Contribute
Register

Will It, Won't It — 7950X & Asus X670E-F gaming Wifi

Will It, Won't It?


  • Total voters
    14
Status
Not open for further replies.
Audio stuttering problem on AM5 platform with Zen 4 has just been solved by @Shaneee!

An awesome effort! Great work Casey. Shaneee's reaction to your Stop the Presses post made me laugh. To quote "Holy crap it actually worked" :lol:
 
An awesome effort! Great work Casey. Shaneee's reaction to your Stop the Presses post made me laugh. To quote "Holy crap it actually worked" :lol:
That was an intense and amazing effort documented in 59 pages! The thread began on September 30. About 6.5 weeks later we appear to have arrived at Golden Status! :)

The challenges with AM5 were much more complex than Alder Lake. We learned a lot.
 
Much more complex indeed. Alder Lake booted and ran fine with just the CPU definition SSDT (CPU-WRAP/CPUR) and E-cores or HT disabled. I expected a more tedious progress of moving small step by small step towards boot. There was no serial debugging involved, and the sole limitation (namely making full use of P and E cores) was solved by the new ProvideCurrentCpuInfo quirk about three weeks after first boot.

Zen4/AMD5 involved new/updated kernel patches to boot, some serious debugging of PCIe behaviour to enable the hardware on Monterey/Ventura and then an investigation into frequency calculations to fix audio. Incidentally, the work went beyond what the "Zen4/AMD5" moniker may suggest since the audio issue already plagued AM4 APUs and some AM4 boards had PCIe issues, introduced by BIOS updates dating from the launch of Zen3—so this work on AM5 actually fixed AM4 as well.
The work is not finished, as the new frequency calculation method has to be thoroughly checked against other architectures and/or carefully restricted to those CPUs which require it before pull requests can be safely accepted into the release versions of OpenCore and/or Clover. The tedious-but-necessary work of quality control and all that…

But this was exciting to watch, and even mere onlookers learned a lot about the inner working of their computers. Kudos to all involved!
 
But this was exciting to watch, and even mere onlookers learned a lot about the inner working of their computers. Kudos to all involved!
And continues to be exciting, I wonder what else will be 'accidentally' fixed!. How was Clover involved?, I read that an old kernel was looked at.
 
Clover and OpenCore use different formulas to calculate FSB and TSC frequencies. Some of the Clover code is reportedly a legacy from the days when AMD hackintoshers compiled custom kernels instead of patching Apple's kernel. Either bootloader produced garbled audio, but mixing code from both eventually got it "right". :crazy:
(To quote Shaneee: "Holy crap it actually worked!")

At this point I think that the only outstanding issue with Ryzen is AppleVTD, and the devices which depend on it such as Aquantia 10 GbE NICs. A skilled programmer may possibly solve it with an intermediate software layer which would present AMD-V as if it were Intel VT-d; I don't think this will come "accidentally", but it might incidentally fix other MMIO issues with AMD and remove the need to manually whitelist address ranges.
 
Some very good news concerning Ryzen 7000 testing!. vit9696 accepted and merged the OpenCore pull request for FSB/TSC, will we see this in 0.8.7?. It was discovered that AMD and Intel use different TSC frequency calculations which can lead to audio stutter on AMD (I think I got that right!). Win Win all round!.

 
Last edited:
Status
Not open for further replies.
Back
Top