I can only think of a few more things that I do, that may or may not help. Make sure crossfire is turned off, disable ULPS in afterburner or in the REG. Check the core clocks, and make sure they are not throttling. Are all your Mobos the same? Check gpuz and see if they are not stuck in pci-e 1 mode. If you are using risers, maybe try swapping them. I am reaching, but i know how frustrating it is to troubleshoot.
Well after thinking about what Vorksholk said I realized since the GPU's are the same the WU's should be fine even if I move cards in the middle of work so I swapped the cards between two machines. Ironically, the cards that were running at half speed now run at full speed and the cards that where running at full speed are now running at half speed. Clearly the issue isn't with the GPU's.
So I suspected it had to be the PCI settings on the motherboard but GPU-Z shows both machines are set to PCI-E 3. Crossfire is off on those machines, ULPS is disabled and the clock speeds seem to stay at 940 which is where I have them set. The GPU's are placed directly on the motherboards so there are no risers.
I really have no idea what else it can be at this point. I'm wondering if I have a bad motherboard. All the motherboards are the same as well. Actually every component is identical in the computers.
Can you post a log with some complete WU's ? What's your GPU utilization like ?
Cosmetically I have the same thing with some of my rigs - only giving 50% of expected PPD and nothing helps. But the GPU's are working away at ~95% utilization so it seems to be a glitch in the F@H software.
I saw your messages about it before as well when I was having the same issue. I actually rebuilt two computers from the same image at that time. One runs at "full" speed and the other "half" speed based on the PPD estimates. This was the case before the rebuild and again after. The GPU utilization shows 100% in GPU-Z on both machines. However, the estimated PPD is half on the machine in question. I don't have logs to show the actual points right now because I've been messing with the machine so much today. I'll have a log in 12 hours from the machine that appears to be acting odd.
I do hope it is just cosmetic but don't you find that odd? I mean one machine has a cosmetic glitch even after a wipe and a GPU swap? That just doesn't sound likely since it is always the same machine. For me it is one machine out of nine. Not all 9 of the machines are identical but 6 are exactly identical and a 7th is pretty close. It is one of the 6 identical machines that shows half expected PPD.
I watched the timing and it seemed to take much longer to do work on that machine but I know that doesn't necessary mean anything because there is so much difference in WU. I need to see how many points are awarded after it submitted work in the log. I'll do that as soon as I can. I don't have any awarded points in the logs for that machine at the moment so I can't post them.
I think it is odd but I have the same thing on 3 rigs out of 18. Also tried everything and nothing helped. Then I checked the log for WU's and it showed 1 WU per card in about 4.5h and ~23 000 points for each of those. This gives me a actual PPD of 122K per card despite the software showing 55K.
I think you are right on what appears to be a very odd client estimate glitch. I have no idea why it would always happen on the same machine but here are what the logs show.
Here are the log for one of the GPU's from the computer that is running at "half" the estimated speed.
08:11:25:WU03:FS02:Upload 6.33%
08:11:31:WU03:FS02:Upload 12.66%
08:11:37:WU03:FS02:Upload 19.00%
08:11:43:WU03:FS02:Upload 25.82%
08:11:49:WU03:FS02:Upload 32.15%
08:11:55:WU03:FS02:Upload 39.46%
08:12:01:WU03:FS02:Upload 45.79%
08:12:07:WU03:FS02:Upload 52.12%
08:12:13:WU03:FS02:Upload 58.45%
08:12:19:WU03:FS02:Upload 64.78%
08:12:25:WU03:FS02:Upload 70.63%
08:12:31:WU03:FS02:Upload 75.50%
08:12:37:WU03:FS02:Upload 80.86%
08:12:43:WU03:FS02:Upload 87.19%
08:12:49:WU03:FS02:Upload 93.04%
08:12:55:WU03:FS02:Upload 99.37%
08:13:06:WU03:FS02:Upload complete
08:13:06:WU03:FS02:Server responded WORK_ACK (400)
08:13:06:WU03:FS02:
Final credit estimate, 80873.00 points08:13:06:WU03:FS02:Cleaning up
Here is a log for one of the GPU's from a machine that is reporting "full" estimated speed.
07:27:01:WU00:FS02:Upload 5.35%
07:27:07:WU00:FS02:Upload 11.19%
07:27:13:WU00:FS02:Upload 18.00%
07:27:19:WU00:FS02:Upload 23.83%
07:27:25:WU00:FS02:Upload 30.16%
07:27:31:WU00:FS02:Upload 36.48%
07:27:37:WU00:FS02:Upload 41.83%
07:27:43:WU00:FS02:Upload 47.67%
07:27:49:WU00:FS02:Upload 53.51%
07:27:55:WU00:FS02:Upload 60.32%
07:28:01:WU00:FS02:Upload 66.15%
07:28:07:WU00:FS02:Upload 72.48%
07:28:13:WU00:FS02:Upload 78.31%
07:28:19:WU00:FS02:Upload 83.67%
07:28:25:WU00:FS02:Upload 89.50%
07:28:31:WU00:FS02:Upload 94.37%
07:28:46:WU00:FS02:Upload complete
07:28:46:WU00:FS02:Server responded WORK_ACK (400)
07:28:46:WU00:FS02:
Final credit estimate, 80997.00 points07:28:46:WU00:FS02:Cleaning up
Both of these machines were down for roughly the same amount of time as I swapped GPU's between them. So their payout was slightly lower than usual but still proportional to each other.
Regardless of what the client says it appears that both machines are earning roughly the same points. I'll watch it close over the next couple of days and see if this remains the same.
Again thanks for the information! It appears to be a very strange client glitch. Not sure what is triggering it but it is consistently misrepresenting the PPD even though the actual work appears to be correct. Maddening but very interesting!
Now only thing I am checking is GPU utilization and if thats 95+% I dont care about the numbers that the client is showing anymore.
I'v got a different problem myself now: Some GPU's get stuck downloading new WU's
Stays on "downloading" forever with no getwork server address available. Restarting the client solves this but is really uncomfortable if you have many rigs to keep an eye on....
I think it is odd but I have the same thing on 3 rigs out of 18. Also tried everything and nothing helped. Then I checked the log for WU's and it showed 1 WU per card in about 4.5h and ~23 000 points for each of those. This gives me a actual PPD of 122K per card despite the software showing 55K.
Hi kqpahv,
I have a question for you: just how do you manage your 18 rigs for folding? If I understand right you have all AMD cards and thus you have to use Windows. I can't still image how difficult it was to maintain more than 3 rigs with Windows. Btw could you share what kind of CPU and how many cards do you use for your rigs?
Thanks in advance,
Regards,
Both of these machines were down for roughly the same amount of time as I swapped GPU's between them. So their payout was slightly lower than usual but still proportional to each other.
Regardless of what the client says it appears that both machines are earning roughly the same points. I'll watch it close over the next couple of days and see if this remains the same.
Again thanks for the information! It appears to be a very strange client glitch. Not sure what is triggering it but it is consistently misrepresenting the PPD even though the actual work appears to be correct. Maddening but very interesting!
Yeah, my experiences taught me don't trust the estimated points which the client gives me. The client and the core have been developed for over 14 years and still very buggy because they won't open source it and only a few handful people are working on this very big project.
Yeah, I am using Teamviewer. It's a pain in the ass but pretty much the only option. Once everything is set up it's easy though - I only check once/twice a day if everything is running. That only takes me about 15 minutes, setting all the rigs up with new drivers and F@H clients was an other story
Stays on "downloading" forever with no getwork server address available. Restarting the client solves this but is really uncomfortable if you have many rigs to keep an eye on.... 
