AMD's Upcoming Zen 4 CPU Delidded by Overclocker

[TheoneandonlyMrK]

I take things personally when it is aimed at me, which your replys to my comments quite often are.

And it's not my precious, it's just a CPU. AMD users on here are far more touchy about their precious than me.


Let's just see how zen4 and AM5 pan out. Intel has been using these socket types for a long let's see how AM5 goes for AMD.

Not a great fan of the IHS design but hopefully it'll be ok.

Well AMD do use LGA on server so it's not new, they should be fine.

I argue against your opinion, btw, it's your opinion, not you that's debatable.

 

[zlobby]

With an IHS that thick, and non-centered Chiplets, I'm wondering if a composite IHS would be 'in the cards'. Either 2 copper sheets sandwiching a sheet of pyrolytic graphite, or 2 copper sheets CVD'd with Graphene and pressed together. Copper conducts heat well, but evenly. 2D and semi-2D carbon allotropes conduct heat across their planar axis considerably better than copper or silver does.

Unlike a decade or 2 ago, (when tech enthusiasts first started 'playing with' allotropic carbons' thermal properties) pyrolytic graphite and/or graphene are common materials nowadays in microelectronics cooling.

But that is (relatively) expensive. They could just slap a slab of metal and call it a day. I guess it all depends on what they want to acheive?

 

[trsttte]

But it makes sense, imo. Unlike monolithic dies, small chiplets concentrate heat into a tiny, non-central spot, which isn't ideal for cooling. You'd rather want to dissipate that heat in order to use as much of your cooler's coldplate area as possible. Zen 2 and 3's problem has always been heat dissipation.

Having an extra-chunky IHS will also add to the thermal mass and make it easier to air-cool. I'm guessing LN2 and high-end water chillers will suffer, but that's such a tiny niche market that custom-build their own retention bracket and probably delid their CPUs anyway.

That only helps if the other source can then dissipate the heat effectively enough

It just doesn't help. Having a bigger IHS means having a bigger thermal resistence between die and heatsink. They can mitigate that by reducing the thermal resistence of the other interfaces (like having thinner dies or a better contact solution between die and ihs) but it's always summing: R(core-die)+R(die-ihs)+R(ihs-heatsink)+R(heatsink-air). By having a bigger resistence in the middle you can either reduce the other ones (other than having fans at higher speed not that easy) or it will run hotter.

I think it would have been better to just have a different height and require a mount adapter (probably also change the hole pattern to avoid incorrect pressure by people not using the correct mount). Another option would be to start designing coolers and socket with variable height in mind and provide a torque wrench (like for the threadripper socket, different problem but same solution). Riskier but much more effective than relying on the screws bottoming out anyway.

 

[InVasMani]

I wonder why they don't make IHS's with built-in water blocks? You know, just hook up the hoses and let it rip!

I for one will gladly pay for such option. With power reaching 250W every degree helps. I'd estimate at least 5-6C lower temps with such solution, which should do miracles here.

I mean the easiest answer to that is probably because you don't need to you can simply mount one. I'm not convinced it would be wise to do anyway because not everyone water cools nor intends to do so. That said I'm doing this custom DIY cooling on my next build. It's a tower cooler with a AIO 240mm's water block mounted on top of the tower coolers CPU block ID-Cooler SE-207 TRX 280w TDP cooler. I think I'm going to put a Grizzly thermal pad between the two blocks surfaces because paste might be too messy and too much hassle given the tight space constraints. I don't know what it'll be like, but I can't wait to find out how it all turns out when Zen4/Raptor Lake arrives.

If it works out well what I'd like to see is a company like Noctua adopt integrating a water block cooler onto the top portion of a tower cooler rather than trying to mount like have it designed to be integrated in the first place. It would really open up possibilities especially in SFF along with high TDP cooling requirements. Also something I had thought up is you could potentially have stacked blocked that each go to a separate AIO radiator loop mounting point within a case if a cooler were design with that concept in mind.

 

[AusWolf]

It just doesn't help. Having a bigger IHS means having a bigger thermal resistence between die and heatsink. They can mitigate that by reducing the thermal resistence of the other interfaces (like having thinner dies or a better contact solution between die and ihs) but it's always summing: R(core-die)+R(die-ihs)+R(ihs-heatsink)+R(heatsink-air). By having a bigger resistence in the middle you can either reduce the other ones (other than having fans at higher speed not that easy) or it will run hotter.

With all due respect, I disagree. If you have a pretty standard tower cooler with 4 direct contact heatpipes, and only one chiplet offset to one side, then all the heat is concentrated below 2 heatpipes while the other 2 run cold. You need to dissipate the heat evenly across the contact surface of the cooler.

I experienced this problem with a R5 3600 that I had to dial down to an actual 65 W PPT from the factory 88 to prevent it from throttling with a be quiet! Shadow Rock LP, while my i7 11700 can easily do 100-120 W with the same cooler. The heatsink was cold to the touch on the 3600 even after several remounts, so it was definitely a contact issue.

Chiplets might be a great thing in some applications, but really awful in others (like SFF builds). I think the thick IHS is supposed to mitigate this issue, though we'll only see for sure when Zen 4 is out for sale. Heck, I might buy one just to test my theory.

 

[zlobby]

I mean the easiest answer to that is probably because you don't need to you can simply mount one. I'm not convinced it would be wise to do anyway because not everyone water cools nor intends to do so. That said I'm doing this custom DIY cooling on my next build. It's a tower cooler with a AIO 240mm's water block mounted on top of the tower coolers CPU block ID-Cooler SE-207 TRX 280w TDP cooler. I think I'm going to put a Grizzly thermal pad between the two blocks surfaces because paste might be too messy and too much hassle given the tight space constraints. I don't know what it'll be like, but I can't wait to find out how it all turns out when Zen4/Raptor Lake arrives.

View attachment 250334

If it works out well what I'd like to see is a company like Noctua adopt integrating a water block cooler onto the top portion of a tower cooler rather than trying to mount like have it designed to be integrated in the first place. It would really open up possibilities especially in SFF along with high TDP cooling requirements. Also something I had thought up is you could potentially have stacked blocked that each go to a separate AIO radiator loop mounting point within a case if a cooler were design with that concept in mind.

I am willing to bet that if someone introduces a factory-mounted WC IHS it will sell well. Of course, many people will still run AIO or air towers but the market is there!

Come on, AMD! Do a trial run and see how market reacts!

 

[TheoneandonlyMrK]

I am willing to bet that if someone introduces a factory-mounted WC IHS it will sell well. Of course, many people will still run AIO or air towers but the market is there!

Come on, AMD! Do a trial run and see how market reacts!

It would be easy to standardise the fitting,s, It would cool better being soldered direct to die too , amazing idea, I think they're actually already used, but not in consumer electronics, great Idea, I would buy.

 

[trsttte]

With all due respect, I disagree. If you have a pretty standard tower cooler with 4 direct contact heatpipes, and only one chiplet offset to one side, then all the heat is concentrated below 2 heatpipes while the other 2 run cold. You need to dissipate the heat evenly across the contact surface of the cooler.

I experienced this problem with a R5 3600 that I had to dial down to an actual 65 W PPT from the factory 88 to prevent it from throttling with a be quiet! Shadow Rock LP, while my i7 11700 can easily do 100-120 W with the same cooler. The heatsink was cold to the touch on the 3600 even after several remounts, so it was definitely a contact issue.

Chiplets might be a great thing in some applications, but really awful in others (like SFF builds). I think the thick IHS is supposed to mitigate this issue, though we'll only see for sure when Zen 4 is out for sale. Heck, I might buy one just to test my theory.

A cold plate/vapor chamber solves that, of course with direct contact heatpipes you'll have a higher resistance between the ones further away but the point still stands, more material in the heat path will mean a higher resistance (like in your single chiplet scenario, the heatpipes further away have more material in between). The "ideal" cooler would be direct contact (like laptops) but with that come mounting and reliability challenges.

 

[AusWolf]

A cold plate/vapor chamber solves that, of course with direct contact heatpipes you'll have a higher resistance between the ones further away but the point still stands, more material in the heat path will mean a higher resistance (like in your single chiplet scenario, the heatpipes further away have more material in between). The "ideal" cooler would be direct contact (like laptops) but with that come mounting and reliability challenges.

You'd have to mount that directly onto the chiplet, not on the centre of the CPU package.

 

[InVasMani]

I am willing to bet that if someone introduces a factory-mounted WC IHS it will sell well. Of course, many people will still run AIO or air towers but the market is there!

Come on, AMD! Do a trial run and see how market reacts!

Perhaps a soldered vapor chamber that you mount a Wraith top down cooler that provides airflow in around the VRM's that has a water block integrated into the top of the heatpipe base or soldered to it in which you can expand cooling via a custom water loop? I could see AMD exploring the idea they already do water cooling options for GPU's why not expand it to CPU side!!?

 

[AusWolf]

Perhaps a soldered vapor chamber that you mount a Wraith top down cooler that provides airflow in around the VRM's that has a water block integrated into the top of the heatpipe base or soldered to it in which you can expand cooling via a custom water loop? I could see AMD exploring the idea they already do water cooling options for GPU's why not expand it to CPU side!!?

Or why not integrate the vapour chamber into the IHS if it's so thick anyway?

 

[Mussels]

I like it, thicker IHS should help with the bendy issues intel is suffering from, as well as sort out the heat density issues the 5800x suffers from

 

[zlobby]

You'd have to mount that directly onto the chiplet, not on the centre of the CPU package.

With an WC IHS they can route the water flow exactly over the chiplets in order to maximize cooling only where it is needed.

 

[InVasMani]

Or why not integrate the vapour chamber into the IHS if it's so thick anyway?

That's what I meant in regard to that, but probably could've been more clear conveying it.

 

[KarymidoN]

230w PPT, thick ihs, smaller process.. these are gonna be hot. I already know how hard it is to cool 230w PPT.

not hard at all, AMD has been testing internally with NHD-15, the same one that has an 12900KS at almost 100ºC so pretty much standard high end cooling

 

[freeagent]

not hard at all, AMD has been testing internally with NHD-15, the same one that has an 12900KS at almost 100ºC so pretty much standard high end cooling

I guess you missed the part where I said I know what it’s like to cool 235w PPT. I can get my 5900X to do those numbers. Now, knowing what I know and then taking into account the smaller process and thicker IHS, it may be a warm one. I also have a high end cooler, it is just as good as a D15. Might even be better slightly.

 

[InVasMani]

So long as the performance per watt is better than the 129000K/129000KS it won't be so pathetic. The whole issue with those two chips in particular relative to the 5950x is they are gaudy in that area at least for multi-thread, but they are also very multi-thread driven CPU's at the same time so jacking power consumption for a pretty minor single thread performance win while like doubling power consumption in full load multi-thread is piss poor trade off and in world where worsening heat is the enemy to the majority at the same time. For the record I don't think CPU's TDP are very out of hand right now GPU's are where things have gone off the rails entirely and getting worse.

 

[freeagent]

what do you think people use to cool the 280W Threadrippers??

We aren’t talking about threadrippers though.

 

[freeagent]

we're talking CPU's who draw a lot of power and generate a lot of heat right?

No.. we are talking AM5..

i would elaborate my point further but your mind is not open to the reality. best of luck.

Lol.. ok.

But so you know, my 5900X is not the highest wattage CPU that I have owned. We were dipping our toes into 300w oced cpus a decade ago.

 

[TheoneandonlyMrK]

No.. we are talking AM5..


Lol.. ok.

But so you know, my 5900X is not the highest wattage CPU that I have owned. We were dipping our toes into 300w oced cpus a decade ago.

Dp by a mod

 

[Count von Schwalbe]

With an IHS that thick, you could machine it into a waterblock if you got the courage... Being LGA would help with that. The main issue would be sealing it, you would need a custom o-ring.

 

[Mussels]

With an WC IHS they can route the water flow exactly over the chiplets in order to maximize cooling only where it is needed.

That's just delidding and direct die cooling with extra steps

We may well see extreme edition CPU's in the future with built in soldered waterblock IHS in all honesty.



I'm surprised i havent seen that mentioned before now, but they simply could have a finned IHS (pay someone like EK for rights to a good design) with screw holes and let companies make custom tops for it.

 

[AusWolf]

That's just delidding and direct die cooling with extra steps

We may well see extreme edition CPU's in the future with built in soldered waterblock IHS in all honesty.



I'm surprised i havent seen that mentioned before now, but they simply could have a finned IHS (pay someone like EK for rights to a good design) with screw holes and let companies make custom tops for it.

The only problem with that would be maintenance. You can swap out a waterblock if it gets gunked up, but what do you do with it if it's soldered to your CPU?

 

[Mussels]

The only problem with that would be maintenance. You can swap out a waterblock if it gets gunked up, but what do you do with it if it's soldered to your CPU?

brush it clean, like you do with any waterblock
To be clear i meant the coldplate specifically, this part:

If that was soldered direct to the die with tight tolerances, you'd be able to deal with heat density on high wattage parts.
I can totally see this happening if these power consumption trends continue.


Partner with a waterblock company like EK with an official top, and let all the other companies make their own ones. You'd have AIO's with a fill port before you could sneeze.

 

[AusWolf]

brush it clean, like you do with any waterblock
To be clear i meant the coldplate specifically, this part:

View attachment 250471

If that was soldered direct to the die with tight tolerances, you'd be able to deal with heat density on high wattage parts.
I can totally see this happening if these power consumption trends continue.


Partner with a waterblock company like EK with an official top, and let all the other companies make their own ones. You'd have AIO's with a fill port before you could sneeze.

Ah! That makes sense.