Contribute
Register

[GUIDE] Injection of AMD Vega Power and Fan Control Properties

Joined
Jan 4, 2014
Messages
32
Motherboard
Gigabyte Z390 AORUS PRO
CPU
i9-9900K
Graphics
Vega 64
Mac
MacBook Air
Mobile Phone
Android
Just updated to 10.14.5, removed all VgTab fixes from config.plist except for PP_DisablePowerContainment = 1
But I got a significantly performance worsening under Geekbench 4 "Compute" benchmark in both OpenCL and Metal tests:

GPU-OpenCL Vega64 10.14.4 vs 10.14.5 score:
193029 / 139176

GPU-Metal Vega64 10.14.4 vs 10.14.5 score:
189589 / 135685

Maybe the powertable injections which I used until 10.14.4 were also giving me a big performance boost because of some overclocked settings? Should I put them back?
 
Joined
Jan 20, 2018
Messages
76
Motherboard
ASUS ROG MAXIMUS X HERO Z370
CPU
i7-8086K
Graphics
Vega 56 - flashed Sapphire Nitro+ 64
Mac
iMac, MacBook, MacBook Pro
Mobile Phone
iOS
Try Luxmark, of Heaven / Valley

I think Geekbench needs an update
 
Joined
Jul 13, 2016
Messages
826
Motherboard
ASUS ROG STRIX Z370-F
CPU
I5-8600K
Graphics
RX580
Mac
Mac mini, Mac Pro
Mobile Phone
iOS
Just updated to 10.14.5, removed all VgTab fixes from config.plist except for PP_DisablePowerContainment = 1
But I got a significantly performance worsening under Geekbench 4 "Compute" benchmark in both OpenCL and Metal tests:

GPU-OpenCL Vega64 10.14.4 vs 10.14.5 score:
193029 / 139176

GPU-Metal Vega64 10.14.4 vs 10.14.5 score:
189589 / 135685

Maybe the powertable injections which I used until 10.14.4 were also giving me a big performance boost because of some overclocked settings? Should I put them back?
That depends did you overclock your card in the power table injection?
 
Joined
Jan 4, 2014
Messages
32
Motherboard
Gigabyte Z390 AORUS PRO
CPU
i9-9900K
Graphics
Vega 64
Mac
MacBook Air
Mobile Phone
Android
Try Luxmark, of Heaven / Valley

I think Geekbench needs an update
Thanks, but I never used Luxmark before this update so I can't compare that much...
what might be a fair score with the OpenCL LuxBall HDR? I just got 30566
 

Attachments

Joined
Jan 4, 2014
Messages
32
Motherboard
Gigabyte Z390 AORUS PRO
CPU
i9-9900K
Graphics
Vega 64
Mac
MacBook Air
Mobile Phone
Android
That depends did you overclock your card in the power table injection?

Nope, I left all settings as VgTab app suggested for a Vega 64, except for default memory voltage set to 1100 instead of suggested 1350mv that was causing some freezes. 1100mv is the setting this card uses with the official Radeon drivers under Windows 10.

I feel it's like VgTab settings are better than Mojave's 10.14.5 default settings for a Vega 64...
 
Last edited:
Joined
Jan 25, 2017
Messages
18
Motherboard
Gigabyte Ga-Z270X-Ultra Gaming
CPU
I7-7700K
Graphics
Vega 64
Mobile Phone
iOS
Nope, I left all settings as VgTab app suggested for a Vega 64, except for default memory voltage set to 1100 instead of suggested 1350mv that was causing some freezes. 1100mv is the setting this card uses with the official Radeon drivers under Windows 10.

I feel it's like VgTab settings are better than Mojave's 10.14.5 default settings for a Vega 64...
More than likely, the VgTab stuff was designed around the windows driver design for the Founders Edition (top binned) cards. It didn't work right at all for my Sapphire Vega64 Nitro+ even after setting all the clocks to the default Windows driver settings that came with the card and 3 or 4 different implementation methods. In some ways it was worse using VGTab.

There's no way Apple is using the reference card design spec VRMs on their iMac Pro motherboards, the heat alone would make Mac soup. You know they're underclocking thier native Vega64 chips as much if not more than on their CPUs. I run my 7700k idle faster than theirs boosts; they simple don't have room for liquid cooling in their AIO design. So their power settings are going to correspond to the VRMs and cooling solution that Apple uses in their pancake computers. They might improve with time as OEM apple users start to adopt more eGPU options but who knows, Apple has never been about getting the maximum performance out of their hardware. They'll underclock and under-cool everything to make it fit in a slim-sexy casing. *I still think my RGB blinking lightshow case is sexier*

I know my card can run at 1660mhz native and upwards of a very nervous 1800mhz overclocked and undervolted in Windows. Once it's in the hackintosh it peaks at around 1400mhz; over the last week. I'm bouncing between a MacPro SMBIOs and 18,3 right now to see which gives me the best performance overall. I doubt I'll ever see Windows level overclocking speeds on the hackintosh without doing something with VgTab or the like. Right now I'm just very happy it's reporting 100% utilization, fans are running at a non-deafening reasonable rate, and my movie render times are down dramatically from the GTX1070. About 30% time savings give or take on a 20 minute movie render with 2 color/luma adjustment layers and a LUT layer, in FCPX. Also get much faster H.264 compression out of Movavi Video Converter (which is designed for CUDA,) so Compressor is in my near future.

For me that calculates to about 3-6 hours a week in time depending on how many videos I have to edit and render.

When I started hackintoshing I mostly gave up on getting 110% performance out of my components, I'm generally happy that my benchmark numbers both on graphics and processing have pretty much always obliterated the comparable machine apple has. If I ran Windows on the exact same box (I don't dual boot) I know I'd get better numbers than I do hackintoshing it. The difference between my i7-7700k and the mid-2017 iMac score is enough to keep me satiated.
 
Joined
Jul 13, 2016
Messages
826
Motherboard
ASUS ROG STRIX Z370-F
CPU
I5-8600K
Graphics
RX580
Mac
Mac mini, Mac Pro
Mobile Phone
iOS
More than likely, the VgTab stuff was designed around the windows driver design for the Founders Edition (top binned) cards. It didn't work right at all for my Sapphire Vega64 Nitro+ even after setting all the clocks to the default Windows driver settings that came with the card and 3 or 4 different implementation methods. In some ways it was worse using VGTab.

There's no way Apple is using the reference card design spec VRMs on their iMac Pro motherboards, the heat alone would make Mac soup. You know they're underclocking thier native Vega64 chips as much if not more than on their CPUs. I run my 7700k idle faster than theirs boosts; they simple don't have room for liquid cooling in their AIO design. So their power settings are going to correspond to the VRMs and cooling solution that Apple uses in their pancake computers. They might improve with time as OEM apple users start to adopt more eGPU options but who knows, Apple has never been about getting the maximum performance out of their hardware. They'll underclock and under-cool everything to make it fit in a slim-sexy casing. *I still think my RGB blinking lightshow case is sexier*

I know my card can run at 1660mhz native and upwards of a very nervous 1800mhz overclocked and undervolted in Windows. Once it's in the hackintosh it peaks at around 1400mhz; over the last week. I'm bouncing between a MacPro SMBIOs and 18,3 right now to see which gives me the best performance overall. I doubt I'll ever see Windows level overclocking speeds on the hackintosh without doing something with VgTab or the like. Right now I'm just very happy it's reporting 100% utilization, fans are running at a non-deafening reasonable rate, and my movie render times are down dramatically from the GTX1070. About 30% time savings give or take on a 20 minute movie render with 2 color/luma adjustment layers and a LUT layer, in FCPX. Also get much faster H.264 compression out of Movavi Video Converter (which is designed for CUDA,) so Compressor is in my near future.

For me that calculates to about 3-6 hours a week in time depending on how many videos I have to edit and render.

When I started hackintoshing I mostly gave up on getting 110% performance out of my components, I'm generally happy that my benchmark numbers both on graphics and processing have pretty much always obliterated the comparable machine apple has. If I ran Windows on the exact same box (I don't dual boot) I know I'd get better numbers than I do hackintoshing it. The difference between my i7-7700k and the mid-2017 iMac score is enough to keep me satiated.
If you had a reference card you might have better results, before 14.5 @pastrychef was running his Vega 56 with a large buff. The iMac pro runs at slower then stock speeds. The the refresh of the iMac pro run the Vega 64x at stock speeds. Not sure on the thermal throttling. I think his reduction in power created enough heat reduction to allow the card to run more efficient thus producing better benchmarks. Vegas are juiced up on power and get super hot it is a wonder they do not melt the slot on the board.
 

pastrychef

Moderator
Joined
May 29, 2013
Messages
9,017
Motherboard
Asus ROG Strix Z370-G Gaming (Wi-Fi AC)
CPU
i9-9900K OC'd @ 5.0GHz
Graphics
Vega 56
Mac
MacBook, Mac Pro
Classic Mac
iBook, Power Mac, PowerBook
Mobile Phone
iOS
Vega 56 tweaked with VGTab created kext.

Screen Shot 2019-05-22 at 8.13.37 AM.png
 
Top