Discussion Thread for the Continuous GW Search known as O2MD1 (now O2MDF - GPUs only)

I was referring to the the previous gravitational wave ones which worked fine (1.0.7 and 1.0.8 if I remember correct), got new gravitational wave tasks yesterday which then all had an error at the end. So I just uncheck the GW waves for this host and only allow Gamma-Ray Pulsar Search tasks?

The 'validate errors' from the new ("O2MD1G2") tasks should now be back to normal (i.e. mostly switched to 'valid'). I can only see 20 'validate error' tasks, which are for good reasons.

How wide is that avoided area? This host didn't get new gpu tasks in a while (there was 'resend' thing going on, which dried out the queue). The host got that single v2.00 cpu task that it was missing and I aborted it. Then the host got fresh bunch of gpu tasks, containing 502.30 - 502.50 Hz stuff.

Some of them seem to run ten times slower than their sisters from the same bunch. I think that's not how it should be, so I'm aborting those slow tasks after looking at their progress long enough to identify them one by one.

The host that had problems with 505 Hz tasks earlier, got three more 505 Hz tasks after I let it download fresh tasks. They errored out just like earlier and I forced NNT. I thought maybe the system had some kind of remnants of the work files which might had caused that.

So I hit 'Reset project'.... and after that, host received 510 Hz stuff . These tasks run fine. I was just wondering if 'Reset project' was actually needed for the 'frequency avoidance' to come into action. Should this other host have received those 502 Hz tasks at all after that change had been already applied ?

edit: "502.30 - 502.50 Hz stuff" ...

I noticed that running them only 1x won't slow  them down. I left them running 1x overnight and they are still going strong. That slowdown problem with 2x is propably not anything generic but related to this specific hardware and those 502 Hz tasks. I will let the remaining tasks finish. Then I'll click "Reset project" and see what frequency new tasks will have and if 2x will work again.

edit #2: I did a "Reset project". The host received then 523 Hz tasks. They run fine with 1x but when I tried 2x the card was again suffocating. The first task that was already running half way when a parallel task was allowed was able to run to end with full speed, though. But that other task was extremely slow from the beginning. And then all tasks after that would be also extremely slow. already right from the beginning.

This is a host-specific thing only and is occurring now... I don't know why. But I'll consider this a mystery and leave this as is. 2x was working on this host before but not anymore. Run times with 1x are alright, so I'll just let it be. For the record, this host has R9 390 + Windows 10. Another host with R9 390 + Windows 7 has been running these tasks happily 2x all the time. Same GPU driver version on both.

Just to confirm that. Host 12768123 received eight 504.85Hz tasks from that O2AS v1.09 search today. All those tasks crashed in 30 secs, this time with Nvidia card. Then again two tasks with 582Hz content run fine and validated succesfully.

Hi! The difference is that your host tried to run the GPU version (2.02) of that application. The other hosts run CPU version (2.00). GPU is supposed to complete tasks much faster than CPU. I quess the run time limit for CPU tasks might be much longer because of that.

Your host has Nvidia GT650M which in theory should have roughly about 25% of the computational power of a  GTX 960 for example. If that would reflect directly on the run time of these tasks, I believe GT650M should be able to complete a task in about max 2 - 3 hours. In that perspective, your GPU was running those tasks already overtime when the progress was at 66 %.

But I believe GTX650M is simply incompatible with this current GW app. That's why the progress is slow or it is slowing down exponentially the further it gets. And that leads to those tasks not being able to complete at all.