folding client count

[porina]

If I look at my stats, it says I have 6 active clients. Thing is, I'm currently running two systems. Each system is using both CPU and GPU, but even if I count those separately that's only 4.

 

Oh, I just had a crash on a 750Ti. Does that imply it is unstable? Or does it happen from time to time anyway?

[TheRandomness]

I can't answer the first question but for the second, yes, yes it's unstable.

[porina]

2 minutes ago, TheRandomness said:

I can't answer the first question but for the second, yes, yes it's unstable.

Damn those gaming factory overclocks. Don't think I ever had a 100% stable GPU out of the box for crunching. Will see if I have time to do an underclock later but just rebooted and let it get on again, and told it not to get new work until I fix it.

[petem]

7 hours ago, porina said:

If I look at my stats, it says I have 6 active clients. Thing is, I'm currently running two systems. Each system is using both CPU and GPU, but even if I count those separately that's only 4.

 

Oh, I just had a crash on a 750Ti. Does that imply it is unstable? Or does it happen from time to time anyway?

Active clients - Best guess: If your folding config has changed during the listed time frame, the identifying information for your GPU (slot id, card type, etc.) might have changed and been considered a different "active client", and been counted as such.  AFAIK cpus are always assigned ID 0 (unless you change it), so the CPUs are probably not the source of the confusion.

 

As for the 750 ti crashing, I have one factory overclocked to 1202MHz (about 18% over reference), and have not had so much as a hiccup out of it... it's a power-sipping work horse - well maybe a "quarter" horse (as in 1/4 the size of a regular horse like the GTX 970).  Been folding on it 24/7 for about a year.

 

- Pete

[porina]

10 minutes ago, petem said:

As for the 750 ti crashing, I have one factory overclocked to 1202MHz (about 18% over reference), and have not had so much as a hiccup out of it... it's a power-sipping work horse - well maybe a "quarter" horse (as in 1/4 the size of a regular horse like the GTX 970).  Been folding on it 24/7 for about a year.

One I have is only running 1150 under load. It is currently on a big unit estimating finishing sometime tomorrow so I might as well let the system keep running.

[petem]

1 minute ago, porina said:

One I have is only running 1150 under load. It is currently on a big unit estimating finishing sometime tomorrow so I might as well let the system keep running.

I think of a crash as bringing down the computer, or at least unceremoniously dumping the application to the desktop.  I've never had that happen from folding on either of my folding rigs. The one with the GTX 750 ti is running windows 7 (it also has a GTX 750), and the other is a linux ubuntu 14.xx LTS system.  They've both been very reliable.

 

It's  normal to have an occasional "bad work unit", "can't reach server", etc. issue - but those are typically folding app/server hiccups, not card hiccups (as indicated by many posts from people regarding problems with the same folding projects).  I've also had the dreaded "slowdown bug", back in November (maybe more than 1, but it's the only one I caught).  A GPU on my linux rig was chugging along at a TPF (time per frame - 1% of a full work unit) of 2 mins 07 secs, and then at 38% complete it dropped precipitously to a TPF of almost 2 hours ><.  The powers that be knew about the bug, and said it was a rare problem with some Core 21 work units.  Mine was Project: 9704 (Run 42, Clone 7, Gen 105), and were working on a fix for it.

 

Hope that info helps. 

 

- Pete

[porina]

3 minutes ago, petem said:

It's  normal to have an occasional "bad work unit", "can't reach server", etc. issue - but those are typically folding app/server hiccups, not card hiccups (as indicated by many posts from people regarding problems with the same folding projects).  I've also had the dreaded "slowdown bug", back in November (maybe more than 1, but it's the only one I caught).  A GPU on my linux rig was chugging along at a TPF (time per frame - 1% of a full work unit) of 2 mins 07 secs, and then at 38% complete it dropped precipitously to a TPF of almost 2 hours ><.  The powers that be knew about the bug, and said it was a rare problem with some Core 21 work units.  Mine was Project: 9704 (Run 42, Clone 7, Gen 105), and were working on a fix for it.

To me a crash is anything that is any state that is not running as expected. The task was no longer folding and there's an error on the screen. To me, that is a crash. The cause is another matter. Is it software, or hardware? That's what I don't understand. Absent other info, I simply restarted the PC and let it go on with the unit.

 

It is a core 21 unit, currently showing 17m21s TPF. Is that bad? Nearly 29% in, ETA 20.5h.

[petem]

12 minutes ago, porina said:

To me a crash is anything that is any state that is not running as expected. The task was no longer folding and there's an error on the screen. To me, that is a crash. The cause is another matter. Is it software, or hardware? That's what I don't understand. Absent other info, I simply restarted the PC and let it go on with the unit.

 

It is a core 21 unit, currently showing 17m21s TPF. Is that bad? Nearly 29% in, ETA 20.5h.

I don't have ready access to my 750 ti for TPF comparison, but it really depends on the project.

 

If you have the project # and the TPF, go to F@H ppd calculator and plug that info into the page to determine your approximate ppd.  I have found it to be very close to what my F@H logs report back from the mother ship.

 

Mine typically runs about 65k-75k ppd.

[porina]

3 minutes ago, petem said:

If you have the project # and the TPF, go to F@H ppd calculator and plug that info into the page to determine your approximate ppd.  I have found it to be very close to what my F@H logs report back from the mother ship.

Great resource. The calc gave within 1% of the estimated PPD shown in the client, which is 71k PPD. Guess it is in the right ball park and will leave it running as is.

[petem]

7 hours ago, porina said:

Great resource. The calc gave within 1% of the estimated PPD shown in the client, which is 71k PPD. Guess it is in the right ball park and will leave it running as is.

Yep, sounds like you're hitting on all cylinders ; )