AMD Explored Vapor Chamber Cooling Design for Zen 4 CPUs

[T0@st]

Gamers Nexus recently visited AMD's headquarters in Austin, Texas—a previous video documented company employees discussing the history of Zen CPUs, and the showcasing of historical prototypes including (unreleased) Ryzen 9 5950X3D and 5900X3D models. The YouTube channel promised that more AMD HQ tour footage would be shared over the next couple of weeks—their latest upload has (host) Steve Burke talking to representatives from various internal labs.

A notable detail extracted from Team Red's thermal laboratory was an old heat spreader concept for Zen 4 processors—the team evaluated whether a concealed vapor chamber would offer improved cooling performance versus conventional metal solutions. Their tests determined that the extra cost (not disclosed) required to integrate a vapor chamber was not worth the resultant 1°C temperature difference, when lined up against a traditional metal design IHS. AMD confirmed that the concept was not developed further since prototype chips were also found to generate heat exceeding expected normal levels, under continuous long-term workload conditions.



Gamers Nexus: "This tour of AMD's Austin Headquarters features labs from around the sprawling campus, allowing us a never-before-seen look at the tools used by a hundred-billion dollar chip designer.

During these tours, we get to see unreleased prototypes - like vapor chamber heatspreaders (at the 29 minutes 50 second mark), first-party direct die plates, and unnamed CPUs - while also learning about the technology and tools used to test and design AMD's Ryzen (Zen) processors. The tour goes through the Bring-Up Lab, the codename "red door" lab, a thermal engineering lab, the device failure analysis lab, and an IHS etching facility."

View at TechPowerUp Main Site | Source

 

[ZoneDymo]

its a really good video, cool stuff all around

 

[ymdhis]

Given how bad AMDs track record is with vapor chambers bulging up, it would've been a disaster if they put them on CPUs.

 

[Konyalee]

I am suprised there was a 1C difference at all, isn't IHS as a single block is more conductive than water between two metal plates ?

 

[dgianstefani]

I am suprised there was a 1C difference at all, isn't IHS as a single block is more conductive than water between two metal plates ?

Not when it's as thick as AM5 IHS is for compatibility with AM4 coolers, and when tiny chiplets produce high density heat.

 

[Arkz]

Given how bad AMDs track record is with vapor chambers bulging up, it would've been a disaster if they put them on CPUs.

Didn't it only happen with one specific SKU and get sorted out pretty quickly?

 

[RamiHaidafy]

Didn't it only happen with one specific SKU and get sorted out pretty quickly?

It did. As far as I'm aware, AMD have never had this problem before that. There is no "track record".

People are just being dramatic.

 

[Vya Domus]

Didn't it only happen with one specific SKU and get sorted out pretty quickly?

It's a general defect for vapor chambers to sometimes swell up like that, it can happen in any product which uses one.

 

[eidairaman1]

Given how bad AMDs track record is with vapor chambers bulging up, it would've been a disaster if they put them on CPUs.

Baseless

 

[Broken Processor]

Baseless

Wasn't there a faulty batch of own brand GPU vapor chambers this gen perhaps that's what they're referring to.

 

[dgianstefani]

Wasn't there a faulty batch of own brand GPU vapor chambers this gen perhaps that's what they're referring to.

The 7900 MBA series, yes.

 

[Chrispy_]

That would be cool. The issue with Zen3 and Zen4 was never outright TDP like Intel's crazy CPUs, but thermal density of those tiny chiplets concentrating ~150W into a 1/3rd the area of an equivalent Intel CPU.

Anything to more effectively transfer heat out to a wider area would be a win, but I'm guessing complexity of getting an effective chamber that slim prevented it being practical. Perhaps for the next socket, whenever that is...

 

[evernessince]

The 7900 MBA series, yes.

A single batch of 7900 XTX reference model cards only. Non-reference 7900 XTXs, 7900 XTs and all other 7000 series cards were not impacted. Regardless, to say that such a small fraction of the cards constitutes a track record is highly misleading.

Mind you the root comment in question from ymdhis mentioned bulging, which wasn't the case in the above example. There are no recent widespread examples of bulging issues on AMD card's vapor chambers. Either they are pulling a very old example, are misinformed, or are making it up.

 

[TechLurker]

Not when it's as thick as AM5 IHS is for compatibility with AM4 coolers, and when tiny chiplets produce high density heat.

For what it's worth, AMD also tested out thinner IHS designs, and found that the temperature difference wasn't as big a deal as cooler compatibility. On the other hand, they've been experimenting with unlidded CPUs and their own direct-die interfaces, but haven't yet found it viable enough to consider. Would be interesting if AMD did eventually sell enthusiast-grade unlidded CPUs, with some warranty caveats. And if they kept pricing the same as regular IHS-covered CPUs (if not a bit more), with the cost instead covering the support bracket and pre-applied protective coating on the SIMDs, would avoid issues of the ignorant just buying it because it was "cheaper".

 

[TheoneandonlyMrK]

I am suprised there was a 1C difference at all, isn't IHS as a single block is more conductive than water between two metal plates ?

Not with a phase change happening within the Heatsink.

 

[InVasMani]

Probably the most practical thing you could do with AM5 atm is get one of the anti-bending brackets for it and put thermal putty on both the CPU IHS and anti bending bracket in tandem with a cooler designed to cover both. Some of that thermal heat would dissipate outwardly hopefully, but how well tough to say. Could probably use the anti-bending bracket and putty with delidd as well.

On a different note it would be interesting if MB makers could coat or dip the CPU socket pins in something type of material like nanotube graphite to kind of cool the pins themselves making contact with the CPU pads themselves. Though idk how that would work out a bit of a obscure abstract thought though which I sort of doubt has been considered.

I could see some kind of nanotube graphite coating being done in between with PCB layers of MB perhaps as well. I'm almost surprised to not having heard of that done yet thinking about it a bit instead we hear 2oz of copper that's the magical amount they all universe have agreed upon it seems.

 

[claes]

Wtf is happening lol