It’s as if the in app setting gets ignored, hope this can get sorted out eventually
2024-08-31-03-04-15-Main.tzlog (6.5 MB)
Try a setting under 50% and see what happens.
I’ll try that tonight, it was around 80% complete when it decided to fail, 8 hours of rendering, poof! I hope the eventually add a function where you can continue even if it errors instead of just deleting the processed output automatically.
still errors at 40%, I’m sure when they get it solved it will be mentioned in the release notes
What’s your GPU?
A 4070 12GB, I do wonder if it’s some sort of conflict with the nvidia drivers, you would think even 80% would be fine, just youtube tops off at 6% vram usage, I’m not doing anything intensive during a render.
The Studio drivers seem to have fewer problems than the game ready ones.
I had BSODs when Interpolating in audio-copy mode
try audio-convert
not a bsod issue
Hi, I originally started a thread more than 2 months ago where I was reporting these errors when running an enhancement of combined Nyx and Aion. I was told by support that the logs were showing out of memory errors which was not the actual truth (even if Topaz AI believed it to be so) and that the problem was because of Nyx.
Well since then I have built a brand new machine literally twice as powerful as that old machine and with a 16 GB TRX4080 Super instead of the previous 8 GB RTX2060 Super. I stability tested this machine thoroughly and run it flawlessly (and I mean FLAWLESSLY) for an entire month before loading Topaz onto it.
As soon as I tried to run those very same jobs again, they failed in precisely the same way they did last time - with supposed “memory” errors. Ridiculous - the VRAM used is 5 GB - 6GB each time and is never exceeded - so even the old machine should never have failed to process these files let alone the new one. Changing the memory slider in Topaz makes zero difference. If you think it does then you are just lucky - for some reason it happened to work that time but that is NOT the answer. Sooner or later you will get errors no matter where you put that slider. The makeshift (and poor) workaround to avoid the errors is the one I mentioned in the thread linked to above. It is of course a useless solution unless you are processing only one job at a time.
I then tried an experiment on my new machine which was to split the jobs into two - meaning one job for the Nyx part of the processing and Aion for the second part. This time around, every single Nyx job worked flawlessly and I can easily repeat this success - I even ran an entire batch of Nyx jobs overnight for hours on end with zero problems. Those Nyx jobs then exported to a lossless FFV1 AVI which I then used as input for the second part of the job - the AION processing. This time around I experienced the same errors galore that I did last time. Typically the first job in the queue would work fine, the next two would error within seconds, the fourth one would work, the 5th and 6th would instantly error out, the 7th would work. You get the pattern. A job would either error out within a handful of seconds to perhaps 30 seconds - exactly like it did on my old machine - or happily run all the way to the end. Again, the VRAM used was the same as before when the jobs were combined into one - less than 6 GB on a 16 GB card and at no time ever exceeding that (not that memory is a problem with 1080p limited Topaz processing on a 16 GB video card).
In that previous thread I also had found a “solution” which you can see I have written down there. Kyle from topaz was even able to reproduce the error at his end and said it would be handed over to the team for investigation. Well that old thread is closed now and obviously there is still no solution.
This is clearly NOT a problem at the user end and it is not a hardware problem. This is also NOT a problem with Nyx as Topaz support erroneously hypothesised. It is a problem with AION and seems to manifest itself when the input files are beyond 576p in resolution (since in all the time I have used Topaz, I have never, ever had an AION based job using 576p input to 1080p fail with any error).
So as you might have guessed I am getting a bit sick of this. An entirely new machine that cost me thousands yet precisely the same errors on precisely the same job in precisely the same way. We are all getting sick of these errors erroneously telling us the problem is at our end when it is clearly is a Topaz problem. There is a bug in Topaz AI causing these errors and it is AION that is the problem, not Nyx.
So set to 3 max jobs, only run 1 job at a time and reboot between runs. I really hope they get this sorted.
Luckily I do not use AION that often but every time I need it I dread having to and weigh up the pros and cons of my workaround (one job at a time between reboots) or processing a batch, re-processing all the errors, re-processing the remaining errors from the second run and re-processing the remaining errors from the third run…
so far so good, going to do a test and suspend ffmpeg in resource monitor and resume and see if it keeps going, it would be a workaround that lets you pause an export
so it errored after about an hour and 10 minutes after I paused and resumed it in resource monitor, I’ll reboot and run it overnight without pausing it and see what it does. I’ll also set max jobs to 4 for good measure
One thing I can say with 100% certainty is that I have run Aion a gazillion times over the last 5 months with 576p25 input material (upscaling and interpolating with Proteus as a single job to 1080p50) and not one single time - ever - have I gotten an error. Not even once. And this is regardless of whether it was just one job that only needed 5 minutes to complete or a whole heap of jobs that kept the machine running for more than 24 hours.
Yet once I go above that input resolution, errors galore. And they are always near instant. Even right after a re-boot. I have only avoided the errors by going with the procedure I mentioned however to be fair, all I can state is that I have never had the errors with input material over 576p in resolution using that procedure. It does not mean it is failsafe. And I have only ever done testing with an output resolution of 1080p. I don’t know what happens at higher output resolutions. In any event, the problem is at the Topaz end of things and there is not much we can really do about it.
For the timer being I just put the jobs in the input queue and randomly come back and click on the option to re-process the ones that errored out. Eventually it gets them all done. I guess I am “lucky” in that when Aion errors out, it is always within 30 seconds and usually instantly so I do not technically lose a lot of processing time.
it went for like 14 hours and then errored, I wanted to stress it good, so 4 jobs max setting, 100% video memory off of a fresh reboot with no pausing seemed pretty stable. Definitely not a vram issue like the logs say
