on my old RTX 3000 series GPU, ive been running TVAI with no issues whatsoever. then i upgraded to a 4000 series card and the problems began.
no matter what TVAI version, what enhancement model, what output settings and what driver version i run, i keep getting GPU errors when running any job in TVAI. the errors show in the event viewer as “nvlddmkm: \Device\Video3 Error occurred on GPUID: 100 0x80000000000000” and every time it happens, the export breaks, usually it shows as TVAI kind of hanging, where it says it’s working on a job but there’s no GPU and CPU usage and the ETA keeps climbing up.
the upgrade was 2 months ago. that’s when i tested TVAI and got these errors. i continued to use this card for playing several AAA games and even running some other frame interpolation tools. there have been completely zero issues, the event log is squeaky clean. i test version 7.0.1 today, same thing again, GPU error about 30 mins into the export. i really doubt i’m having a hardware issue when TVAI is literally the only program having problems.
is this program just not compatible with 4000 series or what?
This happened to my 4090 too. It was a faulty Nvidia adapter type “Octopus” and/or the entire power cord and/or connector. All these cards seem to be shipped with Nvidias mostly faulty cable and adapter. I replaced it with a GPU power cord by beQuiet and since then the issues are gone. Here are some of the available cables:
Ah, okay. Then maybe check if the power cord is faulty. In addition if you have an Intel CPU 13th or 14th generation check if you got the latest BIOS. Intel had serious issues with those, but causing other problems.
it’s a cable i bought just for this card, brand new, so i sure hope it’s not damaged out of the box or anything like that. but just to test, i replugged the PCIe power cables and used the “octopus” adapter my card came with, and nothing seem to have changed. everything else is stable, just TVAI causing GPU errors.
i’m aware of the 13/14th gen Intel issues, but thankfully i do not own such a processor. my current CPU is perfectly stable, i frequency use it for encoding jobs (which would be the kind of load similar to what TVAI does, with ffmpeg) and not once has it failed me. and i’m thinking the same about rest of the hardware. i threw everything i could at this PC, from stress tests like OCCT (both GPU and mixed/power tests), several furmarks, memtest_vulkan, 3DMark, Nvidia’s own GPUStressTest, anything. it just works, but TVAI just continues to complain. that’s why i’m having such a hard time believing the issue is with my hardware.
I spent a lot of time lately to setup my 13900K the best I can. It ran perfectly stable in Cinebench R23 both single core and multi core and also in the cpux.net long term test. But when I ran video ai I got a red cross error. And this is according to my experience most likely a cpu instability. I added a minor +0,010 V via Adaptive Offset in the BIOS and now it works fine. Video AI is a great stability test. I know that Nvidias latest drivers caused problems beginning maybe in March 2025. Please try Nvidia driver 566.36. It is a bit older but told to be one of the most stable ones!
i’m running a ryzen 7 7800x, usually undervolted, but i also tested it on stock settings, even including PBO off. i mean i can try bumping up the voltage offset marginally… but again, on a previous card i have had no issues, and i did not change any of its working parameters. just like how i blasted my GPU with benchmarks and stress tests, i did likewise with my CPU.
i’m aware of the nvidia driver situation, i did stick to version 566.36 for a while, it didn’t help with the TVAI situation though. they fixed their bs (mostly) so right now i’m on the latest driver.
i do appreciate the advice you provided me, even with all the things i tested and done i don’t exclude the possibility of missing something i could try on my side.
ok so i left the program overnight to do some processing again. it didn’t generate any GPU errors, but apparently it repeated the same behaviour as before, where the GPU and CPU load would cease and the program would be stuck infinitely processing a video, but also there’s excessive reading being done on the input file displayed as my drive being used at 100% in task manager.
i looked at the program log and this line shows up several times throughout the log, and also it’s repeated 5 times in a row at around the time when this behaviour began:
File ended prematurely
the rest of the log is filled with lines such as these (that’s the last one):
2025-06-29 08-15-14.317 Thread: 28400 Info FF Process Output: 2 frame= 20609 fps= 0.32205
“frame= 20609” is especially interesting, since the input file only has around 10800 frames and i’m not doing interpolation, just enhancement.
this happens on two separate videos. both of them play just fine, if i pass them through the following ffmpeg command: ffmpeg -v error -i file.avi -f null -
i get no output, meaning no errors, meaning the files should be just fine. i can also convert them without issues using ffmpeg. no idea what’s its problem.
EDIT: i inspected the output file. even though the mentioned behaviour happened at about 30-40% of the file’s progress, the output file showed much more work was done, 5:43 out of 6:30, so more like about 80% of the file was processed. so i guess the job didn’t stall completely, it just slowed down extremely hard - after starting the job, TVAI showed about 3.5 fps speed and around 2 hours ETA, but it only got through that ~80% of the file after over 9 hours.
