Unknown Error when selecting All GPUs

Hi There,

I am using Topaz Video Ai in the following system:

OS Name Microsoft Windows 11 Education
Version 10.0.22621 Build 22621
Processor Intel(R) Core™ i9-9900K CPU @ 3.60GHz, 3600 Mhz, 8 Core(s), 16 Logical Processor(s)
Installed Physical Memory (RAM) 64.0 GB
Total Physical Memory 63.9 GB
Available Physical Memory 54.5 GB
Total Virtual Memory 73.4 GB
Available Virtual Memory 57.7 GB
Page File Space 9.50 GB

I also have two Name NVIDIA GeForce GTX 1070 Ti cards.

When I use one card, I get 0.2fps when up-converting from 1080p to 4K using Gaia model. When I switch to All GPUs , Max Mem% and 4 processes, I get 0.4fps but I also get Unknown Error during the preview process. When I click on the x to see an error the dialog is blank.

How can I improve the performance process?

0.2 fps seems rather low for single GPU.
For comparison my 3070 is getting 2.7fps with same settings.

Maybe the GPU is the bottleneck.

How do I fix it? I have 522 driver and still get an error when using All GPUs. The error dialog does not show any messages.

This is the FFMPEG command:
ffmpeg “-hide_banner” “-nostdin” “-y” “-nostats” “-t” “15” “-i” “C:/Users/Dmitry/Desktop/100MEDIA/DJI_0029.MP4” “-flush_packets” “1” “-sws_flags” “spline+accurate_rnd+full_chroma_int” “-color_trc” “1” “-colorspace” “1” “-color_primaries” “1” “-filter_complex” “tvai_up=model=ghq-5:scale=0:w=3840:h=2160:device=2:vram=1:instances=1,scale=w=3840:h=2160:flags=lanczos:threads=0,scale=out_color_matrix=bt709,split=2[out1][out2]” “-map” “[out1]” “-c:v” “tiff” “C:/Users/Dmitry/AppData/Local/Temp/896155477/temp/_tvai_456738346/%06d.tiff” “-map” “[out2]” “-c:v” “h264_nvenc” “-profile:v” “high” “-preset” “medium” “-pix_fmt” “yuv420p” “-b:v” “0” “-map_metadata” “0” “-movflags” "use_metadata_tags+write_colr " “-map_metadata:s:v” “0:s:v” “-an” “-metadata” “videoai=Enhanced using ghq-5. Changed resolution to 3840x2160” “C:/Users/Dmitry/AppData/Local/Temp/896155477/previews/DJI_0029_1_ghq5.mp4”

Two things, the i9-9900K might be able to handle two at a time, but probably not more than that. I have a GTX 1060 and those speeds sound about right for a 1070 doing HD to UHD. TVAI is really heavy on the CPU, though Gaia uses the GPU more than the other models.

I can’t even use two GPUs, I get unknown error all the time and I do not know what the error is.

I get 0.1fps when using GPU.

From that command, it’s converting to tiff and nvenc h.264. That’s going to fill your drive C really fast, but it’s also doing a preview, so you should have enough space for that.
Does it work if you output to PNG or TIF instead of H.264?
Also, it’s got device set to 2, do you have the integrated Intel graphics enabled?

There is no way to upload an image here, but in the Device Manager under Display Adapters it only show 2 NVIDIA GeForce GTX 1070Ti.

You can paste images if you copy them from inside Paint. Let me check my other machine… it has an iGPU…

image

Yeah, mine doesn’t show the iGPU either, though, I probably made a point of disabling it in the BIOS.
I might be wrong, but I think device=0 is the main GPU, so device=1 should be another like your second 1070. I only think that because if I remote into my other computer, the command with device=0 fails. (I think because that changes device 0 to the remote emulated graphics.)

Sorry, that’s my only lead. I could be totally wrong and maybe device is only used to tell it to use the GPU or CPU.

any other suggestions on why there is unknown error when using All GPUs?

It does need a Monitor on each GPU.

Since Topaz is using DirectX/ML it does only work this way.

You could buy a monitor dongle when you dont have a second monitor.

I did have the same problem over a year ago when i had two RTX Quadro cards.

Thanks, do you have an idea which one to try?

unfortunately no

The need of Directx for TL Software is a little downside.

At the moment the Radeon Instinct is comparable cheap (3500€) and has a lot of FP16 (180 teraflops) power, but it lacks directx hardware, it has no monitor output.

And because of this you cant use it for TL Software.

Or maybe a virtual machine does the job.