Windows GUI client is bone head simple always working well for me but after upgrading my GPU I have a problem with work units not being successfully completed/received. Per the subject, well over 3/4 of the WU that have been processed by my new GPU result in "Server did not like results, dumping". Not really sure what to make of it. I have seen WUs fail over the years from overzealous and/or not quite properly dialed in overclocks but in this instance every time the WU completes sans error in client and logs, seems to take longer than I would expect to upload, and ends with the forum thread subject quoted in the log.
Typically w/ enough errors in the past the client would disable the troubled CPU or GPU slot but in this case the servers continue giving me WUs to crunch and keep rejecting them. I haven't attempted to run WUs on this particular GPU in a few weeks so I am currently generating log data. I will have several WUs worth of results to supply tomorrow but I do want to sort out if this is a problem w/ the GPU or some sort of configurable issue on my end. From memory there is no one particular server that is swatting my results away so I do not believe this to be a problem on f@h's end.
Troubleshooting performed so far:
- I have reinstalled Windows 10 Pro from drive wipe twice since this has cropped up.
I am currently running the latest AMD chipset drivers
Currently on a DDU'd install of nVidia driver 536.23, several prior versions behaved the same
Have tried forcing slower PCIe connection spec to GPU in BIOS
Have tried two different PCIe WiFi adapters + motherboard in-built, and wired to router. Have also tried using mobile Hotspot connection to transfer data. with all three WiFi adapters
Have tested the GPU w. Aida64, OCCT, and FurMark (all > 24 hour continous runs), and the various 3dMark stability tests with no problem found or artifacting as applicable
Have enabled GPU memory ECC
Have swapped powersupplies
Have downclocked system memory to JEDEC DDR4 baseline 2133
Have tried 4 different pairs of known good, reputable manufacture system memory
Have used MSI Afterburner to set negative GPU core and memory offsets down to -500 core, -500 memory with voltage and power limits set to both default and maxed out
Watercooled CPU, GPU, motherboard VRM w/ (2) 560mmx140mmx60mm radiators, excellent flow rate
Asus x570 Crosshair VIII Formula
Ryzen 5800x3d (note, I do not run fah CPU WUs on this system)
Gigabyte 4090 Gaming OC
(2) different Seasonic Prime 1300w Platinum PSUs and a Seasonic Prime Prime 1000W Gold. The daily driver is one of the 1300w Platinums
I think that's everything to say w/o some logs. I will add that this machine games very well with fluid frame times and tight consistency within the capabilities of the CPU. It has yet to render a single visible error in my numerous home video edits which always call for a little bit of GPU work. Save for f@h servers not liking the results and dumping them, I can't tell there is a single thing wrong in the slightest.
I appreciate you and your time in advance.