Question Asus mobos burning x3D CpuS?

[aigomorla]

My 7950x3D working fine with Asrock board with EXPO. Soc at 1.25 and DRAM 1.35, set by Asrock not me.

Thats not how were diagnosing these.

Idle clocks is not what were looking at.
Its what happens when your boosting profile gets loaded, and your board decides to overclock the board in a factory "boost" scenario.

We are all thinking, during this boost profile, when the CPU literally floors the accelerator, it locks heat at X value, and then raises the clock and voltage like how i eat ticktac's at the office when stressed.

We are thinking the extra cache, and "gaming profile" the cpu has, its causing too much power draw on a lote socket which can not handle it, like how old foxconn sockets did on the old Core2's. (Larry's comment.)

Socket 1156 burning issues...is the fire out? - Overclockers

Several months ago, there was some news worthy mentions from many sites about the Foxconn made sockets for the P55 (Socket 1156) platform including Anandtech's article. When the issue was in its prime, one would see many forum users suggesting others to stay away from motherboard "X" because it...

www.overclockers.com

This is why the pad's are burning.
The Board and socket was not designed to handle the X3D's boost profile over the socket.

Easiest fix is to limit voltages, but that comes at the cost of overclocking.
And AMD is not keen on losing the all restrictions off overclocking race with intel.
But this will gimp that.

So the best possible fix... i hate sayIng it.
Move the pins over to spread out voltage load so its not so concentrated at the single point we see.
Which in turn will improve thermals, because you got no sun sitting in 1 spot, but more gas giants spread out.
And less heat means more efficiency, which means less voltage leak, which translates to what we want..

MOAR GHZ~!

 

[moinmoin]

Na, not happening. It's a software issue and board manufactures will get all the blame. AMD won't ditch a platform planned to be long living over some niche OC issue specific to some negligent manufacturers.

 

[aigomorla]

Na, not happening. It's a software issue and board manufactures will get all the blame. AMD won't ditch a platform planned to be long living over some niche OC issue specific to some negligent manufacturers.

Not saying were getting a revision this generation.

Saying it makes no sense to carry over a physical defect to the next generation, while carrying the crutches for it.
So the 7000 series will not get a board revision.
The 8000 series is where i am putting my guess on when we'll see that fix.

 

[tamz_msc]

Looks like board vendors are updating BIOS to limit voltages on X3D CPUs:

https://www.reddit.com/r/Amd/comments/12xmr24

All this could be avoided if there was more transparency - AMD needs to publish detailed datasheets like Intel publicly and tell board manufacturers to adhere to the electrical parameters more strictly, especially with CPUs like X3D which are more sensitive to them.

 

[Mopetar]

I'd be skeptical of that considering there are plenty of people (myself included) who have exceeded those voltages for SOC and IOD/MC for memory overclocking on X3D chips. Though I'm unlikely to be affected now since my recent focus has been minimizing idle power draw so my vSOC is currently 1.1V and IOD/MC is at 1.33V...

It might not be an issue for everyone, but there's going to a be a bar that exceeds the usual spec that some chips can't reach, possibly causing damage as a result. It's not much different from over licking where some people get the bottom barrel CPU that barely exceeds factory specs and other people get one that can guzzle voltage and easily add 400 MHz over what's guaranteed.

 

[reqq]

Thats not how were diagnosing these.

Idle clocks is not what were looking at.
Its what happens when your boosting profile gets loaded, and your board decides to overclock the board in a factory "boost" scenario.

We are all thinking, during this boost profile, when the CPU literally floors the accelerator, it locks heat at X value, and then raises the clock and voltage like how i eat ticktac's at the office when stressed.

We are thinking the extra cache, and "gaming profile" the cpu has, its causing too much power draw on a lote socket which can not handle it, like how old foxconn sockets did on the old Core2's. (Larry's comment.)

Socket 1156 burning issues...is the fire out? - Overclockers

Several months ago, there was some news worthy mentions from many sites about the Foxconn made sockets for the P55 (Socket 1156) platform including Anandtech's article. When the issue was in its prime, one would see many forum users suggesting others to stay away from motherboard "X" because it...

www.overclockers.com

This is why the pad's are burning.
The Board and socket was not designed to handle the X3D's boost profile over the socket.

Easiest fix is to limit voltages, but that comes at the cost of overclocking.
And AMD is not keen on losing the all restrictions off overclocking race with intel.
But this will gimp that.

So the best possible fix... i hate sayIng it.
Move the pins over to spread out voltage load so its not so concentrated at the single point we see.
Which in turn will improve thermals, because you got no sun sitting in 1 spot, but more gas giants spread out.
And less heat means more efficiency, which means less voltage leak, which translates to what we want..

MOAR GHZ~!

Wait you saying socket wasnt design with the 3D spec in mind? That doesnt make sense since 3D chiplet draw less power??? AMD have limit the max temp to 89c from 95c on the 3D cpues.

Anyway i see now my Asrock have new bioses as well, limit the vCore on 3D chiplet to 1.2 always it seems. But my 3d chiplet uses 1.355 on light loads and it been working fine for weeks so yeah i dont know.

 

[In2Photos]

Not saying were getting a revision this generation.

Saying it makes no sense to carry over a physical defect to the next generation, while carrying the crutches for it.
So the 7000 series will not get a board revision.
The 8000 series is where i am putting my guess on when we'll see that fix.

But this is only an issue with the X3D chips. So its not a socket problem. It's a CPU problem.

 

[GaiaHunter]

Something tell me it is not stock settings burning the CPUs. Not even regular expo settings.

 

[coercitiv]

All this could be avoided if there was more transparency - AMD needs to publish detailed datasheets like Intel publicly

Intel publishes detailed datasheets for everyone?! Can you point me to Intel's public resource that documents VDDQ for DDR4 platforms? I'd like to know the official safe voltage for that.

 

[ondma]

It might not be an issue for everyone, but there's going to a be a bar that exceeds the usual spec that some chips can't reach, possibly causing damage as a result. It's not much different from over licking where some people get the bottom barrel CPU that barely exceeds factory specs and other people get one that can guzzle voltage and easily add 400 MHz over what's guaranteed.

"over licking??" that sounds like fun

 

[deathBOB]

Don’t recall seeing physical damage like this from CPU overvolting. Guessing it’s a physical defect in those particular CPUs.

 

[aigomorla]

But this is only an issue with the X3D chips. So its not a socket problem. It's a CPU problem.

its a future socket issue, because it means going forward unless they can majorly optimize the 8000 series to not exceed X3D settings, or do a major pin allocation, they are at what is practically the physical max tollerance the socket can handle.

So no, yes and no.
Its currently a CPU problem, but if we go ahead forward, its going to also end up becoming a socket problem as well.

"over licking??" that sounds like fun

Well i would use the word putting a band aide over.
Or maybe using Zip Ties to hold together since our hobby is also highly dependent on zipties.

But for reasons i stated above, setting limits on a chip that was designed not to have those limits is not good product planning.

 

[In2Photos]

its a future socket issue, because it means going forward unless they can majorly optimize the 8000 series to not exceed X3D settings, or do a major pin allocation, they are at what is practically the physical max tollerance the socket can handle.

So no, yes and no.
Its currently a CPU problem, but if we go ahead forward, its going to also end up becoming a socket problem as well.



Well i would use the word putting a band aide over.
Or maybe using Zip Ties to hold together since our hobby is also highly dependent on zipties.

But for reasons i stated above, setting limits on a chip that was designed not to have those limits is not good product planning.

So how is that one can push the 7950X to use more than 200W without any socket or CPU damage, but the 120W 7950X3D is a socket problem? This really seems like a CPU issue related to SOC voltage. The X3D chips simply can't handle the higher SOC voltage/current that the motherboard's are using. AMD didn't allow any OCing on the 5800X3D and this is likely the reason why, it can't handle it. There's a very good possibility that if you crank SOC voltage on a 5800X3D you'd start seeing failures there too.

 

[aigomorla]

So how is that one can push the 7950X to use more than 200W without any socket or CPU damage, but the 120W 7950X3D is a socket problem?

Because your looking at the entire socket and not individual pins on the said socket.

It doesn't matter how many watts the entire package puts out, what matters is how many AMPS each of those select pins are pushing, and it seems like a collective bunch, so each pin is adding more to each other, melting the socket. Those corresponding pins on the socket obviously demanded more amps because it was a X3D cpu, otherwise we would of seen this a long time ago on the regular.

That is what those pictures shows to me.
In that grid the CPU and Board gave out too much amperage the socket was not able to handle said amperage to that location for prolong periods of time and essentially melted the socket, and burnt the cpu pad.

I am wondering if the CPU is still good, like can he wipe it with ISO, throw it in another board and see if it will boot up.

And lastly your comparison on the 5000 and 7000 series.
It holds no real value or any merit because first off, your on a completely different socket.
Your using thick pins on the cpu soldered onto the pad directly.

The 7000 series is a pin on socket like Intel.

And also the boosting features on both chips are completely different, unless you were to somehow simulate it on the 5800X3D you can't cross compare.

Boosting = Turbo, or how each CPU behaves when factory overclocking.

 

[GaiaHunter]

Didn't AMD say the reason the 5800x3d is clocked lower was due to the 3d-vcache chip having to run at lower voltages? Same for the 7000x3d?

So we do know they have to run at lower voltages than the non 3d chips. That is also why the 3d chips are so power efficient since AMD use higher quality bins.

 

[In2Photos]

Because your looking at the entire socket and not individual pins on the said socket.

It doesn't matter how many watts the entire package puts out, what matters is how many AMPS each of those select pins are pushing, and it seems like a collective bunch, so each pin is adding more to each other, melting the socket. Those corresponding pins on the socket obviously demanded more amps because it was a X3D cpu, otherwise we would of seen this a long time ago on the regular.

So do you believe that the pin layout is different for the non X3D vs the X3D parts? Are the X3D variants not using as many of the available pins? Wouldn't the socket have the same electrical limitations for each CPU? As we know P=IV. Package power fluctuations can be caused by both current and voltage. Lowering SOC voltage on my 7700 reduced package power by not only lowering the SOC voltage (also lowered VDDCR which is what Igor's Lab reports as being damaged), but also the current. So it is still possible that excess SOC voltage is causing too much current. There are so many posts of people applying lower SOC and not having any issues. Also, Der8uer fried a X3D chip by raising vCore to 1.5V. The X3D chips are vary voltage sensitive.

 

[In2Photos]

Didn't AMD say the reason the 5800x3d is clocked lower was due to the 3d-vcache chip having to run at lower voltages? Same for the 7000x3d?

So we do know they have to run at lower voltages than the non 3d chips. That is also why the 3d chips are so power efficient since AMD use higher quality bins.

They also lowered the TJMax indicating that they are more temperature sensitive as well.

 

[IEC]

Because your looking at the entire socket and not individual pins on the said socket.

It doesn't matter how many watts the entire package puts out, what matters is how many AMPS each of those select pins are pushing, and it seems like a collective bunch, so each pin is adding more to each other, melting the socket. Those corresponding pins on the socket obviously demanded more amps because it was a X3D cpu, otherwise we would of seen this a long time ago on the regular.

That is what those pictures shows to me.
In that grid the CPU and Board gave out too much amperage the socket was not able to handle said amperage to that location for prolong periods of time and essentially melted the socket, and burnt the cpu pad.

I am wondering if the CPU is still good, like can he wipe it with ISO, throw it in another board and see if it will boot up.

And lastly your comparison on the 5000 and 7000 series.
It holds no real value or any merit because first off, your on a completely different socket.
Your using thick pins on the cpu soldered onto the pad directly.

The 7000 series is a pin on socket like Intel.

And also the boosting features on both chips are completely different, unless you were to somehow simulate it on the 5800X3D you can't cross compare.

Boosting = Turbo, or how each CPU behaves when factory overclocking.

There's an alternative explanation. You exceed the safe voltage or temperature limit for the 3D stacked cache and this leads to a short-circuit condition resulting in the visible damage.

I'd wager that CPU is dead. Stacked cache is significantly more fragile versus non-3D chips and that's why they have lower voltage and temperature specifications in the datasheets...

 

[IEC]

Statement from ASUS to der8auer regarding removal of old BIOSes:



der8auer of course proceeded to manually set vSOC to 1.5V and run Cinebench loops for 30 minutes to see if unrealistic vSOC would kill his chip. Spoiler: It didn't.

 

[IEC]

Summary of der8auer's video:
1) The plot thickens: At least 2 examples of non-X3D chips (7700X, 7900X) affected by similar damage (though not all consistently in same pins) - as well as at least a few non-ASUS motherboards. Including one chip where the CPU got hot enough on a Gigabyte board to desolder itself from the IHS

2) EXPO seems to be involved somehow

3) To stay on the safe side, he suggests if using EXPO to manually lower SOC voltage (perhaps 1.2V). It's unclear if auto settings are somehow exceeding limits or it requires specific other conditions - as noted in his video he manually set vSOC to an unrealistically high 1.5V without killing his chip

 

[leoneazzurro]

Failures are not always deterministic. That is, if you have failures as a function of voltage, this does not mean that everytime you apply 1,5V you will kill the chip but you have a certain probability that applying 1,5V you will kill the chip. Over a certain value the probability becomes 100% so THAT will be deterministic.

 

[Puffnstuff]

MSI released a statement to update your bios ASAP to keep the bios from overvolting the cpu. ARS technica has an article on this issue that I read earlier today and it mentions several different things that could possibly contribute to this issue.

New BIOS updates attempt to keep Ryzen 7000X3D processors from frying themselves

Motherboard-maker MSI says damage may be "caused by abnormal voltage issues."

arstechnica.com

 

[tamz_msc]

Intel publishes detailed datasheets for everyone?! Can you point me to Intel's public resource that documents VDDQ for DDR4 platforms? I'd like to know the official safe voltage for that.

Sure, here's one for the Rocket Lake platform:



From page 125 of this document:

https://cdrdv2.intel.com/v1/dl/getContent/634648

 

[Kocicak]

A while ago I posted a thread about how 7800X3D possibly could have been made much better product by using the full potential of stacking dies on top of each other, which has been locked by some ill-advised moderator.

Troll thread unlinked

It was titled: Is Ryzen 7800X3D a half baked product nobody should buy?

It turns out, that in some cases 7800X3D can become a fully baked product...

Seriously: What do you do with such a puffed CPU? Do you send it to the baker (I mean MB manufacturer) for replacement, or do you send it to AMD?



3rd Mod callout. Also trying to revive a locked troll thread which got you a previous infraction for trolling the CPU forum.


esquared
Anandtech Forum Director

 

[tamz_msc]

New information from Paul Alcorn at Tom's Hardware.

AMD Ryzen 7000 Burning Out: EXPO and SoC Voltages to Blame (AMD Responds)

Impacts all motherboard makers and all Ryzen 7000 chips.

www.tomshardware.com

TL;DR: Industry sources say that excessive SoC voltage degrade the CPU's thermal sensors, leading to a runaway condition where the CPU draws excess power through the Vcore pins, disregarding thermal limits and killing itself.

EXPO causes the excessive SoC voltage to be set.