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

[kubusikan]

Hi folks, have you managed to get the fans on the Sapphire Vega 56 Pulse to work properly? As in - they do not ramp up to 49% duty 2400rpm and stay there? Thanks

Yes.
If idle, the fans stay at the specified 600rpm. If under load the fans ramp up to 1600rpm i set as target. After the load ended the fans stay at their last speed - usually 1600rpm - for a couple of minutes (4?) and then ramp down to idle speed.
Temps under load are 70ish °C Die and 90ish °C Hotspot.

But as i said, my fan setup is different from a stock Vega 56 Pulse. They push/pull more air and only go up to ~2000rpm.
So maybe the stock fans need (way) more speed to keep the card cool while under load?

 

[mcljot]

Any details of your Vega fan setup? What are they connected to? Thanks

 

[kubusikan]

Any details of your Vega fan setup? What are they connected to? Thanks

The two fans (Noctua NF-A12x25) are connected by a gelid pwm adaptor to the fan header on the card. So the card can control them.
They are located at the lower part of the case and suck the air through the stock heatsink and exhaust it out the bottom.
To aid the airflow i made a shroud that covers the heatsink on two sides.

This first picture shows an earlier revision with a NVidia 1060 and custom heatsink to give you an idea about the
fan placement.
This setup was replaced with the Vega 56 with its stock heatsink, as it quite beefy already.
The second picture shows the current setup with the Vega. Yes, the shroud needs some additional work to look nice.

 

[woozaer]

What might be optimal voltage settings for a Sapphire Vega 64 Nitro+ to set on VGTab?
With default Vega 64 settings I do sometimes experience some awful black screen crashes, which do not happen with the same machine running Windows 10 and radeon drivers, so I expect it might be related with the voltage.

 

[woozaer]

I've checked better my default settings in Windows, the biggest difference was with the memory voltage: VGTabs applies a default 1350 mV for a Vega 64, while the default setting for my card is only 1100 mV! I have now applied it and testing. Hope it will fix my black screen freezes!!!

 

[mcljot]

The two fans (Noctua NF-A12x25) are connected by a gelid pwm adaptor to the fan header on the card. So the card can control them.
They are located at the lower part of the case and suck the air through the stock heatsink and exhaust it out the bottom.
To aid the airflow i made a shroud that covers the heatsink on two sides.

This first picture shows an earlier revision with a NVidia 1060 and custom heatsink to give you an idea about the
fan placement.
This setup was replaced with the Vega 56 with its stock heatsink, as it quite beefy already.
The second picture shows the current setup with the Vega. Yes, the shroud needs some additional work to look nice.

View attachment 390723

View attachment 390720

Nice, congrats! I'm having no luck being able to stop my Sapphire Vega 56 Pulse fans from spinning at 2400rpm constantly. The usual tricks of playing a youtube video or local video file reduce the fan speed, but no combination of the VGTab tool, VegaTab-Merge tool, or manually configuring my config.plist have worked for me. Could I take a look at your EFI folder to see what you're doing differently? Kexts, drivers, and config.plist. The fans are driving me crazy, I'm dual booting into Fedora so it's not the end of the world but the whole point of the machine was to run Mac OS!

Edit: got it working. Manually retrieving the PCI path with gfxutil and then manually updating my config.plist from the VegaTab_56.kext did the trick. I ended up with this:

Spoiler: config.plist Properties entry

        <key>Properties</key>
        <dict>
            <key>PciRoot(0x0)/Pci(0x1,0x0)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x0,0x0)</key>
            <dict>
                <key>PP_DisableClockStretcher</key>
                <integer>1</integer>
                <key>PP_DisablePowerContainment</key>
                <integer>1</integer>
                <key>PP_DisableULV</key>
                <integer>1</integer>
                <key>PP_EnableUploadFirmware</key>
                <integer>0</integer>
                <key>PP_Falcon_QuickTransition_Enable</key>
                <integer>1</integer>
                <key>PP_GfxOffControl</key>
                <integer>0</integer>
                <key>PP_PhmSoftPowerPlayTable</key>
                <data>
                pwIIAQBcAO8GAAAOLAAAGwBIAAAAgKkDAPBJAgAyAAgA
                AAAAAAAAAAAAAAAAAAIBXABAAjcClACPAbQAHgF6AIwA
                rQEAAAAAYwIAAJAAmQJeATQBiAFoNgIAcQICAgAAAAAA
                AAgAAAAAAAAABQAHAAMABQAAAAAAAAABCCADhAO2A+gD
                GgRMBH4EsAQBAeIEAQGEAwAGYOoAAABAGQEAAdxKAQAC
                AHcBAAOQkQEABWywAQAHAQjQTAEAAACAAAAAAAAAHIMB
                AAEAAAAAAAAAAIi8AQACAAAAAAAAAAC07wEAAwAAAAAA
                AAAAgAACAAQAAAAAAAAAAMg/AgAFAAAAAAEAAADIWAIA
                BgAAAAABAAAAGG0CAAcAAAAAAQAAAAAEYOoAAABAGQEA
                ANxKAQAAkF8BAAAACChuAAAALMkAAAH4CwEAAoA4AQAD
                kF8BAAT0kQEABdCwAQAGwNQBAAcACGw5AAAAJF4AAAH8
                hQAAAqy8AAADNNAAAARobgEABQiXAQAG7KMBAAcAAWg8
                AQAAAQQ8QQAAAAAAUMMAAAAAAHARAQABAACAOAEAAgAA
                AQgAmIUAAEC1AABg6gAAUMMAAAGAuwAAYOoAAJQLAQBQ
                wwAAAnj/AABAGQEAtCcBAFDDAAADtCcBANxKAQDcSgEA
                UMMAAASAOAEAkF8BANxKAQBQwwAABdxKAQAAdwEAkF8B
                AFDDAAAGAHcBAAB3AQCQXwEAUMMAAAcAdwEAkJEBAAB3
                AQBQwwAAARgAAAAAAAAAC+QSkAFgCUsACgBUA5ABkAGQ
                AZABkAGQAZABAAAAAAACBDEHpQClAKUALAEAAFkAaQBK
                AEoAXwBzAHMAZABAAJCSl2CWAJBVAAAAAAAAAAAAAAAA
                AAAAAAACAtQwAAACEGDqAAACEA==
                </data>
                <key>PP_PhmUseDummyBackEnd</key>
                <integer>0</integer>
                <key>PP_ToolsLogSpaceSize</key>
                <integer>524288</integer>
            </dict>

 

[jaymonkey]

Edit: got it working. Manually retrieving the PCI path with gfxutil and then manually updating my config.plist from the VegaTab_56.kext did the trick.

@mcljot,

Glad you finally got your fans under control and found the guide helpful.

Cheers
Jay

 

[kubusikan]

Could I take a look at your EFI folder to see what you're doing differently? Kexts, drivers, and config.plist.

Nice you got it working!
If you would still like to have a look at my EFI, i uploaded it here:

iMacPro AsRock X299 e + i9 9940x = MiniMonsterBuild

Hi would you be willing to share the efi files you used so I can try this on my system? Hi, i can upload my efi folder on the weekend as i have no access to my system at the moment.

www.tonymacx86.com

 

[deutscs]

The other way round is more reasonable i think.
Try to set your Mem voltage to the value of a lower P-state at which it should be at it's highest setting.

I set up my Sapphire Vega 56 Pulse in VGTab with these mV values: P0=800, P1=900, P2=925, P3=950, P4=975, P5=1000, P6=1025, P7=1050. Core Clock is stock.
Memory is set to 900MHz and 950mV (mine has Samsung HBM2 and can go a bit higher to 940MHz in Windows but 900 is safe and enough for my use case in MacOS).
As soon as the card is using P3 the memory will be at 900MHz.

Target temp is 50°C and TDC is 50.
My fan setup is not comparable as i am using a custom fan setup with two noctuas. But for the record: idle=600, target=1600, min=400, max=2000

Hello Kubusikan, just intrigued by your custom setup with the fans. any reason why you did this, and which fans did you use. Has it made a difference?

Thanks

 

[kubusikan]

Sorry if this is to offtopic.

Hello Kubusikan, just intrigued by your custom setup with the fans. any reason why you did this, and which fans did you use. Has it made a difference?

I use 2 Noctua NF-A12x25 fans for the GPU (and two of them for the CPU AIO).
The reason i (and many others) am doing this is mainly the very small case i use. It is a NCase M1. The best solution for airflow is said to be to exhaust actively and intake passively which is quite the opposite to bigger cases.
There is a very long thread about the case:

NCASE M1: a crowdfunded Mini-ITX case (updates in first post)

NCASE M1: a crowdfunded Mini-ITX case (was: "Aluminum SG05 evolution by Lian Li") ---------------- This thread documents the origin and development of the NCASE M1, from the original concept of an "aluminum SG05 by Lian Li," through multiple design iterations incorporating community feedback...

hardforum.com

The benefit is that the heat is not circulation inside the case and the air that is fed through the GPU heatsink is quite cool. (The AIO CPU cooler is also exhausting the air from inside.)
Thats one thing but i think that using bigger 12cm coolers with a nicer/lower frequency sound profile would benefit normal installations of GPUs too. There would also be the possibility to use a beefier Morpheus Vega aftermarket cooler for better heat dissipation. But that would restrict me to use slim fans with higher noise.

Basically you get less, more "pleasing" noise and better cooling capacity with bigger and better fans as the fans can spin slower while pushing the same amount of air. Or you get (way) better cooling with equal noise on full blast than with stock fans.

As i never used the stock fans in this setup so i can only speculate but i would assume that they would have to spin fast to keep the Vega cool in this tight space.