Cause of Nvidia Chip Model error discovered - could this be the patch we've been looking for?

[Middleman]

Hi guys,

I have something quite important I found which may interest you guys. I believe I may have found some of the missing properties from the IOReg listing which is the cause of the missing 0GB VRAM in System Profiler for
the Nvidia card.

I was recently installing High Sierra 10.13.6 on my Acer Nitro 5 laptop which has a Nvidia GTX 1050 Ti installed, and I was trying to find out to see whether the VRAM post issue for System Profiler could be fixed.
Along the way I had found a fix for the Intel GPU VRAM values and hardware acceleration using Hackintool, and I wanted to fix my Nvidia card too.

 

So I ran IORegExplorer on it, and at least two other machines including a 2012 MacBook Pro with a GT330M card. I was comparing the IOReg values to see what was amiss. As it happens I also had a Asus Z370

Prime P also fitted with a GTX 1070 Ti that I had installed Sierra 10.13.6 which ran and posted the System Profiler perfectly. As it is the exact same card I hoped to see what could used as a solution.








Between the Acer laptop and the Asus desktop, I found several fields that were missing in the GFX0 values, including a different sub-system vendor ID between the two systems (despite the card being the exact same thing).

Also I found that the GT330M card on the MacBook Pro had declared a function called 'Automatic Graphics Switching'. As Clover has a properties patching function, if we put in the missing data and properties from the Nvidia,
I was wondering could this be the key to solving our ability to use it and enable its hardware acceleration? It would be useful I think for laptop users like myself.

Edited December 29, 2020 by Middleman

[Middleman]

Oops sorry that was a typo - I meant to say I have the exact same 1050 Ti 4GB card in two separate machines (one laptop, one desktop) and I noticed discrepancies between the two IOReg dumps. I'll have to dig out what I found - but the differences were there. It had the same vendor ID but different sub-vendor ID (despite Nvidia being the manufacturer). Also the ROM version wasn't present in the laptop version, but was replaced with a 'Preset 1.0.0' setting. There was also several fields missing in the GFX0 section of the laptop compared to the fully enabled desktop version. In addition to that test programs that I ran didn't fully utilise the dGPU - it ran with the iGPU instead. So I'm thinking there are some parts still in my system that need fixing.

Indeed, nothing else but High Sierra now if we want Pascal support. But even then, somebody I've read managed to get NVidia's drivers working past 10.14.5 through a bit of tweaking.

As for the Macbook Pro GPU switch find/issue, I know that the switching function is only available on official Apple Nvidia Optimus laptops/iMacs. But somebody I found has come up with an app called gSwitch. Perhaps the code could be reverse-engineered to somehow only set the dGPU for Optimus users like myself? Either way I think it's high-time somebody thought of enabling support for dual laptop gpus on slightly older OSes other than Catalina.