ASRock X399-M Taichi Review + Linux test | Level One Techs

Nice power delivery… but ill be completely honest… AMDs platform and I honestly cant believe im saying AMD and efficient in the same sentence… Lmao but AMDs platform is incredibly power efficient even in overclocking that the phases while nice and allow us to push it further arent all as supremely necessary as they have been on Intel counterparts

Is this thing illuminated?

image

image

Nevermind, it’s just a watermark

It’s rgb yes

2 Likes

Well i would say for TR its definitelly needed,
especially for the newer sku´s with more then 16 Cores.
Allthough we still have to see how well those higher core count sku´s will clock.

But yeah honnestlly the entire Zen architecture till now isnt really,
that great wenn it comes to overclocking in the first place.
Tweaking on the memory to get the lowest possible latencies,
seems to be more important then raw clockspeeds for the infinity fabric.
But that was just the main issue with Ryzen, the imc kinda sucks just like with Bulldozer.
With Zen+ they finally seem to have made some improvements on latencies.
Andwe can onlly hope for the same on TR2 sku´s.

1 Like

Honestly I doubt the higher core variants will need any all that extra power. They will be using lower power cores and lower clocked cores to keep them under a certain TDP anyways. That being said I would like to see the IMC’s improve but as with any new architecture it seems to be a reoccuring pattern that when manufacturers complete ground up their architecture that the IMC always has issues… We saw that back when intel moved to Land Grid Array 775 and they worked out the quirks same with Socket 478. I am not sure why IMCs are weak points. It could be that they received less attention? Not sure this is definitely a @wendell because the server side probably deals with this more.

LOL ThreadRipper 2 I really wish they came up with a better name for the revision… I LOVED the threadripper name it was fantastic but to reuse it… EHHhh… like IDK name it Gladiator for the architecture after all the internal code name is Roma at least for the server parts haha

I definitely mispoke on that… Call its Zenith or Mirador LOL it can perform overwatch on intels next big fuck up

Core War II AMD Strikes Back! JK

I think I made an error. :sweat:

I upgraded to the latest X399M 3.10 BIOS (via 1.10 bridge) to see if it fixed the IOMMU groups (which it actually improves somewhat). But now my boot times are worse than watching paint dry. Shutdowns as well…

Running Fedora 28 with kernel 2.17.9-200.
Didn’t seem to be any trouble when running on 1.00E BIOS.

Anyone else seen this issue?

Add pcie_aspm=off to your linux line in grub?

@wendell: Thanks you the suggestion, unfortunatly that didn’t change much. I still have ~8 min boot time, and ~5 min shutdown time.

Could be related to some udev stuff that it compains about.
Got some errors related to e.g. AMD Encryption controller in journalctl.

P.S. I use LUKS on root. But not sure if this means anything. Worked fine before.

Nevermind Seems like there are issues with linux and the latest AGESA included with the new BIOS. Wonder if I can safely roll back?

Someone on Gigabyte forum seems to believe compiling kernel with
CONFIG_CRYPTO_DEV_SP_PSP=n solves the issue…