they also (in some cases) are resending older WUs that had issues with cross validation between nvidia/AMD. one of mine just validated with a Radeon VII task completed on Aug 3rd.
Just got one where a v0.01 BRP7-opencl-ati validated with a v0.03 BRP7-cuda55 (both Linux), and tossed my v0.03 Windows out of the window.
It seems we finally got the CUDA version working. Phew! Comparability of the results (what you see as validation) is still bad, but it seems we got that working, too, behind the scenes. Next version of apps should be better, hopefully early next week.
Twelve validated : no problem with MeerKAT v0.03 () x86_64-pc-linux-gnu here - on a 3060 - in each case when paired with another single x86_64-pc-linux-gnu also running cuda55 ... ;-))
Fourteen invalid : in each case within quorums with two windows_x86_64 !! ;-((
So from my rig it looks to be linux vs windows ?!
On the upside they're awfully quick though @ 17 seconds.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
These are still test tasks, basically trying whether 1. the workunit generator and 2. the app works at all. These should have 1200 templates only, the next, more realistic set will have 48000-50000 templates.
These are still test tasks, basically trying whether 1. the workunit generator and 2. the app works at all. These should have 1200 templates only, the next, more realistic set will have 48000-50000 templates.
WOO HOO!!!
Unfortunately I will be a part of the next test as I'm out of town on a several year Covid cancelled cruise/land stuff for the next 20 days.
[sulks] I want AMD on Windows. Why did you have to program the niche stuff first?
The Windows OpenCL version (for AMD and NVidia) was already tested successfully. And btw, the "niche" is rather AMD, see the sever status page. E@H has 2-3x as many hosts with NVidia cards than with AMD. And I think having more than one card per host is more common with NVidia than AMD, so the card/GPU count will be even more in favor for NVidia.
not only 2-3x more hosts with Nvidia, but also 2-3x daily valid results coming from Nvidia vs AMD.
AMD is the minority. and as Bernd mentioned, the AMD app already proved viable so no further testing necessary. there were some bugs that needed to be worked out to get the CUDA app working properly, and it only makes sense for them to prioritize that since Nvidia makes up most of the pool of compute power.
looking forward to the next version of apps and hope for better validation across platforms (Nvidia vs AMD, and Windows vs Linux).
Richard Haselgrove
)
I have two like that in the opposite, where Windows machines took the win over Linux apps.
Workunit 661203547
Workunit 661202704
_________________________________________________________________________
It seems we finally got the
)
It seems we finally got the CUDA version working. Phew! Comparability of the results (what you see as validation) is still bad, but it seems we got that working, too, behind the scenes. Next version of apps should be better, hopefully early next week.
BM
Got 3 BRP7 tasks on my
)
Got 3 BRP7 tasks on my WIndows 10 PC with GTX 1060 and they take only a limited time, a few dozen seconds.They validate quickly.
Tullio
Twelve validated : no problem
)
On My Dream Machine :
Twelve validated : no problem with MeerKAT v0.03 () x86_64-pc-linux-gnu here - on a 3060 - in each case when paired with another single x86_64-pc-linux-gnu also running cuda55 ... ;-))
Fourteen invalid : in each case within quorums with two windows_x86_64 !! ;-((
So from my rig it looks to be linux vs windows ?!
On the upside they're awfully quick though @ 17 seconds.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
These are still test tasks,
)
These are still test tasks, basically trying whether 1. the workunit generator and 2. the app works at all. These should have 1200 templates only, the next, more realistic set will have 48000-50000 templates.
BM
[sulks] I want AMD on
)
[sulks] I want AMD on Windows. Why did you have to program the niche stuff first?
If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.
Bernd Machenschalk
)
WOO HOO!!!
Unfortunately I will be a part of the next test as I'm out of town on a several year Covid cancelled cruise/land stuff for the next 20 days.
Peter Hucker of the Scottish
)
The Windows OpenCL version (for AMD and NVidia) was already tested successfully. And btw, the "niche" is rather AMD, see the sever status page. E@H has 2-3x as many hosts with NVidia cards than with AMD. And I think having more than one card per host is more common with NVidia than AMD, so the card/GPU count will be even more in favor for NVidia.
BM
Bernd Machenschalk
)
Uh... the "severe status page" ??? I can't find it. Could you direct me to it?
;*)
Proud member of the Old Farts Association
not only 2-3x more hosts with
)
not only 2-3x more hosts with Nvidia, but also 2-3x daily valid results coming from Nvidia vs AMD.
AMD is the minority. and as Bernd mentioned, the AMD app already proved viable so no further testing necessary. there were some bugs that needed to be worked out to get the CUDA app working properly, and it only makes sense for them to prioritize that since Nvidia makes up most of the pool of compute power.
looking forward to the next version of apps and hope for better validation across platforms (Nvidia vs AMD, and Windows vs Linux).
_________________________________________________________________________