Question Asus mobos burning x3D CpuS?

[JoeRambo]

I am running 1.25V SOC on Zen4 16Cs and it does not sound like much to me. But i haven't enabled EXPO either, so no idea what Gigabyte does when it's enabled.
Still i think it is well known by now from XMP and EXPO that motherboard makers will do their best to avoid returns due to not working with some RAM. And with the way AMD DDR5 compatibility was during early days of Z4, one cant really blame them. Vendors are also setting mad CCD/IOD voltages, that could also backfire. I think ASUS defaulted to like 1.1V there. ( <= in retrospect, they probably did so to not have too large differential between SoC and DF voltages ).

Since there are failed motherboards from both Asus and Gigabyte, i don't think it is fair to blame enthusiast motherboard for feeding 400W of power and not tripping OCP. It sure is subtle to all those "00" codes blabla, but 400W of power is not wrong in itself in well built motherboard.
The real culprit is AMDs Zen4 engineering and tolerances, something we saw during Z4 release that suspiciously too many review time chips died, sometimes even on video like happened with Buildzoid. That already means that some chips might be real vulnerable to accelerated degradation. Doesn't really matter if it's on IO side, dead chip is dead chip.
My chips are well cooled, running 5Ghz undervolts, HT disabled, but i guess the I/O side load from 2x32GB and elevated SoC voltage presents unique challenges to longevity. I guess we'll see what happens. IF VSoc limit is 1.3V in BIOS, that does not mean 1.25V is safe

 

[Ranulf]

I'm 2 mins in and "we used themography to prove that yes explosions are hot".

I'm sure this will all be fine. Just fine. No problems at all.

 

[In2Photos]

I wonder if the differences in vSOC are influenced by CAS latency. Would be interesting to see what various motherboards, RAM used, CAS and resulting voltage is. I thought about trying out my daughter's GSkill kit in my PC to see what vSOC is applied by the motherboard.

 

[JoeRambo]

I wonder if the differences in vSOC are influenced by CAS latency. Would be interesting to see what various motherboards, RAM used, CAS and resulting voltage is. I thought about trying out my daughter's GSkill kit in my PC to see what vSOC is applied by the motherboard.

I think the golden rule is "you do not leave VCCSA/VCCIO or VSOC " on auto when overclocking memory. They have "rules" that scale with speed and timings and anything not in profiles can get touched. And i mean anything from voltages to training modes to relaxed tertiaries.
Do note that i do not even consider enabling XMP/EXPO. In the past MB vendors were changing settings that were not even in BIOS to make XMP work, so golden rule was to clear CMOS and to proceed with manual memory OC if XMP was ever enabled on that setup.

 

[Abwx]

but 400W of power is not wrong in itself in well built motherboard.

400W.?..
Seriously?

This amount to roughly 250A, and i dont think that a consumer CPU can stand such current and the induced power continuously, stock 230W should be considered as a maximum since AMD undoubtly made tests at this power applied during a full month, or even more, to check reliability.

Now those who overclock beyond those values do so at their own risks and are no more covered by AMD s warranty, and it s obvious that the ones who first had a problem are exactly in this case and are now whining and noising in the net to get AMD to pay for their risky adventures.

 

[aigomorla]

but 400W of power is not wrong in itself in well built motherboard.

I don't think even the godly 28+ phase digiVRM boards we used to have could handle 250A though that socket without smoking and going poof.

 

[DrMrLordX]

So two thirds of the issue are from AMD, weak io die and weak CCD is what is killing the CPUs.

That's not even remotely true. If you had paid any attention to what GN published in their video, you would know this. In fact, GN has basically said "this is not going to be a problem for the vast majority of you".

Assuming the software reading is accurate and ASUS isn't using some VRM trickery.

GN's conclusions seem to point to Asus boards routinely misreporting vSoC by up to .05v. So it's possibly/probably misreporting on the low side.

Yes, if CPU's where burning up all over the place we would probably have known. But even with 5 boxes you are nowhere near to have a statistically significant amount, to say anything about the probability for it to happen, and if only ASUS boards have been running the vsoc out of spec, then only one of your boxes would be at a higher risk.
But I'm very glad to hear that all your ASRock boards also are doing well

One of the other takeaways from the GN video is that even if you don't get a catastrophic die-cracking failure, there are problems with excess electromigration from running high vSoC through auto settings, prematurely aging the CPU. People may be fine now, but what happens when some of these systems enter their second year of use or later? I was on my 3900x for almost four years. Well technically I'm still on it but whatever.

This amount to roughly 250A, and i dont think that a consumer CPU can stand such current and the induced power continuously, stock 230W should be considered as a maximum since AMD undoubtly made tests at this power applied during a full month, or even more, to check reliability.

Remember that some of these high-dollar boards (like the x670e Hero) do cater to "xtreme" overclockers that might want to be able to push more than 230W chasing some record or another. Not saying the x670e Hero or Extreme are the ideal boards for that particular enterprise, but for the tiny percentage overclocks that invovle LN2 or LHe runs, Asus wants to be front-and-center with their logo everywhere if possible. Possibly that mindset is a bit dated. This occasions marks one of those rare times when you don't necessarily want massively-overbuilt power delivery. Most of us users that wind up buying that board (myself included!) will never want or need some of the "xtreme" features available.

Now those who overclock beyond those values do so at their own risks and are no more covered by AMD s warranty, and it s obvious that the ones who first had a problem are exactly in this case and are now whining and noising in the net to get AMD to pay for their risky adventures.

Ummm I'm not really sure that any of the people featured on GN or in the Reddit megathreads actually did that.

 

[biostud]

That's not even remotely true. If you had paid any attention to what GN published in their video, you would know this. In fact, GN has basically said "this is not going to be a problem for the vast majority of you".



GN's conclusions seem to point to Asus boards routinely misreporting vSoC by up to .05v. So it's possibly/probably misreporting on the low side.



One of the other takeaways from the GN video is that even if you don't get a catastrophic die-cracking failure, there are problems with excess electromigration from running high vSoC through auto settings, prematurely aging the CPU. People may be fine now, but what happens when some of these systems enter their second year of use or later? I was on my 3900x for almost four years. Well technically I'm still on it but whatever.

I agree that it is a major problem that board vendors do not have better quality control on their boards so they aren't full of hidden bugs.

 

[TheELF]

1) If your BIOS sets the voltage for a rail to something wildly out of spec, whether intentional or not -
Is that a "weak" chip, a fault of the BIOS maker who set the wrong voltage, a fault of the chip maker who didn't properly delineate tolerance differences between X3D and non-X3D chips, or some combination of the above?

If anything above 1.35V is considered "wildly" out of spec than isn't that in itself a serious issue?
If you buy a CPU in 2023 you expect it to be of a certain build quality.

400W.?..
Seriously?

This amount to roughly 250A, and i dont think that a consumer CPU can stand such current and the induced power continuously, stock 230W should be considered as a maximum since AMD undoubtly made tests at this power applied during a full month, or even more, to check reliability.

Now those who overclock beyond those values do so at their own risks and are no more covered by AMD s warranty, and it s obvious that the ones who first had a problem are exactly in this case and are now whining and noising in the net to get AMD to pay for their risky adventures.

It's not reaching 400W because somebody set that value, the opportunistic overclock of AMD keeps pushing volts and watts as long as the temperature is below a certain amount, if the temp sensor doesn't work because it was of such a low quality that it blew up then it's always below that temp so the opportunistic overclock keeps pushing volts and watts until something else blows up, which is the mobo.

That's not even remotely true. If you had paid any attention to what GN published in their video, you would know this

I did pay attention to the posted pic that put the blame on the io die the ccd and the bios, two of these three nobody has any influence on except for AMD.

In fact, GN has basically said "this is not going to be a problem for the vast majority of you".

Yes, he said "don't worry since not every CPU will blow up" something like that.
Sarcasm is probably a dead art form.

Edit:
Found the exact quote.
Steve at 5:51:
You don't need to panic completely, only some of you will have this happen.
*Smirk * nudge *laugh *test-picture

 

[sniffin]

Mobo vendors have been doing stuff like this for years. It's a problem now I think because all is not right with these Zen 4 desktop CPUs. Very very curious to see how these chips are going 1-2 years from now, as above posters also suggest.

 

[coercitiv]

i don't think it is fair to blame enthusiast motherboard for feeding 400W

I don't think even the godly 28+ phase digiVRM boards we used to have could handle 250A though that socket without smoking and going poof.

I'm telling you folks, this little fiasco is going to bring out a lot of nasty business about the current state of modern motherboards:



Q: How much current would you like?
A: ALL THE CURRRENT!!!

PS: In case anyone wonders, the default "Auto" Current Limit on this board is 240A, however any attempt to change clock behavior will be prompted with an automatic switch to a so called "Water Cooler" profile. There is no "Cancel" button and pressing Escape only gives you the impression of canceling, as the profile is already applied. The Current Limit is now 512A.

Spoiler: Prompt

 

[JoeRambo]

Q: How much current would you like?
A: ALL THE CURRRENT!!!

And how exactly is that wrong on enthusiast motherboard? Let's ignore the failure mode and focus on good old OCed setups like we saw in this and other forums. For example i've seen @Det0x push ~ 300W in 5950x and obviously to do that motherboards needs to handle load transient currents way above what 300W / volts would demand or else it would be tripping OCP all day long on load changes.
The problem is CPUs that fail due to accelerated degradation mode. Failure of OCP on some vendors to handle CPU that is dead, but not in complete short, but enough resistance to fool motherboard into burning socket is a victim of AMD's failure.

Only thing i wonder is where safe VSoc will land after this debacle

 

[igor_kavinski]

All this current flowing through CPUs makes me wish there was a utility to simulate electric buzzing sounds based on CPU usage so you get an audible clue of what your poor CPU is having to endure

 

[reqq]

That's how I interpreted it. It seems like Steve kind of overpromised with this video. He spent like 2/3 of the video talking about how horrible ASUS is but that doesn't really root cause the issue. He identified a ton of motherboard and BIOS bugs (one big one for Gigabyte too) however, that doesn't fully explain what's going on. It seems like we've got an explanation for how the socket got damaged on ASUS boards but that's all that's explained in this video.

What happens when the CPU blows up but the motherboard is intact? It seems like for that one we've got to wait for the failure analysis lab and we're still weeks away on that one. We've got a theory from Wendell about what's happening on the CPU side but it's just a theory (for now).

Exactly this. More to this. So we have like 8 known dead cpus or something out of thousands, think about that for a second. Asus is the most popular brand so hundreds if not thousands have been using Asus board with a 3D cpus yet only like 4 cases with this combo. You need to be incredible unlucky for that to happened. Also you really think you gonna kill a cpu going from - safe - 1.3 SOC to 1.4, a mere 7% overclock??? I been OC computers a decade and going way beyond that lol without any exploding cpus or graphic cards.

 

[coercitiv]

And how exactly is that wrong on enthusiast motherboard?

It's wrong because it's automatic! It's wrong because it introduces this awful chain of dependencies that is borderline invisible to the average user. Even when you want to maybe lower your boost by 100Mhz, this "feature" kicks in and effectively removes TDP and current limits. You said it yourself that you avoid enabling XMP due to automatic UEFI changes that may hurt the system. How friendly is that for anyone, including enthusiasts?

It's the enthusiast who should know to change a couple of limits and remove safety mechanisms when needed. They shouldn't be removed for you at the first sign of customization, especially considering they are in a different screen within the UEFI.

 

[TheELF]

Exactly this. More to this. So we have like 8 known dead cpus or something out of thousands, think about that for a second. Asus is the most popular brand so hundreds if not thousands have been using Asus board with a 3D cpus yet only like 4 cases with this combo. You need to be incredible unlucky for that to happened. Also you really think you gonna kill a cpu going from - safe - 1.3 SOC to 1.4, a mere 7% overclock??? I been OC computers a decade and going way beyond that lol without any exploding cpus or graphic cards.

Exactly, we have like 8 known dead cpus, we have no idea how many RMAs went to AMD or asus or any of the other mobo makers to make them get off of their lazy buts and come out with "fixes" faster than we ever saw for anything else.
Also we have no idea of how many CPUs are affected but will only degrade slowly.

And it's not about an 7% overclock killing the CPU, it's about the thermal sensor/protection failing with a high degree of possibility at above 1.35V.

 

[eek2121]

A crappy IOD need more than just 1.3v to scale if you haven't play with one.
Whether it's worth pushing on such die is up to one's preference, but your statement is false.

Find me one Zen 4 CPU that needs more than 1.25v-1.3v SOC at stock settings. I will wait.

If you overclock the GPU and/or RAM that may be another story, but this issue is happening at STOCK settings, hence the alarm.

I wonder if the differences in vSOC are influenced by CAS latency. Would be interesting to see what various motherboards, RAM used, CAS and resulting voltage is. I thought about trying out my daughter's GSkill kit in my PC to see what vSOC is applied by the motherboard.

Not likely.

Also, FWIW my 7950x has no issue with 64gb DDR5 6000 32/38/38/32 and a 1.25v SOC.

Exactly this. More to this. So we have like 8 known dead cpus or something out of thousands, think about that for a second. Asus is the most popular brand so hundreds if not thousands have been using Asus board with a 3D cpus yet only like 4 cases with this combo. You need to be incredible unlucky for that to happened. Also you really think you gonna kill a cpu going from - safe - 1.3 SOC to 1.4, a mere 7% overclock??? I been OC computers a decade and going way beyond that lol without any exploding cpus or graphic cards.

8 known so far. Not everyone who has this issue will report it, and had AMD done nothing, months/years down the road a lot more would have failed. Degradation does not happen overnight. Ask any long time overclocker that.

 

[DrMrLordX]

I did pay attention to the posted pic that put the blame on the io die the ccd and the bios, two of these three nobody has any influence on except for AMD.

The problem is that voltages are being run out of spec by default. There's nothing "defective" about the actual silicon. You can't run 400W through the socket and expect anything to survive!

 

[aigomorla]

The Current Limit is now 512A.

someone who translated that bios from chinese to english must of gotten that wrong.
The A must mean GB.... and was intended to for GPU memory allocation, not current. :O
I dont see any other case where we would use a 512, as its divisible by 4, and typically used on RAM until reciently with the inception of 24GB stick.

Oh wait thats even divisible by 4.

They shouldn't be removed for you at the first sign of customization, especially considering they are in a different screen within the UEFI.

Pish everyone knows even your granny can overclock and intel, so AMD said if your granny can, why not?

AMD is trying to 1 up this... saying even your granny can break PCMARK on a AMD Ryzen X3D. *serious face*
*ignoring the fire in the background*

 

[Khato]

8 known so far. Not everyone who has this issue will report it, and had AMD done nothing, months/years down the road a lot more would have failed. Degradation does not happen overnight. Ask any long time overclocker that.

Exactly. 8 known so far within the first 2 months of availability. Time will tell if this is a ticking time bomb with a far higher failure rate or if the mitigations AMD is deploying negate the issue.

 

[TheELF]

The problem is that voltages are being run out of spec by default. There's nothing "defective" about the actual silicon. You can't run 400W through the socket and expect anything to survive!

Yes, that is also why people think that the intel CPUs are hot and using lots of watts, what was your opinion on that?!

Also the 400W is a result of the CPU turboing without a thermal limit because the thermal sensor stopped working, and you would expect any CPU from 2023 to at least be ok up to 1.5V ,not for 24/7 but at least for some time.

If you had crazy expectations you would expect the CPU to shut down before it gets damaged, but I guess nobody expects a good CPU from AMD.

Here, this is what a crazy person would expect from a CPU.

 

[aigomorla]

If you had crazy expectations you would expect the CPU to shut down before it gets damaged

no...
we assumed throttling and voltage were linked like in intel.
So when the cpu throttled, it would drop voltages across dynamically.
The CPU showing signs of throttle would of helped people figure things out faster.

But instead i have no idea how throttling works in a case like this.

 

[SK10H]

Find me one Zen 4 CPU that needs more than 1.25v-1.3v SOC at stock settings. I will wait.

You should state 1.3v vsoc is OK enough for basic EXPO speed at 6000 with fclk at 2000. Yes, EXPO setting is really basic and crap without further tweaking. Stock max support is 5200 if you prefer to nitpick, don't mix and match and classify EXPO as stock here. When talking about vsoc, you definitely need to bring in fclk as well, otherwise go find a cpu that can do 6000mem,2233 fclk cpu, I will wait. 🤨

Your original statement is there's no reason to go beyond 1.3v vsoc, under what scenario???? I can easily name the oc scenario that invalidate your statement as you did not explicitly state it clearly and basically imply oc users are idiots pumping more than 1.3v vsoc for no reasons for some time.

1.3v vsoc now are hard cap on new Agesa. Your post makes it look like 1.3v vsoc is not a big deal and rarely affect anyone. Hard cap is a pain to deal with.

I have made a follow up statement and so have others. But you have not quoted on that, so you know full well what I stated here obviously. Yes, might be nitpicking on my part, but I believe it's also the second time you imply 1.3v hard cap vsoc is not a big deal without considering even the basic 6200 oc, > 2000 fclk people which are plenty.

 

[In2Photos]

Yes, that is also why people think that the intel CPUs are hot and using lots of watts, what was your opinion on that?!

Also the 400W is a result of the CPU turboing without a thermal limit because the thermal sensor stopped working, and you would expect any CPU from 2023 to at least be ok up to 1.5V ,not for 24/7 but at least for some time.

If you had crazy expectations you would expect the CPU to shut down before it gets damaged, but I guess nobody expects a good CPU from AMD.

Here, this is what a crazy person would expect from a CPU.

Good grief, here we go again with the "well it doesn't work like Intel so it must be crap!" sputum. It is entirely possible for different types of chips to behave differently in regards to voltage. And let's not pretend that there aren't 8 Intel 13900K CPUs out there that experienced premature failure, even at stock settings. Sure they may not have done so in a "flashy" way, but they are out there.

 

[DrMrLordX]

Yes, that is also why people think that the intel CPUs are hot and using lots of watts, what was your opinion on that?!

Dude not even Intel chips use that much power. Get a grip. If you pushed 400W through a 13900k it would annihilate the package. Plus you're not even comparing like-to-like when it comes to power density, etc. You'll notice that only small portions of the AM5 socket/pinouts were smoked on the failing ASUS boards.