x399 taichi fails to boot with "33 d3 Ad A2 21 12"

Summary: re-seated CPU; all RAM now recognised; runs fine at 3200 (stable under prime95); but reboots intermittently when pc is idle.

ASRock support response

ASRock support got back to me - my ticket contained all of the information I've posted here and nothing else of note. Their response was really good, imho. In summary:

1. Regarding the initial symptoms they said:

This looks like a faulty CPU (which is where the memory controller is integrated) or a faulty motherboard.

Please remove the CPU from the socket carefully. In clear light check the socket for bent or broken pins. Make sure there is no thermal paste on any socket pin or CPU contact. If there is any problem with a socket pin then please let me know.

If the socket looks okay then if possible borrow another CPU to test with, or try to test your CPU and slot A2/A1 in another motherboard. Maybe your supplier or a computer shop can help with that.

If the CPU is okay (or if there is no way to test it) then please contact the motherboard seller for warranty/RMA.

2. Regarding the Dr debug codes:

The debug code sequence "33 DE AD A2" looking like "EE dead A2" really is just a (remarkable) coincidence as far as I know.

3. Regarding HWinfo reporting the memory clock as 1,065.6 MHz

[This] is not that strange. 1066 MHz = DDR4-2133. This is the default DRAM frequency, maximizing compatibility with different modules. Once the system is detecting all modules reliably you can try loading the XMP profile in BIOS (if your RAM modules offer this) and set the DRAM frequency manually in BIOS. Start with loading the XMP profile, then set the DRAM frequency to DDR4-2400. If that is working fine then try DDR4-2666 , 2933 and 3200.

I think most of this confirms what we'd already said.

What I did next:

1. Removed cooler and took photos of the TIM

2. Unclipped CPU retention frame and rail frame, kept CPU in the frames.

3. Inspected and photographed the socket, pins and the cpu contact surface.

Some TIM had bled out of the sides (tiny, tiny amounts), but certainly hadn't made it into the socket or onto the contact surface underneath the CPU. One pin reflected light strangely on the photos, but didn't look bent to the eye, I intended to check that again but ran out of time.

So here's the weird thing:

I was running out of time so I clipped the CPU back in, screwed it down using the torque wrench, fixed and screwed in the cooler and reassembled the board in the case to keep it safe for the night. The next morning I booted it without thinking. I did nothing different to the original install except I didn't clean the cooler and put new TIM on, I used the old stuff.

You can see where this is going ...

1. All the memory is now reported as usable in windows

2. When I then (a) reset the CMOS; (b) rebooted into UEFI and then set it to XMP Profile 1, selecting 3200 memory speed it (c) booted into windows, and (d) passed prime95 on both blend and then small FFT for 1 hour each.

Temperatures in p95 Blend as reported by HWInfo (ambient is 23 C) are no more than:

1. TDie: 34 C above ambient after running for an hour (2 C when idle)

2. RAM (DIMM 1, 3, 5, 7 respectively): 19, 17, 21, 28 above ambient (3, 3, 5, 9 C when idle).

3. Motherboard: 6 C above ambient (1 C when idle).

Temperatures are more than a little anecdotal as the fans aren't just running at 100% all the time, there is a fan profile which adjust fan speed based on TCtl and mobo temperatures.

Here's the problem though:

It isn't stable. It is stable under p95 but if it sits idle it intermittently reboots. This sometimes happens 65 seconds after showing the windows login screen, without me touching anything, and sometimes doesn't happen before I've done other stuff and rebooted deliberately. It only happens when the PC is idle and doing nothing (i.e. I've stepped away).

I've also noticed that before it reboots the USB mouse and keybard become unresponsive. I've only noticed this before some reboots, other times it reboots without me using them so I don't know if it happened or not. Also sometimes they become unresponsive and it doesn't reboot for a while so I have to reset it myself. This unresponsiveness also happens in the UEFI screen as well as in windows.

(Mouse and keyboard are plugged into the USB ports closes to the flashback button on the back of the board).

What next?

So now I'm stumped.

1. Is it a bad board and/or a bad CPU?

2. Are they ok and the CPU is still seated badly (how is that even possible given the physical engineering of the TR4 slot, the orange carrier frame, the clip to fix into the rails, the blue clips to hold it down, and the provided torque wrench)?

3. Is it seated properly and this is now something else entirely?

And what, therefore, should I do next (I have a second board sitting here now but would rather refund it than use it if at all possible)? None/some/all of:

1. Re-seat the CPU again?

2. Remove the CPU again, check the pins yet again and report back to ASRock?

3. Fiddle with voltages and see if the RAM stabilises when the PC is idle (which voltages would this be)?

4. Try the second board (which will cost me, so I want to avoid it if possible)?

5. RMA this board?

Edited by soniqboom - 25 Jan 2018 at 12:31am

X399 Taichi (BIOS 2.00), 1950X at default, F4-3200C14Q-64GTZR
(full parts list)

I thought I'd post an update.

I opened the second board. On first CPU install (doing nothing any differently to the other times I installed the cpu on the original board) it worked fine. I even re-seated it a couple of times, all with no issues with the A2 slot.

I then went back to the first board and had the A2 slot issue after re-seating the CPU yet again, then it went away again after re-seating it another time again!

ASRock were concerned about a pin on the socket that was reflecting light a little differently in the photos, but after a lot more detailed photos they decided it probably wasn't a problem. In the end due to the issues with re-seating it (that weren't happening in Board 2) they said to RMA it. I was convinced enough by the difference in behaviour between the two boards that this was the right call.

Edited by soniqboom - 30 Jan 2018 at 11:31pm

X399 Taichi (BIOS 2.00), 1950X at default, F4-3200C14Q-64GTZR
(full parts list)