Linux Fixes Issue Where Applying AMD Zen1/Zen2 Microcode Updates Could Slow Boot Times

[TumultuousUnicorn]

You can attempt a genius move like save the uefi settings, write them to the usb where your bios image is and after update reload the settings, well if it works is another matter .....

Did you read my message? "Even if I can save current settings to a file on a USB drive, I cannot restore settings from the file because de UEFI version is different."
I own a MSI B450M MORTAR MAX, and I already tried, but the UEFI refuses to load the configuration file because of version mismatch, which is why it is so annoying.
I think this limitation is present only on a MSI motherboard, maybe I will switch to a different brand next time I need to buy a new motherboard.

[Anux]

My BIOS has a "save user settings" option and those survive a bios flash/reset, I can just load them afterwards. Only my fan curve isn't saved

[stormcrow]

Mine (Asus TUF GAMING X570-PLUS) certainly doesn't. I thought those days were over a long time ago.

Notoriously MSI certainly does. Not only will they wipe the user firmware settings, but they can, and often do, 'forget' keys stored in the fTPM. MSI is one of those vendors I would never recommend at this point between their well known security issues and consumer hostile policies. Above all, never EVER mention you use Linux or anything else but Windows in any CSR communications. They will immediately blow you off (even if you've verified the problem is the same with WIndows).

[erniv2]

I own an AM4 motherboard, and trust me, it is time consuming to update the UEFI. The problem is updating the UEFI reset all settings to factory. Even if I can save current settings to a file on a USB drive, I cannot restore settings from the file because de UEFI version is different.
It is painful each time to reconfigure all settings (PBO, CPU voltage, fan, and so on), it would be better if AMD provides early microcode updates for all CPU (like Intel do).

You can attempt a genius move like save the uefi settings, write them to the usb where your bios image is and after update reload the settings, well if it works is another matter .....

And yes if you update your uefi it will run a atleast 2 post tests and after that boot with the factory defaults, wich actully makes sense.

[XenHat]

same here, I haven't been able to get a POST time below 8 seconds on my Gigabyte Aorus X570 Master, not from the lack of trying. Not sure what's wrong across the whole AM4 socket regarding boot times but I'm hoping we can put this behind us soon.

[jeisom]

I agree with you: UEFI take a while on an AMD mobo.

Startup finished in 17.149s (firmware) + 943ms (loader) + 1.854s (kernel) + 3.584s (initrd)

I do not understand why it takes so much time with default settings, even if I disable some things.

On my board, an Asrock X570-pro4m, there is an ultrafast boot option. It disables usb init and csm, so no legacy boot or keyboard until kernel is loaded. Possibly some other stuff.

Startup finished in 9.061s (firmware) + 3.464s (loader) + 3.949s (kernel). With it off It is around 18/19 seconds for the firmware.

[caligula]

I agree with you: UEFI take a while on an AMD mobo.

Startup finished in 17.149s (firmware) + 943ms (loader) + 1.854s (kernel) + 3.584s (initrd)

I do not understand why it takes so much time with default settings, even if I disable some things.

I think the DDR memory training takes a while, but it does not explain all of the slowdown.

[r1348]

Mine (Asus TUF GAMING X570-PLUS) certainly doesn't. I thought those days were over a long time ago.

My Asus Prime X570 Pro does.

[GreenReaper]

BIOS updates resetting every damn thing is how I learned that my BIOS contains so many duplicated settings that don't have synchronized output. I can set PBO to ON with a -20 offset in one menu and then in another menu the same PBO settings will be set to Auto or Disabled.

Which setting do I trust?

I trust Neither. I go page by page by page and make sure everything that I tweak that has duplicated entries are all set to the same thing.

Probably the best route. On ASUS AM5 B650E-I the ASUS ones seem to override the AMD Auto ones except the GPU Curve Optimiser settings which don't seem to stick unless you use the AMD ones.

[andyhhp]

AMD doesn't usually provide hot-loadable microcode updates for anything but EPYCs, and not every EPYC at that. This isn't anything secret since they disclose it in security bulletins, for example AMD-SB-7014 - SMM Lock Bypass

https://git.kernel.org/pub/scm/linux...5779d3bf3a13b8 is finally the fix for SinkClose, OS loadable on a range of more recent client CPUs.