Folding Forum

I meant 100% of 1 core, rarely 2.
I look at runtime of the associated thread via top on linux. And in general it is within 5% time of WU.
Especially newer core 22 units need a lot of cpu time.

See my response and toTOW's response above. Between them they explain why and how the CPU is used for GPU folding.

If you're folding with a GPU, the sanity checks are done on the CPU. If you are also folding on the CPU, then the two processes will compete for CPU resources, slowing down both of them until either process has been completed. The GPU process uses 1 thread constantly and multiple threads for the sanity check as toTOW has said.

foldinghomealone2 wrote:...my understanding is that sanity checks don't require much time (compared to GPU computing) and therefore overclocking of the CPU doesn't improve GPU folding performance.
Please correct me if I'm wrong.

Officially, F@H doesn't encourage Overclocking. Running the client on your system is very stressful even on stock settings. If you have an unstable overclock, you will experience issues while folding.

You're right that an overclocked CPU will not really improve the GPU folding performance in a realistic manner.

So one thing to note is I did a lot of averaging. Individual work units can and do often perform higher than the average I reported. I remember a few WUs getting close to your PPD, but not often.

Also, my benchmark system is getting a bit old, to put it mildly. The fx-8320e does well enough, but I'm starting to get concerned by my chipset (880g). The pci-e bus is 2.0. While I have read that this doesn't start noticably bottlenecking things until RTX 2080 territory, it could on occasion limit me with fast cards. This, combined with the overall age of the rest of the system (8 years) means it's time for a new benchmark machine.

This is currently half assembled. It's an x570 AM4 based rig with a Ryzen 9 3950x at it's heart...with PCI-E 4.0, 32 GB of 3600 ddr4, and an 970 evo plus NVME drive. I should be good to go in a few days (am working on the cooling solutuon). I might retest the 1080 Ti on this rig to see if there is a difference.

Problem now is I have literally blown the budget for the rest of the year, so it might be a while until I save up enough to test the RTX 2080 Ti I've been dreaming about.

Also I should note that my 1080 Ti is a reference model (blower style). I didn't push the clocks on it...I tend to report "baseline" results. Meaning, something that the average user might see without much tuning or fancy cooling.

I have a GTX 1080ti and an RTX 2080ti and really ave to say I prefer the 1080ti for ease of use. The RTX 2080ti a super producer but from my experience needs way more coddling than any of the the cards I have run from my first 8800GT and almost every NV generation since then. I have recently bought a 1660ti and am interested to see how this plays in my 1080ti rig. I hold out hopes that the next 3000 series will be a sweet one, but have doubts about being an early adopter like I was for the 2000 series.