Folding Forum

Hello, all! I am new to folding and I'm curious about the PPD I am yielding, mainly if it's lower than it should be.
I am folding on my sig rig, but I am running stock profile right now with no overclock on either cpu or gpu. When I launched the FAH client it gave me the client option, and I believe I chose the gpu/smp combined option; my gpu and cpu both show 100% usage. I am currently returning a potential of 7800-8000 PPD, but it seems lower to me that I anticipated.

TL;DR
Is this PPD normal for my hardware at STOCK speeds? (7800-8000PPD on a 1090t + 6950 2gb)
Do you have any tips for me to improve my PPD?
Would it be better to change the client I am using?
Should I switch to only SMP or only GPU, and HOW?

EDIT
Looks like the SMP is doing nothing, while gpu is folding; however, both show 100% usage.
Heres a screen:

Welcome to the forum, DefunctRonin.

The client isn't calculating PPD for your SMP slot yet because the WU hasn't progressed enough for the client to make a prediction (notice the "Unknown" in the Estimated PPD field on the right). The most likely cause of this is that your GPU is using a lot of CPU time. It's a known issue with AMD's GPU drivers; they use almost a full core in situations like these, and SMP folding progresses only as fast as the slowest thread. This is apparently not something the Pande Group can fix, and I don't know when AMD will resolve this (or if it even can be resolved with their current design methodology). You should pause the SMP slot (right-click it in the Folding Slots area and hitting pause in the short menu that pops up instead of hitting the big Pause button at the top, which will pause everything), go to Configure->Slots, Edit the SMP slot, and change the CPUs field from -1 to 5, then save the changes and restart the slot. 5 threads running without much interference is, in fact, notably faster than 6 threads with one of them seeing massive interference.

Note that while normally it's not recommended to fold with a prime number of threads since certain projects have had major issues when running that way (hence why the text in the configuration will say you should stick to a multiple of 2), I do not believe anyone has actually reported any issues with smp:5 (smp:7, definitely). If you ever do see any issues that are determined to be caused by running on 5 threads, you can always bump that down to smp:4, which will certainly work without issue.

Do we play with thread priority any more, or does that not apply in V7?

In v6, we used to set the GPU client to Low, and the SMP client to Idle.

Make sure you are using a passkey

Welcome to foldingforum.org, DefunctRonin.

First, leaving one free CPU-Core for your GPU will speed up SMP considerably.

Second, you do need to get a passkey and it needs to be qualified before you can get bonus points. The first 10 SMP WUs that you return before the Timeout expires using the passkey will not receive a bonus but after that, as long as you maintain a success rate exceeding 80% you'll get Quick Return Bonus points which will exceed the standard points by a rather wide margin.

You may also want to temporarily suspend GPU processing and devote all 6 FahCores to SMP processing since you'll complete the 10 WUs measurably faster folding on 6 cores that 5. Once you're getting bonus points for SMP you can make a more seasoned choice of SMP:6 alone or SMP:5+GPU.

Thanks everyone for the help! I appreciate it. I now have a passkey, and am on my second WU now, about 65% done.
I removed the gpu slot, and focused on SMP. From what I read AMD cards are a bit of a hindrance for folding =(
Thanks, again!

DefunctRonin wrote:From what I read AMD cards are a bit of a hindrance for folding =(

Not a hindrance for folding. A hindrance for SMP folding

Yes, I have a 6950->6970, and i used to fold SMP:3 + GPU. But it used too much power, was hot, slow, and loud. So I switched to SMP:4 only now. I find it a pity though, I like the feeling that my whole computer is being used 100% for science though.

FAH is using all of the CPU resoruces for science which would otherwise have been wasted as idle time. That was the original design goal of FAH some 12 years ago. Those resources are carefully managed by your OS so that any time you want to use your computer, your foreground tasks get priority.

You're not using your GPU for science but that still may be the best option -- to use 98% of the CPU and let the GPU be 98% idle.