to 43 mins 30 secs per task.
Boinc properties box showed a 2.147 % per Min for E@H BRP6 with the new driver and 2.316 % per min for the older driver at max and 2.286 at lowest.
That for me was the clincher!
I suspect that its possible that the earlier non WHQL Vulkan driver works better than the new release because the non vulkan part of the install is better suited to E@H tasks.
Thanks for the report, Cliff. I suspected that the new Vulkan API mechanism in the latest drivers would offer no benefit to distributed computing. Likely no benefit to gaming either until the game developers can hook into the API optimizations that Vulkan offers. Appreciate you being the 'guinea pig' instead of me.
Thanks for the report, Cliff. I suspected that the new Vulkan API mechanism in the latest drivers would offer no benefit to distributed computing. Likely no benefit to gaming either until the game developers can hook into the API optimizations that Vulkan offers. Appreciate you being the 'guinea pig' instead of me.
You're welcome:-)
Yup I suspect the same, and perhaps the more recent NVidia drivers are less then helpful crunching E@H WU than earlier ones. Dunno what they have done with their 'gaming' tweaks, but it certainly seems to impact DComp adversely, at least on AMD platforms. Maybe they do less harm on intel based rigs?
Again being a distrustful soul I suspect competition might be behind any AMD based rig problems.. but who knows..
Now with another card... GTX
)
Now with another card... GTX 760, underclocked, running three BRP6-Beta-cuda55's simultaneously. Again, 356.43 made a clear reduction on the CPU time on every task. These are average results I chose by eye:
361.91:
https://einsteinathome.org/task/546917503
356.43:
https://einsteinathome.org/task/546921590
No gain (gpu) on Nvidia 7xx
)
No gain (gpu) on Nvidia 7xx and newer...
I am running single workunit, with "Nvidia Experience" , maybe theres a small headroom up ?
Just for fun i will run 3 wus , too. And see.
My forecast calculation says : 10200s each wu for the gpu.
I also have manualy this installed:
C:\NVIDIA\DisplayDriver\356.43-vkonly\Win8_WinVista_Win7_64\International\Display.Driver
VulkanRT-1.0.2.0-Installer.exe
u got it too ?
1st:
)
1st: 10296s
https://einsteinathome.org/task/548128636
2nd: 10316s
https://einsteinathome.org/task/548128637
3rd: and the worst 10472s
https://einsteinathome.org/task/548130443
lol, i saved 976 seconds on crunshing 3 wus at the same time, not much.
But what will that tell us ?
RE: VulkanRT-1.0.2.0-Instal
)
No, I don't know what that is. I didn't install anything else than the display driver (and non-avoidable PhysX...) from the 356.43-vkonly_geforce_win10_64bit_international.exe .
well, this exe-file is are
)
well, this exe-file is are the API Runtime Librarys, must be importent for the regular use. I dont know if it changes anything, i installed the drivers, and dont saw anything that said me that the Vulkan devices drivers were installed, then i searched on my own, and got it.
https://developer.nvidia.com/
)
https://developer.nvidia.com/Vulkan
https://developer.nvidia.com/vulkan-driver
Vulkan Driver Support
Windows driver version 356.45 and Linux driver version 355.00.29 provide beta support for Vulkan. Release Updates: March 2nd
Vulkan on NVIDIA Hardware FAQ
.....
4) Will there be an official (non-beta) NVIDIA driver with Vulkan support?
Yes. We plan to include Vulkan in both Windows and Linux drivers in an upcoming driver release. These will be available to all users to run Vulkan content as part of our regular driver releases.
.....
Can wait :-)
I know I am a part of a story that starts long before I can remember and continues long beyond when anyone will remember me [Danny Hillis, Long Now]
RE: Yes. We plan to
)
Driver has now been released by Nvidia 364.47 for Win vista/7/8 and Win10 32 & 64 WHQL
Vulkan as I understand it allows closer usage of the hardware, with less overhead. But I assume that requires that the api be addressed by programming for it with whatever application is being run..
Didn't think any e@h app did so.. unless there is some sort of tie in with Cuda I'm not aware of.. And since I'm most certainly NOT a programmer I leave that to those who are:-)
Yet to try it:-)
Cheers,
Cliff,
Been there, Done that, Still no damm T Shirt.
Hi Folks, Well I tried
)
Hi Folks,
Well I tried Nvidia's latest & greatest:-( And for me it sucks!
I run an Asus sabertooth FX 990 Mobo, Amd FX 9590 cpu and 2 GTX 980ti GPU's.
1 x 980ti is dedicated to E@H
1 x 980ti is dedicated to MW@H
cpu runs 1 x A@H cpu task.
Both GPU's have NVI corrected p2 state memory timings to 3505MHz
Cpu has OC FSB of + 11 Mhz.
Under the new driver I observed a reduction in time to complete [TTC] of 5/6 secs for MW@H tasks down from 1 min 55 secs to 1 min 49/50 secs.
For E@H I had 3 min plus added to TTC for BRP6 tasks!!
For A@H no significant change.
I have reverted to a prior driver by simply installing it over the new one.
And saw an immediate improvement in BRP6 times. Back to 43 mins 30 secs per task.
Boinc properties box showed a 2.147 % per Min for E@H BRP6 with the new driver and 2.316 % per min for the older driver at max and 2.286 at lowest.
That for me was the clincher!
I suspect that its possible that the earlier non WHQL Vulkan driver works better than the new release because the non vulkan part of the install is better suited to E@H tasks.
My tuppence worth, YMMV.
Cheers,
Cliff,
Been there, Done that, Still no damm T Shirt.
Thanks for the report, Cliff.
)
Thanks for the report, Cliff. I suspected that the new Vulkan API mechanism in the latest drivers would offer no benefit to distributed computing. Likely no benefit to gaming either until the game developers can hook into the API optimizations that Vulkan offers. Appreciate you being the 'guinea pig' instead of me.
Hi Keith, RE: Thanks
)
Hi Keith,
You're welcome:-)
Yup I suspect the same, and perhaps the more recent NVidia drivers are less then helpful crunching E@H WU than earlier ones. Dunno what they have done with their 'gaming' tweaks, but it certainly seems to impact DComp adversely, at least on AMD platforms. Maybe they do less harm on intel based rigs?
Again being a distrustful soul I suspect competition might be behind any AMD based rig problems.. but who knows..
Regards,
Cliff,
Been there, Done that, Still no damm T Shirt.