Disappointing Vegas performance with new AMD RX6600XT

Comments

Kosta-Stanic wrote on 10/12/2022, 7:28 AM

Vegas PRO 20 subscription.

Playback with 7950X + 6950XT is great.

Playback with 5950X + 6950XT was aweful.

Is there a way to benchmark? I want to benchmark 6950XT now and 4090 in few days.

Former user wrote on 10/12/2022, 7:56 AM

@Kosta-Stanic @RogerS can point you to the benchmark test 👍

jetdv wrote on 10/12/2022, 8:18 AM

Perhaps the "red car" test?

https://www.vegascreativesoftware.info/us/forum/ok-where-s-all-the-red-car--127346/#ca792083

 

Former user wrote on 10/12/2022, 8:40 AM

AMD still the speed demon of H.265 hardware encoding

Kosta-Stanic wrote on 10/12/2022, 9:22 AM

AMD still the speed demon of H.265 hardware encoding

Thanks, this is what I needed.

Kosta-Stanic wrote on 10/12/2022, 9:27 AM

So it looks like 7950 XT might be best upgrade? AMD GPU this is coming in November.

RogerS wrote on 10/12/2022, 3:47 PM

Intel ARC did very well for HEVC transcodes here: https://techgage.com/article/intel-arc-a750-a770-workstation-review/ Not that I'd buy Intel at this point... too many driver issues.

To return to Vegas, here's a benchmark project. It would be great to get new numbers for the latest GPUs and see how they perform. Please see the instructions and form here and feel free to direct message me if you have questions or problems:

https://docs.google.com/forms/d/1Exbi4K3hbxw6snJuisR1ble-0tCPVNcIcNnx0BAtSIM/

Former user wrote on 10/12/2022, 5:11 PM

@Kosta-Stanic You should do that benchmark test with existing card and whatever you replace it with so you can compare the GPU processing power of the cards in Vegas. Transcoding such as what you see with that handbrake test is only testing the hardware encoder, it's not testing the processing power of a GPU which is how a $250 RX6600 can beat a $1600 RTX4090 in a handbrake transcode

Kosta-Stanic wrote on 10/13/2022, 3:07 AM

@Kosta-Stanic You should do that benchmark test with existing card and whatever you replace it with so you can compare the GPU processing power of the cards in Vegas. Transcoding such as what you see with that handbrake test is only testing the hardware encoder, it's not testing the processing power of a GPU which is how a $250 RX6600 can beat a $1600 RTX4090 in a handbrake transcode

I used to have problems with playback of HVEC 265 before on 5950X and 6950XT.

Yesterday I tested "Vegas pro subscription" with my new 7950X and 6950XT and I was working with 100gb size files that were HVEC 265. No problem what so ever. I was suprised.

Now I would rather wait for AMD GPU and see what to order. There is a Benchmark solution that @RogerS posted. That is to complicated for me, or I need some time to do that.

But I will get 4090 or 7950XT in november and I can do some benchmarks for comparing.

I will post results here.

What I care about is playback quality while editing big HVEC 265 files.

 

walter-i. wrote on 10/13/2022, 3:16 AM
There is a Benchmark solution that @RogerS posted. That is to complicated for me, or I need some time to do that.
Please see the instructions and form here and feel free to direct message me if you have questions or problems:

@Kosta-Stanic
I would take up the offer from @RogerS

Former user wrote on 10/13/2022, 4:54 AM

I used to have problems with playback of HVEC 265 before on 5950X and 6950XT.

Yesterday I tested "Vegas pro subscription" with my new 7950X and 6950XT and I was working with 100gb size files that were HVEC 265. No problem what so ever. I was suprised.

@Kosta-Stanic That is interesting, was the file using the GPU decoder?

Now I would rather wait for AMD GPU and see what to order. There is a Benchmark solution that @RogerS posted. That is to complicated for me, or I need some time to do that.

That is a problem, the perceived complexity, and also people are human, they make mistakes in testing and recording accurate data. If the user could run a script or benchmark executable such as the Puget Benchmarks for Davinci Resolve and Premiere Pro there would be many more results and the accuracy much higher

Kosta-Stanic wrote on 10/13/2022, 5:19 AM

Yes, I would easy do a benchmark that is simple today. But complexity of doing a benchmark that @RogerS asked for is kinda big.

Anyway I will spend few hours actualy doing it before I get 4090 or 7950XT.

Kosta-Stanic wrote on 10/13/2022, 5:22 AM

I used to have problems with playback of HVEC 265 before on 5950X and 6950XT.

Yesterday I tested "Vegas pro subscription" with my new 7950X and 6950XT and I was working with 100gb size files that were HVEC 265. No problem what so ever. I was suprised.

@Kosta-Stanic That is interesting, was the file using the GPU decoder?

Now I would rather wait for AMD GPU and see what to order. There is a Benchmark solution that @RogerS posted. That is to complicated for me, or I need some time to do that.

That is a problem, the perceived complexity, and also people are human, they make mistakes in testing and recording accurate data. If the user could run a script or benchmark executable such as the Puget Benchmarks for Davinci Resolve and Premiere Pro there would be many more results and the accuracy much higher

I had clean install of Vegas Pro Subscription and everything worked perfect. Playback quality editing 100GB 265 HVEC files was 0 problem. I was shocked. I think it have something to do with 7950X cpu.

I did same 1 month ago with 5950X cpu and it was not possible to edit 265 HVEC with playback. I had to downgrade or change 265 HVEC files before editing.

Reyfox wrote on 10/13/2022, 5:53 AM

@Kosta-Stanic I am surprised at the differences you've mentioned between the 5950X and the 7950x. It might also come down to the DDR5 RAM and motherboard throughput also compared to X570.

I am still considering upgrading to the 5950x, which should improve performance compared to my 3900x. the cost of going (and being an early adopter) of new hardware is something I prefer to wait a few months for BIOS/driver updates.

Newbie😁

Vegas Pro 22 (VP18-21 also installed)

Win 11 Pro always updated

AMD Ryzen 9 5950X 16 cores / 32 threads

32GB DDR4 3200

Sapphire RX6700XT 12GB Driver: 25.3.1

Gigabyte X570 Elite Motherboard

Panasonic G9, G7, FZ300

Kosta-Stanic wrote on 10/13/2022, 6:04 AM

Maybe it have to do something with DDR5 or 7950X. But I can finaly edit 265 HVEC with no trick or special options.

RogerS wrote on 10/13/2022, 6:51 AM

Yes, I would easy do a benchmark that is simple today. But complexity of doing a benchmark that  RogerS asked for is kinda big.

Anyway I will spend few hours actualy doing it before I get 4090 or 7950XT.

It's pretty simple. Download the project from the link I sent.

Highlight the two markers and hit play. Estimate average FPS (see under the preview window). Write it down.

Do a test rendering the whole timeline with Magix AVC and Mainconcept (CPU) in UHD and record the time. Settings are right on the form.

Do a test rendering of the whole timeline with Magix AVC with NVENC or VCE in UHD and record the time.
Settings are right on the form.

Enter the info into the form twice.

That's it.

 

Kosta-Stanic wrote on 10/14/2022, 3:11 PM

Here is someone who compared 4090.

Encoding is much better with AMD

Former user wrote on 10/14/2022, 4:17 PM

Difficult to work out what's going on there. It's not a transcode otherwise The 3050 would have the same render time as the other 30 series, so there is GPU processing, but what makes the 4090 faster then the 3090, from what I had heard dual encoders only works for HEVC and AV1, and H.264 is the same as 30 series, so the 4090 is actually processing faster. Maybe that comes down to the larger memory cache

 

Kosta-Stanic wrote on 10/24/2022, 5:09 PM

I have sent in numbers with 7950X, 64GB ram and 6950XT.

What I can say from expereince is that playback FPS is much better with 7950X or new platform.