P8101 - R7 C7 G160

[Qinsp]

Stable machine, no errors, running cool, SMP16, no tasks, 2.9ghz.
All the 8101's have been 32:00 TPF so far, the one last night was 31:45 here.

This 7/7/160 is 38:xx TPF.

Is it my equip or the WU?

edit - 100% on all cores, 20% of RAM.

[Qinsp]

Sidebar, it's only at 15% progress, and I have faster equip. Is there a way to move it so I don't burn it? If so, PM me please. I hate not making deadlines.

I could tear it apart and swap CPU's to SMP24x2.6ghz, but that means tearing down 2 machines.

[P5-133XL]

Project 8101, Run 7, Clone 7, Gen 160

No data back from query: So no one else has returned the WU.

I do not believe that you can conveniently move WU's to another machine without losing the bonus points. The servers monitor the install that was issued the WU and refuse to give bonus points if it isn't the same install that returns the WU.

[Nathan_P]

Try a reboot, it has been posted on several forums that sometimes that speeds the WU up - No I don't know why

[Qinsp]

Project 8101, Run 7, Clone 7, Gen 160

No data back from query: So no one else has returned the WU.

I do not believe that you can conveniently move WU's to another machine without losing the bonus points. The servers monitor the install that was issued the WU and refuse to give bonus points if it isn't the same install that returns the WU.

Thanks Nathan, I just rebooted, but it will take a few hours to know.

Now ... Not that I encourage this, as it can be abused, but all 3 of this "family" are the same mobo, same Linux install, same version FAH.
The only difference is CPU's and RAM speed.

I have a hunch I could just swap HDD's and nothing would freak out.

What stunned me, was I was in the middle of a job folding SMP16, shut off the machine, put in 24 cores, and it didn't skip a beat. Neither Linux or FAH. Just speeded up.

[bruce]

Is it my equip or the WU?

It could be either, but my first guess is progress in the WU.

I'm going to assuming that your protein starts out unfolded and that it progresses toward a folded state. (That's not always true.)

Take a snapshot of the protein and add up a grand total of the number of atoms that are "near" each individual atom. Now add up all the forces from those nearby atoms. Since a later WU will generally have more "nearby" atoms than an earlier WU, more significant forces will have to be calculated and added together to get the total force on each individual atom. As the protein becomes more densely packed, processing gets slower because it has to do more calculations.

Changes like that are gradual and relatively small.

What stunned me, was I was in the middle of a job folding SMP16, shut off the machine, put in 24 cores, and it didn't skip a beat. Neither Linux or FAH. Just speeded up.

That's the way it "should" work although it's not recommended. Resuming work with a different number of CPUs can do funny things, like causing the WU to abort. I do that sort of thing, too, sometimes but I recognize the risk and if it fails to resume, there's no point in complaining about it. The recommendation will still be wait for the WU to finish, even though crashes from changing the SMP number are pretty rare, especially with nice even numbers like you used.

[Qinsp]

OK, rebooting fixed it. 32:15 so far, that should still make it.

Thanks All!

Bruce, yeah, I fully expected it to crash when I swapped CPU's. They weren't even the same "family", went from Shanghai to Istanbul? 4Px4C > 4Px6C, right at about 50% completion. I was impressed with both the OS and FAH. And the motherboard (Supermicro).