Ryzen 7000 benchmarks out

https://www.pugetsystems.com/labs/artic ... -Gen-2362/

" LongGOP (H.264/HEVC) - Intel Core with Quick Sync is the stronger option (unless you are using a flavor without hardware decoding support in DaVinci Resolve, in which case, it often will be bad no matter what hardware you have).

Intraframe (ProRes, DNx, etc.) - AMD Ryzen and Intel Core should be very comparable at similar price points, with only the Ryzen 7950X showing an appreciable performance lead over Intel.

RAW (BRAW, RED, etc.) - AMD has a large 22-35% performance lead with the Ryzen 7900X and 7950X. Lower-end models will be comparable between Intel and AMD. "

Sad, Nvidia 40 series didn't add support for h265 4:2:2. And while Ryzen now has added graphics to all of their CPU's, it is basically worthless and you still want an Nvidia card.

QuickSync is basically iGPU hardware encoding which has it's drawbacks when it comes to image quality.

Most of the encoding will be done by the dGPU's hardware encoders from Nvidia or AMD... i don't really see a point in using the CPU's iGPU encoders for that matter.

You either go with full software encoding which uses the full CPU power or dGPU encoding which uses the Nvidia/AMD GPU encoders.

Then again you have your particular need for h265 4:2:2 which is a very solid issue for people needing that.

Looking into their raw benchmark data I see nothing suggesting that Intel's Quick Sync is of any advantage...

Also the puget Resolve benchmark is a bit too heavy sided on exporting for my taste and most of the rest depends on the GPU - so I don't know if this is really the way to assess CPU performance for DVR work...
It's difficult to benchmark here, I'd say.

And if you look at their Premiere Pro benchmark you can clearly see that there must be some Premiere code problem to be ironed out with the new CPU - the playback speed drop in multicam h264 test for example where the 7950X drops to 17.52 fps while the 5950X gets 45.49 fps, the 5900X gets 59.43 fps and the 5600X tops out at 59.54 fps. Looks like Adobe needs to optimize their code.

QuickSync is basically iGPU hardware encoding which has it's drawbacks when it comes to image quality.

Most of the encoding will be done by the dGPU's hardware encoders from Nvidia or AMD... i don't really see a point in using the CPU's iGPU encoders for that matter.

You either go with full software encoding which uses the full CPU power or dGPU encoding which uses the Nvidia/AMD GPU encoders.

Then again you have your particular need for h265 4:2:2 which is a very solid issue for people needing that.

I am only interested in hardware decode. I export to dnxhr.

I believe the specifications for decoders assures that any given file will be decoded identically regardless of if it was done on Mac, Nvidia, etc.


My 5800h laptop can software decode 4k H265 422 from my Canon r5 at almost 48 fps. So timeline work is a bear. :'(

Looking into their raw benchmark data I see nothing suggesting that Intel's Quick Sync is of any advantage...

Also the puget Resolve benchmark is a bit too heavy sided on exporting for my taste and most of the rest depends on the GPU - so I don't know if this is really the way to assess CPU performance for DVR work...
It's difficult to benchmark here, I'd say.

And if you look at their Premiere Pro benchmark you can clearly see that there must be some Premiere code problem to be ironed out with the new CPU - the playback speed drop in multicam h264 test for example where the 7950X drops to 17.52 fps while the 5950X gets 45.49 fps, the 5900X gets 59.43 fps and the 5600X tops out at 59.54 fps. Looks like Adobe needs to optimize their code.

The advantage of quicksync is shown here.

https://www.pugetsystems.com/labs/artic ... udio-2122/

If you edit h265, 4:2:2 Intel 11/12 gen and M1/M2 mac are the only game in town for hardware acceleration.

Prores, Raw, Dnxhr, yep Quicksync is meanlingless.

Export to h264/265. Any GPU can do that, and no doubt as speeds faster than your project is rendering, so not a bottleneck.


I agree with your assessment that the benchmark cares more about GPU performance than it does CPU. But I would say that is consistent with my experience actually editing in Resolve. My exports the GPU is at 100% and the CPU at 30%. So if I got a faster CPU, I would not expect any change in performance.


You see a much bigger spread in scores on the GPU comparisons, as I expect.

https://www.pugetsystems.com/labs/artic ... ance-1990/


If I was making money at Resolve, I would be way more concerned about what GPU I have than CPU.

Looking forward to tests with the ARC770 which may be a good companion for the AMD 7xxx series if games are not on your requirements and the hardware decode /encode are the top of the list.