Contribute
Register

AMD Radeon Performance Enhanced SSDT

@mattystonnie, what steps should I follow with my rx 470 to get those changes and have my card appear in the PCI section?
 
@mattystonnie
I really appreciate the fantastic work you've done here.
I run an iMac 17,1 in headless configuration with a Sapphire Vega 64 +Nitro
This is what I did:
1, installed your SSDT.RX Vega 64-Version 2.3.aml in EFI/CLOVER/ACPI/patched folder
2, installed dAGPM.kext in EFI/CLOVER/Kexts/Other folder
Then rebooted and did a Gb5 Metal test, but saw no improvement in performance over previous tests.
So I went back to Clover Configurator/ACPI and moved the SSDT to the top of SortedOrder in my config.plist (see picture)
Then retested and wow! I saw a 30% performance jump in Metal tests! Scores went up to almost 80k from an average of 60k
BruceX Test - 5K.fcpxml down to 10sec from about 20 sec

Thanks again!


View attachment 467581

Hey! Thankyou! This trick got me +- 5000 more metal points on Geekbench!
My score went to 79126 with a Sapphire Radeon RX 5700 XT Pulse
 
After some tweaks and overclocking (including this SSDT, v2.3), my GB4 score went up from ~170k to 210k on my (seriously) overclocked Vega 56 (watercooled, custom loop), but only a very slight improve on GB5, from ~59k to ~61k.

Nice! :)

PS: MacPro7,1 SMBIOS! :headbang:
 
Random question on this.
I saw this in the release notes for the latest WEG:

"Add support to injecting Force_Load_FalconSMUFW from OpenCore"

What does this mean for this effort? Anything? nothing?
Does it simplify the solution?
 

Attachments

  • Screen Shot 2020-05-06 at 3.47.11 PM.png
    Screen Shot 2020-05-06 at 3.47.11 PM.png
    478.1 KB · Views: 101
Random question on this.
I saw this in the release notes for the latest WEG:

"Add support to injecting Force_Load_FalconSMUFW from OpenCore"

What does this mean for this effort? Anything? nothing?
Does it simplify the solution?

It means bye bye Radeon Bo**t kext .
 
....
BruceX Test - 5K.fcpxml down to 10sec from about 20 sec

are you sure about the fcpx speedup?
for me neither ssdt nor kext speed up fcpx. ok, bruceX maybe 0.3s out of 8s faster, margin of error with that test though

geekbench DOES show better scores than without, so ssdts/kexts working
 
@qbe @slipanot

yeah.. pretty much what I was thinking the result would be...

artificial benches are mostly NOT a good indicator or real world performance any more.

too many benches are written highlight hardware.. and too much hardware is optimized to pass benches... its a self licking ice cream cone

I think things may change a bit if apples engineers think they can squeeze more out of some of the more modern cards that MIGHT go into the 7,1... but they didn't really care about what people were shoeing into their aging 5,1s and probably TOLD to make the iMac pro look better...

for that.. there is definitely a thermal issue with optimizing the powerful cards for the iMac since its space constrained and the AMD bugs internal to the graphics card are not apples to fix.. so I think they optimized that vega to get the most out of it for FCP and called it a day
 
Random question on this.
I saw this in the release notes for the latest WEG:

"Add support to injecting Force_Load_FalconSMUFW from OpenCore"

What does this mean for this effort? Anything? nothing?
Does it simplify the solution?

I would be happier if WEG would get the VEGA to boot into multi monitors without having to dickdance sleep/wake/hotplug/sacrificeacat/stand on one leg BS
 
Back
Top