News Intel’s next-gen desktop CPUs may run even hotter than current ones — chipmaker allegedly extends maximum temperature for Arrow Lake CPUs

[TheHerald]

I didn't say in the CPU die, I said on the CPU, it's not like anybody runs (can run) their CPUs without a cooling block or cooling in general.

Yes, but at 5C more you have 5% more time at the same amount of cooling until that time, or your cooling can run 5% lower and the saturation will happen at the same time.
At 5% lower running cooling your room, or lap, will be 5% cooler for that duration of time and if you use your laptop for light workloads. that duration of time can be indefinite.

Again, it's not how it works. A cpu that draws 125w will have to dissipate 125w no matter what temperature it's running at.

 

[TerryLaze]

Again, it's not how it works. A cpu that draws 125w will have to dissipate 125w no matter what temperature it's running at.

Only once the cooling is fully saturated.
Go into your bios and look at you fan curve, it doesn't associate rpm( % ) with wattages, it associates rpm ( % ) with temperature barriers.
As long as the CPU is below X degrees the fan will run at Y %
If X increases by 5% Y can decrease by 5% for the same final amount.

 

[TheHerald]

Only once the cooling is fully saturated.
Go into your bios and look at you fan curve, it doesn't associate rpm( % ) with wattages, it associates rpm ( % ) with temperature barriers.
As long as the CPU is below X degrees the fan will run at Y %
If X increases by 5% Y can decrease by 5% for the same final amount.

The heck are you talking about man. Stop. You are wrong. Just stop.

If the cooler isn't dissipating 125w on a cpu that is pulling 125w then the temperature of the cpu will keep rising. Forever. Until it physically melts when it reaches sun surface temperatures.

 

[bit_user]

Yes, but at 5C more you have 5% more time at the same amount of cooling until that time, or your cooling can run 5% lower and the saturation will happen at the same time.

Those numbers are completely fictitious. Any increase in turbo time you'd get would come from the first of the two benefits listed below, both of which have already been discussed.

1. Increasing the temperature delta across a thermal solution increases its efficiency (i.e. the rate at which it's able to transfer heat).
2. A smaller process node naturally means increased thermal density, so a higher TJMax could be unavoidable without sacrificing performance.

As @Notton said, many laptop makers will probably just exploit # 1 to get away with lighter, cheaper, and smaller thermal solutions, rather than allowing it to provide better boosting.

As for # 2, you've previously ridiculed Zen 4's high die temperatures, but their small size is an intrinsic part of the problem. Now, Intel gets hit with the same phenomenon. The following tweet compares thermal densities between Alder Lake, Raptor Lake, and Ryzen 7000:

Source:

https://twitter.com/x/status/1564409602933792770

View: https://x.com/harukaze5719/status/1564409602933792770

​

 

[Silas Sanchez]

Although a larger temperature gradient means a stronger driving force which in general means more power per unit area, the flow of heat in solid is also dictated by the thermal resistance of that material. You want the metal and anything solid in the path to be as thin as possible. A large heatsink has a poor heat capacity because its designed to have the least thermal resistance and max surface area with the air. Although metal (copper) is a great conductor, once the thermal resistance gets high enough the heat flux becomes truly awful as the heat spreads out randomly under diffusion unlike say the flow of current under the force of the electric field, the thermal resistance captures this. And is why Newtons law of cooling applies more to estimating say how long until a warm drink cools down in a fridge. With metals, the medium for energy flow is the sea of electrons within the crystal lattice, there are so many and this is also why conduction (radom particle movement) works so well that defeats heat flow.

Heat is also released through other parts of the computer, like the rear of mobo, the amount of heat coming off the back of my mobo when the 7950x is bencmarking was enough that i put a fan system in place, which is why these large dual compartment cases are good. The mobo conducts heat from cpu and this shows up in nvme temps. The GPU is another pig and contributes alot of heat to my mobo. Any heat inside the PC or Laptop will make its way out just at a slower rate, its not all released by the cooling system, unless you have some insulation...A key factor in making fridges work well in summer.

Heat flow happens primarliy first via conduction or diffusion (solids interaction with solids &/or liquids), diffusion within liquids (air) that under the influence of gravity and in turn from buoyancy forces establish convection currents, which tend to move the bulk of the heat away to the sides and roof of case/openings. And finally forced convection (active cooling) which is by far the largest mode of heat transfer pound for pound due to how efficient the interaction is between air and a solid, a tiny amount of power to run a fan that removes alot of heat, thats the second law of thermodynamics. This is why you have a large case with large diameter fans, that amount of volume can reach all the places like VRMs, chipsets, nvmes etc. Finally radiative cooling plays a very small role
Now in laptops, it is aboslulety true that much of the heat can be removed from the case/body itself, for example my P72 when benchmarking at ambient roomtemp. The cpu sits at 80-84 degrees C when there is a 1.5Inch air gap under it, vs 90-94C when its siting on a flat surface. And blowing air on the bottom underside or even just the top body will be even better at removing heat, conduction happens at a frantic rate.

However it doesnt do much at all for the perf or times to finish a heavy task, only the actual cpu temps. So the temps dont give much indication of perf and heat dissapation, a low powered cpu will run hot under a heavy load. my P70 cpu is sitting at 84C just to play a 4k 60fps h.265 video, yet the P72 in same setting is like 60C!

As far as Intel goes, the insider consensus has always been chip makers test the die at 125C for 6 months under extreme conds or soemthime like that, and from that derive some model to make predictions, this it important or else they would have no way of knowing what their product can do. Which by the sounds of it means these cpus are good for 5-6years under full load at least but after that the uncertainty and lacking applicablity kicks in.

 

[TerryLaze]

The heck are you talking about man. Stop. You are wrong. Just stop.

If the cooler isn't dissipating 125w on a cpu that is pulling 125w then the temperature of the cpu will keep rising. Forever. Until it physically melts when it reaches sun surface temperatures.

It will only keep rising until it hits the temp barrier set in the fan curve of your bios, are you even reading the things you reply to or are you just ranting now?!
Your cooler doesn't dissipate 125W when it's running at 0% or at 20-30% which are set up in your fan curve.

 

[TerryLaze]

Those numbers are completely fictitious. Any increase in turbo time you'd get would come from the first of the two benefits listed below, both of which have already been discussed.

1. Increasing the temperature delta across a thermal solution increases its efficiency (i.e. the rate at which it's able to transfer heat).
2. A smaller process node naturally means increased thermal density, so a higher TJMax could be unavoidable without sacrificing performance.

As @Notton said, many laptop makers will probably just exploit # 1 to get away with lighter, cheaper, and smaller thermal solutions, rather than allowing it to provide better boosting.

As for # 2, you've previously ridiculed Zen 4's high die temperatures, but their small size is an intrinsic part of the problem. Now, Intel gets hit with the same phenomenon. The following tweet compares thermal densities between Alder Lake, Raptor Lake, and Ryzen 7000:

​

​

Source:​

https://twitter.com/x/status/1564409602933792770

View: https://x.com/harukaze5719/status/1564409602933792770​

​

​

Who talked about turbo time?!

 

[TheHerald]

It will only keep rising until it hits the temp barrier set in the fan curve of your bios, are you even reading the things you reply to or are you just ranting now?!
Your cooler doesn't dissipate 125W when it's running at 0% or at 20-30% which are set up in your fan curve.

What the heck are you even talking about. Fan curves are irrelevant. The cooler HAS to dissipate as much power as the cpu is consuming regardless of the actual temperature of your cpu. Even if your cpu is at 300c the cooler has to dissipate All of the wattage the cpu is pulling. That's like physics 101 man. Just stop.

 

[TerryLaze]

What the heck are you even talking about. Fan curves are irrelevant. The cooler HAS to dissipate as much power as the cpu is consuming regardless of the actual temperature of your cpu. Even if your cpu is at 300c the cooler has to dissipate All of the wattage the cpu is pulling. That's like physics 101 man. Just stop.

What the heck are you talking about?!
You have never seen your CPU get hotter than it was before?!

 

[TheHerald]

What the heck are you talking about?!
You have never seen your CPU get hotter than it was before?!

The cpu getting hotter is irrelevant. The cooler has to dissipate all the power consumed by the cpu. The heat capacity of the cpu itself is negligible, even at a 300c thermal throttle limit it cannot store heat for more than half a second.

 

[TerryLaze]

The heat capacity of the cpu itself is negligible, even at a 300c thermal throttle limit it cannot store heat for more than half a second.

Nobody said anything about any of that being the case though.

The cooler has to dissipate all the power consumed by the cpu.

It has to when the CPU has reached it's thermal limits because otherwise the CPU will get slower, so technically even then it doesn't have to, we just prefer to be able to run the CPU at full power as long as we want to.

The cpu getting hotter is irrelevant.

I presume you mean hotter than an other, so 105 instead of 100.
So I ask you "Using the same amount of power, heating the same thing up to 100 and to 105 degrees, will it take the same time or will 105 take a bit longer? "
And in reverse "to get the same thing to 105 instead of 100 degrees in the same time, will you need the same amount of power or will you need more? "

 

[TheHerald]

Nobody said anything about any of that being the case though.

It has to when the CPU has reached it's thermal limits because otherwise the CPU will get slower, so technically even then it doesn't have to, we just prefer to be able to run the CPU at full power as long as we want to.

I presume you mean hotter than an other, so 105 instead of 100.
So I ask you "Using the same amount of power, heating the same thing up to 100 and to 105 degrees, will it take the same time or will 105 take a bit longer? "
And in reverse "to get the same thing to 105 instead of 100 degrees in the same time, will you need the same amount of power or will you need more? "

Yes, heating it to a 100 or 105 will take the same time. The difference will be in the realm of milliseconds. Basically the question is how much heat is needed to get a cpu from 100 to 105c. The answer is an imperceptible amount. It won't make any difference

 

[bit_user]

Who talked about turbo time?!

That's a natural consequence of what we're talking about. The reason you can have a separate PL2 is that the thermal solution has an initial capacity to absorb heat at a rate higher than it can transfer it, end-to-end. When you're talking about the higher die temperature providing some extra time, that only works because the higher thermal delta across the heatsink interface means more heat can be transferred into it, before it starts to saturate, and the time it takes to reach saturation should be reflected in the combination of PL2 and Tau.

So, with a higher TJMax, the same cooling solution should be able to use a higher PL2, a longer Tau, or a little bit of both. Note, that assumes we're holding the die area constant, which is the other wildcard, here. And I'm still curious whether TJMax is just a point measurement or averaged over what area.

 

[bit_user]

The cooler HAS to dissipate as much power as the cpu is consuming regardless of the actual temperature of your cpu. Even if your cpu is at 300c the cooler has to dissipate All of the wattage the cpu is pulling.

I think the key words you're missing are "steady state". On average, the heatsink is dissipating the same amount of heat as the CPU generates.

Terry was talking about transients. They don't change anything, in the long run. Both at steady state and when you average out the transients, the amount of thermal energy being produced by the CPU must be dissipated by the heatsink.

 

[hotaru251]

Also it's the official recommendation for the 5950x and 7950x.

yes recommended, but at rate intels going its going to be required not just recommended.

 

[vijosef]

Tecs are terrible. When you need them the most (high power draws) they completely fail.

Just because lazy, bad engineering. And they didn't tried cooling the radiator fins.

 

[TerryLaze]

Yes, heating it to a 100 or 105 will take the same time. The difference will be in the realm of milliseconds. Basically the question is how much heat is needed to get a cpu from 100 to 105c. The answer is an imperceptible amount. It won't make any difference

What circumstances are you talking about here?! Taking a CPU and throwing it in the fire?! Because then I agree, it won't make any difference, but we are talking about PCs here.
If the CPU uses 125W and the cooler can only cool 120W then getting that extra 5 degrees hotter will take some time.
If the CPU uses 350W and the cooler can only cool 60W then yes, no time at all.

 

[TerryLaze]

I think the key words you're missing are "steady state". On average, the heatsink is dissipating the same amount of heat as the CPU generates.

Terry was talking about transients. They don't change anything, in the long run. Both at steady state and when you average out the transients, the amount of thermal energy being produced by the CPU must be dissipated by the heatsink.

And you are missing the point that CPUs don't start up with Tjmax already reached...they start up at room temperature and have to heat up to Tjmax.

With everything else being the same heating up 5 degrees more will take some extra time.

 

[bit_user]

And you are missing the point that CPUs don't start up with Tjmax already reached...they start up at room temperature and have to heat up to Tjmax.

I'm not missing the point. I did account for that in post #29, but the caveats and qualifiers I mentioned there are important.

With everything else being the same heating up 5 degrees more will take some extra time.

Ah, but as I also mentioned in that post, we cannot assume "everything else being the same", because thermal density on Intel 20A will definitely be higher, for reasons I explained.

The long and short of it is that we don't know if this is necessary (or even adequate) to deal with higher thermal density. Therefore, it'd be premature to say whether it will result in more, less, or about the same efficiency from the same thermal solution.

At this time, all we can conclude is that it's good they didn't merely keep it the same, but it will be something to keep an eye on, as these products reach the testbench.

 

[ex_bubblehead]

Alright children this is your one and only warning. KNOCK OFF THE PETTY BICKERING RIGHT NOW! If you can't stay on topic and get along this thread will be closed and sanctions WILL be handed out. The ONLY acceptable response here is "I hear and will comply" ANYTHING else will bring down sanctions. This ENDS NOW.

Thank you to those that actuallystayed on topic.

 

[ThomasKinsley]

Lunar Lake and Arrow Lake both sound rather promising, based on what Intel has published about them. Both are due out this year (Lunar in Q3, then Arrow in Q4).

They're trying some rather unique changes, including ripping out hyperthreading on Arrow Lake. It could either pay dividends or it could be shades of Bulldozer all over again. I certainly hope I'm wrong and Intel does well.

 

[PCWarrior]

Ivy Bridge did actually run hotter die temperatures than Sandybridge. Part of that was their unfortunate decision to replace solder with thermal paste, between the die and IHS. Another part of it was due to the size reduction of Ivy Bridge's cores. Smaller cores need to have proportionately smaller TDPs, or else the amount of thermal energy per unit of area goes up. The era of constant thermal density ended approximately in 2006, so it's a given that a lithographic shrink of effectively the same microarchitecture would run at higher thermal density. And that can result in hotspots that are more difficult to cool.

This is one reason Zen 4 is thought to run rather hot, in fact - the thermal density with TSMC's N5 node is quite high. So, given these CPUs' use of Intel's 20A node (and 18A, for Panther Lake?), it seems to me like it might be more of a necessity to tolerate higher TJMax temperatures.

At stock speeds 3rd gen didn’t run hotter than 2nd gen. Not perceivably anyway. The 22nm process was more power efficient compared to 32nm so 3rd gen cpus were consuming less power compared to 2nd gen while also performing better. This reduction in power consumption compensated for the use of polymer TIM instead of solder TIM and the higher heat density as there was less heat to extract in the first place.

Now when overclocked to 10-20% speeds higher than stock the polymer TIM became a limitation and 3rd gen cpus were indeed running hotter than 2nd gen. Even then running hotter didn’t mean reaching TJmax and throttle thermally or anywhere near TJmax. An OCed 3770K at 4.5-4.6Ghz was still running well below TJmax (80-85C on a good air cooler). But again, that was not stock clocks. On stock clocks there was barely a difference and we are talking about sub 60C anyway. Also the main contributor to the higher OC temps was the use of polymer TIM, not the higher heat density. This was verified with tests with delidded cpus using liquid metal on both 2600K/2700K and 3770K. The 3770K was about the same thermally as the 2nd gen when clocked at the same OC speeds (e.g. 4.5-4.6Ghz) and still well below TJmax. Also for OCing with water cooling it was hitting an electrical stability limit (read needing voltage above 1.4V for stability) before cooling was becoming an issue (reaching temps close to TJmax).

 

[XxDarkMario20xX]

They should spend more time making CPU 14900k is pefect example was waste to bring it out because not much difference between 13900k nearly same specs. They could created 14900k more powerful

 

[Lucky_SLS]

I might be wrong but I think this is the case cuz its Intel's first gen design of GAA in a CPU architecture.