As my computers are finishing WU's new units are NOT being sent.
Is there a problem as the server status suggests over 200 long WU's are ready to send.

Same here. It appears to have happened due to the development of the new ACEMD3 app.

My Win machines (Pascal GPU) with 3xx drivers can download ok, my Win Machines (Pascal GPU) with 4xx drivers cannot.

I have downgraded one Win machine from 4xx driver to 3xx driver, and it can now download tasks.

hopefully only a temporary glitch.

App 2.04 should support cuda10, if the scheduler collaborates.
Expect hiccups...

T

Ouch. It must be because of the conditions for offering cuda100 vs cuda80. I promote "new" hosts to cuda100 but there is no such app for acemd2 (there is for acemd3).

Will try to find a solution.

t

Ouch. It must be because of the conditions for offering cuda100 vs cuda80. I promote "new" hosts to cuda100 but there is no such app for acemd2 (there is for acemd3).

Will try to find a solution.

t

That's easy: create a CUDA10 Windows app. :)

Please check if solved

Ouch. It must be because of the conditions for offering cuda100 vs cuda80. I promote "new" hosts to cuda100 but there is no such app for acemd2 (there is for acemd3).

Will try to find a solution.

t

That's easy: create a CUDA10 Windows app. :)

We'll probably do it, but for acemd3. In the meantime, acemd2 should not stop working.

I see 844 tasks on server-status but nothing for the GPUs my computers on Windows 10 !?

Please check if solved

Yes, new tasks ready to crunch !

Please check if solved

Yes, new tasks ready to crunch !

My CUDA10/Windows 10 host received 2 CUDA8.0 tasks at 10:08 CET (08:08 UTC)

:)

The "legacy" apps will stay cuda 80 only.

All hosts now with tasks.
Thanks everyone :)

It looks like cuda 6.5 is no longer available for Kepler cards, with Windows 7 and 3xx.xx drivers.


http://www.gpugrid.net/results.php?hostid=494023


Is this by accident or intent?

Might be fixed. Please check.

Hi guys, I have Linux system with Nvidia Pascal graphics card, Driver 418.56
Can't get any new tasks when was trying to connect this week.

---

Update:
Apparently in Ubuntu 18.04 you have to install CUDA separately in addition to drivers.

Was able to install nvcc from default repository and after restart started to get tasks: "New version of ACEMD 2.04 (cuda100)".

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2017 NVIDIA Corporation
Built on Fri_Nov__3_21:07:56_CDT_2017
Cuda compilation tools, release 9.1, V9.1.85

Hope to get some long runs too

---
Update2

It didn't last for long, got only 6 tasks and silence again, it's really confusing to figure out if something wrong on user side or the project is just being picky :)

It's not you -- the tasks you received are 'test tasks' testing a new version of the Linux application. The regular/long work units right now are only for Windows hosts until the new application is deployed. (The new app is for Linux hosts at the moment.)

http://www.gpugrid.net/forum_thread.php?id=4935

As my computers are finishing WU's new units are NOT being sent.
Is there a problem as the server status suggests over 200 long WU's are ready to send.

ok, thanks for clarification !

Might be fixed. Please check.

It's not fixed.


http://www.gpugrid.net/result.php?resultid=20990388

Might be fixed. Please check.

It's not fixed.


http://www.gpugrid.net/result.php?resultid=20990388

Made yet another attempt. Is there a reason for not upgrading the driver? Supporting old versions is really a problem.