folding as a stability test.

[mkorpal]

How close is folding for 24 hours to running prime 95 for 24 hours. And will F@H give you a warning as soon as it becomes unstable or will you have to wait for it to finish the wu to find out if it has errors.

[z0n3]

If it has errors it will end the WU.

[TalRW]

use prime95. You don't wanna run folding on an unstable machine because sometimes it will produce bad results but not show up in the logs if it is a minor instability. This can be bad for the project because the results aren't accurate. Also prime95 in the stability testing mode is made for that so I would use that.

[Captain Newbie]

If it has errors it will end the WU.
If it has _detectable_ errors, such as failed sanity checks or physical impossibilities, it will terminate the WU and bunge the client into an idle loop. Simple inaccuracies or instabilities are not guaranteed to result in a failed sanity check, however, and that's why...

FOLDING IS NOT A STABILITY TEST!

Sorry, just had to get that out of my system. Validate your stability with Prime95 _first_.

[mkorpal]

Thanks for the replyies. Sorry that I didn't know that. Anyway, I did test with prime 95 before folding and it was stable. I just wanted to know if it could be used. Although I would still like to know how close folding comes to the stress put on the system by prime 95.

[pscout]

Thanks for the replyies. Sorry that I didn't know that. Anyway, I did test with prime 95 before folding and it was stable. I just wanted to know if it could be used. Although I would still like to know how close folding comes to the stress put on the system by prime 95.

Folding stresses things differently than p95 or orthos. Orthos actually includes stresscpu which is based on the gromacs core so you can select it for a test in addition to p95 if you want.
Also, different cores will change the stress on components. I found that QMD's were the most stressfull wu's but they aren't around anymore ... although a few turned in ones are passed around for testing (I have one).

Folding shouln't be used instead of other stability testers, but since it becomes your defacto final stability test anyway if you are a folder, you need to be careful when setting up for folding on a new rig.

Here is a post I wrote in the benchmarking forum on final folding stablity testing. I am sure there are other approaches that work as well but it describes my approach -> http://www.ocforums.com/showpost.php?p=4631951&postcount=15

[tenchi86]

Though I don't believe it should be used as your only stabilty test, I do believe it can pick up on things others cant. I ran Prime95 before for around 4 or 5 days using all different modes and it passed everything fine. I ran FAH and within 15 hours I got errors.

[Silver]

I have had the same thing happen running a single instance of Prime. I run two large and two small primes all at the same time. After running all four at the same time for an hour I have never had fah fail on anything. If one of the four instances of prime fails then the system is not stable and must be adjusted.