Is AMD Zen 4 Too Hot for a Box Cooler? Ryzen 7600X + Wraith Spire Tested

[wiyosaya]

Eh? This isn't how thermodynamics works. A CPU encased in a theoretical infinite-sized block of copper would be cooled far better than any possible combination of air-cooled heatsinks.

If you consider water-cooling, it may not be true, as water's thermal conductivity is around 0.6 W/m.k, versus air's 0.025. Copper is still several hundred times higher, but given a high enough temperature differential, water's convective aspect could exceed copper's pure conduction.

He's talking about thermal interface materials such as those made by Arctic. If you took the time to check the Arctic web site, they state the same thing. In fact, they state that thinner applications of thermal interface material result in better temps. That's why no one wants to gob on thermal interface paste to their CPU expecting that it will work better - because it won't.

While what you state about a block of copper is definitely true, in this case, however, your statement is more of a straw man rather than a decent argument.

I'm rather surprised just how many of the "expert enthusiasts" on this site seem to not understand thermal considerations.

It doesn't matter that the temperature is, what matters is the heat dissipated (which equals power consumption). If this doesn't make sense, consider gaming on your phone. If you checked you'd find the SoC temp often hits ~90-95C when gaming, yet you don't worry about the phone heating up the house as its only consuming 3-5W of heat at max load. It is the 3-5W of heat that matters, not the die temperature.

If you are worried about overheating VRMs and the like, an air cooler is often better than an AIO because most AIO result in no air flow over the VRMs. If you are worried about heat in general, just do what the review testing did, either select Eco mode, or set your own custom power limit or temp limit.

What you state is true to a point, but if the maximum opertating temperature of the component's silicon is exceeded, you will end up with a dead chip - end of story. Its misleading, IMO, to state that the junction temperature does not matter and the only thing that matters is pulling away the "power" put into the junction. True, you have to pull away that power, but why you have to do that is so that maximum junction temperature of the device is not exceeded. If the maximum junction temperature is exceeded, the device will fail because its junction temperature exceeded the maximum operating temperature.

 

[Makste]

I appreciate these just for science tests. They're very informative.

 

[ubronan]

I still do not get all the fuss about the high temp of the new series cpu's
AMD has explained that the cpu is tested for very high temps which is allowed for these cpu's
In the paste we had cpu's which got similar hot, from which non ever died untill they got discarded as usefull for the ongoing improvements of these ever faster becoming systems
For me over 5 Ghz means hot temperatures as all who do overclocking know very well.
It has been as long as I remember been told going above 5 Ghz pumps up the temps if you like it or not
There is no other way to get lower temps unless some solution is found to overcome this issue
So when you want lower temps THE ONLY SOLUTION IS LOWER THE CPU SPEED if you hate seeing high temps
Especially above 5.2 Ghz it becomes hot, now the new models are creeping up to the known limit of 6 Ghz.
The constant smaller dies does make this possible but I am not sure that it will go on as eacht time people expect a faster cpu. AMD and Intel seek constant to break this limits time will tell if they succeed
At one point it is no longer possible to add more cores and upping the frequency unless new materials or a different approach will work
So do not whine when the cpu is hot AMD tells you it is made todo so, Intel has todo the same count on that as their products will use more power and deliver much more heat as well.
Higher temps says nothing when it is made for these temps so deal with it or step back to a old cpu
For me adding more cores is simply and give that the highest clocks is for me insanity
I want to see a cpu with higher clocks with much less cores, not the other way around
But now you see only the most expenssive cpu gets them so you have a cpu which does at least 96% of the die nothing at all.
In reality the useless extra cores are probably much higher as I do not use any programs which can address more than 4 cores, nope not any game I ran does use more for real unless you count the 0.001 to 0.1 % usage of a core that it is working
Yes I had a cpu with 16/32 cores and it never ever really used the rest besides these 4 cores the same goes for my current intel which is a 8/16 cpu and that uses still nothing more than 4 cores for real.
So I hate the silly way cpu selling is for years now making these products as they do now, the opposite of what would be logical and only greed related.

 

[mrvco]

The perf loss from running in 65w mode really begs the question why that isnt the STNADARD mode. Is that 1~% really worth all the extra wattage? I'd think itd be a much more impressive launch if a 65 watt part came out and still beat the 200+ watt core i9.
Not blind only the perf graphs are there.

Because 'Synthetic Benchmarks Arms Race'.

By all rights Eco-mode should be enabled by default on Zen4, but all comparisons by Team Blue friendlies would run synthetic benchmarks with Zen4's Eco-mode enabled (ie 'stock' configuration) and Intel in full flame thrower mode (ie 'stock' configuration), while ignoring real-world use-cases and declare victory in their click-bait headlines.

 

[Fastturtle]

No matter how good thermal conductivity a material has (the IHS) it will fail the energy transfer after a certain thickness is reached and act more like an insulator.
This is why we see the 2-8 seconds thermals rising from 30C to 90C, not enough heat is transfered from die to cold plate.
It's the same issue from Zen1 to Zen4. But Zen 4 has more issues because the extra thickness.
Just bad design, if they made a central square formation from the I/O die and CPU chiplets I'm sure those capacitors would fit under the IHS.
Also keeping AM4 cooler compatibility was pretty wrong done. They could just adjust the bracket height for coolers with 2 point mount.
This would aslo mean all cooler's need to have new brackets.
It was a compromise done for a new design that should never pass the internal QC.

This is one thing I agree about. AMD shouldn't have worried about AM4 Cooler Compatibility and gone straight to something new. Long term, it would actually be better as the CPU's could then hit higher clocks while having a decent aftermarket and yes, all of those aftermarket coolers with backplates could easily be designed with the socket retention mounting screws. They do it already for aftermarket AM4 coolers with their own backplates, so what's so special about the AM5?

 

[Marco Mint]

The reason is due to the IHS being thicker to accommodate AM4 coolers. Der Bauer has a good video where he delided I think a 7900x/7950x and noted the thickness of the IHS. As soon as he delided and applied cooling directly to the die, he saw thermal drops of 20+ degrees while running everything at stock settings (voltage, frequency, etc.).

I read a comment somewhere that the IHS was thicker on these initial cpu's due to forward planning by AMD and that the future X3D cpu's will be taller due to the extra cache. So the IHS on those X3D cpu's will therefore be thinner and still remain compatable with the older coolers.
Whether this is true or pure speculation on the person who commented on this, I don't know. But it would make sense.

 

[RaXoR]

I read a comment somewhere that the IHS was thicker on these initial cpu's due to forward planning by AMD and that the future X3D cpu's will be taller due to the extra cache. So the IHS on those X3D cpu's will therefore be thinner and still remain compatable with the older coolers.
Whether this is true or pure speculation on the person who commented on this, I don't know. But it would make sense.

That'd be an improvement. The hope would be that they thin out the IHS in general and make up for the height some other way. JTC has a video where he sanded down the IHS and got a 10+ degree drop in temps.

 

[m3tavision]

This is one thing I agree about. AMD shouldn't have worried about AM4 Cooler Compatibility and gone straight to something new. Long term, it would actually be better as the CPU's could then hit higher clocks while having a decent aftermarket and yes, all of those aftermarket coolers with backplates could easily be designed with the socket retention mounting screws. They do it already for aftermarket AM4 coolers with their own backplates, so what's so special about the AM5?

I think a professionally lapped Ryzen 7000 series, is going to be a market on to itself. For majority of Gamers/enthusiasts (ie:85%~) are going to be happy with AM5's plug & play platform (w/expo), will be hard to beat.

But for those who will go the nTh degree and need that extra 10%~ of performance (no cost option). Direct Mounting, or delidding, and/or lapping will eek out the inefficiencies of individual enthusiast's dies...


...leaving nothing on the table... every enthusiast likes a well working machine that is ultra efficient... how far are you willing to go...?

 

[Arcmanov]

AMD really suckered me with the 'AM4 cooler compatibility' spiel...so much so that I invested in a second AIO as a backup to my current one...only to find out that the 'compatibility' comes with a big, fat caveat.
So now, for me to adopt AM5, I would need board, CPU, RAM and a new cooler. Not ready for that kind of outlay.

Think I'm gonna just drop in a 5800X3D and call it a day.