Single Radeon VII faster than dual Radeon VII ?!

Get answers to your questions about color grading, editing and finishing with DaVinci Resolve.
  • Author
  • Message
Offline

SylvestCorp

  • Posts: 87
  • Joined: Sat Feb 01, 2020 8:19 pm
  • Real Name: Hans Christian Sylvest

Re: Single Radeon VII faster than dual Radeon VII ?!

PostFri Mar 06, 2020 3:03 pm

Single card ok for resolve but dual Radeons VII is a no go. End of this thread.


Sent from my iPhone using Tapatalk
Offline

Carsten Sellberg

  • Posts: 1084
  • Joined: Fri Jun 16, 2017 9:13 am

Re: Single Radeon VII faster than dual Radeon VII ?!

PostFri Mar 06, 2020 5:05 pm

SylvestCorp wrote: Single card ok for resolve but dual Radeons VII is a no go.


Hi.

I am sorry, but I am not sure that I agree with you. Yes dual Radeons VII with Windows currently
have some problems. I don't know if the problem is with Windows or Resolve?

But Lucius Snow wrote in this thread, that he is running dual Radeons VII on CentOS 8.1. Pretty stable with AMD GPU drivers 19.50 for Linux.

And Barefeats run VII DUO in his Apple 2019 Mac Pro:

https://barefeats.com/mac-pro-2019-four-gpus.html

So I will expect it also will run without problems in a Hackintosh.

But you are right. For the moment is dual Radeons VII in Windows a no go.

Regards Carsten.
URSA Mini 4.6K
Offline

mclarenf1

  • Posts: 18
  • Joined: Fri Jan 03, 2020 6:16 pm
  • Real Name: Philip Coltart

Re: Single Radeon VII faster than dual Radeon VII ?!

PostFri Mar 06, 2020 5:31 pm

I have found that two cards make minimal difference but here is what I use the second card for. NeatVideo. I set NeatVideo to one card and Davinci to the other card. I have done lots of variations and that yields the best results. I can put 6k H265 on a 4K timeline with color grading and NeatVideo applied to almost max settings. I get around 7ps export. Pretty stinking good for NeatVideo. Two cards in Davinci with NeatVideo applied causes crashes. Split them up and it's solid and fast.
Ryzen 3960x ThreadRipper
64GB 3600 Memory
Dual Radeon Pro VII 16GB
Titan Ridge
Dual LG UltraFine 5k Monitors
TRX40 Aorus Master MB
BeQuiet cooling tower
Geekbench 5 21,350
Cinibench R20 14,000
Offline

SylvestCorp

  • Posts: 87
  • Joined: Sat Feb 01, 2020 8:19 pm
  • Real Name: Hans Christian Sylvest

Re: Single Radeon VII faster than dual Radeon VII ?!

PostSat Mar 07, 2020 7:29 am

Was contemplating the same. I disable the one when using resolve
In windows 10 pro. Sorry I agree that CentOS looks to be an option. And for any other tasks they are fine. I am only disappointed I can’t get it running. I use windows because of other applications and I have no clue regarding Linux.

For puget bench 8k I found that r3d had to be on to complete.




Sent from my iPhone using Tapatalk
Offline

Carsten Sellberg

  • Posts: 1084
  • Joined: Fri Jun 16, 2017 9:13 am

Re: Single Radeon VII faster than dual Radeon VII ?!

PostSat Mar 07, 2020 9:21 am

SylvestCorp wrote: I use windows because of other applications and I have no clue regarding Linux.


Hi.

I think the solution can be a DUAL BOOT, where you can run some applications in Windows and Resolve in Linux or Hackintosh. If your knowledge are better for MacOS than for Linux, can you consider a Hackintosh?

Regards Carsten.
URSA Mini 4.6K
Offline

Sulo Kokki

  • Posts: 202
  • Joined: Thu Sep 28, 2017 11:39 pm

Re: Single Radeon VII faster than dual Radeon VII ?!

PostSat Mar 07, 2020 10:22 am

Um, you don't "just" turn a computer into a Hackintosh (and under the hood, macOS and Linux are very much alike; UNIX genealogy ;) ). While virtualization and GPU passthrough have vastly increased overall stability and ease of upgrade, it's still a bugger to build, and it takes some research, select components and a good deal of testing.

Worth it? You betcha. :mrgreen:
Linux Mint 19.1 | DaVinci Resolve Studio 16.1.2 | 2700x 32gb Vega 56 | macOS Mojave
Offline

Frank Engel

  • Posts: 439
  • Joined: Thu May 17, 2018 11:09 pm
  • Real Name: Frank Engel

Re: Single Radeon VII faster than dual Radeon VII ?!

PostTue Mar 10, 2020 12:57 pm

Carsten Sellberg wrote:
SylvestCorp wrote:If your knowledge are better for MacOS than for Linux, can you consider a Hackintosh?


Installing macOS on non-Apple hardware violates the EULA and therefore is illegal.
Offline

Glenn Venghaus

  • Posts: 1200
  • Joined: Wed Jan 01, 2014 9:56 pm
  • Location: Amsterdam , The Netherlands

Re: Single Radeon VII faster than dual Radeon VII ?!

PostTue Mar 10, 2020 6:29 pm

Image
Beatstep & APC-40 Resolve Edition Custom Controllers https://posttools.tachyon-consulting.com
Test rig : Dual Xeon | UNRAID VM OSX10.13/14 | 48 vcore | GTX980ti/Vega 56 | 128GB | DRS 16.x
V: SDI > HDlink > Eizo + Scopebox
A: 5.1RME-FireFace
Offline

SylvestCorp

  • Posts: 87
  • Joined: Sat Feb 01, 2020 8:19 pm
  • Real Name: Hans Christian Sylvest

Re: Single Radeon VII faster than dual Radeon VII ?!

PostSat May 23, 2020 10:07 am

I was completly wrong !!!!!! latest gaming drivers mgpu set to on. In resolve deselect the none display card.
Could anyone confirm the Candelmark 15-17 FPS 66 nodes with dual "Radeon VII" some driver improvements have already helped the non pro card.

puget system benchmark results are absolutely massive:

-----System Specs-----
Computer Name: SYLVESTCORP
CPU: AMD Ryzen Threadripper 3960X 24-Core Processor
GPU: AMD Radeon VII (26.20.15029.27016)~AMD Radeon VII (26.20.15029.27016)
RAM: 64GB
Storage: Force MP600 (X:)~NVMe Samsung SSD 970 SCSI Disk Device (C:)~ST4000LM024-2AN17V (Y:)
OS Version: Windows 10 Pro (Ver. 1909)
----------

Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K,Overall Score,1420,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K Test Average,Basic Grade,197.2,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K Test Average,OpenFX - Lens Flare + Tilt-Shift Blur + Sharpen,117.4,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K Test Average,Temporal NR - Better 2 Frames,135.6,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K Test Average,3x Temporal NR - Better 2 Frames,72.8,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K Test Average,Optimized Media,187.1,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K CinemaRAW Light,Codec Average,106,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K H264 150Mbps 8bit,Codec Average,125.1,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K ProRes 422,Codec Average,155.4,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K ProRes 4444,Codec Average,155.1,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K RED,Codec Average,168.5,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K CinemaRAW Light,Basic Grade,64.81,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K CinemaRAW Light,OpenFX - Lens Flare + Tilt-Shift Blur + Sharpen,20.17,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K CinemaRAW Light,Temporal NR - Better 2 Frames,21.17,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K CinemaRAW Light,3x Temporal NR - Better 2 Frames,0,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K CinemaRAW Light,Optimized Media,128.63,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K H264 150Mbps 8bit,Basic Grade,92.77,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K H264 150Mbps 8bit,OpenFX - Lens Flare + Tilt-Shift Blur + Sharpen,23.23,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K H264 150Mbps 8bit,Temporal NR - Better 2 Frames,23.99,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K H264 150Mbps 8bit,3x Temporal NR - Better 2 Frames,0,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K H264 150Mbps 8bit,Optimized Media,151.34,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K ProRes 422,Basic Grade,106.62,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K ProRes 422,OpenFX - Lens Flare + Tilt-Shift Blur + Sharpen,24.82,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K ProRes 422,Temporal NR - Better 2 Frames,25.71,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K ProRes 422,3x Temporal NR - Better 2 Frames,9.54,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K ProRes 422,Optimized Media,260.95,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K ProRes 4444,Basic Grade,64.00,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K ProRes 4444,OpenFX - Lens Flare + Tilt-Shift Blur + Sharpen,23.59,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K ProRes 4444,Temporal NR - Better 2 Frames,22.16,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K ProRes 4444,3x Temporal NR - Better 2 Frames,9.06,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K ProRes 4444,Optimized Media,143.70,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K RED,Basic Grade,46.47,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K RED,OpenFX - Lens Flare + Tilt-Shift Blur + Sharpen,15.12,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K RED,Temporal NR - Better 2 Frames,20.00,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K RED,3x Temporal NR - Better 2 Frames,8.90,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),4K RED,Optimized Media,110.57,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K,Overall Score,1432,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K Test Average,Basic Grade,215,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K Test Average,OpenFX - Lens Flare + Tilt-Shift Blur + Sharpen,127.8,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K Test Average,Temporal NR - Better 2 Frames,138.6,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K Test Average,3x Temporal NR - Better 2 Frames,59.5,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K Test Average,Optimized Media,175,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K H265 100Mbps to 8K,Codec Average,105.2,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K H265 100Mbps to UHD,Codec Average,109.6,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K RED to 8K,Codec Average,157.8,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K RED to UHD,Codec Average,200.2,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K H265 100Mbps to 8K,Basic Grade,35.63,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K H265 100Mbps to 8K,OpenFX - Lens Flare + Tilt-Shift Blur + Sharpen,5.73,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K H265 100Mbps to 8K,Temporal NR - Better 2 Frames,7.72,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K H265 100Mbps to 8K,3x Temporal NR - Better 2 Frames,0,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K H265 100Mbps to 8K,Optimized Media,61.33,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K H265 100Mbps to UHD,Basic Grade,55.46,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K H265 100Mbps to UHD,OpenFX - Lens Flare + Tilt-Shift Blur + Sharpen,22.88,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K H265 100Mbps to UHD,Temporal NR - Better 2 Frames,21.52,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K H265 100Mbps to UHD,3x Temporal NR - Better 2 Frames,8.91,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K H265 100Mbps to UHD,Optimized Media,61.31,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K RED to 8K,Basic Grade,25.54,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K RED to 8K,OpenFX - Lens Flare + Tilt-Shift Blur + Sharpen,5.15,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K RED to 8K,Temporal NR - Better 2 Frames,7.25,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K RED to 8K,3x Temporal NR - Better 2 Frames,0,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K RED to 8K,Optimized Media,62.70,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K RED to UHD,Basic Grade,55.46,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K RED to UHD,OpenFX - Lens Flare + Tilt-Shift Blur + Sharpen,22.52,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K RED to UHD,Temporal NR - Better 2 Frames,21.52,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K RED to UHD,3x Temporal NR - Better 2 Frames,8.91,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),8K RED to UHD,Optimized Media,62.70,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),Fusion,Overall Score,1296,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),Fusion Test Average,Score,129.6,Score
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),Fusion_3DBacklitText,Reference FPS,7.06,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),Fusion_PhoneComposite,Reference FPS,0,FPS
Puget Systems DaVinci Resolve Studio Benchmark V0.6 BETA (DR 16.2.2.11),Fusion_TurbulantParticles,Reference FPS,3.75,FPS
Offline

SylvestCorp

  • Posts: 87
  • Joined: Sat Feb 01, 2020 8:19 pm
  • Real Name: Hans Christian Sylvest

Re: Single Radeon VII faster than dual Radeon VII ?!

PostSat May 23, 2020 6:46 pm

20.4.2
Now with duall Radeons with mgpu perform as one in candle mark and some tests in puget 0.61 resolve bench mark are x3. And stable.


Sent from my iPhone using Tapatalk
Offline

joncoy

  • Posts: 80
  • Joined: Wed Jun 21, 2017 4:27 pm
  • Location: Columbus, OH

Re: Single Radeon VII faster than dual Radeon VII ?!

PostSat May 23, 2020 6:56 pm

SylvestCorp wrote:20.4.2
Now with duall Radeons with mgpu perform as one in candle mark and some tests in puget 0.61 resolve bench mark are x3. And stable.


Sent from my iPhone using Tapatalk


Okay, that's what I thought. I'm going to try this driver with your suggestions and report back. I'm also running dual Radeon VII on Windows 10 Pro 1909.
Offline

joncoy

  • Posts: 80
  • Joined: Wed Jun 21, 2017 4:27 pm
  • Location: Columbus, OH

Re: Single Radeon VII faster than dual Radeon VII ?!

PostSat May 23, 2020 8:04 pm

With a complex grade of mine that uses three OFX plugins (including 3 frames of TNR set to Better), 5.7K braw clip on a 2K timeline, my playback improves from 13-14.5fps to 20-21fps when I enable MGPU and uncheck the second Radeon card in Resolve.

Strange.
Offline

SylvestCorp

  • Posts: 87
  • Joined: Sat Feb 01, 2020 8:19 pm
  • Real Name: Hans Christian Sylvest

Single Radeon VII faster than dual Radeon VII ?!

PostSun May 24, 2020 6:43 am

Last time I checked it was a sure crash. Cancel mark and some other benchmarks are better or the same as a single card with the latest drivers. Retested my rtx 2080 and there is no change, so must be and drivers.
When you now look at the util of the gpus they sometimes both go to 99% which never happened before.

Sent from my iPhone using Tapatalk
Offline

MishaEngel

  • Posts: 1300
  • Joined: Wed Aug 29, 2018 12:18 am
  • Real Name: Misha Engel

Re: Single Radeon VII faster than dual Radeon VII ?!

PostSun May 24, 2020 12:07 pm

Frank Engel wrote:
Carsten Sellberg wrote:
SylvestCorp wrote:If your knowledge are better for MacOS than for Linux, can you consider a Hackintosh?


Installing macOS on non-Apple hardware violates the EULA and therefore is illegal.


A lot is illegal, that why companies like Apple(eavesdropping on siri https://www.forbes.com/sites/kateoflahertyuk/2019/07/28/apple-siri-eavesdropping-puts-millions-of-users-at-risk/, https://www.idropnews.com/news/5-apple-scandals-youll-never-forget/38414/, etc...) are filthy rich(and they also don't pay their fair share of taxes with the help of the goverments from the US, Netherlands and Ireland).
Offline
User avatar

Marc Wielage

  • Posts: 6693
  • Joined: Fri Oct 18, 2013 2:46 am
  • Location: Hollywood, USA

Re: Single Radeon VII faster than dual Radeon VII ?!

PostMon May 25, 2020 1:34 am

MishaEngel wrote:A lot is illegal, that why companies like Apple(eavesdropping on siri https://www.forbes.com/sites/kateoflahertyuk/2019/07/28/apple-siri-eavesdropping-puts-millions-of-users-at-risk/, https://www.idropnews.com/news/5-apple-scandals-youll-never-forget/38414/, etc...) are filthy rich(and they also don't pay their fair share of taxes with the help of the goverments from the US, Netherlands and Ireland).

Actually, that was true until 2018 when Apple brought most of their $250 billion in overseas cash back to the U.S., and they agreed to pay $38 billion dollars in income tax:

Why a $38 Billion Tax Payment Is a Good Deal for Apple
https://www.wired.com/story/why-a-dolla ... for-apple/

Apple is, in fact, the largest U.S. corporate taxpayer in the world, paying far more American taxes than Amazon or Netflix or Microsoft or any of the various oil companies.

Here's Apple's statement on the matter:

The facts about Apple’s tax payments
https://www.apple.com/newsroom/2017/11/ ... -payments/

I dislike Apple as a company for a lot of reasons, and I don't think they're generally on the user's side, but in this specific case I think they're acting honorably. I buy and own a lot of Apple products, but not a day goes by when I don't curse them -- loudly -- for one thing or another.
marc wielage, csi • VP/color & workflow • chroma | hollywood
Offline

Carsten Sellberg

  • Posts: 1084
  • Joined: Fri Jun 16, 2017 9:13 am

Re: Single Radeon VII faster than dual Radeon VII ?!

PostMon May 25, 2020 3:58 am

joncoy wrote: my playback improves from 13-14.5fps to 20-21fps when I enable MGPU and uncheck the second Radeon card in Resolve.

Strange.


Hi.

First will I congratulate you with your findings.

Yes, AMD had fixed a lot of issues in their GPU drivers. Here is a link from March:

'AMD Driver Update: Company Responds to Complaints With a Whopping 40 Fixes'

https://www.tomshardware.com/news/amd-d ... ts-3579197

Quote: 'The team acknowledges that there is still work to be done' and here is a link to the latest AMD Radeon VII Driver, the 20.4.2 :

https://www.amd.com/en/support/graphics ... radeon-vii

But in a blog AMD writes: 'Often, many problems can be resolved with a clean installation of our driver. In order to perform a clean installation, you must select the “Factory Reset” option at the beginning of the installation process.' Here is where to click for the “Factory Reset” option:

Image

Else did I in a YouTube Video see an idea, that DDU in safe mode not always work, and a rekommandation of, to reinstall Windows if you go from a nVidea driver and to and AMD driver.


And finally will I suggest you try to help AMD with their good work. And if you still experiencing any issue, and you can consistently reproduce it, then fill one of their new bug reports:

https://www.feedback.amd.com/se/5A1E27D27812944E

Regards Carsten.
URSA Mini 4.6K
Offline

SylvestCorp

  • Posts: 87
  • Joined: Sat Feb 01, 2020 8:19 pm
  • Real Name: Hans Christian Sylvest

Re: Single Radeon VII faster than dual Radeon VII ?!

PostMon May 25, 2020 5:17 am

Carsten Sellberg wrote:
joncoy wrote: my playback improves from 13-14.5fps to 20-21fps when I enable MGPU and uncheck the second Radeon card in Resolve.

Strange.


Hi.

First will I congratulate you with your findings.

Yes, AMD had fixed a lot of issues in their GPU drivers. Here is a link from March:

'AMD Driver Update: Company Responds to Complaints With a Whopping 40 Fixes'

https://www.tomshardware.com/news/amd-d ... ts-3579197

Quote: 'The team acknowledges that there is still work to be done' and here is a link to the latest AMD Radeon VII Driver, the 20.4.2 :

https://www.amd.com/en/support/graphics ... radeon-vii

But in a blog AMD writes: 'Often, many problems can be resolved with a clean installation of our driver. In order to perform a clean installation, you must select the “Factory Reset” option at the beginning of the installation process.' Here is where to click for the “Factory Reset” option:

Image

Else did I in a YouTube Video see an idea, that DDU in safe mode not always work, and a rekommandation of, to reinstall Windows if you go from a nVidea driver and to and AMD driver.


And finally will I suggest you try to help AMD with their good work. And if you still experiencing any issue, and you can consistently reproduce it, then fill one of their new bug reports:

https://www.feedback.amd.com/se/5A1E27D27812944E

Regards Carsten.
I always do a clean install. :-) hoping for more improvements as the Radeon pro VII comes to marked. Just happy that two cards are now the same or faster. I am still back on a single rtx 2080 as it is still faster for me in general. Just could not resist testing it.


Sent from my iPhone using Tapatalk
Offline

MishaEngel

  • Posts: 1300
  • Joined: Wed Aug 29, 2018 12:18 am
  • Real Name: Misha Engel

Re: Single Radeon VII faster than dual Radeon VII ?!

PostMon May 25, 2020 12:07 pm

Marc Wielage wrote:Why a $38 Billion Tax Payment Is a Good Deal for Apple
https://www.wired.com/story/why-a-dolla ... for-apple/

Apple is, in fact, the largest U.S. corporate taxpayer in the world, paying far more American taxes than Amazon or Netflix or Microsoft or any of the various oil companies.

Here's Apple's statement on the matter:

The facts about Apple’s tax payments
https://www.apple.com/newsroom/2017/11/ ... -payments/

I dislike Apple as a company for a lot of reasons, and I don't think they're generally on the user's side, but in this specific case I think they're acting honorably. I buy and own a lot of Apple products, but not a day goes by when I don't curse them -- loudly -- for one thing or another.


They still use the Irish/Dutch sandwich and $38 billion (with their profit margins and revenue's) is not a lot of money when you haven't paid near to nothing for the last 20 years. Big corp don't pay taxes, they evade it and they get all the help they can get from western goverments US, UK, Netherlands, Ireland, Qatar, etc..
Offline

joncoy

  • Posts: 80
  • Joined: Wed Jun 21, 2017 4:27 pm
  • Location: Columbus, OH

Re: Single Radeon VII faster than dual Radeon VII ?!

PostMon May 25, 2020 12:46 pm

Carsten Sellberg wrote:
joncoy wrote: my playback improves from 13-14.5fps to 20-21fps when I enable MGPU and uncheck the second Radeon card in Resolve.

Strange.


Hi.

First will I congratulate you with your findings.

Yes, AMD had fixed a lot of issues in their GPU drivers. Here is a link from March:

'AMD Driver Update: Company Responds to Complaints With a Whopping 40 Fixes'

https://www.tomshardware.com/news/amd-d ... ts-3579197

Quote: 'The team acknowledges that there is still work to be done' and here is a link to the latest AMD Radeon VII Driver, the 20.4.2 :

https://www.amd.com/en/support/graphics ... radeon-vii

But in a blog AMD writes: 'Often, many problems can be resolved with a clean installation of our driver. In order to perform a clean installation, you must select the “Factory Reset” option at the beginning of the installation process.' Here is where to click for the “Factory Reset” option:

Image

Else did I in a YouTube Video see an idea, that DDU in safe mode not always work, and a rekommandation of, to reinstall Windows if you go from a nVidea driver and to and AMD driver.


And finally will I suggest you try to help AMD with their good work. And if you still experiencing any issue, and you can consistently reproduce it, then fill one of their new bug reports:

https://www.feedback.amd.com/se/5A1E27D27812944E

Regards Carsten.


Hi Carsten,

thanks for checking in on this. I have never performed a factory reset with the graphics driver since I always use DDU before updating. Based on your recommendations from AMD, I cleaned the drivers again with DDU and then installed the latest 20.4.2 drivers and also clicked "Factory Reset."

My system seems more stable, and when I test a complex grade with 3 frames of TNR on a 2K timeline with 5.7K braw footage, I get 100% playback, and both of my GPUs display about 30% workload in Windows Task Manager. I even get 100% playback with 4 or 5 frames on some clips. This is an enormous leap in performance for my system!
Offline

joncoy

  • Posts: 80
  • Joined: Wed Jun 21, 2017 4:27 pm
  • Location: Columbus, OH

Re: Single Radeon VII faster than dual Radeon VII ?!

PostMon May 25, 2020 1:08 pm

I should also note that I'm NOT currently using MGPU in the Radeon Settings, and I have BOTH GPUs selected under my Resolve preferences.

I also get better performance with Resolve's UI, such as loading OFX Plugins for the first time and loading Stills for performing comparison wipes. These types of actions used to present a delay the first time I would use them in a session.
Offline

joncoy

  • Posts: 80
  • Joined: Wed Jun 21, 2017 4:27 pm
  • Location: Columbus, OH

Re: Single Radeon VII faster than dual Radeon VII ?!

PostMon May 25, 2020 3:54 pm

joncoy wrote:I should also note that I'm NOT currently using MGPU in the Radeon Settings, and I have BOTH GPUs selected under my Resolve preferences.

I also get better performance with Resolve's UI, such as loading OFX Plugins for the first time and loading Stills for performing comparison wipes. These types of actions used to present a delay the first time I would use them in a session.


I retract parts of this. I WAS having snappy response times regarding applying OFX plugins to nodes and opening stills for the first time, but now, without changes, I'm back to having to wait about three seconds before a still will load for the first time, and about the same time for some OFX plugins when first applied. Still better than things used to be.

Playback is still excellent with TNR, but the snappiness that I saw in the UI isn't quite there.

Going to play around with settings. :roll:
Offline

MishaEngel

  • Posts: 1300
  • Joined: Wed Aug 29, 2018 12:18 am
  • Real Name: Misha Engel

Re: Single Radeon VII faster than dual Radeon VII ?!

PostMon May 25, 2020 4:43 pm

joncoy wrote:
joncoy wrote:I should also note that I'm NOT currently using MGPU in the Radeon Settings, and I have BOTH GPUs selected under my Resolve preferences.

I also get better performance with Resolve's UI, such as loading OFX Plugins for the first time and loading Stills for performing comparison wipes. These types of actions used to present a delay the first time I would use them in a session.


I retract parts of this. I WAS having snappy response times regarding applying OFX plugins to nodes and opening stills for the first time, but now, without changes, I'm back to having to wait about three seconds before a still will load for the first time, and about the same time for some OFX plugins when first applied. Still better than things used to be.

Playback is still excellent with TNR, but the snappiness that I saw in the UI isn't quite there.

Going to play around with settings. :roll:


Slashcam.de has an explenation for this:

Die Performance unter DaVinci Resolve 16

Unsere früher erwähnten Probleme, bei denen Resolve mit AMD-Karten immer lange Umschaltzeiten zwischen den Räumen benötigt ist mittlerweile aus der Welt. Dafür fiel gegenüber Nvidia Karten auf, dass die Clips bei AMDs Treiber irgendwie noch einmal gecached werden, bevor sie ruckelfrei von der Grafikkarte abgespielt werden. Und so bleiben Vorschaubilder im Programm beim Start erst einmal ein paar Sekunden leer. Wir führen dieses Verhalten mittlerweile auf unterschiedlichen Code zwischen OpenCL und CUDA zurück. Denn noch ein weiteres Speicher-Verhalten fiel uns auf: Während unser 8K Test-Projekt im CUDA-Modus mit einer RTX2080Ti trotz 11 GB DDR6-RAM immer wieder wegen Speicherüberläufen abbricht, passiert dies mit der 6GB-AMD Karte seltsamerweise nicht. Diese agiert zwar deutlich träger, aber bricht nicht wegen vollem Speicher die Bearbeitung ab. Das Speichermenangement unter OpenCL dürfte daher grundsätzlich anders funktionieren als unter CUDA.


Google Translate:

The performance under DaVinci Resolve 16

Our previously mentioned problems, in which Resolve with AMD cards always requires long switchover times between rooms, is now out of the world. However, compared to Nvidia cards, it was noticed that the clips in AMD's drivers were somehow cached again before they were played smoothly on the graphics card. And so preview images in the program remain empty for a few seconds at first. We attribute this behavior to different code between OpenCL and CUDA. Because we noticed another memory behavior: While our 8K test project in CUDA mode with an RTX2080Ti keeps aborting due to memory overflows despite 11 GB DDR6 RAM, this doesn't happen with the 6GB AMD card. Although this is much slower, it does not abort processing due to full memory. The storage management under OpenCL should therefore function fundamentally differently than under CUDA.


https://www.slashcam.de/artikel/Test/AMD-XFX-Radeon-RX-5600-XT-Grafikkarte---Durchschnittliche-Mittelklasse--XFX-Radeon-RX-5600-XT-Thicc-II---Die-P.html
Offline

joncoy

  • Posts: 80
  • Joined: Wed Jun 21, 2017 4:27 pm
  • Location: Columbus, OH

Re: Single Radeon VII faster than dual Radeon VII ?!

PostMon May 25, 2020 6:03 pm

MishaEngel wrote:
joncoy wrote:
joncoy wrote:I should also note that I'm NOT currently using MGPU in the Radeon Settings, and I have BOTH GPUs selected under my Resolve preferences.

I also get better performance with Resolve's UI, such as loading OFX Plugins for the first time and loading Stills for performing comparison wipes. These types of actions used to present a delay the first time I would use them in a session.


I retract parts of this. I WAS having snappy response times regarding applying OFX plugins to nodes and opening stills for the first time, but now, without changes, I'm back to having to wait about three seconds before a still will load for the first time, and about the same time for some OFX plugins when first applied. Still better than things used to be.

Playback is still excellent with TNR, but the snappiness that I saw in the UI isn't quite there.

Going to play around with settings. :roll:


Slashcam.de has an explenation for this:

Die Performance unter DaVinci Resolve 16

Unsere früher erwähnten Probleme, bei denen Resolve mit AMD-Karten immer lange Umschaltzeiten zwischen den Räumen benötigt ist mittlerweile aus der Welt. Dafür fiel gegenüber Nvidia Karten auf, dass die Clips bei AMDs Treiber irgendwie noch einmal gecached werden, bevor sie ruckelfrei von der Grafikkarte abgespielt werden. Und so bleiben Vorschaubilder im Programm beim Start erst einmal ein paar Sekunden leer. Wir führen dieses Verhalten mittlerweile auf unterschiedlichen Code zwischen OpenCL und CUDA zurück. Denn noch ein weiteres Speicher-Verhalten fiel uns auf: Während unser 8K Test-Projekt im CUDA-Modus mit einer RTX2080Ti trotz 11 GB DDR6-RAM immer wieder wegen Speicherüberläufen abbricht, passiert dies mit der 6GB-AMD Karte seltsamerweise nicht. Diese agiert zwar deutlich träger, aber bricht nicht wegen vollem Speicher die Bearbeitung ab. Das Speichermenangement unter OpenCL dürfte daher grundsätzlich anders funktionieren als unter CUDA.


Google Translate:

The performance under DaVinci Resolve 16

Our previously mentioned problems, in which Resolve with AMD cards always requires long switchover times between rooms, is now out of the world. However, compared to Nvidia cards, it was noticed that the clips in AMD's drivers were somehow cached again before they were played smoothly on the graphics card. And so preview images in the program remain empty for a few seconds at first. We attribute this behavior to different code between OpenCL and CUDA. Because we noticed another memory behavior: While our 8K test project in CUDA mode with an RTX2080Ti keeps aborting due to memory overflows despite 11 GB DDR6 RAM, this doesn't happen with the 6GB AMD card. Although this is much slower, it does not abort processing due to full memory. The storage management under OpenCL should therefore function fundamentally differently than under CUDA.


https://www.slashcam.de/artikel/Test/AMD-XFX-Radeon-RX-5600-XT-Grafikkarte---Durchschnittliche-Mittelklasse--XFX-Radeon-RX-5600-XT-Thicc-II---Die-P.html


Thanks Misha, I kind of figured as much regarding these issues. This seems to line up with what BMD’s tech support told me in terms of stability on OpenCL and AMD cards and speed on CUDA/Nvidia.
Offline

SylvestCorp

  • Posts: 87
  • Joined: Sat Feb 01, 2020 8:19 pm
  • Real Name: Hans Christian Sylvest

Re: Single Radeon VII faster than dual Radeon VII ?!

PostTue May 26, 2020 9:27 am

Fyi- i am out of the Dual Radeon VII investigation as my 12 year Old needed a card for 300 fps Minecraft . :-)


Sent from my iPhone using Tapatalk
Offline

joncoy

  • Posts: 80
  • Joined: Wed Jun 21, 2017 4:27 pm
  • Location: Columbus, OH

Re: Single Radeon VII faster than dual Radeon VII ?!

PostTue May 26, 2020 1:04 pm

Can anyone explain why going into the Radeon Settings for 20.4.2 and resetting the shader cache under global graphics would make all OFX plugins, when applied to a node, display their effect immediately rather than the usual couple-second delay I get?

Is it really just about how OpenCL caches information? I know all of this probably isn't even worth the headaches, but at this point, i really want to know why I can get really snappy performance in Resolve only when I fool around with weird knobs and buttons.

*cue headache*
Offline

josuelems

  • Posts: 1
  • Joined: Wed May 27, 2020 6:53 pm
  • Real Name: Josue Lemus

Re: Single Radeon VII faster than dual Radeon VII ?!

PostWed May 27, 2020 6:53 pm

Hi Guys,

Been following this thread with the same proble, dual radeon 7 never got more than 14-17 fps in render nor real time playback during editing, also while rendering the system never used more than 10% of second card nor 50% of first one.

I tried every single driver with clean install since the 19x version up the the lastest 20x version, re-install microsoft and nothing, Radeon settings did not overcloked my GPU but one.

The solution:
UNINSTALL DRIVER AND CHECKED THE OPTION TO PREVENT WINDOWS UPDATE TO DONWLOAD THE NEW DRIVER.

(I'm using the Radeon Pro Drivers which I believe are more power efficient, seems like the gaming drivers drain more power which I noticed with my 1000w and 750w power supply, The gaming drivers don't work with my 750w but the Pro Drivers does)

Uninstall Resolve (clean)
Remove Blackmagic Design Folders from
C:/Program Files
C:/ Program Files 86
C:/Users/YourUser/AppData/Roaming
Regedit
HKEY_CURRENT_USER
HKEY_LOCAL_MACHINE
Under Software Option
And any other folder related to resolve.

The problem with no achieving real time playback with resolve is due the miss configuration or manual tweeking in both resolve and Radeon Settings.

0. Make sure your system is up to date
1. Install your Radeon Pro Drivers 2020Q2 (again you could you the regular gaming drivers but Pro are more power efficient).
2. Restart
3. Install Resolve
4. In Resolve/Preference check Use display GPU for compute & leave in auto
REsolve will use your second GPU (the one with non monitor conected to render at 100% and use the main GPU to share its compute power, this one will use 50% of load)
5. Decode Options / Decode h.264 / h.265 with AMD
6 Do not overclock or tweek anything in Radeon Settings,
(fan noise will adjust pretty good and will not ramp up too much during render time


My results:
Real Time 24fps with H.264 GH5 LongGOP 3 Power Windows, Noise Reduction, blur & grain NO DROPPED FRAMES!!!!
Export for 7 MInutes Video

3 Power Windows
TNR 2 Better, at 5.5%
Film Grain
GBlur
2 Luts applied
Heavy Grade

RENDER TIME 4K UHD = 8:37 more or less at 22fps avarage
150% both gpu load.

for 1080p same proyect took 1:33 AT 112fps avarage.



I DON'T OVERCLOCK NOR CHANGE WITH MANUAL SETTINGS ANYMORE.
Offline

joncoy

  • Posts: 80
  • Joined: Wed Jun 21, 2017 4:27 pm
  • Location: Columbus, OH

Re: Single Radeon VII faster than dual Radeon VII ?!

PostThu May 28, 2020 4:20 pm

Update:

Updating my install of Resolve Studio 16.2.2 build 011 to the sneaky new 012 version fixed the initializing lag I was having between Resolve and my Tangent Wave 2 panel as well as the initializing lag I was having any time I added an OFX plugin to a node for the first time in a grading session.

I still have the slow media loading issue that will most likely never go away on my build, but my Radeon VIIs play back LOTS of OFX and TNR together in real time beautifully, so I'm going to stop complaining for now and NOT CHANGE A SINGLE THING.
Offline

joncoy

  • Posts: 80
  • Joined: Wed Jun 21, 2017 4:27 pm
  • Location: Columbus, OH

Re: Single Radeon VII faster than dual Radeon VII ?!

PostThu May 28, 2020 7:20 pm

joncoy wrote:Update:

Updating my install of Resolve Studio 16.2.2 build 011 to the sneaky new 012 version fixed the initializing lag I was having between Resolve and my Tangent Wave 2 panel as well as the initializing lag I was having any time I added an OFX plugin to a node for the first time in a grading session.

I still have the slow media loading issue that will most likely never go away on my build, but my Radeon VIIs play back LOTS of OFX and TNR together in real time beautifully, so I'm going to stop complaining for now and NOT CHANGE A SINGLE THING.


CORRECTION: After shutting down the system, starting back up, opening Resolve again (build 012) and opening a project, my system is back to taking a while to initialize my first move on the Tangent panel as well as the first time using an OFX plugin on a serial node.

Honestly, I feel like the combination of Resolve 16 and my hardware/OS config is turning into pure garbage.
Offline

ylnk01

  • Posts: 9
  • Joined: Mon Aug 12, 2019 7:54 am
  • Real Name: Kumar Yajamanam

Re: Single Radeon VII faster than dual Radeon VII ?!

PostFri May 29, 2020 2:01 am

@Josue Lemus's solution above worked for me :) . My dual Radeon VII with Davinci Resolve Studio 16.2.2 is snappy, stable and is working great with Radeon Pro 20q2 drivers. The dual Radeon VII setup does generate enough heat that I can save money in winter :D So, it appears that we need to install the drivers first (after thorough cleanup of Resolve and old drivers) and only then install Resolve. I followed those steps from Josue and so far ... so good!!!!
Offline

joncoy

  • Posts: 80
  • Joined: Wed Jun 21, 2017 4:27 pm
  • Location: Columbus, OH

Re: Single Radeon VII faster than dual Radeon VII ?!

PostFri May 29, 2020 2:59 am

ylnk01 wrote:@Josue Lemus's solution above worked for me :) . My dual Radeon VII with Davinci Resolve Studio 16.2.2 is snappy, stable and is working great with Radeon Pro 20q2 drivers. The dual Radeon VII setup does generate enough heat that I can save money in winter :D So, it appears that we need to install the drivers first (after thorough cleanup of Resolve and old drivers) and only then install Resolve. I followed those steps from Josue and so far ... so good!!!!


I'll try the above solution, though I'm skeptical for my system. Are you on the 20.Q2 Pro drivers released 5/13/2020?
Offline

joncoy

  • Posts: 80
  • Joined: Wed Jun 21, 2017 4:27 pm
  • Location: Columbus, OH

Re: Single Radeon VII faster than dual Radeon VII ?!

PostFri May 29, 2020 3:25 am

ylnk01 wrote:@Josue Lemus's solution above worked for me :) . My dual Radeon VII with Davinci Resolve Studio 16.2.2 is snappy, stable and is working great with Radeon Pro 20q2 drivers. The dual Radeon VII setup does generate enough heat that I can save money in winter :D So, it appears that we need to install the drivers first (after thorough cleanup of Resolve and old drivers) and only then install Resolve. I followed those steps from Josue and so far ... so good!!!!


Updated to Radeon Pro 20q2 Enterprise drivers--no real change in terms of media loading nor in terms of "snappiness" of OFX initialization. Most OFX like FilmConvert Nitrate, Glow, Lens Reflections, etc take about 3-7 seconds to initialize after adding to a node. In other words, it takes about 3-7 seconds for me to be able to adjust the parameters in real time. After initializing a couple OFX plugins, all of the rest seem to initialize immediately. As well, if I switch to another project from my projects manager (AFTER performing these initialization actions), my project's media opens instantly and the UI is snappy and responsive.

So here it is--Resolve on my system is simply annoying until I punch a few buttons on my Tangent panel and apply a few OFX plugins to "initialize" it all, and after that things work just fine. Playback is also excellent.

I don't know...maybe I'm complaining about nothing at this point.
Offline
User avatar

Marc Wielage

  • Posts: 6693
  • Joined: Fri Oct 18, 2013 2:46 am
  • Location: Hollywood, USA

Re: Single Radeon VII faster than dual Radeon VII ?!

PostFri May 29, 2020 10:16 am

MishaEngel wrote:They still use the Irish/Dutch sandwich and $38 billion (with their profit margins and revenue's) is not a lot of money when you haven't paid near to nothing for the last 20 years. Big corp don't pay taxes, they evade it and they get all the help they can get from western goverments US, UK, Netherlands, Ireland, Qatar, etc..

You are mistaken. Apple got rid of the Double Irish with a Dutch Sandwich some years ago:

https://fortune.com/2018/01/18/apple-bo ... xes-trump/

For those who want to know what this legal tax avoidance practice is, here's the explanation:

https://dukeundergraduatelawmagazine.or ... -sandwich/

The list of companies that funnel money through Panama and the Caribbean is staggering.

Again, the recent 2018 article revealed that Apple agreed to bring their overseas profits back into the U.S., expand their American factories, pay their taxes, and employ more Americans. I'm not saying they aren't an evil corporation -- I'm saying they now pay U.S. taxes, far more than Microsoft or Dell or HP or Amazon or Netflix. There are degrees of evil in business. Again: I generally love their products, hate the company.

In the meantime, back to the topic: I'm very curious if the Radeon Pro Duo GPU will run as fast two Radeon Pro Vega II's. I hope to find out and report back in a few weeks.
marc wielage, csi • VP/color & workflow • chroma | hollywood
Offline

joncoy

  • Posts: 80
  • Joined: Wed Jun 21, 2017 4:27 pm
  • Location: Columbus, OH

Re: Single Radeon VII faster than dual Radeon VII ?!

PostFri May 29, 2020 11:18 am

Marc Wielage wrote:
In the meantime, back to the topic: I'm very curious if the Radeon Pro Duo GPU will run as fast two Radeon Pro Vega II's. I hope to find out and report back in a few weeks.


I’m very curious about this as well. Pro Vega II looks very interesting.
Offline

joncoy

  • Posts: 80
  • Joined: Wed Jun 21, 2017 4:27 pm
  • Location: Columbus, OH

Re: Single Radeon VII faster than dual Radeon VII ?!

PostFri May 29, 2020 1:57 pm

ylnk01 wrote:@Josue Lemus's solution above worked for me :) . My dual Radeon VII with Davinci Resolve Studio 16.2.2 is snappy, stable and is working great with Radeon Pro 20q2 drivers. The dual Radeon VII setup does generate enough heat that I can save money in winter :D So, it appears that we need to install the drivers first (after thorough cleanup of Resolve and old drivers) and only then install Resolve. I followed those steps from Josue and so far ... so good!!!!


Following up on my install of Radeon Pro 20q2 enterprise drivers, I also re-enabled HBCC for both cards, which fixed my lagging initializing of OFX plugins and Tangent panel completely. They all function immediately after applying to a node, and the panel responds immediately.

Playback performance is still excellent. Really happy about this.
Previous

Return to DaVinci Resolve

Who is online

Users browsing this forum: Google [Bot] and 70 guests