Folding Forum

A small update. The rig is working well, been folding 100% 24/7 since purchase, it has been really stable with new amd-drivers. Two weeks of uptime, no reboots and averaging 380kppd lately.

I totally jinxed with that post yesterday, had to reboot this morning.

PPDs have been down lately, is it only me that find some WU troublesome?

With no additional information, we can't say anything ...

toTOW wrote:With no additional information, we can't say anything ...

Nothing much to say really. A couple of months ago most WU were resulting in 150k every 3x3h period. No its mostly 30k each 3h or locking up. Will post log next time it stops at 99.99% for a day...

JonasTheMovie wrote:Will post log next time it stops at 99.99% for a day...

This part indicates that you actually have an issue ...

I'm sure you'll find that according to the log, the WU that apparently got to 99.99% really quit working some time before that. As toTOW said, we need to diagnose WHY it quits working.

Are you overclocking?

The 99,99% has happened three times the latest month from what I can see. Will makes sure to post logs next time. But the "big issue" that is significant here is the 25%-30% drop in ppd. Without any errors. The Fury X is really bad at overclocking so I have stayed away from that.
Only thing that has changed really are driver updates from AMD, which makes me really interested in dumping Team Red when it comes to folding. All those tflops, but not suitable for folding...
To test I have tonight reverted back to 16.1 drivers. One could wish that AMD spent some efforts on fah, but, no...

Got myself a faster a10-7870K processor last weekend, tested some folding on the iGPU R7 and Fury X simultainously but it was not working very well at all, think fah mixed the slots up, with the R7 showing 280kppd and the FuryX 20kppd.
But the point of a new CPU is to be ready to go Green, around crhistmas, to what I believe is going to be an EVGA GTX 1070 hybrid cooler card, passing the Fury X off to someone who will game on it instead.

FAHClient 7.4.4 does sometimes mix up the GPUs. I understand that's supposed to be fixed in the next version of FAHClient.

Ordered a EVGA GTX 1080 Hybrid 6288 (FTW PCB), tired of the poor PPD performance of the Fury X.
If I would of known this a year ago I would of built a dual slot tower, but hey...

JonasTheMovie wrote:Ordered a EVGA GTX 1080 Hybrid 6288 (FTW PCB), tired of the poor PPD performance of the Fury X.
If I would of known this a year ago I would of built a dual slot tower, but hey...

You'll probably find that power consumption is less as well

Hopefully some cooler summer days...

Im a bit curious to, after a while, change back to the A4 processor (from A10) to compare the performance impact with a truly high end GPU like the GTX1080.

So, power draw from the wall is the same as for the Fury X, but the ppds are lifted from 300kppd to around 750kppd, really nice increase there.
The standard fan on the hybrid goes full trottle and is noisy, so swapped it for a 1100rpm model, temps went from 48C to 65-68C. (AMD 64C with the same case config).
Have removed the obstructive front plate to improve air flowto the 200mm front fan, which with the 1100rpm fan on the 1080 radiator lowers temps 10C.

When it comes to the topic, low end CPU with high end GPU, with the 1080 the A10 is utilized 30% by fah_core, more than twice what the AMD needed, if I remember correctly. Not quite lineary to the actual production (not ppd), so I guess the saying that Nvidia takes more CPU has something to it. I don´t think the A4 would of had much time over with the 1080, good thing I upgraded to the A10.

Trying to facilitate the A10-7870K iGPU again, the R7. It shows correctly in fah now, no mix of slots as with the FuryX, the GTX is unaffected. But.....

Before you ask me for logs.... there is no logs... that´s the problem.
The R7 slot chews through percentage faster than it should and nothing shows in the logs.
An estimated ppd for the slot of 900kppd i bonkers, so, what is really happening here?
When reaching ca 13% it finally logged 1% and the est. ppd normalized to 21kppd.

The percentage counter tries to start incrementing straight away, based on previous performance on the same project.
In this case, it's probably assuming 1080 performance until the core reports back its actual progress, then the counter and ppd reflect that.