How make really working Dual RTX GPUs

I have a question referred to the work with a Dual GPU setup. I can see in my trial version that both GPUs are recognized and working in 100%, but the treates frames FPS are not doubled with 2 GPUs in.
The conversion speed is the same with 1 than with 2 RTX GPUs.
If i dont get more speed with 2 GPU cards, why are both on 100%? Dont understand this, its equall what version i test it, even with v3 its the same.
Does TVAI not support multi GPU setup?

Have the same working speed on setting AUTO, then on 1 GPU or all GPUs.

And what is the benefit of using 1 or 2 or more instances?

CPU is Ryzen 9 7900. Thx in advance

With multiple instances you can process multiple videos at the same time. If you want to work on one video with two GPU simultanously you might let one instance work on the first half of the video and the 2nd instance on the 2nd half exporting the video to frames. Keep sure you select a different GPU for each instance.

1 Like

Hello, seen the video. Why you change the wide resolution, does it not function with original resolution?
Is png the best format to store the result? How to convert to video again?

Id like an HEVC 10bit file at last, how do i get there? Do i need other programs more?

Hello, I just changed the size because I wanted to make an upscale. It works with the original width and height too. png stores with lossless compression. That is a benefit over the jpg format but png makes larger files. If you are low on memory or storage space maybe use jpg instead. To make a video file out of the images after processing use the tool of your choice. Vegas Pro can do this for example.

And how is the rule to use an upscale mode, changing wide size?

What I find curious is that if I set TVAI’s preferences to “All GPUs,” running process will load both GPUs. And there is some effect on speed, because if you have multiple GPUs that don’t match, the processing speed will be somewhere in between what each will do individually. So what is the app doing with both GPUs on that setting?

Maybe it’s like SLI. You wouldn’t get 2X the frames per second. Somewhere around 1.4X to 1.6X though.

It seemed to me that a long time ago 1.x or 2.x worked that way with the RX550s I had at the time, but the folks at Topaz tell me it doesn’t, and mandeunamail up above says that processing is the same speed. So it’s a mystery.

I think the “All GPUs” setting is total crap.
When processing 4 videos (2 in beta / 2 in release version) in parallel, every video has a slightly different filesize.
No clue where this instability is comming from.
Doing the same thing with the beta program set to gpu 1 and the release version set to gpu 2 all 4 videos have the exact same filesize.
The processing speed is about the same in both cases…In fact with “All GPUs” the total system load is about 1660W and about 1850W with each program running on just 1 GPU, so a bit faster.
It really feels like both GPUs just process the same thing without any benefits.

Yes thats the point, on ALL GPU, even the IGpu of my CPU is recognized and trottle al 100% (so 3 GPUs working) while encoding, but the FPS are identical, its the same to have preset AUTO or ALL GPUs or 1 GPU. But all are at 100%, so i dont understand if all are working 100%, why the speed dont change. What are they doing on full trottle?
If this not works, why is the option availible? It seems its only to waste more energy …
Make your own tests, if you consider interesting to report also

I plugged all the spare GPUs I had around the house into one workstation, and here’s what I got (Artemis benchmark only to save space):

Topaz Video AI  v4.2.2
System Information
OS: Windows v11.23
CPU: Intel(R) Xeon(R) Gold 6132 CPU @ 2.60GHz  190.71 GB
GPU: AMD Radeon (TM) Pro WX 7100 Graphics  7.9451 GB
Processing Settings
device: 0 vram: 0.8 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	01.66 fps 	2X: 	00.91 fps 	4X: 	00.26 fps 	
Topaz Video AI  v4.2.2
System Information
OS: Windows v11.23
CPU: Intel(R) Xeon(R) Gold 6132 CPU @ 2.60GHz  190.71 GB
GPU: AMD Radeon Pro W5700  7.9596 GB
Processing Settings
device: 2 vram: 0.8 instances: 0
Input Resolution: 1920x1080
Benchmark Results

Artemis		1X: 	04.86 fps 	2X: 	02.47 fps 	4X: 	01.02 fps 	
Topaz Video AI  v4.2.2
System Information
OS: Windows v11.23
CPU: Intel(R) Xeon(R) Gold 6132 CPU @ 2.60GHz  190.71 GB
GPU: AMD Radeon PRO W6600  7.9558 GB
Processing Settings
device: 1 vram: 0.8 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	04.83 fps 	2X: 	02.52 fps 	4X: 	00.96 fps 	
Topaz Video AI  v4.2.2
System Information
OS: Windows v11.23
CPU: Intel(R) Xeon(R) Gold 6132 CPU @ 2.60GHz  190.71 GB
GPU: AMD Radeon (TM) Pro WX 7100 Graphics  7.9451 GB
GPU: AMD Radeon PRO W6600  7.9558 GB
GPU: AMD Radeon Pro W5700  7.9596 GB
Processing Settings
device: 3 vram: 0.8 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	09.03 fps 	2X: 	05.25 fps 	4X: 	01.75 fps 	

All CPUs does seem to boost FPS significantly, at least for AMD GPU benchmarks.

Next I tried running a simple Artemis HQ 1:1 on a 1080p video.

WX7100: 1.0 fps
W5700: 4.7 fps
W6600: 4.9 fps
All CPUs: 7.9 fps

There is a speed boost, but it’s smaller than what the benchmark saw. I’d get better results by splitting the video in two and running two instances with each having one GPU all to itself (as the folks at Topaz recommend).

Interesting test, but i see instances zero?
Also would be interesting same GPUs with a modern CPU, think would be much faster.
The IPC of older XEONs are moderate compare to modern sistems, only a personal notice.

This is your CPU
https://www.cpubenchmark.net/cpu.php?cpu=Intel+Xeon+Gold+6132+@+2.60GHz&id=3227

This is my CPU
https://www.cpubenchmark.net/cpu.php?cpu=AMD+Ryzen+9+7900&id=5167

It’s actually this:
https://www.cpubenchmark.net/cpu.php?cpu=Intel+Xeon+Gold+6132+%40+2.60GHz&id=3227&cpuCount=2
TVAI benchmark doesn’t display second CPU on dual processor workstations but TVAI does use both of them.
I think these GPUs would just bottleneck a newer CPU. They’re hitting 95-100% utilization with these two now.
And I’ve never figured out what the instances number is supposed to mean…

1 Like