On one of my computers (ID: 8059007 KEYLIME-PIE Windows 10 PRO x64) I installed a nVidia GTX 750 a few days ago. Since this computer has an AMD A8-5500 APU it also has a Radeon HD 7600 GPU integrated into the CPU. I have both enabled. I installed the nVidia 391.35 drivers. I used the lunatics_win64_v0.45_beta6-for-sog_setup.exe to install the optimized binaries.

Almost immediately I was getting errors in the BOINC Event Log of:
4/12/2018 12:50:54 AM | SETI@home | task postponed 180.000000 sec: Cuda runtime, memory related failure, threadsafe temporary Exit

The Windows System Event Log also has a bunch of errors starting with:
Event ID 13 from Source nvlddmkm saying:
\Device\Video6
Graphics SM Warp Exception on (GPC 0, TPC 0): Misaligned Address

I've already increased the TdrDelay to eliminate the problems that was causing with the Radeon.


Can someone tell me how to fix this?

Should I delete the app_info.xml and try the default apps?

Does this indicate some hardware error? I got the nVidia GTX 750 used on ebay and shipper used no padding so the board could be damaged?

Should I try disabling the onboard Radeon HD 7600?

---

Just rerun the Lunatics installer and choose the SOG app for nvidia also.
The CPU and AMD GPU seems to run fine.

---

With each crime and every kindness we birth our future.

OK. I installed the SOG app.... It seems to have got rid of the
task postponed 180.000000 sec: Cuda runtime, memory related failure, threadsafe temporary Exit

error but now I'm getting alot of: task postponed 30.000000 sec errors. I think every SETI task has had several of these.

4/13/2018 12:54:05 AM | SETI@home | task postponed 30.000000 sec:
4/13/2018 12:57:11 AM | SETI@home | task postponed 30.000000 sec:
4/13/2018 1:00:57 AM | SETI@home | task postponed 30.000000 sec:
4/13/2018 1:02:17 AM | SETI@home | task postponed 30.000000 sec:
4/13/2018 1:04:09 AM | | Suspending computation - CPU is busy
4/13/2018 1:04:29 AM | | Resuming computation
4/13/2018 1:04:41 AM | SETI@home | task postponed 30.000000 sec:
4/13/2018 1:05:53 AM | SETI@home | task postponed 30.000000 sec:
4/13/2018 1:06:41 AM | SETI@home | task postponed 30.000000 sec:
4/13/2018 1:08:12 AM | SETI@home | task postponed 30.000000 sec:

Also I've notice the MB8_win_x86_SSE3_OpenCL_NV_SoG_r3557.exe is using about 1 core of CPU all the time while it's using the nVidia. A strange thing, I used the x64 installer but it installed the x86 app.

Is this the way it is suppose to work?

Thank you.

---

4/13/2018 1:04:09 AM | | Suspending computation - CPU is busy
4/13/2018 1:04:29 AM | | Resuming computation

If you set BOINC to stop crunching when the CPU is doing other things, then that is what it will do.
In your account, Computing Preferences
Make sure these are not selected.
Suspend when computer is in use
Suspend GPU computing when computer is in use

Leave this one blank, or set it really high (say 85% or higher).
Suspend when non-BOINC CPU usage is above --- %

Save the changes, Press Update in the BOINC Manager on the computer, and see if it then stops suspending processing.

Also I've notice the MB8_win_x86_SSE3_OpenCL_NV_SoG_r3557.exe is using about 1 core of CPU all the time while it's using the nVidia. A strange thing, I used the x64 installer but it installed the x86 app.

Is this the way it is suppose to work?

Yep.
It's result of how NVidia supports Open CL, 1 CPU is required to each GPU WU being crunched.

---

Grant
Darwin NT