News Ryzen 7000 I/O Die Cracks Into Five Tiny Chiplets During Experiment

bit_user

Titan
Ambassador
AMD should sell a variant of the Ryzen CPUs without the stupid IHS with a limited use warranty.
For anyone who's willing to take the time & trouble to do direct-die cooling properly, I think removing the IHS isn't a significant hurdle. A lot of those who can't be bothered to remove it are probably the same ones likely to break their CPUs by doing direct-die cooling incorrectly.

A few broken CPUs quickly turns into a huge PR problem, because the tech press loves to sensationalize and too many readers overlook key details and nuances. Let's see how long it takes for someone to post something negative, overlooking the fact that the CPU was running without any IHS or heatsink, at the time of the failure.
 
  • Like
Reactions: Elusive Ruse
For anyone who's willing to take the time & trouble to do direct-die cooling properly, I think removing the IHS isn't a significant hurdle. A lot of those who can't be bothered to remove it are probably the same ones likely to break their CPUs by doing direct-die cooling incorrectly.

A few broken CPUs quickly turns into a huge PR problem, because the tech press loves to sensationalize and too many readers overlook key details and nuances. Let's see how long it takes for someone to post something negative, overlooking the fact that the CPU was running without any IHS or heatsink, at the time of the failure.
You can't compare the two in good faith... I don't even know why you did.

Removing the IHS completely voids your warranty. Period. No buts. That's not even counting the extra cost and associated risk. As a consumer, why would you want to risk your warranty and spend a lot of extra money to do something AMD wouldn't cost more money at all?

If AMD sells CPUs without IHS (like they do for laptops) and include a limited use warranty, at least, then it's a different situation altogether.

Don't forget both AMD and Intel used to sell CPUs without IHS for a long time. There's other better ways nowadays to protect the exposed die via the ILM that make including an IHS not really needed.

Regards.
 

bit_user

Titan
Ambassador
You can't compare the two in good faith... I don't even know why you did.
Exactly what are you claiming is in bad faith?

Removing the IHS completely voids your warranty. Period. No buts. That's not even counting the extra cost and associated risk. As a consumer, why would you want to risk your warranty and spend a lot of extra money to do something AMD wouldn't cost more money at all?
The article shows why it's too dangerous for AMD to support or encourage, in any way.

Don't forget both AMD and Intel used to sell CPUs without IHS for a long time.
Don't forget the thermal density of CPUs is way higher than it was, back then.

Back in the 1970's hard drives weren't even sealed. I worked with a guy who told me how you'd physically even sand down a platter, after a head crash, and put it right back into service!

What they could get away with doing, years ago, is not necessarily a viable option with today's tech. It's ever more highly-tuned, with ever slimmer margins for error.

There's other better ways nowadays to protect the exposed die via the ILM that make including an IHS not really needed.
Explain.
 
Last edited:

thisisaname

Distinguished
Feb 6, 2009
939
522
19,760
I wonder if he broke it putting the CPU cooler on, the speed he need to move to put it on in two to three seconds would be like hitting it hammer.
Plus the CPU was hot and I would think it would be weaker and a sudden increase is pressure would shatter it.
 
Exactly what are you claiming is in bad faith?
Putting a de-lid at the same level as buying a CPU without the IHS. They're not comparable.

The article shows why it's too dangerous for AMD to support or encourage, in any way.
No, it shoes someone doing something really stupid for science and clicks.

Don't forget the thermal density of CPUs is way higher than it was, back then.

Back in the 1970's hard drives weren't even sealed. I worked with a guy who told me how you'd physically even sand down a platter, after a head crash, and put it right back into service!
Yes, things change as now HDDs are packed with special gasses that need them to be completely sealed and such. Well, as I recall.

While the thermal density of CPUs has increased and power consumption along side it (kind of), you will still get better thermal transfer with an exposed die and have the same risks as back then. Physics doesn't change with time, so the risk for both Intel or AMD of selling exposed dies is the same. Hence why I also said "limited use warranty".

Well, if you would only watch der8auer, a YT'er, you would know what I mean in more detail. He has created several parts to complement his de-lidding tools for CPUs and to great success in thermals and ease of use. Imagine if AMD or Intel would offer those instead with a warranty of any kind. Laptops use de-lidded CPUs as well, so I do not see why (going back to your previous argument) is such an outlandish thing to say or ask; at least, it reads like you're gasping for air really hard for even proposing it...

I'll stop here; not worth discussing further.

Regards.
 

bit_user

Titan
Ambassador
Putting a de-lid at the same level as buying a CPU without the IHS. They're not comparable.
De-lidding isn't too hard, with the right tools. Effectively cooling a de-lidded processor is more challenging and delicate than cooling one with the IHS. If someone can do the latter, removing the IHS shouldn't be much of a problem for them. If removing the IHS is too burdensome, then they're probably better off leaving it intact.

I think your criticism is misguided. Instead of criticizing their IHS solution, you want them to do something you haven't asked Intel to do. That's inconsistent, at best.

No, it shoes someone doing something really stupid for science and clicks.
The article didn't indicate what clocks he was running at. I assume he tried to reduce clock speeds, or had other reasons to believe his experiment was safe.

My point was that if anything is wrong with the cooling solution, running a de-lidded CPU has much less margin for error than if you've at least got an IHS. The IHS will help even out heat distribution, lessening thermal stresses.

As for the "science" part, his photographs tell us more about how these CPUs are assembled and operate.

While the thermal density of CPUs has increased and power consumption along side it (kind of), you will still get better thermal transfer with an exposed die and have the same risks as back then.
No. Increased thermal density means greater thermal stresses, and the die shaving they do these days means the dies are more fragile.

Physics doesn't change with time, so the risk for both Intel or AMD of selling exposed dies is the same.
Physics doesn't change, but technology improves by pushing ever closer to the limits. That reduces margins for error. You can't just blindly apply whatever was true of tech from 20 years ago and expect it to apply equally, today.

Well, if you would only watch der8auer, a YT'er, you would know what I mean in more detail. He has created several parts to complement his de-lidding tools for CPUs and to great success
People have broken their de-lidded CPUs, even using all of his tools and shims. It's more risky, no matter what you think.

The whole EXPO thing has showed that these companies run into a lot of trouble, when they try to limit warranties. From their perspective, it's simply not worth the risk to sell de-lidded CPUs.

It's similar to why people can tune cars for better performance than they come from the factory. There were good reasons why the automotive engineers "held back", including safety, reliability, and warranty liabilities.

It's perfectly fine for there to be an aftermarket for CPU cooling, delidding, etc. What AMD should do is design a better IHS solution, to lessen the benefit of de-lidding.

Laptops use de-lidded CPUs as well, so I do not see why (going back to your previous argument) is such an outlandish thing to say or ask;
Laptops also run at lower power and thermals. That might make it much less risky than high-performance desktop CPUs, where the whole point would be so that people can overclock them to extremes!

For all we know, the laptop CPU dies aren't shaved (as much), specifically to make them more durable.

Also, last I checked, end-users aren't building laptops. Even modular laptops, like Framework, probably have the thermal solution pre-installed on their CPU modules.

I'll stop here;
I doubt it.
 
Last edited:
There are a lot of things that can be done to minimize the damage to a CPU that doesn't have an IHS. Let's face it the likely reason there is an IHS has nothing to do with individuals breaking CPUs and everything to do with OEM building being easier with one. Personally I'd love it if Intel and AMD sold a SKU line without an IHS even though it'd require lapping a cooler or getting a custom one for best cooling. That being said I certainly understand why they don't and likely won't ever do so.
 
Let's see how long it takes for someone to post something negative, overlooking the fact that the CPU was running without any IHS or heatsink, at the time of the failure.
So what is the positive view on a CPU not shutting down when the shutdown temp is reached but keeping running until it breaks into 5 pieces?!

The theory was that the thermal sensor breaks down with too much V and causes the recent issues, but maybe the thermal sensor just doesn't work right outside of a way too small range, so going too low on voltage, like he did, also disables the thermal sensor.

Because the only other reason for this to happen would be what thiisaname mentioned, that the guy slammed the cooler on the chip so hard that it fractured.
But I'm no expert in physics to know that this could happen from slamming a large surface onto another.
 

bit_user

Titan
Ambassador
So what is the positive view on a CPU not shutting down when the shutdown temp is reached but keeping running until it breaks into 5 pieces?!
With no IHS, there's nothing to mitigate the thermal stresses caused by the hotspots. It's not obvious the thermal throttling can either act quickly enough or scale back drastically enough to avoid the kind of stresses triggering the fracture. It certainly wasn't designed to do protect the chip when de-lidded.

In facto, I'll bet there's a good chance of it happening to current Intel CPUs, too.

Because the only other reason for this to happen would be what thiisaname mentioned, that the guy slammed the cooler on the chip so hard that it fractured.

But I'm no expert in physics to know that this could happen from slamming a large surface onto another.
I don't understand why they said that or you're repeating it, because the article clearly stated that it fractured 3 seconds before he could get the heatsink back on. That's such a precise figure that I wouldn't disregard it.

Anyway, @-Fran- , do you see how someone even doing something blatantly unsupported and warranty-violating can be used to attack AMD? Imagine if they sold de-lidded CPUs and there were a few more such mishaps. AMD's detractors would have an absolute field day with each and every one of them!
 

Elusive Ruse

Estimable
Nov 17, 2022
459
597
3,220
Tom's clickbait titles at full throttle! What's next?

7950X Catches Fire During Heat Treatment Trial!
Well-known pyrotechnician doused the flagship CPU in kerosene and lit it up with a flamethrower!
 
With no IHS, there's nothing to mitigate the thermal stresses caused by the hotspots. It's not obvious the thermal throttling can either act quickly enough or scale back drastically enough to avoid the kind of stresses triggering the fracture. It certainly wasn't designed to do protect the chip when de-lidded.
The thermal stresses of between room temperature and 95C ?
Those that would not change even with TIM and IHS?
The heat would be transferred away but the amount and levels of stress would be the same.
And he stated that he got thermal throttling, or that it reached 95C anyway.
M_57UOi99K0mUGQ_zPz8m8R_g2aNlzws00FUst_bOfO4zgUIMd7jjhK1dqBMUALn3hVhlZDPu-jycqbdB17Z8IOZ5dl2htzvBsfhUUaU

I don't understand why they said that or you're repeating it, because the article clearly stated that it fractured 3 seconds before he could get the heatsink back on. That's such a precise figure that I wouldn't disregard it.
It crashed then he put the cooler on then it burned out.
So maybe it was the stress of having a 95C sourface touching the cold cooler. Although even for that it seems a very extreme reaction.
 

bit_user

Titan
Ambassador
The thermal stresses of between room temperature and 95C ?
No, the thermal stresses between different areas of the die.

Those that would not change even with TIM and IHS?
Hotspots within the die get mitigated, because some of the heat can go into the TIM. Even though some of that heat flows back into the die, that still spreads out the hotspots, which means less thermal stress.

And he stated that he got thermal throttling,
As I said, its thermal throttling was not designed to be drastic enough to protect a bare die.
 
No, the thermal stresses between different areas of the die.


Hotspots within the die get mitigated, because some of the heat can go into the TIM. Even though some of that heat flows back into the die, that still spreads out the hotspots, which means less thermal stress.


As I said, its thermal throttling was not designed to be drastic enough to protect a bare die.
There is clear discoloration under the IOdie area in cases where CPUs blew up with the heat sink and tim in place.
There is something in these CPUs not working right, not preventing extreme overheating, and lack of tim or heatspreader ain't it.
b4fY43f.jpg

One of the first cases was a 7900x where the idium melted away from under the heat spreader...
View: https://www.youtube.com/watch?v=34VbutE-Qss
 
Last edited: