F@h & Oc

[Ultra Nexus]

Hey there!

I was wondering... I almost fold gromacs units but my question is: Can WUs get return completely OK (no restarts or errors in fahlog) but be corrupted cause of excesive OC? Or can I rest assured that if no problems are found during the entire WU folding, the result will be consistent as well?

I ask this cause I`m always trying to squeeze the last Mhz from my proc but I dont wanna return garbage results that wont count anyway.

Any feedback on this?

Thanks!

[edcentric]

I have had one of two out of 1,000 that looked like the finished OK but were rejected when I tried to send results. My guess is that the erroes were in the log files.
It is the risk that we take.

[muddocktor]

Stanford has put in some rigorous checks into the client to keep from having bad data be accepted by them. If you push it too hard you will lose the work and the credit but it won't affect their results as it will be filtered out. You will find that with Gromacs, your maximum stable overclock will likely be lower than what even Prime95 will tell you.

[mmonnin]

Yeah gromacs will find most instabilities in the CPU and RAM. Prime does not test SSE and other extensions like gromacs do.

[Ultra Nexus]

muddocktor said
Stanford has put in some rigorous checks into the client to keep from having bad data be accepted by them. If you push it too hard you will lose the work and the credit but it won't affect their results as it will be filtered out. You will find that with Gromacs, your maximum stable overclock will likely be lower than what even Prime95 will tell you.

You know I thought the same... that no software could stress more than folding a gromac, but I found that the latest version of Prime95 (ver22.12) its even MORE stressing... I need to up my vcore up by 0.05v to do hours of stable prime95 while before I could fold on and on without any problem.

[Ultra Nexus]

mmonnin said
Yeah gromacs will find most instabilities in the CPU and RAM. Prime does not test SSE and other extensions like gromacs do.

I believe it does now...

[muddocktor]

Ultra Nexus said
muddocktor said
Stanford has put in some rigorous checks into the client to keep from having bad data be accepted by them. If you push it too hard you will lose the work and the credit but it won't affect their results as it will be filtered out. You will find that with Gromacs, your maximum stable overclock will likely be lower than what even Prime95 will tell you.

You know I thought the same... that no software could stress more than folding a gromac, but I found that the latest version of Prime95 (ver22.12) its even MORE stressing... I need to up my vcore up by 0.05v to do hours of stable prime95 while before I could fold on and on without any problem.

Thanks for the update on Prime95, Ultra Nexus.:) I didn't know they had a new version out.