AMD Radeon Performance Enhanced SSDT

[Quesadilla]

Tested SSDT-RX580- Version 1.0

Catalina 10.15.4 (iMac17,1 6700k rx 580)

Before:
OpenCL: 44354
Metal: 51257

After:
OpenCL: 44633
Metal: 53981

So it did something, Thanks Mattystonnie!

 

[mobi]

Early SSDT for the rx 580, works best with Catalina, for Mojave I don't know.

My Sapphire Nitro RX 480 8GB has a clock of 1276 Mhz, so RX 580 should get a higher score.

How to install : use the SSDT into Clover/acpi/patched + WEG.

Also test after a sleep/ wake cycle, for me worked without.

For the second stage boot loading bar use the first display port of your GPU.

More good things are coming, trying to inject properties from the Radeon Pro 580.

On Mojave 10.4.6 & RX590 get a black screen

 

[mattystonnie]

I ran with and without the new RX580 SSDT and don't see much difference. Could this be because I have Clover graphics options enabled (config.plist) as described below?

My GeekBench 5 test results with and without Radeon SSDT v1.0 are as follows:

Without SSDT

• Metal: 49930
• OpenCL: 45057

With SSDT

• Metal: 51914
• OpenCL: 45188

My system is as follows:

• Socket 1156 / H55 / Xeon X3460 / 16GB DDR3-1600
• SMBIOS MacModel: MacPro5,1
• Sapphire Pulse RX580 4GB
• Clover (Legacy) r5115 (boot-arg: -no_compat_check)
• Catalina 10.15.4
• Lilu 1.4.3, WEG 1.3.8

My CLOVER r5115 graphics config is as follows:

• FB Name: Orinoco
• Vendor ID: 0x1002
• Product ID: 0x67DF
• Inject ATI
• RadeonDeinit

Try without clover injection.

same to me:

Without SSDT, but with another SMBios
View attachment 466091

With SSDT and current SMBios (iMac19,2)
View attachment 466093

Ioreg pls and test after a sleep/ wake cycle.

 

[mattystonnie]

Those scores are not good, probably something wrong.
Will post my RX 480 geekbench 5 scores.

 

[Jsalexjs]

Those scores are not good, probably something wrong.
Will post my RX 480 geekbench 5 scores.

Surely this is not good for a Vega 64? It gets 80k in windows!

 

[dragonmel]

I have been having a devil of a time getting an DX58so with MSI Vega 56 airboost stable on mojave .6 (2020-002)

this rig was rock solid on el cap and a 660 but upgraded to a Vega that was supposed to be native and has been a disaster.

have a 4k at DP1 using HDMI adapter and a second monitor on HDMI.

can't boot with the HDMI monitor powered on, and it will lock up and black screen the video card if turned on after boot.. I have to do a sleep/wake to get second monitor to work. hot plugging doesn't help

the dx58so had a reference intel UEFI that was half baked but with a w3680 this hardware is very similar to a real Mac Pro 5,1

I use a 5,1 smbios and iMac-boardID in boot args

my DSDT has been heavily edited and I boot clover with almost no auto-patching turned on

I can post whatever people need..

THANKS

 

[macnb]

Thank you for posting this and thank you for your hard work! I disassembled your RX580 SSDT and see that you call DTGP. For those of us not using DTGP, I assume we need to add the Clover DTGP patch or edit your SSDT to eliminate DTGP.

I could not get V1.0 SSDT for RX580 to work because of the DTGP method.
Attached are the @mattystonnie's modified SSDT .DSL & .AML

It now changes the frame buffer to Orinoco.
But for there's was no significant change in either GB 4 nor GB 5 scores.

BUT a big change in IDLE Power consumption. It went down from ~100W to 36W

 

[mattystonnie]

I could not get V1.0 SSDT for RX580 to work because of the DTGP method.
Attached are the @mattystonnie's modified SSDT .DSL & .AML

It now changes the frame buffer to Orinoco.
But for there's was no significant change in either GB 4 nor GB 5 scores.

BUT a big change in IDLE Power consumption. It went down from ~100W to 36W

Ioreg pls.

 

[mattystonnie]

Geek5

 

[deeveedee]

EDIT: I attached IO Reg dumps with and without CLOVER Injection of RADEON attributes.
----------------------------------------------------------

Try without clover injection.

I don't see much performance difference without Clover injection. For my rig, it seems that the Clover injection is sufficient. My IOReg dumps (with and without clover injection) are attached. Results with Clover injection are below for comparison.

With SSDT (without Clover Injection of FB / Vendor ID / Device ID)

• Metal: 51446
• OpenCL: 44621

With SSDT (with Clover Injection of FB / Vendor ID / Device ID)

• Metal: 51914
• OpenCL: 45188

My system is as follows:

• Socket 1156 / H55 / Xeon X3460 / 16GB DDR3-1600
• SMBIOS MacModel: MacPro5,1
• Sapphire Pulse RX580 4GB
• Clover (Legacy) r5115 (boot-arg: -no_compat_check)
• Catalina 10.15.4
• Lilu 1.4.3, WEG 1.3.8

My CLOVER r5115 graphics config (when operating with Clover injection) is as follows:

• FB Name: Orinoco
• Vendor ID: 0x1002
• Product ID: 0x67DF
• Inject ATI
• RadeonDeinit