Folding Forum

I'd like to choose, too. I expect that option to be available in FAHCore_22 which will reportedly be available for CUDA, though that won't help folks who fold with ATI/AMD GPUs or with CPUs. Until 0x22 is released, though, we won't know.

I'm currently running on a system with 2 GPUs. I'd happily set one GPU for UI responsiveness and the other for performance -- and switch to 2x responsiveness when I go to bed.

Anyone know on Linux if in NVIDIA X Server Settings under GPU/PowerMizer changing Preferred Mode from Auto or Adaptive to Prefer Maximum Performance has the same boost as it does under Windows??? When I change it to PMP it doesn't persist after a reboot like in Windows. I can't see anyway to make it remember to stay on the PMP setting. Maybe it works better with Linux and Auto is just fine.

I usually just use coolbits and manually set the clock to as fast as I think it can go, rdp in and set powermizer to max performance, and you're off to the races. It's not like your linux machine will ever reboot . I haven't figured out how to get it to stay :/ so this is my method of madness.

I hear the miners are complaining that the Linux Meltdown/Spectre fix slowed their hashrate down and they're removing it. I don't feel like it's hurt my folding rate. Anyone taking a hit? I did the Nvidia driver updates and the MS Windows updates, not sure if MS patched Meltdown or not.

See the topic: How will Patches for Meltdown and Spectre Affect Folding?

Aurum wrote:I hear the miners are complaining that the Linux Meltdown/Spectre fix slowed their hashrate down and they're removing it. I don't feel like it's hurt my folding rate. Anyone taking a hit? I did the Nvidia driver updates and the MS Windows updates, not sure if MS patched Meltdown or not.

If you are running anything GPU on multiple machines and they are completely dedicated and seg'd from the rest of your network (connecting via squid or some other filtering proxy/firewall), I personally wouldnt bother patching it. But if they are all plugged directly into a gateway and not seg'd, I probably would. Since this exploit seems to require rogue software and you to be actually using the machine for day to day things, it is irrelevant on a dedicated gpu rig. What can the exploit do? Look at parts of kernel memory. Annoying if you are doing your banking on it. Not so much if the only thing running(and connecting to the gateway) is FAHClient. What are they going to spy on, on that machine? The data/results being folded is freely available.

Aurum wrote:Anyone know on Linux if in NVIDIA X Server Settings under GPU/PowerMizer changing Preferred Mode from Auto or Adaptive to Prefer Maximum Performance has the same boost as it does under Windows??? When I change it to PMP it doesn't persist after a reboot like in Windows. I can't see anyway to make it remember to stay on the PMP setting. Maybe it works better with Linux and Auto is just fine.

auto automatically adjusts it.
Neither one of the 3 settings affect FAH, except within the first few tenths of a second between WU loads and saves where the card GPU frequency spools down and up. But there should be no difference in PPD selecting any of the 3 for FAH.

I am presuming that these settings are more aimed towards games that can switch between high and low loads quickly.
Performance mode allows the card to remain in high clockrates a little longer; and throttles down a little slower than adaptive.
I would put performance when gaming, adaptive when doing office jobs. And auto, when folding. (it makes not a single PPD difference).