If changing/modifying a GPU causes a WU to error, is it possible to have the specific WU resent for re-crunching?

Thanks

____________

In a word, No.

A WU is never sent to the same host twice. Just make sure you exit BOINC before changing/modifying graphics card.


____________
Radio Caroline, the world's most famous offshore pirate radio station.
Great music since April 1964. Support Radio Caroline Team -
Radio Caroline

In a word, No.

A WU is never sent to the same host twice. Just make sure you exit BOINC before changing/modifying graphics card.

Hi, GPUGrid Folks:

I have recently brought on line a new PC with an AMD FX-8350 8 core CPU and two ASUS GTX 650 Ti GPUs on board. I have successfully processed three tasks - two long and one short. However, two other short run tasks have failed as shown below. Is anyone else having similar problems? It's very frustrating to have failures so close to the end of the run time.

Thank you,

John



Status Run time
(sec) CPU time
(sec) Credit Application
6990596
4547226
26 Jun 2013 | 9:55:03 UTC 26 Jun 2013 | 15:32:02 UTC Completed and validated 19,860.15 19,429.04 20,550.00 Short runs (2-3 hours on fastest card) v6.52 (cuda42)
6987452
4544848
25 Jun 2013 | 0:38:24 UTC 25 Jun 2013 | 23:21:30 UTC Completed and validated 66,621.91 65,430.47 133,950.00 Long runs (8-12 hours on fastest card) v6.18 (cuda42)
6987227
4544701
25 Jun 2013 | 0:38:24 UTC 25 Jun 2013 | 19:03:43 UTC Completed and validated 65,295.42 64,314.58 133,950.00 Long runs (8-12 hours on fastest card) v6.18 (cuda42)


6991878
4548183
27 Jun 2013 | 1:41:49 UTC 27 Jun 2013 | 8:47:04 UTC Error while computing 25,124.96 24,758.14 --- Short runs (2-3 hours on fastest card) v6.52 (cuda42)
6990599
4547228
26 Jun 2013 | 9:55:03 UTC 26 Jun 2013 | 14:48:37 UTC Error while computing 17,187.14 16,752.82 --- Short runs (2-3 hours on fastest card) v6.52 (cuda42)

Is your CPU heavy loaded?
____________
DSKAG Austria Research Team: http://www.research.dskag.at

Hi John,

- I see you are having quite a few failures, here is my tuppence:

The first glaringly obvious thing would be to upgrade your drivers to the latest WHQL directly from the nvidia - 320.18 were awful. Note personal experience is NOT to uninstall and use driver cleaners etc as many on the web will recommend but let nvidia do their job and just upgrade the drivers directly via their install.

Secondly I would ask are you overclocking? what are your temps? Specs and age of your PSU and any other apps / tasks are running.

If you have not already done so; with BOINC disabled I would suggest you test your systems stability with MemTestG80 (http://folding.stanford.edu/English/DownloadUtils) for at least 12hrs and OCCT (http://www.ocbase.com/index.php/download) for about 10mins (OCCT is particularly punishing hence set the max temp to a safe 85C)

Have noticed GPUGrid to be particularly CPU sensitive - no so much in performance but in availability inorder to feed the GPU hence would recommend you leave one core free for the OS.

As you are running windows 7 I would disable CPU parking and would also recommend you prioritise the acemd2865P.exe process using something like eFMer priority (http://www.efmer.eu/boinc/download.html) or as you have eight cores ProcessLasso (http://bitsum.com/processlasso/).

If the above are successful - as a test I would pause any other apps/tasks and just run GPUGrid tasks to get an idea of throughput, then one by one enable any other BOINC tasks.

Hope this helps, have fun

Are you using the app_config.xml file and if so, what is its contents?

____________

Hi, Gandalf:

I am not using any the app_config.xml file to try and keep things understandable.

Thanks,

John

Hi, GoodFodder:

Thanks for the lengthy response.

1. I am using the 320.18 drivers from NVIDIA. I will try to update these without causing a crash.........I am very leery about touching drivers!

2. My system runs at factory clock speeds - in the past I have experienced extreme instability with overclocking.

3. I will run GPUGrid tasks only as a test: the sharing of the GPUs with Folding at home Alzheimer's tasks may be an issue on the short runs. I did not share on the long runs.

It may take a while, but I will get around to using your suggestions

Thanks, again,

John

Hi John,

- I see you are having quite a few failures, here is my tuppence:

The first glaringly obvious thing would be to upgrade your drivers to the latest WHQL directly from the nvidia - 320.18 were awful. Note personal experience is NOT to uninstall and use driver cleaners etc as many on the web will recommend but let nvidia do their job and just upgrade the drivers directly via their install.

Secondly I would ask are you overclocking? what are your temps? Specs and age of your PSU and any other apps / tasks are running.

If you have not already done so; with BOINC disabled I would suggest you test your systems stability with MemTestG80 (http://folding.stanford.edu/English/DownloadUtils) for at least 12hrs and OCCT (http://www.ocbase.com/index.php/download) for about 10mins (OCCT is particularly punishing hence set the max temp to a safe 85C)

Have noticed GPUGrid to be particularly CPU sensitive - no so much in performance but in availability inorder to feed the GPU hence would recommend you leave one core free for the OS.

As you are running windows 7 I would disable CPU parking and would also recommend you prioritise the acemd2865P.exe process using something like eFMer priority (http://www.efmer.eu/boinc/download.html) or as you have eight cores ProcessLasso (http://bitsum.com/processlasso/).

If the above are successful - as a test I would pause any other apps/tasks and just run GPUGrid tasks to get an idea of throughput, then one by one enable any other BOINC tasks.

Hope this helps, have fun

Hi dskagcommunity:

Thanks for taking the time to respond. Please see my responses below.

John

Can you post an image of your Device Manager?

____________

Which part? the entire thing or a section?

Which part? the entire thing or a section?

I'm actually running the same CPU that you are on a 990FX chipset and had issue's in the beginning. Main one being the BIOS on some of those boards sets the voltages to low for the memory (1.46v), I upped mine to 1.575 and things smoothed out, memory can take 1.6v no problem. Make sure all power saving settings are turned off along with turbo core - all set to disabled. Turn on HPC (High Performance computing), Those CPU's can run really hot!! Use a program called AMD overdrive to get accurate readings from the chipset like temperatures, voltages and memory speed.

http://sites.amd.com/us/game/downloads/amd-overdrive/Pages/overview.aspx

John:

re your comment " the sharing of the GPUs with Folding at home Alzheimer's tasks may be an issue on the short runs. I did not share on the long runs."

Running F@H and GpuGrid on the same machine is asking for trouble as they will be heavily competing for resources. You need to choose one or the other - no guessing which one on this forum people would recommend.

John:

re your comment " the sharing of the GPUs with Folding at home Alzheimer's tasks may be an issue on the short runs. I did not share on the long runs."

Running F@H and GpuGrid on the same machine is asking for trouble as they will be heavily competing for resources. You need to choose one or the other - no guessing which one on this forum people would recommend.

Note Gpugrid and Collatz are both managed by BOINC - F@H is not.

Note Gpugrid and Collatz are both managed by BOINC - F@H is not.

Thanks to all:

My system builder has diagnosed a Win7 problem which he will fix ASAP. In the meantime, I will continue processing F@H Alzheimer's tasks.

John

In the meantime, I will continue processing F@H Alzheimer's tasks.