[SOLVED] Radeon Polaris poor performance after wake - Dell XPS 8930 Big Sur...

[brousseau6933]

Thanks @Feartech ,

I'll try all of this once I'll be home (at work now, pause...). I hope those quirks, SMBIOS replacement and igpu headless mode will kick the AMD once it wakes.

On a cloned drive obviously and a spare usb stick as the machine works quite ok (apart noise) with the GT740 for now.

I also have a couple of other things to try if it doesn't work:

- boot with Clover and disable "plug" CPU PM and goes all Kernel PM, inject P-States like it was on my old Optiplex 9020. It worked fine with CPU being used from 800 mHz to 3.9 gHz. I know, not "real" PM but it worked fine.

- With Clover, load extracted RX460 VBIOS at boot.

Please note that this machine can't have the CFG-lock disabled in any way, the entry doesn't even exist in the extracted BIOS... It was there for the 9020 but for this one, no BIOS (via UEFI Shell) mods could be done.

 

[Feartech]

Thanks @Feartech ,

I'll try all of this once I'll be home (at work now, pause...). I hope those quirks, SMBIOS replacement and igpu headless mode will kick the AMD once it wakes.

On a cloned drive obviously and a spare usb stick as the machine works quite ok (apart noise) with the GT740 for now.

I also have a couple of other things to try if it doesn't work:

- boot with Clover and disable "plug" CPU PM and goes all Kernel PM, inject P-States like it was on my old Optiplex 9020. It worked fine with CPU being used from 800 mHz to 3.9 gHz. I know, not "real" PM but it worked fine.

- With Clover, load extracted RX460 VBIOS at boot.

Please note that this machine can't have the CFG-lock disabled in any way, the entry doesn't even exist in the extracted BIOS... It was there for the 9020 but for this one, no BIOS (via UEFI Shell) mods could be done.

the config.plist changes should make a difference ie with usb ports as well, as you can test before making the change, open up ioregistry and see where it says your system definition, it probably only says iMacPro1 when it should say iMacPro1,1

 

[brousseau6933]

the config.plist changes should make a difference ie with usb ports as well, as you can test before making the change, open up ioregistry and see where it says your system definition, it probably only says iMacPro1 when it should say iMacPro1,1

No, it says iMacPro1,1. Maybe I made a mistake with EFI copy when removing serial, MLB and UUID? Anyway, time for testing!

 

[Feartech]

No, it says iMacPro1,1. Maybe I made a mistake with EFI copy when removing serial, MLB and UUID? Anyway, time for testing!

i meant when you run ioregistry explorer app

 

[brousseau6933]

i meant when you run ioregistry explorer app

 

[brousseau6933]

for Dell machines, you need to set in your config.plist:
CustomSMBIOSGuid -> true
UpdateSMBIOSMode -> custom

you could also set your machine up in "headless" mode:

An iDiot's Guide To Lilu and its Plug-ins

An iDiot's Guide To Lilu and its Plug-in's Last Update: 22nd Jan 2021 (Add note about OpenCore config.plist modifications) About this Guide Note: This guide was written when Clover was the boot loader of choice as such the config.plist examples in this guide are mostley Clover specific ... if...

www.tonymacx86.com

it will mean changing your smbios to iMac19,x (replace x with maybe 2) read above

you would then remove:
-wegnoigpu boot argument

also setup a proper device properties for the intel graphics for computing tasks only:

0300913E

also a boot argument that should be required:
-disablegfxfirmware

and also if you remove these ones (now you have a bootable system):
debug=0x100 keepsyms=1
(will also result in a quicker boot up)

Changes applied... and no changes with Radeon card RX460. Performance drops more than a half after sleep. 1st result after sleep: 9877 Basemark Metal Medium test. 24463 before sleep. Let's not forget that the display lags in the second test, after sleep...

Also (cosmetic, I know), no GPU in system infos when adding the iGPU as headless. I could probably add it via device properties but it's a moot point as my goal is to find why Polaris cards are behaving like this on this XPS.

At least, you helped me clean up my config . Now, I'll try Clover without "proper" CPU PM, only relying on KernelPM, AppleIntelCPUPM and injecting P-States. I'll doubt it'll work but I still could try next:

- Load VBios from Clover;
- wiring a spare PSU as maybe, as those cards (RX460 and WX5100) aren't auxiliary powered (only PCIe slot power), it could make a difference. This one, however, will not be tested today (tonight)...

Honestly, I begin to think that Polaris cards (and probably all Radeon cards, I do not have high hopes with the soon coming RX5500!) have a problem with this specific model Dell XPS8930 and MacOS, any version upward Mojave as I've experienced


Remember, no graphic performance loss in Windows and with NVidia in MacOS.

 

[Feartech]

Changes applied... and no changes with Radeon card RX460. Performance drops more than a half after sleep. 1st result after sleep: 9877 Basemark Metal Medium test. 24463 before sleep. Let's not forget that the display lags in the second test, after sleep...

Also (cosmetic, I know), no GPU in system infos when adding the iGPU as headless. I could probably add it via device properties but it's a moot point as my goal is to find why Polaris cards are behaving like this on this XPS.

At least, you helped me clean up my config . Now, I'll try Clover without "proper" CPU PM, only relying on KernelPM, AppleIntelCPUPM and injecting P-States. I'll doubt it'll work but I still could try next:

- Load VBios from Clover;
- wiring a spare PSU as maybe, as those cards (RX460 and WX5100) aren't auxiliary powered (only PCIe slot power), it could make a difference. This one, however, will not be tested today (tonight)...

Honestly, I begin to think that Polaris cards (and probably all Radeon cards, I do not have high hopes with the soon coming RX5500!) have a problem with this specific model Dell XPS8930 and MacOS, any version upward Mojave as I've experienced


Remember, no graphic performance loss in Windows and with NVidia in MacOS.



View attachment 561673
View attachment 561674

yes, using headless will mean adding the information in your config.plist as mentioned in the thread I linked before
but you will probably only see your AMD graphics in your settings

another thing, open up hackintool, go to USB section and post a screen shot of that window

 

[brousseau6933]

yes, using headless will mean adding the information in your config.plist as mentioned in the thread I linked before
but you will probably only see your AMD graphics in your settings

another thing, open up hackintool, go to USB section and post a screen shot of that window

I will do in a couple of minutes but before, success (kind of)!

Clover 5148 with "plug" unchecked and PStates and CStates checked: no more GPU lag or performance drop after sleep. Really counterintuitive but maybe it has to do to the fact that it's impossible to disable CFG-lock on this BIOS/Motherboard?

However, instead of lowest CPU frequency being 900 mHz (I do not understand this one anyway as this CPU could go to 800 mHz, not a big deal), it's now 1 gHz the lowest. Turbo didn't changed, 4.3 gHz.

I will try some experiment, removing Plug.aml from OC and try to drop some SSDT tables (CpuPm and others) in OC. Is it still possible to drop tables in OC BTW?

**I'll post a screenshot of USB in Hackintool as requested soon.

 

[brousseau6933]

Screenshots, everything appears ok to me. Important to note that the 5 gbps reported for ASM1142 is a known bug because the SSD plugged there measures in the 800-900 mb/s as it should (USB 3.1 gen 2 WD external SSD). It was the same with a PCIe ASM1142 card I had in my old Dell. Windows reports it correctly.

Two Port USB 3.1 GEN II PCI Express Card

Anyone else ordered a 2 Port USB 3.1 GEN II (10Gbps) PCI Express Card 1 USB Type C and 1 USB Type A Port (USB 3.1 A + C) from AliExpress.com? Seem like they have one on sale for $26.99 in 19 hours …

pikeralpha.wordpress.com

 

[Feartech]

Screenshots, everything appears ok to me. Important to note that the 5 gbps reported for ASM1142 is a known bug because the SSD plugged there measures in the 800-900 mb/s as it should (USB 3.1 gen 2 WD external SSD). It was the same with a PCIe ASM1142 card I had in my old Dell. Windows reports it correctly.

Two Port USB 3.1 GEN II PCI Express Card

Anyone else ordered a 2 Port USB 3.1 GEN II (10Gbps) PCI Express Card 1 USB Type C and 1 USB Type A Port (USB 3.1 A + C) from AliExpress.com? Seem like they have one on sale for $26.99 in 19 hours …

pikeralpha.wordpress.com

View attachment 561693
View attachment 561694
View attachment 561695

you may want to look into your USB ports setup:

The New Beginner's Guide to USB Port Configuration

Important Update: June 2023 From Big Sur 11.3 on the built-in OpenCore XhciPortLimit quirk no-longer worked as described. This became a known "bug", though in reality it was Apple changing their internal code, making a simple "peek and poke" to keep it operating, a difficult task. However, with...

www.tonymacx86.com

should only have 15 ports enabled on your XHC controller