AMD Ryzen 7000X3D Processors Prone to Physical Damage with Voltage-assisted Overclocking, Motherboard Vendors Rush BIOS Updates with Voltage Limiters

[AusWolf]

Based on more recent reports of it potentially affecting non-X3D chips, I am thanking my lucky stars that I have never turned on any of the auto-overclock features like PBO and EXPO. I've also let my boss know since he just finished building a 7950X3D build.

I've only gotten back into the new PC building game recently, have new platform launches always been this "interesting"? And if AMD keep making the news it's going to make it harder to playfully rib my Intel/Nvidia enthusiast workmates, sheesh.

AM5 has been along for about half a year now. Somebody fried their 7800X3D, and only now "suddenly", we get sporadic reports (Der8auer?) from "people" with their non-3D chips dying. I'd say it's not just a little suspicious. I'm smelling Youtube sensationalism again.

OK, so I disabled EXPO and these are the results...

Enabling EXPO automatically added .21v (1.245 - 1.035 = .21) for a maximum of 1.245 volts on the SOC. That's two tenths of a volt here that my motherboard added by default just by enabling EXPO.

What's really stupid here is that outside of benchmarks, turning off EXPO has no discernable change to how my system feels. I don't notice any kind of slowdown in how long it takes to boot Windows, there's no slowdowns in how web pages are rendered, the program that I compile from C++ source code using Microsoft's Visual C++ Compiler takes very nearly the same amount of time that it took with EXPO enabled, nor has it really affected Cinebench R23 scores (multi-core score of 19179).

So, I'm going to keep EXPO off for the time being until new UEFI versions come out. So far, Gigabyte hasn't come out with a new version for my board yet.

Yep. Memory OC is overrated.

 

[trparky]

I'd say it's not just a little suspicious. I'm smelling Youtube sensationalism again.

True, but that doesn't mean that people shouldn't still be a little cautious until we know more.

 

[1stFalloutboy]

I am wondering how safe my Under/Overclock will be. I brought this 7600Mhz memory as an experiment - quite an expensive one granted but it might now be paying off.

Although I setup AMD Expo with memory timings for a 6000Mhz memory module which I then overclocked to 6200Mhz ( or underclocked depending on how you look at it ), I set the voltages manually.
Vsoc 1.2v
MemVDD 1.28v
MemVDDQ 1.18v
It looks by the looks of ZenTimings below that I may be able to lower the starting Vsoc voltage even lower as it is currently .0150v lower than what I specified I just have to confirm this voltage is accurate with other software and perhaps a multimeter.

What are the chances though that this will cause the damage discussed?

 

[AusWolf]

True, but that doesn't mean that people shouldn't still be a little cautious until we know more.

I agree, though based on the fact that we only have one report from one user, and some "updates" from Der8auer, who's notoriously famous for "reporting" on everything AMD being damaged, faulty, breaking, etc, and conducting biased tests to support his claim, I'd say the issue is blown way out of proportion.

We've had literally zero such reports since AM5 released, and now we "suddenly" have this. It smells fishy to me.

 

[jigar2speed]

It is not the first time that manufacturers have to respond to AMD's blunders. It is clearly a mistake in the design of the socket and the protection of the processor. Hard to destroy a processor these days. He doesn't actually let you kill him.

Gica, you okay there little buddy ? You keep insisting its AMD's fault where as everyone knows AMD has blocked OCing on X3D chips. If motherboard makers are bypassing recommended protocols, its on motherboard makers and not on AMD.

What is so difficult to understand ? Please explain.

 

[JustBenching]

LGA1700 - issues with the socket bending and first-gen non-heterogenous cores requiring numerous bios fixes, patches, and scheduler re-writes to iron out all the problems.

That's just unfair. I had a 12900k from day one,, never updated the bios or the chipset drivers or anything. Didn't find any issues whatsoever besides a single game that didn't want to play ball (ac valhala) so I had to press literally 1 button on my keyboard to disable ecores and launch the game.

The socket bending is an issue with every single motherboard and cpu in existence. They all do that. Others more, others less. I never experienced it myself

Gica, you okay there little buddy ? You keep insisting its AMD's fault where as everyone knows AMD has blocked OCing on X3D chips. If motherboard makers are bypassing recommended protocols, its on motherboard makers and not on AMD.

What is so difficult to understand ? Please explain.

That's nonsense. Do you understand the difference between a CPU dying cause of electromigration and the cpu frying itself? The latter should not happen just by "overclocking" or god forbid enabling expo, lol

 

[Bomby569]

Gica, you okay there little buddy ? You keep insisting its AMD's fault where as everyone knows AMD has blocked OCing on X3D chips. If motherboard makers are bypassing recommended protocols, its on motherboard makers and not on AMD.

What is so difficult to understand ? Please explain.

It had to be some insane OC to melt solder, unheard of territory. And EXPO should not kill a CPU, the CPU should have protections in place to avoid that. That's just apologising for for the sake of apologising, no real reasoning behind it.

 

[1stFalloutboy]

For reference I just installed the new "mitigated" bios and my O/C is still working.

 

[Berfs1]

In your example if FORD show in an advertisement that their cars can run submerged and when you do it, it kills the engine, that is false advertisement isn't it

oh i didn't mean submerged in water, just meant driving on an area that is below the average sea level haha

Anyways, you do bring a good point, AMD even told reviewers to use 6000 MHz for their reviews, so in AMD's eyes, that should be perfectly fine. Unless they were like Intel telling that one person to last minute overclock that 28 core xeon on a chiller lol

 

[R0H1T]

It's false only if you can prove it, the US has pretty good consumer protection but I'm sure manufacturer have gotten away with worse even there!

 

[Vayra86]

I wouldn't pin it on the user so soon. DIY motherboards often run things out of spec even by default.

Correct, I remember this one vividly

Don't Run Z490 Motherboards with Default Settings: Thermals, Power, Boosting, & MCE for 10th Gen CPUs

It’s difficult to differentiate motherboards, at least from a marketing perspective. There are definitely better and worse boards, and you can check any of the roundups or reviews Buildzoid has produced for this channel for explanations as to why, but “better” doesn’t mean “higher FPS in games”...

www.gamersnexus.net

I could easily fish for another handful of generations/sockets where this has occurred at one or more mobo mfgrs.

So again... user due diligence is needed, has been needed and will always be needed with DIY building. Its why we call it DIY. People need to get a life and a good bit of common sense. And AMD/mobo partners need to lock this down on X3D's just as well, because there is literally no purpose to leave it unlocked. And if unlockable, it needs to come with warning on the UEFI toggle.

Frankly if you fried an X3D, you're just a n00b. And you just learned a thing. RMA it / get new one
/thread

 

[JustBenching]

Correct, I remember this one vividly

Don't Run Z490 Motherboards with Default Settings: Thermals, Power, Boosting, & MCE for 10th Gen CPUs

It’s difficult to differentiate motherboards, at least from a marketing perspective. There are definitely better and worse boards, and you can check any of the roundups or reviews Buildzoid has produced for this channel for explanations as to why, but “better” doesn’t mean “higher FPS in games”...

www.gamersnexus.net

I could easily fish for another handful of generations/sockets where this has occurred at one or more mobo mfgrs.

So again... user due diligence is needed, has been needed and will always be needed with DIY building. Its why we call it DIY. People need to get a life. And AMD/mobo partners need to lock this down on X3D's just as well, because there is literally no purpose to leave it unlocked. And if unlockable, it needs to come with warning on the UEFI toggle.

Yes and no. Z490 pushed insane sa voltages (basically imc) whenever you tried to run DDR4 at 4400 and upwards like 1.5 - 1.6 sa voltage was quite common at auto settings. That is almost insta death territory for the imc. But it would never lead to that catastrophic of a failure we are seeing with the am5 here. Your cpu would just die, not cook itself and the mobo.

 

[Vayra86]

Yes and no. Z490 pushed insane sa voltages (basically imc) whenever you tried to run DDR4 at 4400 and upwards like 1.5 - 1.6 sa voltage was quite common at auto settings. That is almost insta death territory for the imc. But it would never lead to that catastrophic of a failure we are seeing with the am5 here. Your cpu would just die, not cook itself and the mobo.

But does it cook itself and the mobo, or are we looking at a big bunch of lies?
We're still talking about an extremely rare occurence and a world full of people looking for ad revenue with cool videos
If this was a board default thing even on ONE board of ONE vendor, we would have seen many more issues. Shit just doesn't line up

And honestly I challenge myself to read articles from any of the random tech sites online, and every time I'm literally astounded by the lack of actual article in each of them. Basically its regurgitating everything every gen and looking for storms in teacups 99% of the time. Sure, things also truly, rarely happen, but boy are we making it big and boy are we bad at fact checking.

Similarly, look at the Nvidia space invaders, the supposed faulty memory from vendor X or Y, the 12 pin horror show et al. While truly issues, they're also blown way out of proportion. I look at myself too - I'm not immune to those hype trains either, but in retrospect, every time, its a lot smaller than we are led to believe.

 

[JustBenching]

But does it cook itself and the mobo, or are we looking at a big bunch of lies?
We're still talking about an extremely rare occurence and a world full of people looking for ad revenue with cool videos
If this was a board default thing even on ONE board of ONE vendor, we would have seen many more issues. Shit just doesn't line up

Well - what I said was assuming the rumors are true. For all we know it could just be Asus doing Asus things again, not the first time their super high end mobos end up being trash. My z690 apex.... yikes

 

[Klemc]

Well - what I said was assuming the rumors are true. For all we know it could just be Asus doing Asus things again, not the first time their super high end mobos end up being trash. My z690 apex.... yikes

Asus and all others, x3d and non-x3d, read more pages

 

[Bomby569]

Asus and all others, x3d and non-x3d, read more pages

asus, gigabyte and asrock, according to derbauer

 

[Klemc]

asus, gigabyte and asrock, according to derbauer

BIOSTAR too...

 

[R0H1T]

With how many chips failing? Is there a scientific conclusion that can be drawn from any of these "outrageous" observations or just click-bait/fluff videos!

 

[Klemc]

I lost 10°c at stock by using MX5 instead MX4 btw

 

[1stFalloutboy]

I lost 10°c at stock by using MX5 instead MX4 btw

MX5 and MX4, I'm still using Arctic Silver.

 

[Bomby569]

With how many chips failing? Is there a scientific conclusion that can be drawn from any of these "outrageous" observations or just click-bait/fluff videos!

a lot, but it's not so much the quantity but how catastrophic the failure is
no scientific conclusion, the best you can get is a theory from derbauer

 

[1stFalloutboy]

Well I've upgraded to the mitigated BIOS so barring any further information I am hoping that is safe enough, I will not be happy if things go POP.

 

[phanbuey]

Alrighty, here's the deal with warranty. If you are in the US, you are covered under the Magnuson Moss Warranty Act, because applying EXPO also changes other settings that the motherboard auto defined. Because the MOTHERBOARD entered in the unsafe settings and not you, you wouldn't be liable for this kind of damage, the motherboard manufacturer would. While yes the manufacturer of a product has to prove you caused the damage, in this particular case since the motherboard was auto defining unsafe values as safe, it would be the motherboard manufacturer that would be liable for this. Also why did AMD not test these chips beforehand? If over 1.3V SOC voltage instantly kills CPUs, AMD should have known about this prior to launching them. This is mainly AMD's fault for not testing their products and telling motherboard manufacturers what values to NOT auto define. Imagine Ford shipped you a car that should work in ideal conditions, but when driving below sea level, the engine blows up because of the higher air density. Why would that be YOUR fault because you drove it in a place that was 200ft below sea level? It's not like you are visiting a volcano or going to the core of the earth which would need a special kind of car, it's 200 feet below sea level.

Ford ships engines that blow up on a regular basis - but I like your analogy.

 

[R0H1T]

Moral of the story ~ don't run AMD near a volcano

 

[Gica]

Gica, you okay there little buddy ? You keep insisting its AMD's fault where as everyone knows AMD has blocked OCing on X3D chips. If motherboard makers are bypassing recommended protocols, its on motherboard makers and not on AMD.

What is so difficult to understand ? Please explain.

i5-13500
SA Voltage: locked
Overclocking: locked
All the voltages provided by the processor are within the parameters imposed by Intel.
If with Alder they had a small escape with overclocking (non-k processors with very expensive motherboards), with Raptor not even God can overclock a non-k processor.
This means the processor is blocked!

However, this is not the problem. The serious problem is that the socket and/or the processor deteriorates quickly and without warning. I remain in my opinion that it is a design error, aka insufficient pins allocated to power the processor.
Of course, I'm not an expert, you don't have to listen to me and try to break world overclocking records with AM5. God be with you!