Question Asus mobos burning x3D CpuS?

[moinmoin]

For detailed CPU hardware data on Ryzen chips under Linux the out-of-tree modules zenpower (latest fork) and especially ryzen_smu were popular. Unfortunately neither ever got official support by AMD and volunteer development for newer chips like Zen 4 hasn't happened yet.

 

[tamz_msc]

Things are about to get interesting:

https://twitter.com/x/status/1652098512706838530

We have been able to reproduce a catastrophic failure resulting in the motherboard self-immolating while we were running external current logging, thermography, and direct VSOC leads to a DMM. The issue involves incompetence on many levels. Video script being finalized now. This video has taken us all week to do. We stepped off a plane from AMD HQ (unrelated, but convenient timing) on Tuesday and immediately set to work. It has been a time-consuming, expensive, physically exhausting process and we've loved every minute of it. The team is thrilled to have a real result and successful root-cause analysis that goes layers beyond "it's SOC voltage," because it's not that simple. We appreciate your support in viewing the video when it launches, because that's the real reward for our tired team. Keep an eye out and share it wide.

 

[IEC]

My guess:

Incompetence from AMD in AGESA with some conditions with EXPO leading to vSOC and other rails being set to match (or exceed) vDIMM
Incompetence from motherboard vendors in blindly applying non X3D profiles to X3D chips

Meanwhile I will watch with popcorn using my comfy manual 1.1vSOC.

 

[A///]

Gamers nexus is the guy doughy guy with the long 80s metal hair, right? I like his videos. Very informative stuff! looking forward to seeing what their findings are. i can't help but feel as if am5 specifically zen 4 was rushed or misworked on due to the pandemic and having to relieve staff once the economic contraction began.

 

[AdamK47]

My guess:

Incompetence from AMD in AGESA with some conditions with EXPO leading to vSOC and other rails being set to match (or exceed) vDIMM
Incompetence from motherboard vendors in blindly applying non X3D profiles to X3D chips

Meanwhile I will watch with popcorn using my comfy manual 1.1vSOC.

I can't even pass 5 seconds on Prime95 with 1.15V SOC.

 

[deasd]

This dead cpu story looks more and more like a random Internet "chilling effect" though......sorry for my bad english but I couldn't find a word describing this funny situation.

 

[A///]

This dead cpu story looks more and more like a random Internet "chilling effect" though......sorry for my bad english but I couldn't find a word describing this funny situation.

Probably not the correct word. Your English is fine but what would you describe it in your language? Would you describe it as a tall tale? Someone exaggerated the cause of it?

 

[deasd]

Probably not the correct word. Your English is fine but what would you describe it in your language? Would you describe it as a tall tale? Someone exaggerated the cause of it?

Well yeah lol. I have a feeling it's exaggerated. I already saw many dead cpus from both sides, and couldn't find a single rule of cause.

OTOH
What is coincidence is there's already a similar thread from Apr.1, I don't know whether it is related or not.

Info - Motherboard software bug can accidentally kill AMD Ryzen X3D CPUs

Might want to be careful with the voltage settings for your X3D overclocking. Motherboard software bug can accidentally kill AMD Ryzen X3D CPUs Apparently, the problem has been found in MSI, ASUS, Gigabyte, and ASRock boards. The BIOS seems to be mis-identifying the X3D chips as the normal...

forums.anandtech.com

The BIOS seems to be mis-identifying the X3D chips as the normal variant and allowing voltages to be set too high, which kills the CPU.

"Motherboard software bug allows users to manually exceed safe voltage limits for X3D chips"

 

[A///]

Well yeah lol. I have a feeling it's exaggerated. I already saw many dead cpus from both sides, and couldn't find a single rule of cause.

OTOH
What is coincidence is there's already a similar thread from Apr.1, I don't know whether it is related or not.

Info - Motherboard software bug can accidentally kill AMD Ryzen X3D CPUs

Might want to be careful with the voltage settings for your X3D overclocking. Motherboard software bug can accidentally kill AMD Ryzen X3D CPUs Apparently, the problem has been found in MSI, ASUS, Gigabyte, and ASRock boards. The BIOS seems to be mis-identifying the X3D chips as the normal...

forums.anandtech.com

It might fall under confirmation bias. A similar event happened when the 4090 cables began melting at the connection point. Every fault was over amplified. When whoever did the advanced testing later figured it was people not properly connecting their cards. a lot of people mess up seating their 24 pin connector fully and the findings didn't surprise me too much. I can't remember what the 3090 issue was or if it was legitimately a problem. People get too excited and keep repeating what they read as you say and it makes it seem as if there are more issues present than there really are.

 

[Timmah!]

I am on the older rig now, so cant do :-D
Will look into it over the weekend.

OK, so i checked and per HWinfo:

SOC is 1,25V, pretty much constantly
VDD is fluctuating, minimum is 1,136V and max one i saw being reported 1,44V. Only tried Cinebench and Google Earth though, no other stress inducing apps or games installed yet.

is that OK or should i lower it manually(especially the second 1,4V+ voltage)? I have the RAM that applies 1,4V with DOCP on (not 1,35V)

 

[IEC]

I can't even pass 5 seconds on Prime95 with 1.15V SOC.

Silicon lottery applies. I can't go below 1.20V on one of my 7700X. (For 6000 MEMCLK)

 

[aigomorla]

We are back in T-Bird days again...

Man... there is no way anyone can say this is OK.

Steve says he has a sample of one that had no OC and no EXPO. :X

You guys aren't safe... keep and eye on that voltage like your staring down a fat kid and a jar of white chocolate macadamian nut cookies.
Both you want to keep LEAN if you know what i mean.

Steve is also saying, its both Board and AMD's fault.

Pin Layout From GN:



@17:19 in the video steve says what i theorized.
The over protection on the board could not keep up with the socket.
Hence it pushed too much voltage after the cpu degraded from the high SOC, causing the board to overclock more, like Sisyphus rolling the boulder up the hill, until eventually the cpu or the socket physically fails.

Steve says offical ASUS statement, and i am assuming its with AMD, that max safe VSOC is 1.3
I would probably keep it under 1.2875 for possible +/- Vdroop.

tldr;
Issues found with ProcHOT, vSOC, OCP.
Issues with both AMD and Board Makers.


MAH... disclaimer...
This is ONLY the X3D chips.
If you do not have one, don't let this thread freak you out.

 

[DrMrLordX]

So a key takeaway is that lowering vSoC isn't necessarily the only fix that's needed (Asus and Gigabyte issues aside). What is good news is that it looks like most of this can be fixed via BIOS/UEFI updates.

By this point I figured we'd be out of the "paid beta tester" phase of AM5, but I guess not!

 

[aigomorla]

By this point I figured we'd be out of the "paid beta tester" phase of AM5, but I guess not!

welcome to 2023, where we beta test PC hardware ontop of 0-day AAA+ Game launches.
At least this way EA can say:
"Jedi Survivor is having hardware error because your 7000X3D cpu is churning up heat hotter then hell... So its TOTALLY not our fault we did a bad port, but your fault you used a 7000X3D processor and not a regular 7000 series. Oh and all the RTX4090 users who had problems are running this game on an i3 and need more cores. "

What next? Beta testing my retail car?
Oh wait i have a tesla... i do that too when i use FSD.

PS: im not kidding about what EA said to us RTX4090 users btw... they say im running it on a i3.

https://twitter.com/x/status/1651990800862183426

 

[A///]

will watch the video but that ea statement is so stupid. they must take their consumers for fools.

 

[coercitiv]

So this is 100% the fault of AMD & mobo makers. The few users who may have triggered this using manual OC settings will be a blip on the RMA radar, everyone else will be average JOEs who pushed the shiny EXPO button. Whatever electrical vulnerability AMD introduced, mobo makers made it multiple times worse by screwing around with settings and protections in every possible way.

Asus is on another level, they did this intentionally. Gigabyte is just incompetent (imagine this being the better excuse...). Asus sets 1.35V VSOC just from enabling EXPO, and then delivers 1.4V for good measure. On a $700 board. Like Steve said, this is not incompetence, it's just laziness. They want to obtain the juicy memory OC results while using as few engineering hours as possible.

It still feels like we haven't really gotten to the bottom of this, I expect more to come.

 

[A///]

Gigabyte is volt happy on z790 from what I have read, too. utter mess

 

[eek2121]

This is absolutely not an X3D specific issue, though it is definitely much more rare on non-X3D chips. I have seen at least 3 different reports of non-X3D chips having the exact same type of damage.

Lower your SOC voltage. There is no reason for it to be above 1.3V.

AMD continues to fail in the software department. If Intel can fix their perf/watt issues with Arrow Lake, my next machine will most certainly be powered by team blue.

 

[Puffnstuff]

AMD continues to fail in the software department. If Intel can fix their perf/watt issues with Arrow Lake, my next machine will most certainly be powered by team blue.

Validation issues like this one are why I went back to intel with my current build.

 

[DrMrLordX]

The few users who may have triggered this using manual OC settings will be a blip on the RMA radar, everyone else will be average JOEs who pushed the shiny EXPO button.

At least one of the reported users lost his CPU and he wasn't even using EXPO.

Lower your SOC voltage. There is no reason for it to be above 1.3V.

That seems to be the only fix available at this time; however, bear in mind that a VSoC setting of 1.3v may produce an actual vSoC of up to 1.35v depending on your board.

 

[A///]

If Intel can fix their perf/watt issues with Arrow Lake, my next machine will most certainly be powered by team blue

as fat a chance as docs developing hair transplant using your bum hairs.

 

[coercitiv]

AMD continues to fail in the software department. If Intel can fix their perf/watt issues with Arrow Lake, my next machine will most certainly be powered by team blue.

Validation issues like this one are why I went back to intel with my current build.

You poor little unsuspecting souls.

 

[DAPUNISHER]

This is another opportunity to give all of the tech tubers and reviewers involved props for the consumer advocacy. Until this SNAFU is fixed, at least awareness of it, and the ability to get a replacement or refund, will be much higher than otherwise.

I am increasingly comfortable with my decision to sit out Zen 4. I knew it would be rough sledding, but this is like skiing with Gwyneth Paltrow. ☠️

 

[eek2121]

At least one of the reported users lost his CPU and he wasn't even using EXPO.



That seems to be the only fix available at this time; however, bear in mind that a VSoC setting of 1.3v may produce an actual vSoC of up to 1.35v depending on your board.

Yep, ASUS does this. Watch the video.

It is mind boggling TBH.

 

[CakeMonster]

Asus released the 1410 BIOS after this video was shot, that was the first BIOS for me that got 1.30v SOC right on Auto, before that DOCP/EXPO would still be at 1.35v.