CPU cant get assignment - GPU works fine

[MrCov]

Hey guys, New to folding here.

I'm currently running a Core i7 6800K OC @4.4GHz with nvidia GTX 1080 FE and I seem to be hitting a problem getting work orders for the CPU. The GPU gets jobs just fine, but for some reason when the CPU finishes a job, I get an error message:

00:29:27:WARNING:WU01:FS00:Failed to get assignment from '171.67.108.45:8080': Empty work server assignment
00:29:27:WU01:FS00:Connecting to 171.64.65.35:80
00:29:28:WARNING:WU01:FS00:Failed to get assignment from '171.64.65.35:80': Empty work server assignment
00:29:28:ERROR:WU01:FS00:Exception: Could not get an assignment

However if I completely stop the folding, then restart it, the CPU receives a job again.

Any thoughts on what could be going on here?

Thanks for the help

[JimboPalmer]

When you have a GPU, FAH uses the number of CPUs minus one to fold on the CPU slot. However FAH 'hates' prime numbers larger than 3. so you want to set your CPU count to number of CPUs minus two. that will fold.

In the Advanced client, configure the CPU slot to 2 less than the CPU count you have.

[MrCov]

Thanks for the reply! I just changed it to 2 less like you said. (10 threads, down from 12) and started a work load. I'll let you guys know it if restarts after this one is done!

[bruce]

With no GPU, 12 would work fine. With a single GPU, allocating 1 to the GPU leaves 11 out of 12 or 7 out of 8 and both 11 and 7 are bad numbers. Ten or six work out a lot better.

[MrCov]

Yup lowering the thread count to 10 solved the issue. Thanks for the help!

[Ricky]

Primes of 5 or lower is the key and multiples of these primes up to 24.

[wickedproxy]

When you have a GPU, FAH uses the number of CPUs minus one to fold on the CPU slot. However FAH 'hates' prime numbers larger than 3. so you want to set your CPU count to number of CPUs minus two. that will fold.

In the Advanced client, configure the CPU slot to 2 less than the CPU count you have.

I had GPU working and CPU not able to get a WU. My CPU is 6 core and the slots was set to 5. I changed the CPU slots to 4 and now it seems to be running! YAY!!! Thanks for the help!

[Crucnched]

I had the exact same issue for a week. All good now after setting it from 7 to 6.
Shame the cores are no longer running at 100% though. Any possible downsides to setting the CPU slots to 8 (ie one higher than I have?)

[Joe_H]

I had the exact same issue for a week. All good now after setting it from 7 to 6.
Shame the cores are no longer running at 100% though. Any possible downsides to setting the CPU slots to 8 (ie one higher than I have?)

No downsides to setting to 8 if you do not have a GPU slot. If you do have a GPU slot, then setting the CPU slot to 8 will both slow down folding on the GPU, and the CPU usage of the folding core for the GPU will probably negatively affect CPU folding by more than you will gain through the extra CPU thread.

[bruce]

We're talking about different issues.

Some numbers of CPUs have had numerous WU failures -e.g: 11, 13, 17 ...) and the servers will not distribute WUs to those systems. Five has not been in this group, but some projects don't do well on 5, so the project owner will block the distribution of that project. When that exclusion is applied, it's not clear to anyone if there are other projects which are allowed to assign to CPU:5 systems. If this produces a shortage, you'll see the condition wickedproxy reported. (There may be other projects for CPU:5 on another day.)

CPU:7 is in the same category, but I suspect that all projects now exclude those assignments, so it might as well be considered a permanent restriction.

Assignment to large systems can also be problematic. If a protein is too small to be subdivided into 32 sub-regions they won't be assigned. For the most part 24 is fine. Greater than 24 is problematic.

There's a heurictic "rule" that you need at least 500 atoms per CPU, so only proteins with significantly more than 12000 atoms will be assigned to systems with more than 24 CPUs. (You still have to follow the "no large prime factors" rule, too.)

Development has been attacking this problem. I can't promise how soon a fix will be forthcoming but it may go something like this:

Suppose your system is set for CPU:7. It asks for an assignment and is told "none available right now" The software overrides your choice and says "how about CPU:6?" Maybe you get an assignment ... maybe you don't, depending on which projects are active, which servers have WUs, and what combinations of restrictions are in effect. If still no assignment, "how about CPU:5?" etc.

In other words, the setting of CPU:7 has traditionally meant "I will only accept projects that can run with 7 CPUs" the servers will treat it as "Give me a project that can run with <=7"

If you're set to CPU:32 and the largest protein available at the moment uses 24, you'll see 8 unused CPUs, but you will get an assignment.

See also the discussion in the beta forum beginning here and continuing onto the next page.