MP 1,1-5,1 Activate AMD hardware acceleration

[itdk92]

Metal support is one thing, Metal acceleration is another. Adobe significantly added to their "Metal engine" months ago. It's good to see Apple finally doing something for their own software. This is intended for the MP7,1 but trickle down benefits a lot more.

It is not self evident to me that Metal acceleration was previously missing.

Here’s part of the official text:

“Today’s update to Final Cut Pro delivers a new Metal-based engine that improves playback and accelerates graphics tasks including rendering, real-time effects and exporting on Metal-compatible Mac computers. Video editors with 15-inch MacBook Pro, the most popular system for Final Cut Pro users, will benefit from performance that’s up to 20 percent faster, while editors using iMac Pro will see gains up to 35 percent.”

To me, that sounds like things got faster, not that acceleration was missing.

But that’s my interpretation - I can definitely see it is a bit open to other interpretations.

Although I do not se why they would talk about a “new” metal engine, if there wasn’t a previous one.

Not really caring about what Adobe is doing as this is FCPX related.

 

[crjackson2134]

I also updated FCPX to 10.4.7 and ran BruceX a few times (cMP spec as per my signature)

11s for exporting ProRes 422

18s for exporting H264

I am really happy with this update 😁

Seems to launch faster on my machine too... Nice update.

 

[antonlinus]

I also updated FCPX to 10.4.7 and ran BruceX a few times (cMP spec as per my signature)

11s for exporting ProRes 422

18s for exporting H264

I am really happy with this update 😁

I've read through a couple threads now but can't really a find definitive answer to my question:
considering all latest updates and and gpu acceleration are there actual real world performance differences between the Pulse RX580, the Vega 56 and the VII in a 5.1 MacPro when used mainly for video editing? I don't want to hardware mod my PSU but downvolting the VII seems easy enough and, as far as I understood, the Vega 56 and pulse 580 don't need Pixlas mod either. Also they all seem to be natively supported in macOS now if I'm correct. I could afford a VII but I'm not willing to pay close to double the price of a Vega 56 if the performance isn't significantly better either. I'm not interested in geekbench scores, I won't run windows on my Mac and I don't game either. I just can't find any comparable results with latest updates and drivers for these cards. are you willing to make a recommendation? Thanks I advance for the support!

 

[tommy chen]

but pixlas mode will be allways the better solution

 

[antonlinus]

but pixlas mode will be allways the better solution

Why? all 3 cards seem to run just fine on a single 6pin cable according to multiple posts in this forum.

 

[MIKX]

Why? all 3 cards seem to run just fine on a single 6pin cable according to multiple posts in this forum.

I don't think so.
Underpowered GPUs will shut down a cMP under stress : eg : Using Furmark benchmark.

The two motherboard Mini 6 pin sockets are the "basic" power for 2009 ~ 2012 cMPs.
SATA ports can be used for auxiliary power with a "2 X SATA to 6 pin " cables above that the Pixlas mod or an external PSU would be required

HD 5870 6 pin + 6 pin
HD 7950 8 pin + 6 pin
HD 7970 8 pin + 6 pin
RX280 8 pin + 6 pin

In my case I power my MSI Armor RX580 8gb this way . . I have an EVGs PowerLink in the power chain to the RX 580
( EVGa Powerlink https://www.amazon.com/EVGA-PowerLink-Support-Founders-600-PL-2816-LR/dp/B01MQ1YEYH
==============================================
Power input to EVGs Powerlink
1.
From 2 Mini 6 pins to PowerLink 8 pin ( 2 x Mini 6 pin to 8 pin cable )

2.
TWO x SATA to ONE x 6 pin to Powerlink
This way my RX 580 8 gb gets combined, balanced power from the Mini 6 pins + two SATA ports + the PCIe slot
===========================================
Power output to GPU
From Powerlink 8 pin out put to RX 580 8 pin socket.

I never experience shutdowns through GPU under-powering.

 

[itdk92]

Why? all 3 cards seem to run just fine on a single 6pin cable according to multiple posts in this forum.

Power distribution, and not wearing off connections. "Runs fine" is not really a valuable observation when physics are involved, especially in the long term.

 

[IanK MacPro]

I've read through a couple threads now but can't really a find definitive answer to my question:
considering all latest updates and and gpu acceleration are there actual real world performance differences between the Pulse RX580, the Vega 56 and the VII in a 5.1 MacPro when used mainly for video editing? I don't want to hardware mod my PSU but downvolting the VII seems easy enough and, as far as I understood, the Vega 56 and pulse 580 don't need Pixlas mod either. Also they all seem to be natively supported in macOS now if I'm correct. I could afford a VII but I'm not willing to pay close to double the price of a Vega 56 if the performance isn't significantly better either. I'm not interested in geekbench scores, I won't run windows on my Mac and I don't game either. I just can't find any comparable results with latest updates and drivers for these cards. are you willing to make a recommendation? Thanks I advance for the support!

I am also interested in what the differences are between these cards now apple seem to be utilising them better.

I have 3x 12 core MacPro 5.1's, all with 80gb Ram & Pixlas Mod.

2x of them are used for Cinema 4D work using Redshift and both are kitted out with 2x Nvidia 1080ti's and have performed flawlessly running on High Sierra.

My other machine is my main work machine with a Sapphire 580 Pulse which I use mainly for the Adobe suite along with Final Cut pro and Motion, this was also still running High Sierra as I had no reason to upgrade until this latest Final Cut Pro update.

Yesterday I installed the latest fresh Mojave install on a spare SSD to test, Final Cut Pro and Motion are now much faster!

Bruce X went from 25 seconds to 17 seconds.

I outputted another smallish Final Cut Pro project, the old version took 02:13 to export, new version 01:24

Apple Motion also showed a decent speed increase, a project which took 03:31 to export now takes 02:15.


So this now leaves me wondering how I can speed things up further ow that the software is utilising the hardware better, replace the Saphire 580 with a Vega 56 or VII?

And if either of these cards are better, would I see a further benefit from installing two of them (assuming blower style) and would the PSU be ok with 2x AMD cards? I know its fine with 2x 1080ti's and they're not exactly low power cards.

Any help greatly appreciated.

Ian

 

[MIKX]

IanK MacPro

WOW ! Three 5,1 Pixlas cMPs 🤑 . . .I'm insanely jealous.

h9826790 seems to have had notable success with his VII.

 

[goMac]

Metal support is one thing, Metal acceleration is another. Adobe significantly added to their "Metal engine" months ago. It's good to see Apple finally doing something for their own software. This is intended for the MP7,1 but trickle down benefits a lot more.

Metal requires changes for Catalina to support the Vega Duo. This is one thing Apple added, and it’s very unlikely Adobe has done Metal work for Catalina yet.

 

[itdk92]

I am also interested in what the differences are between these cards now apple seem to be utilising them better.

I have 3x 12 core MacPro 5.1's, all with 80gb Ram & Pixlas Mod.

2x of them are used for Cinema 4D work using Redshift and both are kitted out with 2x Nvidia 1080ti's and have performed flawlessly running on High Sierra.

My other machine is my main work machine with a Sapphire 580 Pulse which I use mainly for the Adobe suite along with Final Cut pro and Motion, this was also still running High Sierra as I had no reason to upgrade until this latest Final Cut Pro update.

Yesterday I installed the latest fresh Mojave install on a spare SSD to test, Final Cut Pro and Motion are now much faster!

Bruce X went from 25 seconds to 17 seconds.

I outputted another smallish Final Cut Pro project, the old version took 02:13 to export, new version 01:24

Apple Motion also showed a decent speed increase, a project which took 03:31 to export now takes 02:15.


So this now leaves me wondering how I can speed things up further ow that the software is utilising the hardware better, replace the Saphire 580 with a Vega 56 or VII?

And if either of these cards are better, would I see a further benefit from installing two of them (assuming blower style) and would the PSU be ok with 2x AMD cards? I know its fine with 2x 1080ti's and they're not exactly low power cards.

Any help greatly appreciated.

Ian

Yes, Mojave, Catalina and the new FCPX are better. Also, Radeon VII would be much better as well.

You could run two VEGA 56 depending on how well those Pixlas mods have been made, but I would not run two Radeon VII.

 

[itdk92]

Any news about HW accel and Catalina 10.15.0 and 10.15.1b?

 

[Ludacrisvp]

You could run two VEGA 56 depending on how well those Pixlas mods have been made, but I would not run two Radeon VII.

why would you not run 2xVII?
i seem to recall that @h9826790 is running his single VII without pixla setup and is undervolting it for better performance and lower power usage and more stable thermals.
one would have to think that you could run 2 of them (one pixla style and one without or both from pixla style) and be fine Especially if using undervolt kext.

 

[tsialex]

why would you not run 2xVII?
i seem to recall that @h9826790 is running his single VII without pixla setup and is undervolting it for better performance and lower power usage and more stable thermals.
one would have to think that you could run 2 of them (one pixla style and one without or both from pixla style) and be fine Especially if using undervolt kext.

If you use the mod to 240W with load balancing, it's 15W over what is considered a responsible usage of MP5,1 PCIe slot power plus AUX BOOST A+B. Some people with VEGA 56 have shutdowns when running benchmarks, not even Furmark, and it's only 236W. You will have to use the PowerPlay mod to 204W.

Guru3D Radeon VII review shows that loaded power consumption of Radeon VII is 299W for the GPU only, this is not safe to use without Pixla's mod when running without the PowerPlay kext mod. Dual Radeon VII without the 204W PowerPlay mod will get you very near of the total capacity of the MP5,1 power supply when considering total system usage.

 

[itdk92]

If you use the mod to 240W with load balancing, it's 15W over what is considered a responsible usage of MP5,1 PCIe slot power plus AUX BOOST A+B. Some people with VEGA 56 have shutdowns when running benchmarks, not even Furmark, and it's only 236W. You will have to use the PowerPlay mod to 204W.

Guru3D Radeon VII review shows that loaded power consumption of Radeon VII is 299W for the GPU only, this is not safe to use without Pixla's mod when running without the PowerPlay kext mod. Dual Radeon VII without the 204W PowerPlay mod will get you very near of the total capacity of the MP5,1 power supply when considering total system usage.

^This. Even undervolting might not bring the requirements down enough.

 

[Ludacrisvp]

I see. For some reason I thought the undervolting brought it down to 204 watts not 240watts.

 

[IanK MacPro]

If you use the mod to 240W with load balancing, it's 15W over what is considered a responsible usage of MP5,1 PCIe slot power plus AUX BOOST A+B. Some people with VEGA 56 have shutdowns when running benchmarks, not even Furmark, and it's only 236W. You will have to use the PowerPlay mod to 204W.

Guru3D Radeon VII review shows that loaded power consumption of Radeon VII is 299W for the GPU only, this is not safe to use without Pixla's mod when running without the PowerPlay kext mod. Dual Radeon VII without the 204W PowerPlay mod will get you very near of the total capacity of the MP5,1 power supply when considering total system usage.

it’s difficult to know what the real world power usage would be. The 1080TI isn’t far behind the Radeon VII on that chart and yet my two machines running dual 1080ti’s have been completely solid.

However I know that using redshift for rendering on them won’t pull anywhere near the maximum wattage on them, probably under 200w each in the real world.

As my other machine is for FCPX and motion work, I’m not sure if I had 2x Vega 56’s or 2x Radeon VII’s how much real world power FCPX etc would pull. I realise that Benchmarks like Furmark etc could push them over the edge but that never equates to real world usage in my experience.

Difficult to know without trying I suppose.

Also difficult to know how well FCPX etc will actually utilise a dual gpu setup without trying.

 

[VaZ]

Anyone know of a good app where it shows the frequency and speed of the GPU kinda like GPU-Z for Windows?
Wanted to compared the 2nd bios switch on my card to see if there's any difference.

 

[h9826790]

Anyone know of a good app where it shows the frequency and speed of the GPU kinda like GPU-Z for Windows?
Wanted to compared the 2nd bios switch on my card to see if there's any difference.

Which GPU?

 

[VaZ]

Which GPU?

GIGABYTE HD 7970 WindForce 3X OC - My assumption is one of the settings is the OC

 

[h9826790]

There is no GPU-Z equivalent in macOS.

However, if you know the exact GPU model, it should be very easy to get the ROM info TechPowerUp, and compare them.

For 7970, AFAIK, in most case, should be no difference between the two BIOS

 

[VaZ]

There is no GPU-Z equivalent in macOS.

However, if you know the exact GPU model, it should be very easy to get the ROM info TechPowerUp, and compare them.

For 7970, AFAIK, in most case, should be no difference between the two BIOS

Cool i'll check it out

 

[jscipione]

Definitely NOT natively. One report about this method also work in 10.13.6 https://forums.macrumors.com/threads/activate-amd-hardware-acceleration.2180095/#post-27365018
h9826790 Post #264 Jun 18, 2010

Does anyone know if this mod is confirmed to work with the AMD RX580 running under High Sierra 10.13.6? I need my Mac Pro to run High Sierra for some software. Thanks!
iluvmacs99 Post #358 Jul 23, 2019

Hi ! for compatibility reasons I need to stick with High Sierra. Does hardware acceleration work natively with 10.13.6 ? If not can I use this method ?
Rolotomasi0212 Post #263 Jun 18, 2019

would mean that the RX580 does not work under 10.13.6 I'll watch that when I have time
IndioX Post #21 May 15, 2019
Activate AMD hardware acceleration
AMD hardware acceleration works also in high sierra 10.13.6 with security update 2019-003 and bootROM 144.0.0.0.0 !?!


=========
I can confirm that the Lilu.kext mod definitely does not work on 10.13.6 on the same 5,1 + RX580 hardware that it does work on 10.14.6 encoding a BluRay video to h.264 VideoToolbox in Handbrake 1.2.2. On HIGH Sierra Handbrake freezes immediately and makes the entire computer unresponsive until you reboot while on Mojave the Encoding work fine.

 

[Fish1313]

Hey Folks,
thanks for all the helpful contributions in here.
I went with Lilu and WEG because of High Sierra and a Cuda eGPU.
If i test the jellyfish 10bit h265 all the video play smooth in VLC but the CPU went 500% every time!?
Checked with VideoProc states HEVC is not enabled
Why?!
The Vega56 is properly announced by the system but not in VideoProc.
Could anybody help?!

Thanks in advance
Sebastian

 

[devon807]

I also updated FCPX to 10.4.7 and ran BruceX a few times (cMP spec as per my signature)

11s for exporting ProRes 422

18s for exporting H264

I am really happy with this update 😁

I just re-ran BruceX as well and got 16 seconds Sapphire RX VEGA 64