Asus X299 - Support

[mm2margaret]

So, with 10.14.6 Beta 1 testing, a slight improvement with audio glitch and scratch with the Radeon VII. Not perfect, but so far, better....of course, ymmv.....I'm using Audio Device ID 7.....haven't tested a lot yet.....

 

[izo1]

It's fast to boot - really fast. I'm just tidying up my configs now, but it wasn't a huge amount of work to move over from Clover - maybe an evening.

Do you see purple lines during boot with 10.14.5

 

[izo1]

So, with 10.14.6 Beta 1 testing, a slight improvement with audio glitch and scratch with the Radeon VII. Not perfect, but so far, better....of course, ymmv.....I'm using Audio Device ID 7.....haven't tested a lot yet.....

Keep sending Apple this bug along with purple lines via Feedback App. Just don’t attach any system diag files. They got back to me on one of them. Just mention it’s eGPU on a imac Pro or something with a Radeon vii

 

[flmmkr]

That actually doesn't look too bad. In that other benchmarking site by Puget, they saw performance especially in Resolve that surpassed even the RTX 2080Ti.

I posted some rough results earlier on here before a CPU swap here. (This was in Resolve 16 Public Beta).

You should open up the Activity Monitor and open the GPU window to see if the debayering process is 100% on the GPU instead of the CPU.

Also you should try a regular playback in the REDCINE-X PRO. It's highly optimized for RED footage even though Resolve and Premiere are not, you can see where we are in terms of current gen GPUs (Especially with the GPU and CPU utlization via Activity Monitor).

The 2080Ti with CUDA under Windows can do 8k realtime playback of RED RAW footage. The Radeon VII needs a lot of optimizations and it still won't be there, but 4k/6k should be good, at least with half res playback on 6k.

Yeah, I looked at Puget's results and was kind of expecting higher performance than what I'm getting. However, they're testing in Windows and with a different CPU. Different computer and OS, so different results. And when I really think about how much better the 1080TI was than the Vega FE I tried, I'm not surprised that the 2080TI is still faster than Radeon VII. AMD still needs to catch up, but they are getting better. I'm happy with it now that I can be on the current OS version and have at least the same performance (and actually a little better) than before.

I typically don't use RedCine-X Pro, so I didn't bother testing with it before, but I just tried it. Results are similar. With no adjustments and just using camera metadata I don't get quite realtime playback at Full or 1/2, but I do get realtime at 1/4. Full is like Full in Resolve, and 1/2 is like Half Good in Resolve. I don't know what Half Premium is in Resolve. Same results when I do make some color correction adjustments. So I'm not really noticing anything different in RedCine-X Pro vs. Resolve.

Resolve does use the RED SDK, so the performance shouldn't be that much different.

Here is the CPU and GPU activity from Activity Monitor during playback at Full with some correction:

I did see the CPU meters go up to nearly 100% on all cores/threads during the playback.

 

[tonyarnold]

Do you see purple lines during boot with 10.14.5

Yes, I do. There’s a little bit of boot text right before the FileVault login, too, but I haven’t had time to debug that much yet.

 

[juantrix]

Did someone load AppleALC kext for a S1220 of Rampage VI Extreme? (Tried 3 latest version with id 7 with no look).

 

[djlild7hina]

Did someone load AppleALC kext for a S1220 of Rampage VI Extreme? (Tried 3 latest version with id 7 with no look).

Try 2

 

[izo1]

Yeah, I looked at Puget's results and was kind of expecting higher performance than what I'm getting. However, they're testing in Windows and with a different CPU. Different computer and OS, so different results. And when I really think about how much better the 1080TI was than the Vega FE I tried, I'm not surprised that the 2080TI is still faster than Radeon VII. AMD still needs to catch up, but they are getting better. I'm happy with it now that I can be on the current OS version and have at least the same performance (and actually a little better) than before.

I typically don't use RedCine-X Pro, so I didn't bother testing with it before, but I just tried it. Results are similar. With no adjustments and just using camera metadata I don't get quite realtime playback at Full or 1/2, but I do get realtime at 1/4. Full is like Full in Resolve, and 1/2 is like Half Good in Resolve. I don't know what Half Premium is in Resolve. Same results when I do make some color correction adjustments. So I'm not really noticing anything different in RedCine-X Pro vs. Resolve.

Resolve does use the RED SDK, so the performance shouldn't be that much different.

Here is the CPU and GPU activity from Activity Monitor during playback at Full with some correction:
View attachment 406188
I did see the CPU meters go up to nearly 100% on all cores/threads during the playback.

If during playback CPU is still hitting at 100% then the debayering process is still tasked to the CPU rather than GPU.

Take a look at this video and notice how during his test, when NVIDIA GPU is enabled, it goes to 100% GPU load and much much less CPU load. I get they are different operating systems, but something to benchmark with when comparing performance.

There needs to be optimizations done by either Red or Apple...

Radeon VII is nowhere near the 2080Ti performance. It's closer to 2080 and loses especially in 3d performance. However compute wise it's an interesting beast.

GCN is a terribly optimized architecture and AMD just slaps on more memory bandwidth instead of optimizing their architecture like NVIDIA does.

They went to 7nm and the perf is still worse than Turing 12nm from NVIDIA.

It's also worth testing the same thing in Windows just to see how bad the perf hit is in macOS...I will give it a shot soon, as I'm curious too.

Yes, I do. There’s a little bit of boot text right before the FileVault login, too, but I haven’t had time to debug that much yet.

Seems like a macOS issue related to AMD then. Many users are reporting the same issue.

 

[eiriasemrys]

I downloaded the 3 Helium 8K S35 24fps clips. City/Moon shot Redcode 8:1 is #1, Animal shot Redcode 7:1 is #2, City traffic shot Redcode 7:1 is #3. Resolve 15.3.1, 8K timeline, image scaling/output scaling set to HD for playback. Camera Raw settings set to Camera Metadata.

Playback with Full Res Debayer, with or without basic correction:
#1 12.5 fps
#2 10.5 fps
#3 10 fps

Playback with Half Res Premium Debayer, with or without basic correction:
#1 12.5fps
#2 10.5fps
#3 10fps

Yes, Full Res and Half Res Premium were the same. I even went back and ran both again.

Playback with Half Res Good Debayer, with or without basic correction:
#1 19 fps
#2 15 fps
#3 14.5 fps

Playback with Quarter Res Debayer, with or without basic correction:
#1 23.976 fps
#2 23.976 fps
#3 23.976 fps

Playback with Full Res Debayer, with basic correction & TNR:
#1 6.5 fps
#2 6.5 fps
#3 6.5 fps

Playback with Half Res Premium Debayer, with basic correction & TNR:
#1 7.5 fps
#2 7.5 fps
#3 7.5 fps

Playback with Half Res Good Debayer, with basic correction & TNR:
#1 8 fps
#2 8 fps
#3 8 fps

Playback with Quarter Res Debayer, with basic correction & TNR:
#1 11 fps
#2 11 fps
#3 11 fps

With the TNR on I did hear the GPU fans get louder than I ever heard the 1080TI. Kind of a whining sound because of the blower type of cooler. Typically I would leave the TNR off if possible while doing any actual grading (adding the TNR at the end) so the noise wouldn't bother me anyway since that would be something I leave on only for final rendering. That is unless a shot is really noisy, but then to get the realtime playback you have to render cache anyway, in which case playback with render cache wouldn't tax the GPU as much. Either way, I don't see the noise being a huge issue. If a model of Radeon VII comes out with a different cooling system that isn't as noisy then maybe I would upgrade, but I don't see it as a big deal right now.

Quarter Res Debayer looked great on my HD color grading monitor and on my projector (as it should), so I'd have no problem leaving it set at Quarter Res Debayer to enjoy the realtime playback while grading, then turn scaling output to match timeline when doing the final render.

Now for the 4K timeline (no one has been asking me for 8K outputs).

Playback with Full Res Debayer, with or without basic correction:
#1 19 fps
#2 14.5 fps
#3 14.5 fps

Note, there's really no reason for Full Res Debayer of 8K footage on a 4K timeline, but I wanted to see the numbers anyway.

Playback with Half Res Premium Debayer, with or without basic correction:
#1 19 fps
#2 14.5 fps
#3 14.5 fps

Playback with Half Res Good Debayer, with or without basic correction:
#1 19 fps
#2 15 fps
#3 14.5 fps

This time Full Res, Half Res Premium, and Half Res Good were the same. Ran tests twice to be sure.

Playback with Quarter Res Debayer, with or without basic correction:
#1 23.976 fps
#2 23.976 fps
#3 23.976 fps

Playback with Full Res Debayer, with basic correction & TNR:
#1 11.5 fps (fluctuated)
#2 11 fps
#3 12 fps

Playback with Half Res Premium Debayer, with basic correction & TNR:
#1 ~16.5 fps (fluctuated)
#2 ~13.5 fps (fluctuated)
#3 ~13 fps (fluctuated)

Playback with Half Res Good Debayer, with basic correction & TNR:
#1 ~17 fps (fluctuated)
#2 ~14 fps (fluctuated)
#3 ~13.5 fps (fluctuated)

Playback with Quarter Res Debayer, with basic correction & TNR:
#1 23.976 fps
#2 23.976 fps
#3 23.976 fps

Rendering the 4K timeline to a 4K ProRes 422HQ .mov results in 8.5 fps at Half Res Premium debayer. Full Res Debayer was 7.5 fps. That is with correction and TNR.

With the 4K timeline, the fan on the Radeon VII did not get loud like with the 8K timeline. With Red 8K footage I'd probably work in a 4K timeline (my clients right now are always 4K outputs or lower) and would use Quarter Res Debayer until final output, and that would even allow me to do some NR. I'm pretty happy with that.

I believe Resolve may be doing some auto debayer settings with Red footage, and it must be based on the timeline or on the output scaling settings. Otherwise I don't know why the results of the full res and half res premium debayer settings are always the same.

Thanks for the speed tests!! That’s awesome info. That is a head scratcher on both of those being the same. Any difference in metal vs opencl?

 

[mm2margaret]

Keep sending Apple this bug along with purple lines via Feedback App. Just don’t attach any system diag files. They got back to me on one of them. Just mention it’s eGPU on a imac Pro or something with a Radeon vii

@izo1 - I respond much better to polite requests than I do orders from you....