macOS 12.3 Update Causes Problems for 5700/6800/6900 Graphics Cards

[tedyun]

What are your results in 12.4?

I haven't really checked in the same manner as I did the previous tests. Performance is essentially back to normal. I don't think the sPPT makes a difference with the correct framebuffer injected. I don't run serious games on the Mac side, and I don't do serious video work, so I haven't really bothered to squeeze any extra performance out of it.

 

[zirkaiva]

I haven't really checked in the same manner as I did the previous tests. Performance is essentially back to normal. I don't think the sPPT makes a difference with the correct framebuffer injected. I don't run serious games on the Mac side, and I don't do serious video work, so I haven't really bothered to squeeze any extra performance out of it.

But how about the GB5 tests? I have BS11.6.6 on another drive and when comparing GB5 results they are better in 11.6.6 compared to 12.4. Also 12.3.1 was better than 12.4. For me it dropped again.

 

[tedyun]

But how about the GB5 tests? I have BS11.6.6 on another drive and when comparing GB5 results they are better in 11.6.6 compared to 12.4. Also 12.3.1 was better than 12.4. For me it dropped again

I can't remember when I upgraded my video card, but I think I was on Monterey, so I don't have a good comparator to BS. But I do think that when I upgraded from Catalina to BS, my Vega 64 took a hit, then when I went from BS to Monterey, it took another hit.

I find that GB is not very reliable as a benchmark program. The variation between scores is very high. I think you need to do multiple tests, in the order of 20-50 repeats to get statistically significant comparisons. I read somewhere that the problem is that GB pauses between tests, so that can affect the scores.

If the scores have gone down in 12.4 vs 12.3.1, I am not sure I've noticed it. I can try running the GB test again when I get a chance. But I didn't measure it in 12.3.1, after they fixed the framebuffer issue, so the only comparison I can make is with 12.3, when we were manually injecting the framebuffers in OC.

 

[tedyun]

But how about the GB5 tests? I have BS11.6.6 on another drive and when comparing GB5 results they are better in 11.6.6 compared to 12.4. Also 12.3.1 was better than 12.4. For me it dropped again.

Come to think of it, I remember at the Catalina > Big Sur transition, the change blocked the "hacks" of the SSDT or kexts that injected the correct framebuffers for the Vega and Navi cards. But those hacks only seemed to affect GB scores. With other benchmarking programs like Heaven and an OpenGL one (I'm blanking on the name right now -- it basically renders an image of a marble) there wasn't much difference, or if there was, real world performance wasn't drastically affected.

I wouldn't worry too much about differences in GB. I can get scores ranging from 160K - 200K in any given run. It was useful when trying to troubleshoot that framebuffer issue introduced in the 12.3 update, but even then, I couldn't really rely on its accuracy in assessing whether the tweaks I was making were having impacts. I don't think GB is a very good benchmarking program, or rather, its use is limited. I say this as an owner of the product too.

 

[plucasa]

I can't remember when I upgraded my video card, but I think I was on Monterey, so I don't have a good comparator to BS. But I do think that when I upgraded from Catalina to BS, my Vega 64 took a hit, then when I went from BS to Monterey, it took another hit.

I find that GB is not very reliable as a benchmark program. The variation between scores is very high. I think you need to do multiple tests, in the order of 20-50 repeats to get statistically significant comparisons. I read somewhere that the problem is that GB pauses between tests, so that can affect the scores.

If the scores have gone down in 12.4 vs 12.3.1, I am not sure I've noticed it. I can try running the GB test again when I get a chance. But I didn't measure it in 12.3.1, after they fixed the framebuffer issue, so the only comparison I can make is with 12.3, when we were manually injecting the framebuffers in OC.

Hey tedyun - i just bought a second hand RX 5500 XT. Right now i'm in 12.4 do i have to manually inject the framebuffer in OC? or it is not needed anymore in 12.4?

 

[tedyun]

Hey tedyun - i just bought a second hand RX 5500 XT. Right now i'm in 12.4 do i have to manually inject the framebuffer in OC? or it is not needed anymore in 12.4?

@plucasa - you're good to go! The problem only exists in 12.3 and was fixed in 12.3.1 and beyond.

 

[faraway]

Ever since I've upgraded to Monterrey 12.4 from Big Sur, I have been experiencing the occasional Kernel Panic or black screen after the Apple logo on boot with a reference 6900 XT

This is the crash log:

panic(cpu 0 caller 0xffffff7f92b0c9ff): "GDDR6 Long Training Failed !!!
" @AmdRadeonController.cpp:1998
Panicked task 0xffffff87250ec670: 232 threads: pid 0: kernel_task
Backtrace (CPU 0), panicked thread: 0xffffff87262baaa0, Frame : Return Address
0xffffffa114a938f0 : 0xffffff8003881c8d mach_kernel : _handle_debugger_trap + 0x41d
0xffffffa114a93940 : 0xffffff80039e1596 mach_kernel : _kdp_i386_trap + 0x116
0xffffffa114a93980 : 0xffffff80039d0963 mach_kernel : _kernel_trap + 0x4d3
0xffffffa114a939d0 : 0xffffff8003821a70 mach_kernel : _return_from_trap + 0xe0
0xffffffa114a939f0 : 0xffffff800388205d mach_kernel : _DebuggerTrapWithState + 0xad
0xffffffa114a93b10 : 0xffffff8003881816 mach_kernel : _panic_trap_to_debugger + 0x2b6
0xffffffa114a93b70 : 0xffffff8004115163 mach_kernel : _panic + 0x84
0xffffffa114a93c60 : 0xffffff7f92b0c9ff com.apple.kext.AMDRadeonX6000Framebuffer : __ZN34AMDRadeonX6000_AmdRadeonController10doGPUPanicEPKcz + 0x1b9
0xffffffa114a93d70 : 0xffffff7f92ac3116 com.apple.kext.AMDRadeonX6000Framebuffer : __ZN39AMDRadeonX6000_AmdRadeonControllerNavi219doGddr6LongTrainingEv + 0x7c
0xffffffa114a93d90 : 0xffffff7f92ac2f72 com.apple.kext.AMDRadeonX6000Framebuffer : __ZN39AMDRadeonX6000_AmdRadeonControllerNavi25startEP9IOService + 0x8e
0xffffffa114a93dc0 : 0xffffff8004019ffe mach_kernel : __ZN9IOService14startCandidateEPS_ + 0x10e
0xffffffa114a93e20 : 0xffffff8004019abe mach_kernel : __ZN9IOService15probeCandidatesEP12OSOrderedSet + 0xdce
0xffffffa114a93ef0 : 0xffffff8004018b60 mach_kernel : __ZN9IOService14doServiceMatchEj + 0x3e0
0xffffffa114a93f50 : 0xffffff800401bb87 mach_kernel : __ZN15_IOConfigThread4mainEPvi + 0x157
0xffffffa114a93fa0 : 0xffffff800382119e mach_kernel : _call_continuation + 0x2e
Kernel Extensions in backtrace:
com.apple.kext.AMDRadeonX6000Framebuffer(4.0.8)[2FE08341-C2AC-391F-91AD-104EB632986C]@0xffffff7f92abf000->0xffffff7f92d45fff
dependency: com.apple.AppleGraphicsDeviceControl(6.5.7)[53C8B9FD-723F-3BE0-A78D-E0DB7D06F9BD]@0xffffff7f9b128000->0xffffff7f9b12bfff
dependency: com.apple.iokit.IOACPIFamily(1.4)[3224D91D-C81D-33E2-A8B3-1A168235F4E7]@0xffffff8005f7b000->0xffffff8005f7cfff
dependency: com.apple.iokit.IOGraphicsFamily(597)[BBEE13AB-2291-37D9-AAE4-9E565E8950A5]@0xffffff7f9c93f000->0xffffff7f9c96dfff
dependency: com.apple.iokit.IOPCIFamily(2.9)[654ABF9D-EB24-3B64-AB1C-EC2AC8357192]@0xffffff8006424000->0xffffff800644ffff

Process name corresponding to current thread (0xffffff87262baaa0): kernel_task
Boot args: keepsyms=1 debug=0x100 alcid=13 -wegnoigpu agdpmod=pikera chunklist-security-epoch=0 -chunklist-no-rev2-dev

Mac OS version:
21F79

Kernel version:
Darwin Kernel Version 21.5.0: Tue Apr 26 21:08:22 PDT 2022; root:xnu-8020.121.3~4/RELEASE_X86_64
Kernel UUID: 3C587984-4004-3C76-8ADF-997822977184
KernelCache slide: 0x0000000003600000
KernelCache base: 0xffffff8003800000
Kernel slide: 0x0000000003610000
Kernel text base: 0xffffff8003810000
__HIB text base: 0xffffff8003700000
System model name: MacPro7,1 (Mac-27AD2F918AE68F61)
System shutdown begun: NO
Panic diags file available: YES (0x0)
Hibernation exit count: 0

System uptime in nanoseconds: 52274656160
Last Sleep: absolute base_tsc base_nano
Uptime : 0x0000000c2bcffbf9
Sleep : 0x0000000000000000 0x0000000000000000 0x0000000000000000
Wake : 0x0000000000000000 0x00000024dff82b09 0x0000000000000000
Compressor Info: 0% of compressed pages limit (OK) and 0% of segments limit (OK) with 0 swapfiles and OK swap space
Zone info:
Zone map: 0xffffff80bb6f7000 - 0xffffffa0bb6f7000
. PGZ : 0xffffff80bb6f7000 - 0xffffff80bf6f8000
. VM : 0xffffff80bf6f8000 - 0xffffff858ba2b000
. RO : 0xffffff858ba2b000 - 0xffffff8725091000
. GEN0 : 0xffffff8725091000 - 0xffffff8bf13c4000
. GEN1 : 0xffffff8bf13c4000 - 0xffffff90bd6f7000
. GEN2 : 0xffffff90bd6f7000 - 0xffffff9589a2a000
. GEN3 : 0xffffff9589a2a000 - 0xffffff9a55d5d000
. DATA : 0xffffff9a55d5d000 - 0xffffffa0bb6f7000
Metadata: 0xffffff806f6e7000 - 0xffffff808f6e7000
Bitmaps : 0xffffff808f6e7000 - 0xffffff809b6e7000

Anyone experiencing anything similar to this and would spoorfing the W6900 XT Framebuffers help?

 

[anikeevproduction]

Sup everyone.
Help me figure it out if this numbers good for 5700XT?
Did that fix with masking GPU, not sure did it right.
Working kinda good, but still windows feels like 5-10% better in video proccessing, blender cycles, etc.
Thanks.
Monterey 12.5

 

[zappy39]

Sup everyone.
Help me figure it out if this numbers good for 5700XT?
Did that fix with masking GPU, not sure did it right.
Working kinda good, but still windows feels like 5-10% better in video proccessing, blender cycles, etc.
Thanks.
Monterey 12.5

looks ok I get about 67500 through Razer x egpu same 5700 xt nitro

 

[anikeevproduction]

looks ok I get about 67500 through Razer x egpu same 5700 xt nitro

Thanks.