Sony Takes a Step Backwards with PlayStation 5, Cuts on Cooling Capacity in the new Revision

[Zuli_Muli]

"That is just measuring the air coming out" - you know that the air exit temperature being hotter indicates the cooling solution is working better, not worse, right? Getting rid of the heat is what the cooling solution does. An indication the CPU is running hotter would be if the air exit temperature was cooler, because the cooling solution wouldn't be removing the heat as efficiently.

Exit air can not be hotter then the hardware. Or another way of putting it is in a passive system like this (heat pipes and fins with a fan) if the air coming out is hotter then it use to be then the chip is also running hotter.

Now most likely the fan is the bigger issue, if it really is a downgrade in airflow (L/min) from the old fan like they seem to have implied then what's happening is that there isn't as much air to carry away the heat as before so there is a build up of heat and there for higher temps.

I disagree, given that both are air cooling so reaching steady state is fast say 2min, and the consoles produce a similar amount of heat say 200w, yet one console is exhausting more heat than the other, so either the new model is doing a better job, or the old model is dissipating heat through another way than the exhaust via the body of the console. Either way the new model seems more efficient.
Here's a more serious article than the misleading nonsense of the YouTuber.

New PS5 model’s weight-loss mystery solved: A smaller, likely better heatsink

Parsing this weekend’s informative, but not entirely authoritative, revision teardown.

arstechnica.com

Exit air can not be hotter then the hardware. Or another way of putting it is in a passive system like this (heat pipes and fins with a fan) if the air coming out is hotter then it use to be then the chip is also running hotter.

Now most likely the fan is the bigger issue, if it really is a downgrade in airflow (L/min) from the old fan like they seem to have implied then what's happening is that there isn't as much air to carry away the heat as before so there is a build up in heat and the new equilibrium is now higher.

 

[mtcn77]

Exit air can not be hotter then the hardware. Or another way of putting it is in a passive system like this (heat pipes and fins with a fan) if the air coming out is hotter then it use to be then the chip is also running hotter.

Now most likely the fan is the bigger issue, if it really is a downgrade in airflow (L/min) from the old fan like they seem to have implied then what's happening is that there isn't as much air to carry away the heat as before so there is a build up of heat and there for higher temps.

Most think this indicates efficiency when it is actually down the drain due to the presence of heat sump compounding the overall heat load. No component can cool off when the chip and the heatsink are at the same temperature.

 

[eidairaman1]

This thread is friggin late, giving credit to @lynx29 for posting like 3 days ago on this

 

[Totally]

I mean, less and smaller heatpipes, less and smaller fins and a lot less copper volume on the baseplate (as it turns out it was never a vapor chamber as I thought, just looks like one, might still help with thermal mass particularly for the memory on the other side but meh), I think it's hard to argue that the original wasn't better (might still not be but to me looks very much like it) and they definitely found a way to make it cheaper.

It might have been too good and had margins to be reduced but you're getting a worse product even if it's still fine (or the new might perform better, I doubt it but will wait on testing that is not just measuring an exaust temp.)


On topic but surprisingly left out of the discussion, the wifi module also changed but since the only info we have is it now uses 2 antenas instead of 4, it's better to leave it out anyway

The extra wires are just redundant antennas to reduce "blindspots" nothing special.

 

[lexluthermiester]

I mean, less and smaller heatpipes

Take a closer look.

 

[Chomiq]

"smaller heatpipes"

ok...

 

[Valantar]

"smaller heatpipes"
View attachment 215060
ok...

Looks to me like they have streamlined the layout of the heatpipes from the previous version. There might be fewer of them as well, but it's difficult to tell from this angle. I'm assuming there are two (or at least one) in the middle fin stack as well. Probably simplified manufacturing significantly, with only one relatively loose bend per heatpipe (or two for the 1-2 pipes that pass into the small heatsink on the other side of the cold plate) rather than 2-3 quite tight ones on the previous design. The copper plate is also mostly redundant given that there are heatpipes running along the bottom of the fin stack (and from memory, GN's original teardown showed that no part of the cold plate was contacting anything but the core, with the steel midframe/RF shield being the only contact for RAM and VRMs). Also interesting to note the changes in the RF shield - looks like there have been noticeable changes to the peripheral VRMs (not SoC VRMs) of this updated version, with fewer/smaller parts of the shield raised to clear chokes. It also looks like there might be some improved VRAM "cooling" (as much as a steel plate can provide that) to the right of the heatsink, with a larger depression visible there.

Given the simplicity of this new design, I think I'll revise my previous guesstimate of cost savings - this is definitely more than a few dollars cheaper in total, at least when factoring in production costs, speed and complexity (reject rates for bent heatpipes specifically). The overall reduction in heatsink size isn't that huge though - it's mainly that bottom corner that's about 2x the size of the previous middle gap (though that is probably partially made up for by there now being heatpipes along the top and bottom of the entire leftmost fin stack, rather than half of it only having bottom heatpipes), and the middle section is larger. The leftmost portion is partially narrower, but ... meh. That rightmost part is 1/3 the size, but I always thought that looked like it would do nothing but impede airflow anyhow, creating a three-deep stack of dense fin stacks and a very sub-optimal flow path. The creation of the bottom left gap as such likely makes this part of the heatsink more useful, reducing flow impedance. Enough to make up for the smaller size? No idea, I'd guess not, but it's less wasteful regardless.

I still don't doubt for a second that this is a weaker cooler overall, but it also has clear signs of being significantly optimized both for cost and for making the most of the flow path and characteristics of the design, rather than just stuffing everything full of weirdly placed fins.

 

[lexluthermiester]

"smaller heatpipes"
View attachment 215060
ok...

Same size(width) heatpipes, but MUCH longer and more of them running along the lateral cooling fins. This cooling solution is an example of applied science. They made a design and learned from it. They revised the design to get the same cooling effect but with materials that cost less, weigh less and are easier to build. Nickel plating Aluminium is less expensive, less time consuming and easier to do than Nickel plating Copper and yet provide very similar thermal performance properties. Once again, much-ado-about-nothing. See below.

Looks to me like they have streamlined the layout of the heatpipes from the previous version. There might be fewer of them as well, but it's difficult to tell from this angle.

See below image;

The red lines are heatpipe pathways, approx. The pink circled areas are where they moved cooling fins to be more effective.

The heatpipe pathways are now much more efficient as they take better advantage of the airflow within the chassis. The circled pink areas are cooling fins that were moved because they were not as efficient where they were placed.

These changes clearly reflect research and understanding being put to work in a way that makes the design more effective. The materials changed only reflect that the Sony engineers learned about the properties of one metal VS another as applied to this specific design and opted to make a choice that would improve manufacturing time, manufacturing complexity and cost, all without sacrificing the overall cooling capacity of the heatsink.

The complaints made by Austin and those who agree with him are without merit. This redesign is an improvement is every way possible. People need to quit complaining.

 

[Valantar]

Same size(width) heatpipes, but MUCH longer and more of them running along the lateral cooling fins. This cooling solution is an example of applied science. They made a design and learned from it. They revised the design to get the same cooling effect but with materials that cost less, weigh less and are easier to build. Nickel plating Aluminium is less expensive, less time consuming and easier to do than Nickel plating Copper and yet provide very similar thermal performance properties. Once again, much-ado-about-nothing. See below.


See below image;
View attachment 215093
The red lines are heatpipe pathways, approx. The pink circled areas are where they moved cooling fins to be more effective.

The heatpipe pathways are now much more efficient as they take better advantage of the airflow within the chassis. The circled pink areas are cooling fins that were moved because they were not as efficient where they were placed.

These changes clearly reflect research and understanding being put to work in a way that makes the design more effective. The materials changed only reflect that the Sony engineers learned about the properties of one metal VS another as applied to this specific design and opted to make a choice that would improve manufacturing time, manufacturing complexity and cost, all without sacrificing the overall cooling capacity of the heatsink.

The complaints made by Austin and those who agree with him are without merit. This redesign is an improvement is every way possible. People need to quit complaining.

Your drawing isn't quite accurate. Going from iFixit's teardown shots, the bottom left heatsink has one pair of pipes through it, which are the same ones that bend up and run along the top of the top left heatsink. The top left heatsink also has a second pair running along its bottom, which presumably run in a diagonal line through the middle heatsink.

Edit: hey, would'ya look at that, they've got an X-ray. Makes this simple. What we can tell: that heatsink had a lot of bends in its heatpipes. Unnecessarily so. That likely made production pretty wasteful. The X-ray also shows us how unnecessarily complicated the flow path for that original heatsink is, and how it likely impedes flow noticeably with that huge fin stack in the middle of the chassis with its fins perpendicular to the air coming off the fan.

The air coming off the heatsink being hotter is still indicative of higher internal temperatures, but alongside reports of it being quieter I'm now leaning towards lower fan speeds being the cause of this, with the increase in internal temperatures being down to Sony seeing that a more relaxed fan curve is okay for the vast majority of users. As for the heatsink itself, its total theoretical dissipation is undoubtedly lower, but given the significant flow impedance optimizations and better heatpipe coverage, I'm tempted to think that the net result in actually cooling PS5s in the PS5 chassis is probably very similar. At the very worst it's still close enough that they could still reduce fan speeds without causing throttling.

 

[lexluthermiester]

Your drawing isn't quite accurate.

It was close enough for everyone to see the point.

Edit: hey, would'ya look at that, they've got an X-ray. Makes this simple.

Yeah there we go. Now we need an XRay of the other one.

 

[trsttte]

Take a closer look.

"smaller heatpipes"

ok...

I don't have them to measure but in the video Austin said they were smaller and they look smaller (in width obviously!)

Same size(width) heatpipes, but MUCH longer and more of them running along the lateral cooling fins. This cooling solution is an example of applied science. They made a design and learned from it. They revised the design to get the same cooling effect but with materials that cost less, weigh less and are easier to build. Nickel plating Aluminium is less expensive, less time consuming and easier to do than Nickel plating Copper and yet provide very similar thermal performance properties. Once again, much-ado-about-nothing. See below.

(...)

The complaints made by Austin and those who agree with him are without merit. This redesign is an improvement is every way possible. People need to quit complaining.

The new design might be an improvement in manufacture in every way possible, but it certainly isn't for the end product. Might still be fine but less heatpipes, less thermal mass, less heatsink area is not better, absolutely it can work perfectly fine but it's not better

 

[lexluthermiester]

I don't have them to measure but in the video Austin said they were smaller and they look smaller (in width obviously!)



The new design might be an improvement in manufacture in every way possible, but it certainly isn't for the end product. Might still be fine but less heatpipes, less thermal mass, less heatsink area is not better, absolutely it can work perfectly fine but it's not better

We'll just wait for the benchmark testing then.

 

[TheoneandonlyMrK]

We'll just wait for the benchmark testing then.

Best plan, it's hard to be definitively pissed without benches personally.

 

[Valantar]

We'll just wait for the benchmark testing then.

The problem is there's no real way to measure SoC temperatures (a thermocouple on the SoC might kill it due to poor contact after all). Best you can do is a thermocouple next to/underneath the SoC, which is less than ideal, though it will give an indication. Hoping GN picks up where their previous teardown+testing left off (including memory thermals).

 

[RH92]

GN just showed that the cooler didn't really touch the memory at all, using a steel plate to contact some of the memory chips. That hasn't changed, memory cooling is likely to be identically bad to previously, they just reduced the heatsink area for what it actually does cool meaningfully - the SoC.

Yes they showed memory being the big issue but not only that . SOC is HOT aswell , they show 73 degrees but keep in mind this is just measuring underneath the SOC ( they didn't remove the heatsink ) so the silicon itself is probably in the low to mid 80 degrees . With the new heatsink being massively downscaled you can expect the SOC to be in the high 80 if not 90 degrees , knowing PS5 boost algorithme is very variable you can be sure this is going to impact performance on demanding games .

 

[Chomiq]

The best part about this "controversy" is that at least in Poland people were trying to get the new revision because it runs quieter.

 

[BorisDG]

The best part about this "controversy" is that at least in Poland people were trying to get the new revision because it runs quieter.

It's from unit to unit. Usually fans are quiet. The problem is the vibrations over the plastic chassis and the coil whine which is super common.

 

[Valantar]

Yes they showed memory being the big issue but not only that . SOC is HOT aswell , they show 73 degrees but keep in mind this is just measuring underneath the SOC ( they didn't remove the heatsink ) so the silicon itself is probably in the low to mid 80 degrees . With the new heatsink being massively downscaled you can expect the SOC to be in the high 80 if not 90 degrees , knowing PS5 boost algorithme is very variable you can be sure this is going to impact performance on demanding games .

Silicon is perfectly fine running at 90°C essentially forever, so this isn't really an issue. It's only an issue if a) the console becomes noisier, b) the console becomes unable to avoid thermal throttling (which could be anywhere from 100-110°C) increased SoC thermals cause other components close by to overheat.

As for the boost algorithm, that's a misrepresentation. The PS5 does not boost in the same way a PC does (i.e. in direct relation to thermals and power draw), but follows a DVFS curve set from a "model SoC" depending on the load scenario. In other words, performance is adjusted not through an answer to the question "how hot am I, and how much power am I consuming?", but through an answer to the question "what software am I running, and what level of performance do I need to deliver in this scenario?". From what I've seen it won't throttle, but rather ramp up its fan if needed. (Obviously it will throttle at some point, but only when it actually overheats. And being a console, it's likely to throw an overheating error and shut down/freeze to protect itself at that point.) And those fans have a lot left in the tank in normal conditions.

 

[Crackong]

ok ,
HardwareBusters did the test.
And the new cooler isn't a cut down
It just trade memory temps for CPU core temps.
New version's core temp is 10C cooler than the old ones.

 

[lexluthermiester]

ok ,
HardwareBusters did the test.
And the new cooler isn't a cut down
It just trade memory temps for CPU core temps.
New version's core temp is 10C cooler than the old ones.

And there we are, more info showing an improvement. The higher exhaust temp? That's a good thing. It means the cooling system is extracting heat better. This is why engineers get paid the big money.

 

[Valantar]

And there we are, more info showing an improvement. The higher exhaust temp? That's a good thing. It means the cooling system is extracting heat better. This is why engineers get paid the big money.

Lol, it's so confusing that there are two threads on this. We've been discussing this for quite a while over in the other thread. The methodology is ... a bit iffy (especially measuring SoC temperatures on the outside of the RF shield on the opposite side of the PCB, which means the thermal pathway is core->package->BGA->PCB->(likely various SMDs+some air->)rear cover->(air->)retention bracket->(air->)RF shield), but probably good enough for a like-for-like comparsion. I'm also very curious why HWB's "old" VRM temps are ~half of what GN found. Still, if anything this demonstrates that the original cooler design must have been pretty terrible overall. Most of us have been assuming that the baseline was a good cooler and discussing how this might improve (or not) on that, which is quite different from starting from a bad cooler and discussing how that might improve. I still struggle to understand how a too restrictive design can result in lower exhaust temperatures (with the same thermal load and slower air movement the air should be heating up more, not less, as it spends more time close to the heat source), but there's probably some other factor in play here that I'm not understanding.

(One possible explanation from looking at some teardown pics: if the original heatsink was sufficiently restrictive, that might have caused more air to leave through the (much smaller, but less restricted) back-of-the-motherboard vents, causing a significant part of the airflow to just bypass the heatsink entirely. This is pure guesswork of course, and it also begs the question of how this would affect temperatures measured on this side (given that everyone is using disc-less PS5s for this there's no ODD to block airflow behind the SoC). A steel RF shield isn't a good heatsink, so the difference from this is likely far less than any difference from more air actually going through the heatsink. But it would be interesting to see some experiments around this, for example by taping over the exhaust vents coming from the rear of the motherboard on the old design and seeing if/how this affects airflow and exhaust temperatures. (Of course that would likely cause the VRMs to overheat, so not exactly a fix for anything )

 

[BorisDG]

ok ,
HardwareBusters did the test.
And the new cooler isn't a cut down
It just trade memory temps for CPU core temps.
New version's core temp is 10C cooler than the old ones.

If the memory is now more hot than previous, that's really bad news, since PS5's memory ran already toasty. The APU was never a big issue.

 

[Valantar]

If the memory is now more hot than previous, that's really bad news, since PS5's memory ran already toasty. The APU was never a big issue.

What they label as 'memory' there is one of the SSD's NAND chips. They didn't measure RAM temperatures for some reason.

 

[BorisDG]

I rather have more hot SOC than any memory chip.

 

[Tardian]

Engineering World Cup?