Good point... running as a zombie **, then when it reaches the internal exceed time point it self destructs. A hard look at security software logs may be in order, system temps and some memtest86 diagnostics running too.

** The zombie part is dubious... tasks processes have to have a 'keep alive' connection with BOINC client at least 1 time every 30 seconds or they reset (heartbeat loss, zero status resets).

[ot], grats to your reentry into the Societé Club Max... I'm not expecting to get the renewal card until later next week, on my crunching account... hunting for other objectives and taking whatever comes by in the mix[/ot]

Just to add a bit more detail to my last post. With a computing preference set to store 0.7 days work, I received some 80 Zika WUs with an estimated completion time of 19 mins 53 secs each. So multiplied by 40 (from SekeRob* post), that would be approximately 13 hours 20 mins for completion. As I said earlier most WUs complete in about 8 hours which is fine, and at present, for the 4 WUs currently running, the Elapsed time plus Remaining time for those WUs is just less than 8 hours. However I have had 2 WUs which, with somewhere between 10 and 13 hours elapsed time, showed a “Remaining time” of 2 – 3 days, and these are the ones which errored out.

FYI, the laptop is not throttled so it's running at full speed, but occasionally WUs are suspended for a minute or two when I am doing other tasks. However, the laptop runs 24 * 7, so there should be enough time over night, when nothing else is running, for the current Zika WUs to finish.

Anyway when I have completed the 80 WUs and receive some with a more realistic estimated completion time, the problem should go away.

Either that, or perhaps there are some WUs which will never complete due to looping or another problem. and these will still error out, even though the completion time is more realistic?

Regards
Chris

Smells like some on-device research might bring things to light, or, it is a particular compound type where things fall over... to name another possibility. Whilst we know there is a variable number of jobs in a task [a task with tasks in them], to us outside mortals there's no telling what compounds are being run against the [what i presume] is the same ligand for all tasks in a task [why couldn't these subs have been called jobs or some other label instead?]

Yes, that was me. As a temporary solution I stopped the PC from getting GPU tasks from other projects, as that throttles down the processor and that seems to be what is causing the project to result in errors. Also, I second this as my i7-4700MQ machine, when running a GPU project, it also produced 3 error workunits with the time limit exceeded error, so this seems to be related to PCs that throttles the CPU due to thermal issues, like running GPU projects. After disabling the GPU project none of the machines has produced errors.

By the way, I have plenty of Atom machines: 3 Atom Z3735G, 1 Atom X5-Z8500 and yesterday I got a machine with 1 Atom X5-Z8300 which is performing great with this project, taking 13 hours approximately per workunit.