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

Level1Techs
1

http://www.amazon.com/dp/B079Z8W61Z/?tag=level1techs-20

This is a companion discussion topic for the original entry at https://level1techs.com/video/asrock-x399-m-taichi-review-linux-test
2 Likes
2

+1 grant messed up on the edit, but not a big deal, just put wrong info on the screen between video and bios screens

3

Where is the memory latency test (testmem?) from?

4

It’s on GitHub, it’s the Linux memory latency tester for copying 32mb. Note the ns latency is what aida64 reports and the memtest program is more of a homegrown test of framebuffer copies.

5

mmm … Thanks Wendell.

6

Great video, almost exactly what I wished to see :slight_smile:

One thing though I really, really wish to see is the board + the Noctua NH-U14S TR4 and the Fractal Design Meshify C Mini TG, is there any possibility that could just take some pictures to showcase the clearance between cooler + RAM and cooler to side panel @wendell ?

Also could you dump the MMU groupings here?

7

Can do! The 14s is like sheet of paper levels of clearance with the side panel. . The 12s would be safer? But the real question is does the wraith ripper fit :smiley:

8

I was never able to be truly stable at 3200 or above, but managed to get 8 sticks of Super Talent F24EA8GS (which are samsung B-die) to this:

OC%203066%2014-15-15-14-32-48%20procODT%2048%20v2
OC 3066 14-15-15-14-32-48 procODT 48 v2.png1922x798 165 KB

They also have some 16gb versions that from what I hear work exactly the same, so I could see those working well for getting a 4 slot m-atx board to 64gb of overclocked ecc memory.

Edit Also, for anyone tempted to try asrock’s [Beta] 1.91E bios for x399 boards, be aware that it completely breaks ecc reporting functions. Errors will still be corrected, but not reported.

9

Could you provide the link to the repo?

10
1 Like
11

Thanks. The output looks different than in the video. Here are my results (2700X, C7H, 2X16GB F4-3200C14D-32GTZ @ 3200 DOCP):

$ ./tm64 32
32 MB = 0.790488 ms
-Compare match (should be zero):  0 

$ ./tm32 32
32 MB = 2.809878 ms
-Compare match (should be zero):  0
12

that’s pretty zippy.

the video output is just a for loop from 1 to 32 mb (it should be a linear increase but sometimes odd things happen). the 32 bit version being slower is an odd consequence of some things, but you can switch memcpy for memcpy_sse for 32 bit and it will work well again.

1 Like
13

Only bad part with the board for me has been memory support, on my old B350 ASrock board I could just load XMP with the latest BIOS and it would work.

On this one even with the latest BIOS loading the XMP profile on a basic 16GB Ripjaws V 3200mhz dual channel kit makes my system crash after about 30 minutes no matter what I do it seems, have to lower the XMP profile to 2400mhz.

Not sure how much to blame the board, and how much to blame the RAM.

In the testing I can do before it crashes I gain about +200cb in Cinebench, and like +40fps in Dota 2 with faster RAM, makes me want to buy a quad channel kit, but they cost just so damned much.

14

Ermagerd, could it really be? The Streetguru is alive? He is back in the forum? I’m so happy I can cry.
How’s it going, buddy? Where have you been?

15

Attempting to save people from themselves on linus’s forum

1 Like
16

You have abandoned us for Linus?

17

Yeah the Asrock bios can be a littlebit wonky wenn it comes to memory support.
Or lets say memory timings and clock tweaking.
But thats something they might improve on with future bios updates maybe.

18

Just considering the price of the board you’d think they’d maybe speed it up or try a little harder

This forum isn’t too active?

19

Vrm info:

Asrock X399 Taichi vrm overview.
Also counts for the X399M Taichi.

Main cpu Vcore vrm.

  • pwm: IR35201 in 8+0 phase mode, so 8 true phases.
  • powerstages: IR3555 60A fully intergrated powerstages.
  • Inductors: 60A.
  • output caps: Tantalum´s.

Soc vrm:

  • pwm: a second IR35201 pwm is providing 3 phases for controling the SOC vrm.
  • powerstagesa: IR3555 again.

Memory vrm:

  • pwm: 2x uP1647P pwm’s providing 2 phases for each set of 4 dimm slots.
  • mosfets: Sinopower SM7341EHKP dual-N Mosfets.
20

@wendell Was memory support/overclocking w/e mentioned in the review? I’m not sure I recall it without a rewatch, probably something important for future Ryzen/TR content in general anyways.