PantherX wrote:
That would be the correct method to add the CPU Slot. Regarding the free CPU per GPU rule, it varies on AMD system due to drivers. Thus, what you can do is to open up Task Manager and see how much CPU cycles is FahCore_17 using (exception is at the starting of the WU and during checkpoints). If it generally uses very little CPU cycles, you can use all CPUs for CPU folding. If you notice that it is using a significant amount of CPU cycles, then a free CPU would be sufficient for it.
Thanks for the info!
PantherX wrote:
Okay, so you were folding on both GPUs and the CPU, thus, had three active folding slots when the issue occurred. This seems to the initial issue where 2 GPUs within the same system wouldn't fold, is that right?
Yes that is right. This is the same situation it has always been but much much worse. I have not seen things fail repeatedly like that literally within seconds of startup. Makes me wonder about the GPU but to see if it is the GPU I'll swap it with the GPU running now to see if the "suspect" GPU will run alone. In all honesty I expect it to run fine as the lone GPU.
I also had a strong suspicion about the WU itself since it seemed to be simply impossible to process it so I scanned my other machines to see if a WU from project 10466 was running anywhere. Sure enough I see that project being processed on a machine with two GPU's right now and it is processing fine. Which again makes me wonder if I have a GPU failing. I'll find out soon.
PantherX wrote:That is weird. Did you make sure that FAHClient was running since Advanced Control (AKA FAHControl) can only display those tabs once it is connected with FAHClient? Generally, manually editing the config.xml isn't recommended since a typo might cause issues. However, once you deleted the config.xml, FAHClient re-detected the hardware and created the matching Slots.
I didn't know FAHControl needed FAHClient. In fact, I thought they were one in the same because I don't launch the FAHClient directly. I have seen what I described before though when I had an issue where Windows seemed to lose a GPU. I had the same issue and FAHControl would appear to hang. I can't remember if it said updating or connecting but it never seemed to get anywhere.
I guess my question would be why would the FAHClient not launch as it always does when I launch FAHControl? It is something I can look for in the future though.
PantherX wrote:That is good to hear that your CPU and GPU carried on folding WUs which were primarily assigned to it. Regarding the 2 GPU's WU, it will be either discarded or will be shifted to GPU 1's Slot once the current WU finishes. Do note that Project 10466 to Project 10469 were recently released to full F@H (viewtopic.php?f=24&t=26459).
In this case it looks like it was discarded. At the pace it was on with that machine it would have taken 10,000 years to finish anyway.
PantherX wrote:
Okay, so with a single GPU, you have been folding without issues on the CPU and GPU simultaneously? If so, that sounds good. Hopefully, it will continue like this for a long time.
That's correct. I ran tests before on a different machine where I ran one GPU only and then switched. They each ran fine. After that I added the CPU's back to folding because I only removed them to see if things became more stable. They didn't, so I added them back in. 5 of the 6 machines I'm still running 2 GPU's and a CPU. I also dropped the slider to Medium to see if it would make a difference. The machines appear more stable but they still have a few issues a day spread among them. If I dropped to a single GPU and used the CPU for folding I'm sure I would run without incident and that may be what I end up doing.
Right now I'm still trying to find a way to make what I have work but I think I've reached a point where it is safe to say running 2 GPU's in the 6 machines I've been talking about will never be smooth. The question is can I find a way to keep more than 3 of them running for 24 hours a time because if I can I'll still do better than I would if I just ran 6 GPU's in those machines total. It appears I can generally keep 4-5 machines running with 2 GPU's for a 24 hour period of time. So it is a net gain in PPD and WU completed. I just don't know if that will be maintainable when I'm not available as much.
I can use TeamViewer to remote in to the machines via my phone but as I've mentioned often times the machines become so corrupt after a driver failure that I have to hit the reset button. If I catch it soon enough the machine will reboot fine on its own but if the video driver crashes multiple times the machine is not able to reboot on it's own. So we will see how it goes. If it isn't doable then I'll just have to sell some GPU's.
PantherX wrote:
BTW, this is only 1 out of 6 systems. Are the other 5 systems folding fine with 1 GPU or are you only testing on a single system and once satisfied, will make the appropriate changes to the other 5 systems?
See the post above this regarding the current state of my machines, but to be very clear I don't expect to actually get the systems running smooth at this point if I use two GPU's. I have now seen that running with a single GPU on a system even with CPU folding enabled is solid and would be incident free. I may even be able to turn up the clock speeds on the GPU if I only used one. So that is something I need to look at.
Someone in the EVGA forum told me they had another folder that ran into a similar situation where it just appeared to be impossible to get two GPU's to fold consistently with the persons hardware. Apparently that person stopped folding after fighting with it for awhile and not being able to find a solution. They want me to try overvolting my CPU and the PCI-E lanes, but I ran into the stability issues even with the CPU folding disabled so it would have to come down to the bus lanes if that were the case. I need to look in the BIOS to see if I can even find settings to do it but I'll try it if I can.
In any event, I intend to keep folding. I have no intention of going back to "mining" even if the hardware can do it.
Edit: So I wrote an explanation as to what I was doing and I was going to try to keep two GPU's per system running and right after I posted the message I did a check of my machines and 4 out of 6 machines were down. LOL See I think the machines just like messing with me.
In case you are wondering the machine that is using one GPU was still working flawlessly. So it is looking more and more likely that I'll drop them all to one GPU just so I don't have to check every 10 minutes. Of course I would prefer to keep the machines using two GPU's but it just doesn't seem possible if I want to keep my sanity.
On the machine that is now running one GPU I set the clock speeds of the GPU back to it's defaults which were Core: 1040, Memory 1300. It was running at Core: 947, Memory 1000. I'll see if the GPU is still running in the morning.
PantherX wrote:
Unless Intel has changed stuff, it seems that you don't have OpenCL packages installed. Since I have installed it, this is what appears in the Programs and Features list:
Intel® SDK for OpenCL - CPU Only Runtime Package
Yeah I do not have that entry in my programs list so I think it is safe to assume I don't have this installed.
