Been wracking my brain over a problem with TVAI and was hoping perhaps someone might know what is going on.
I have two identical motherboards, same CPU (an intel i5, 4-core), memory (16mb 2133mhz DDR4), SSD drives and both are running Windows 10 and both have identical BIOS settings.
With a 3060ti video card, one will process at around 3.1fps, while the other runs at around 1.8fps. So I tried a different GPU, a 3080ti and again, the one machine will run at about 3.5fps while the other still won’t go over 1.8 to 1.9fps.
To troubleshoot further I swapped SSD’s and the problem swapped with them, the slow machine was now fast, the fast one is now slow (with identical FPS numbers)
So I cloned the fast machine’s SSD to another and now both machines are working at around 3.1fps…
Then I cobbled together another machine, this one using a much newer gaming MB (i9 10 core (20 threads total, 32gb ram, much faster m.2 drive) with a clean install of Windows 10 with current drivers installed and again, any video card I put on it processes at around 1.8-1.9fps.
This appears to be a windows problem and I just can’t seem figure it out.
Has anyone run into this and have a possible solution? If I did this right I’ve attached my logs and DxDiag files.
All 3 machines, the 2 identical ones (the fast and slow ones) and the gaming machine are using the same settings, and were rendering a copy of the same file (each machine has a copy of the file on their respective SSD’s.)
Those settings are:
input: 1920x1080 @23.976fps
output is 3840x2160 @60fps
Motion Deblur turned on
Frame interpolation is set to Chronos Fast
Enhancement is set to Progressive, Proteus with the parameters set to auto.
Oh, and all 3 have the GPU selected with max memory and 1 process (I tried them on 2 and it didn’t seem to make a difference.)
If you mean they are maxed out speed wise, that is something I had not considered. I’ll run a disk drive benchmarking program on them to see of there is a difference in speed.
Space wise, they all have plenty of free space, each machine pretty much just has windows, topaz and a few misc programs on them.
I noticed that this option in the BIOS significantly affects GPU utilization. At least in my case (M.2 SSD, B450 motherboard, RTX 3090) GPU usage in the Topaz Video AI program jumped from 35-40% to around 85-95%. (verified, for the same video with the same upscaling settings). So maybe you guys should check it out and experiment as well.
It seems to me that there is an acceleration. Since my motherboard is B450, it does not support PCI Express 4.0, only 3.0. FPS has definitely jumped and is more stable, there are not as many frame drops. Before, the load is higher on the processor, around 90%, and lower on the graphics card, now the graphics card is more involved while the processor is less.
just tried it following your post. for background my 6800xt already has 99.9% gpu utilization in topaz at all times. changing this value from auto to x4x4x4x4 has lowered the speed by around 15%, at the same gpu power consumption, but whats important is that it means this setting definitely has an effect on topaz speed! thank you for sharing.
My CPU is definitely a bottleneck for my graphics card. (The Ryzen 1700 from 2017 is too weak for the RTX 3090) On top of that, the motherboard only supports PCI-Express 3.0, which also leads to some performance loss. I recently upgraded to a KC3000 M.2 SSD, and while the overall performance is slightly better, it’s not a drastic improvement. I only see a few extra frames per second, and running benchmarks in Geekbench shows almost no difference between the two setups (the scores are nearly identical).
@zelenooki87 I don’t think the PCIe lanes would have any effect on Topaz performance.
Most of us anyway have the graphics card installed on the PCIe x16 slot. they don’t transmit so much bandwidth data (over 16BG/s) to need or take advantage of Gen.4
You shouldn’t worry to much about not being Gen.4
You’re definitely right. However, in my case, the RTX 3090 is utilized at almost double the rate proportionally, and it lowered the CPU load. Just wanted to share my observation. Regards.
usually i see about 3-6 fps going from 1080p30 to 4k60 using proteus and apollo and encoding to nvidia av1 10 bit @ 30mbps, reading the video off a 990 pro ssd and encoding it onto a different 990 pro
im on a 7950x3d (affinity set to ccd1) and a 4090 plus 64 gigs 6600mhz ddr5… i think i should be seeing higher numbers based on what im reading
