Perhaps people are already aware of this but I have noticed tasks seem to be prone to error when first loading into gpu memory. Unless the gpu memory is mostly empty when starting there is a very high chance for error, some tasks worse than others. NATHAN seems to be the worst as i can tell. Take a look at this random computer I came across for example http://www.gpugrid.net/results.php?hostid=43714 (not mine lol!). Is there nothing that can be done? Sorry if the topic has already been covered..

Interesting observation. Perhaps you are on to something and Nate's tasks could do with clearing the memory before starting?

Generally when tasks fail quickly, it's less of a dissapointment than if they fail mid-run; all being well you start a queued task or download another and in a few minutes are up and crunching again. A mid run failure on the other hand typically means the loss of many hours of GPU time. So, if it's going to crash, it's better to crash early on :))

---

FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Fortunately they do seem to run pretty solid once they get going however for me it can be a problem midway as well and I have lost a couple this way. Sometimes I suspend the GPu to play a game and when I resume the memory is not clear and it errors. Luckily I have found a way to recover and it looked like someone else may have also figured this out in another post (but I do get credit for figuring this out idependently ;):

Before closing the windows dialog box that informs you of the error first exit boinc, or just simply suspend the gPU. Then reopen boinc or re-enable GPU use after closing the dialog and it will attempt to restart instead of aborting the task. I have rescued quite a few this way. Sometimes it takes a couple tries to clear the memory but it is quick and you may need to close some open programs including simple things like a picture.

---