Contribute
Register

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

That is a great score for a Vega 56. Here is mine for the same card. Mine is overclocked as well, it is the Gigabyte OC. I am using the iMac Pro SMBIOS
 

Attachments

  • Screenshot 2019-05-22 at 13.44.19.jpg
    Screenshot 2019-05-22 at 13.44.19.jpg
    490.1 KB · Views: 117
That is a great score for a Vega 56. Here is mine for the same card. Mine is overclocked as well, it is the Gigabyte OC. I am using the iMac Pro SMBIOS

In my testing, I have found that the most important thing is to keep the card cool enough so that it never throttles. The max temps my card ever reaches is 75C but it's rare and, most of the time, it maxes around 72C.

Secondly, the I have found that the most gains in performance come from HBM clocks, not core clock.
 
In my testing, I have found that the most important thing is to keep the card cool enough so that it never throttles. The max temps my card ever reaches is 75C but it's rare and, most of the time, it maxes around 72C.

Secondly, the I have found that the most gains in performance come from HBM clocks, not core clock.
My card never runs hot. I get 50c-53c on idle with the fans off. Stressed I get 75c-76c as an absolute maximum. Well done, great score.
 
I have the Sapphire Vega 64 Liquid Cooled edition card,

With the 10.14.5 update it is still hitting the max 1750Mhz die clock speed and 945Mhz Memory Clock without using VGTab or SoftPowerPlay Table injection so i don't think Apple is under clocking 3rd Party OEM Vega cards.

The Fan/Pump runs much lower at idle ~900RPM, maxes out at 2400 RPM / 55 Degrees C under furmark stress test.

Cheers
Jay
Hi Jay.

I have an XFX Vega 64 Liquid Cooled running in 10.14.5 no VGTab, and I noticed that I'm only getting very low Geekbench 4 results compared to Windows results.

Is there a problem with Geekbench 4 and this card and or 10.14.5?

Can I still run VGTab with 10.14.5 and if so, do I have to remove Whatevergreen?
 
I have an XFX Vega 64 Liquid Cooled running in 10.14.5 no VGTab, and I noticed that I'm only getting very low Geekbench 4 results compared to Windows results.


@jb007,

Your not alone, myself and everyone else with a Vega based GPU are seeing lower GeekBench compute scores in MacOS 10.14.5 .. So nothing to worry about ...

I haven't seen an official reason as to why, some speculate that it could be connected to the Intel Zombieland CPU fixes that where introduced in 10.14.5 and GeekBench is especially susceptible to it, my own theory is that prior to 10.14.5 MacOS would use IGPU + dGPU for Open-CL / Metal compute but now just uses dGPU.

Cheers
Jay
 
Last edited:
Just for comparison:
This is Luxmark 3.1 running on 10.14.5. with no tweaks whatsoever. Just plain macos, no settings in config.plist and no kext. Running on iMacPro SMBIOS.
Hackintosh is about 7 years old, just switched out the GFX card for Vega 56

408350
 
Just for comparison: This is Luxmark 3.1 running on 10.14.5. with no tweaks whatsoever.


@sebideluxe,

As has already been reported most people are reporting that Luxmark scores are not effected by the 10.14.5 update.

You can not compare a LuxMark score against a GeekBench 4 compute score as they are two totally different tests and can only be compared against themselves before and after an MacOS update.

Cheers
Jay
 
So to answer my previous wonderings surrounding the injection of PowerPlayTable, fan behavior does follow the one injected through clover config instead of mojaves now native curve.

I can also confirm same results as previous posts regarding Geekbench scores, however my LuxMark scores increased from around 28k to 29,8k.
 
Back
Top