Current Recommended Configuration

[ShaidarHaran]

As per this thread I am back folding for the time being. Recommended settings for CPU + GPU folding? I have a 4.4GHz 5820k with 32GB RAM and a 2.1GHz GTX 1070. Right now my estimated PPD is a meager 390k, which seems low... System info in the client says my GPU is unsupported, guessing that's why... I'll try a driver update and see if that helps. Otherwise, do I need to fun any flags like in the old days? Big WUs, advmethods, what not.

[ShaidarHaran]

Update: updated driver and deleted gpus.txt file and restarted client, PPD seems to be climbing but still sub 600k. It's early yet, I'll check back later tonight or maybe tomorrow morning before I go to work. Definitely want to hear if there's any flags for bonus PPD for using more RAM or any other flags I should be running.

[jima13]

Welcome back 'techguy'

I just went with the recommended install since I'd been out of touch for awhile. I think there's still a beta test still available, but you might want to check the forum. I was running the 368 drivers when I installed the 1st 1070 so no problem; 2nd one I had to update the drivers before it started to ramp up.

[Turbo_Rage3d_]

Welcome Back!!!

I've tried a couple of those flags when I first started back, but ultimately I set everything back at default and things seem to be doing just fine.

-Turbo

[ShaidarHaran]

Thanks for the insights guys. Have either of you checked the advanced client control system info tab under the GPU section and seen the same line "UNSUPPORTED: NV3 [PCI]"? I feel like this card should be closer to 700k PPD with my OC under water and it seems like 500-600k is all its putting out. I mean, I'm folding @ 2126MHz @ 34C

[jima13]

Running 2 980ti's on this system and don't see that. I'll check the others when I get to the shop.

[absolutefunk]

Thanks for the insights guys. Have either of you checked the advanced client control system info tab under the GPU section and seen the same line "UNSUPPORTED: NV3 [PCI]"? I feel like this card should be closer to 700k PPD with my OC under water and it seems like 500-600k is all its putting out. I mean, I'm folding @ 2126MHz @ 34C

I've noticed that when my Nano is folding, I need more than one CPU core available for it (nVidia GPUs that I've dealt with normally are fine with one CPU core, but the Nano needs about 40% CPU resources for max output). Try disabling CPU folding for a couple days and see how your GPU output changes, if at all.

-Brian

[ShaidarHaran]

I've noticed that when my Nano is folding, I need more than one CPU core available for it (nVidia GPUs that I've dealt with normally are fine with one CPU core, but the Nano needs about 40% CPU resources for max output). Try disabling CPU folding for a couple days and see how your GPU output changes, if at all.

-Brian

Thanks for the tip, CPU folding actually wasn't running because the project it picked up is disabled with > 6 threads. Jima actually helped me get the CPU client up and running with that tip. I'll keep an eye on production for now.

I've got a couple servers just sitting here, each with dual six-core X5670 Xeons (Westmere generation). Worth running?

[Turbo_Rage3d_]

Thanks for the insights guys. Have either of you checked the advanced client control system info tab under the GPU section and seen the same line "UNSUPPORTED: NV3 [PCI]"? I feel like this card should be closer to 700k PPD with my OC under water and it seems like 500-600k is all its putting out. I mean, I'm folding @ 2126MHz @ 34C

No I haven't checked, I will when I get home though.

[ShaidarHaran]

One more thing to add, I had a GT 210 in this system prior to the 1070. This machine was originally only intended for media server duties but at one point I decided to make it into a dual-purpose machine for Flight Sim so I added the 1070. I imagine it's possible this could be causing a conflict. The GT 210 is not currently installed in the machine and when I changed cards I did a clean install of the driver.

[absolutefunk]

I've got a couple servers just sitting here, each with dual six-core X5670 Xeons (Westmere generation). Worth running?

Yes

[ShaidarHaran]

Yes

Sure, 48 threads but how "worth it" is it? (PPD)

[absolutefunk]

Sure, 48 threads but how "worth it" is it? (PPD)

Should be around 120k PPD or so cumulative, based on the Passmark score of those CPUs, which tends to be rather consistent. Energy usage may be a bit iffy though, but for pure PPD, not bad for CPU folding

[Turbo_Rage3d_]

Should be around 120k PPD or so cumulative, based on the Passmark score of those CPUs, which tends to be rather consistent. Energy usage may be a bit iffy though, but for pure PPD, not bad for CPU folding

Sounds fairly worth it to me!

[jima13]

Well, just ran around and checked 'system info' on the 3 1070's and they all show>
GPU:2
GPU 0 Unsupported
GPU 1 Nvidia 1070

one of them had gpu 0 & 1 reversed but that's all I know. Might look around the forum to see if I can dig up some more info.

[absolutefunk]

Well, just ran around and checked 'system info' on the 3 1070's and they all show>
GPU:2
GPU 0 Unsupported
GPU 1 Nvidia 1070

one of them had gpu 0 & 1 reversed but that's all I know. Might look around the forum to see if I can dig up some more info.

Based on your PPD I'd say all systems go!
The unsupported GPU may be the integrated intel graphics.

-Brian

[ShaidarHaran]

Based on your PPD I'd say all systems go!
The unsupported GPU may be the integrated intel graphics.

-Brian

I know you were addressing Jima with this remark but I can say with a certainty that this is not the case for me, as my 1070 is in an x99 system which has no onboard graphics.

[Flyordie]

Pull up GPUz and see what the bus load is as well as GPU usage? If its 90-95% then you are pretty much maxed out.

I have to leave 2 cores free on my FX8320 to feed the 780Ti above 95%. [email protected] GPU requires an FPU to be free 100%. (Windows 10 knows which "cores" are linked to a module though, so I was able to lock it to one module.)

[ShaidarHaran]

Pull up GPUz and see what the bus load is as well as GPU usage? If its 90-95% then you are pretty much maxed out.

I have to leave 2 cores free on my FX8320 to feed the 780Ti above 95%. [email protected] GPU requires an FPU to be free 100%. (Windows 10 knows which "cores" are linked to a module though, so I was able to lock it to one module.)

I haven't checked GPU utilization but I know my CPU's not being over-utilized since I can see the performance of each FAH process. The CPU workload is being limited to 6 threads (so 50% of my 5820k) and the GPU workload takes only about 7-8%.