B850M-X [WiFi] quick review > it killed my 9800X3D

As Xaltar wrote, I would've posted something if I got ANY useful information...
unless they made me sign an NDA or something to get a replacement,
which did not happen.

So nothing at all. They merely sent me a new-in-box replacement, which means, I got another sticker (laugh).

Old one: 2448PGY. New one: 2538PGE.

Since then, ASRock also replaced the motherboard, which came in a generic white box. I don't even know if it works because I never did anything with it.

I did build a new system with ASRock X600W USB4 Edition (tiny DeskMini), with a new R5-8600G. Love the 7-8W idle power consumption, with two NVMe drives in it. Weird that ASRock is so far behind with BIOS update on these little guys. Everything else is at AGESA 1.3.0.1b Patch A, but X600M-STX variants are still at 1.2.0.3g, which will be a year old in a few months from now. It's a shame because these little systems are really neat, and way more customizable than the usual "mini-PC" stuff out there.

Your timeline and troubleshooting details are extremely thorough, which makes me wonder about one thing: after AMD replaced your CPU, did they provide any failure analysis or indicate whether the returned chip showed signs of electrical degradation, CCD failure, or another specific hardware defect? That information could help determine whether this is primarily a CPU manufacturing issue or whether something in the platform is contributing to these failures.

VSOC and EXPO will strain the IMC more than JEDEC and conservative timings. Looking
at various forums where failures have been reported there is a disproportionately
larger number of EXPO users who have experienced failures (unlocked VSOC too in
many cases). The thing is, people taking precautions have also experienced
failures, like eccential the OP but these tend to be cases where the failure
occurred after 6+ months of use.

What you mention isn't something to be ruled out, it supports my assessment.
Lower VSOC and conservative RAM profiles (JEDEC) will place less strain on the
IMC and that tallies with the failures I have seen reported (baring in mind many
of these sources are unreliable). The majority of reported failures were using
EXPO and no limits on VSOC. There are still failures reported with locked VSOC
under 1.15v and no EXPO/XMP loaded (using JEDEC or conservative custom profiles).
4.10 seems to change a number of secondary timings for various RAM kits
even at the JEDEC level, this is one of the primary reasons I suspect the IMC
is degrading faster than AMD expected.

Bare in mind, I am just a community moderator, I don't have access to actual failure
rates, statistics and other data that AMD/board manufacturers would. This is
purely my personal opinion based on observations drawn from keeping an eye
on failure reports from various unofficial sources (forums like this one).

My point is, VSOC, EXPO, aggressive timings, misconfigured sub timings, all of
this puts strain on the IMC.

It's an incredibly difficult issue to pin down because everyone with a dead CPU
is reporting it as though their board killed it. This includes user error, overly
aggressive overclocking, DOA CPUs (QC issue by AMD), EXPO profiles that failed
to POST and a huge number of other causes. Basically, any time someone gets a
"00" code or DRAM and CPU debug lights are lit, they think the board killed their
CPU.
I have seen multiple cases where all that was needed was a proper CMOS clear
with battery out.
All this muddies the data. There is an issue but what symptoms are real and which
are misreported failures with completely unrelated causes?

We now have a lot more data, muddy as it may be, to guess based on. To me
"00" and DRAM + CPU LEDs lit indicate a failed IMC or IOD based on carefully
sifting through the data. I am not an electrical engineer however so my speculations
should be taken as just that.

With the recent uptick in failures on CPUs in use for over 7 months I am suspecting
more and more that the IMC is the issue. It also tracks with ASRock boards being
hit harder early on. ASRock has always pushed the RAM side of things in their BIOS
to get a competitive edge over other brands. It has worked very well for them in
the past so why wouldn't they continue doing what works for them. They had the
experience, knowhow and baseline data to work from.

IMO, and this is purely speculative on my part, AMD did exactly the same thing
with the 9k series, pushing the RAM to it's absolute limit to widen the performance
gap between the 7k series and the 9k series. Combine the already aggressive settings
from AMD via their AGESA and ASRock's usual performance tweaks and you have overloaded
IMCs. The thing is, I think AMD pushed the aggressive settings further than they
should have. This is based on the number of failures we are seeing 7 months+ out
from initial build. When a CPU fails within a month it's almost always either
a badly binned CPU or a serious issue with a BIOS version. When a CPU fails after
6 months+ of reliable use, it's degradation. Yes, BIOS settings can cause said
degradation, I used to see this quite often in my heavily overclocked systems, but
many of these reported failures are happening on other brands now. This tells me
that either AMD's binning is not aggressive enough or their RAM/IMC related settings
are overly aggressive and causing degradation.

Batch numbers would certainly be a factor here, if a batch was binned less aggressively
then it will have a higher failure rate. I imagine AMD has been steadily tightening
up their binning as the platform matures and yields improve as well as to prevent
the failure rate from climbing beyond acceptable industry failure rates. Basically,
they are doing what they are supposed to be doing.

My opinion is that the initial timings, voltages and related settings were
slightly too aggressive and this resulted in clustered failures (all IMC related).
The thing is, these failures still fall well within acceptable margins so we are
not talking some massive mess-up or critical error. Unfortunately the fact that
so many of the failures happened quickly and were reported on the same platform
all at once (redit) made the issue appear far more alarming than it is.

TLDR, I think IMCs in the 9k series are degrading faster than they did on the 7k
series. All IMCs will degrade, this is the Achilles heel of pretty much all modern
CPUs. AMD was a little too aggressive with their RAM performance push on the 9k
series and it just pushed the IMC a little too hard on less stringently binned
CPUs/batches. Given the physical IMC (contained in the IOD) in the 9k series is
identical to the 7k series I think they pushed it right to the edge with the 9k
series and it resulted in somewhat higher failure rates.

I could naturally be entirely wrong but that is my take for whatever it's worth.

Edited by Xaltar - 07 Apr 2026 at 3:55pm

Thanks for the info eccential, hopefully someone at ASRock or AMD will find it useful.

This issue has really got me scratching my head. I have seen other brands report
failures too but far fewer. I don't know if it's because ASRock sells more boards
in this segment or if their boards are actually more susceptible. I know both AMD
and ASRock are honoring RMAs without too much issue but the cause is what's bothering
me.

I don't have an AM5 system to test with myself so all I can do is look at reports
like yours and ponder.