Toni, it seems to work well - both for Windows 10 and Windows XP :-)))

Many thanks for the efforts put in by everybody at GPUGRID !

...and this thread is back on topic ;)

Got one on a Windows box:

7/27/2018 6:35:48 AM | GPUGRID | Aborting task e37s19_e36s5p0f20-ADRIA_FOLDT1019_v2_predicted_pred_ss_contacts_50_T1019s1_4-0-1-RND2194_2: exceeded elapsed time limit 7288.00 (250000000.00G/34302.98G)

What did you guys have to do to fix the application?

Toni, it seems to work well - both for Windows 10 and Windows XP :-)))
Many thanks for the efforts put in by everybody at GPUGRID !

Iam afraid I was to early with my above statement :-(

The task on the Windows 10 machine broke off after 8.963 seconds with:

197 (0xc5) EXIT_TIME_LIMIT_EXCEEDED
exceeded elapsed time limit 8947.10 (250000000.00G/27942.01G)

for more details, see here: http://gpugrid.net/result.php?resultid=18262430

The same here.

http://www.gpugrid.net/result.php?resultid=18261977

The same here http://www.gpugrid.net/result.php?resultid=18262508

<core_client_version>7.6.33</core_client_version>
<![CDATA[
<message>
exceeded elapsed time limit 9498.88 (250000000.00G/26318.88G)
</message>
<stderr_txt>

other task http://www.gpugrid.net/result.php?resultid=18262784

I think something (either failures, or likely recent short tasks) made some machine over-optimistic about its own fp-ops. As a consequence, BOINC estimated that tasks could be run in a few hours, which is untrue.

Try re-running the benchmarks. https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4273

I think something (either failures, or likely recent short tasks) made some machine over-optimistic about its own fp-ops. As a consequence, BOINC estimated that tasks could be run in a few hours, which is untrue.

Try re-running the benchmarks. https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4273

Can't be true, my machine have only run GpuGrid long WU's and have the same problem.

Anyway GpuGrid was my last BOINC project and I have decided to hang up my BOINC boots.
The satisfaction of contributing has just left me.

Good Luck to all.

---

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

Try re-running the benchmarks. https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4273

I guess I gathered what you mean, but this machine has not run any BOINC tasks in the meantime. So there should not be any (too short) runtime values somewhere deep in BOINC.

Anyway, followed this advise (shown in your link) with a newly downloaded GPUGRID task:
You can help yourself out of this situation by increasing <rsc_fpops_bound> of Sixtrack tasks 1000 times larger or possible even more
However, I increased the value by the factor 10, guess this should be sufficient.

So, I'll see what happens

Yeah! Two WUs downloaded. I love the sound of my GPU fans spinning up.

Anyway GpuGrid was my last BOINC project and I have decided to hang up my BOINC boots.The satisfaction of contributing has just left me.

It's a pity...

Have you also tried selecting the "re-run benchmarks" (or something) menu option?

[
I guess I gathered what you mean, but this machine has not run any BOINC tasks in the meantime. So there should not be any (too short) runtime values somewhere deep in BOINC.

You are right. Don't know what so say except that it's frustrating on this side too. There is an excess of hidden state and undocumented checks. My hope is that it will resolve by itself at some point (maybe resetting the project).

Do you require help from some of the BOINC devs? They're pretty responsive on the BOINC Projects email group, and if there is some sort of transparency problem, they'd want to hear about it.

Do you require help from some of the BOINC devs? They're pretty responsive on the BOINC Projects email group, and if there is some sort of transparency problem, they'd want to hear about it.

I've been sitting in the same conference room as about 25 BOINC developers for the last three days. If someone had called, we could have answered...

But today was the group walk in the Oxfordshire countryside, and we meet in an hour for our final group meal before they get their 5 am flights home. I have a simple 200 mile drive home before I'm reunited with my GPUs - I'll look at it Sunday, report Monday,

Running benchmarks won't solve it. because they measure the CPU speed, and this is a GPU app. But you're on the right lines - the initial speed estimate will be low, and the quickest workround will be to increase the

<rsc_fpops_bound>

for the new v9.22 app by a factor of at least 10 and perhaps 100. You may have to generate new workunits with the uprated bound. Runtime estimates will almost certainly appear to users as vastly inflated in the initial stages, but hang in there - they will become 'accurate' (-ish) after the first 11 completed tasks.

More when I can eyeball it.

Thinking about it in the shower, that's the wrong way round - apps faster than expected shouldn't cause a problem.

Jacob, while I'm drinking/eating/drinking/sleeping/travelling/sleeping, can you pull the guts out of the <app_version> for 9.22 and a matching WU&task - just the BOINC metadata, not the file references - and post them for me to look at before I get home. Even better if you could subsequently run it and point me to the outcome online. I'm wondering if the project might have slipped half-a-dozen orders of magnitude in <rsc_fpops_est>.

Now I've got a bus to catch.

No, I just started a 4 day vacation, sorry.

The interesting thing is that this problem does NOT come up in the cuda65 app (for Windows XP), but only in the cuda80 app (for Windows10).

I've downloaded 2 Long run tasks on my Windows 10 PC and one is running. It seems to run OK but,according to the Task manager, it seems to use both the CPU and GPU (GTX 1050 Ti), very scarcely compared to SETI@home GPU tasks on the same host.
Tullio