Hi, catalyst 14.8 now gives me also invalid results when running two wus with my 290 ...
13.12 and 14.4 give valid results for two wus but the runtime is worse than running just a single wu and a newer driver.
i understand where you're coming from. you have something that is a good value based on theoretical maximum performance and efficiency. on this project any amd card above a tahiti class isn't going work well with multiple wu per gpu. i don't know when i'll get around to trying linux again. i can't be sure if the problem is with the hardware and/or the drivers. if you just got the card maybe you can return it, but some nvidia cards cost real money
The same problem has been happening with Linux m/c's using the Hawaii/Tonga cards. Drivers don't seem to be the problem. It may be something in the structure of the app, which the developers will have to sort out...
FYI the problem happening in SETI to, so is not E@H related it´s something diferent.
For all i was ablle to dig it´s related to the driver itself, but the older drivers seems to be slower so the only real option for now it´s running 1 wu at a time.
What i did not try is what happening when you start 2 compleately separate boincs on the same host (not just run 2 wu at a time), may that works but i´m unable to test that option remotely maybe you could try and tell us what happeing.
Hi, catalyst 14.8 now gives
)
Hi, catalyst 14.8 now gives me also invalid results when running two wus with my 290 ...
13.12 and 14.4 give valid results for two wus but the runtime is worse than running just a single wu and a newer driver.
Try clenup gpu driver with
)
Try clenup gpu driver with DDU and install Catalyst 13.12. It is very fast and stable driver to use with E@H.
Sounds like newer driver and
)
Sounds like newer driver and one wu per gpu is the way to go for now
Another option might be
)
Another option might be linux?
but for 1 wu average gpu load
)
but for 1 wu average gpu load is 67% only. i spent so much money not for 67% load =)
i understand where you're
)
i understand where you're coming from. you have something that is a good value based on theoretical maximum performance and efficiency. on this project any amd card above a tahiti class isn't going work well with multiple wu per gpu. i don't know when i'll get around to trying linux again. i can't be sure if the problem is with the hardware and/or the drivers. if you just got the card maybe you can return it, but some nvidia cards cost real money
RE: Another option might be
)
The same problem has been happening with Linux m/c's using the Hawaii/Tonga cards. Drivers don't seem to be the problem. It may be something in the structure of the app, which the developers will have to sort out...
Gord
then linux isn't the answer
)
then linux isn't the answer either.
for now einstein is a backup project for me and i just run one wu per gpu on 15.7.1/15.9.1 drivers.
some might wish that the app would more load the gpu more but then heat could become an issue
how could we report to the
)
how could we report to the developers about this problem?
FYI the problem happening in
)
FYI the problem happening in SETI to, so is not E@H related it´s something diferent.
For all i was ablle to dig it´s related to the driver itself, but the older drivers seems to be slower so the only real option for now it´s running 1 wu at a time.
What i did not try is what happening when you start 2 compleately separate boincs on the same host (not just run 2 wu at a time), may that works but i´m unable to test that option remotely maybe you could try and tell us what happeing.