7700X lapping, replace IHS or direct die cooling?

[ChrisM255]

Got a few hours to devote to this Saturday morning if anyone wants to catch the results for stock with Kryonaut vs. IHS and cooler lapping results vs the stock with Kryosheet I previously posted. Did some heat cycles with the new Kryonaut so it will be ready for me to test first thing on Saturday.

 

[ShrimpBrime]

Sorry Shrimpie, you totally called it earlier, please don't get mad...
Sooo, I calculated the difference in mass and thermal resistance from AM4-AM5 IHS's

The IHS top plate, like just the top not including the side mass, it went from:
AM4 5800X - TDP/PPT 105/142w, 1300 sq mm @ 2mm thick = 2600 cu mm
AM5 7700X - TDP/PPT 105/142w, 910 sq mm @ 3.7mm thick = 3367 cu mm
That is a 29.5% increase in mass with an 85% increase in z-height
Using the calculations for the thermal conductivity of copper using the dimensions:
AM4 IHS results has a thermal resistance of 0.0038484 degrees C/W
AM5 IHS results has a thermal resistance of 0.01018414 degrees C/W

Its 2.6463309 times the thermal resistance and 30% less contact area to transfer to the cooler.

I'm not saying your wrong about any of your cooling solutions, or that there are no other ways to take care of it or even that it can't run at 95C.
Just for the love of Pete, stop saying that the IHS is "adequate" and "fine"
ALL I'm saying here is:
They F'd it up, come on, ADMIT IT!!! They put a bad IHS on it...

The resistance of copper goes up ever degree of heat added. Do note this feature of copper.

My point about it being adequate was the difference of running a cpu full boost all the time vs it's base clocks.

Now if they released the cpu at base clocks and let you OC the processor, you'd have similar results.

So I will stand firm the IHS plate is adequate. You may go back and review the thread where others with the same exact chip are NOT close to having a thermal issue. That's your proof.

You'd think the 13700K I use with a thinner plate, yet more surface area would be difficult to cool. This is correct. It boosts to 253w.

If I use the same chip, which I've already demonstrated in another thread, I can tune it to run cooler.

That's 253w on a Wraith Prism just sitting on the Intel CPU.

If you want to hear it..... fine, I'll approach the conversation like this.

The AMD IHS plate is not adequate for cooling and overclocked processor, which is already overclocked.

AMD boosts to throttle point. Intel boosts to throttle point.

Neither manufacturer has adequate IHS plates to cover all types of cooling, case flow and ambient temps not to mention partially other variables such as types of fans and humidity. They should release processors at a slightly lower power package, and we'd have less issues cooling our chips.

 

[ChrisM255]

If I use the same chip, which I've already demonstrated in another thread, I can tune it to run cooler.

That's 253w on a Wraith Prism just sitting on the Intel CPU.

Can you link the thread? I am very interested

A question came to me while I was reading, have you ever tried a vapor chamber in your ventures? Direct die? Or is that just crazy...

 

[KedsDead]

I am pretty sure all air coolers use a vapor chamber design in their heat pipes.

 

[ChrisM255]

In high end graphics cards, don't vapor chambers sink heat into heat pipes?

That's a different ballgame in thermal load though I digress

The AMD IHS plate is not adequate...

 

[ShrimpBrime]

Can you link the thread? I am very interested

A question came to me while I was reading, have you ever tried a vapor chamber in your ventures? Direct die? Or is that just crazy...

Yes. I even have a picture of the cooling.

13700k - Gigabyte B760M DS3H MB - E-Cores running at half speed, Benchmarks at half expected?

I have a new Gigabyte B760M DS3H DDR4 Motherboard and i7-13700k CPU. The long and short of it is I appear to be getting half or less of the benchmark values I should be. System performance is stable at this point and the games I'm running work much better than the prior system I was using (both...

www.techpowerup.com

As @KedsDead said, heat pipes are vapor chambers, essentially. Instead of a flat plate design, they are tubes. But work in the same fashion. I have not used a vapor chamber plate, but I have seen them in the wild. Didn't really tickle my fancy to ever try one.

At certain temperatures, copper is effecient with thermal movement. Aluminum as well, but the material is not as dense.

But it's all about surface area.
The IHS plate - surface area
Edit in: Integrated !!! Heat !!! Spreader !!!
Water block - surface area
Heat sink - surface area
Radiators, also - surface area.

You want as many thermal touching as much surface area as possible.

Water delta in a water loop. After radiators, you want the water delta as close to the ambient air temp before returning liquid to the heat source. If not, you just recycle heat through the system.

Have you ever tried .... ...Direct die? Or is that just crazy...

I didn't fully reply.

Yes. I've probably direct die cooled.... I dunno maybe 100 processors? Give or take 20. Dozens of them soldered de-lids. Dozens more just pasted old school processors.

It's not crazy to direct die. It's actually smart to eliminate the plate. With a nice waterblock. I've found with most cooling, more copper is better.

Right, so I mean I'm using a 220w cooler on a 253w chip. It's not exactly easy. Lol. But tuning the chip was easier results than a delid or spending money on a bigger cooler.

AMD lacks surface area.

I can cool my cpu 2c by lowering my room temp by maybe 5c.

I mean look at a NH-D15 air cooler. If a CPU IHS plate cannot match the surface area of coolers that supply a large surface area, that's where AMD actually went wrong.

 

[Chrispy_]

Vapor chambers are crushed more easily than heatpipes so their use has fallen out of favour with CPU cooling since mounting pressures are much higher than on GPUs. Typically an air cooler is a heavy 500g-1kg chunk of metal hanging off the socket with a good ~80-90mm of leverage from the center of mass (heatsink middle) to the fulcrum (lowest edge of the cooler's baseplate).

 

[ChrisM255]

I've found with most cooling, more copper is better.

Was there anything that surprised you that you have ran into?

 

[ShrimpBrime]

Was there anything that surprised you that you have ran into?

Yes in fact.
In 20 years, moving 125w of energy has not once changed.

What surprises me is that people are clouded by terms thrown in like "density".

Yeah 7nm more dense than 45nm. If each chip is 125w, that all I'm moving all these years. The 7nm is more dense Yes, BUT is that much power power effecient. It's still..... 125 fuckin watts.

All the big brains behind moving heat know you have a certain rate of thermal transfer possibility with certain metals.

Lapping works simply because you removed the nickel plating. Have a better contact.

My first surprise on my first waterblock AND cpu lap job, the cpu stuck to the cooler WITHOUT thermal paste. Just like the old timers said it would do!!!
You now can use oh so just a slight amount of thermal grease and that's what you want. As much metal to metal contact as possible.

You won't need ink and a flexible glass pane to know your cpu and cooler are flat. You'll know when you place the cooler on the cpu and lift it off the work bench without paste.

 

[Coyote_ar]

Hi, i found this thread while googling and i found it interesting, since i used to play the "cool the little nuclear reactor" game back in the late 90s/early 00s.


My 2 cents on the cooling deal.


1) More copper is indeed better, but with diminishing returns. Sure conductivity in a straight line diminishes with thickness ... but that is assuming the Area remains constant on both ends. Now what if you use that thickness to spread that heat into a bigger area on the output side? Suddenly the penalty of lower conductivity is compensated by a bigger area, which takes us to the next point.
2) Surface surface surface ... more is always better, as long as you dont mind the delay on moving heat from A to B. So distance is really critical and bad for cooling only when removing heat from the die ... after that its just a matter of how good you are transfering heat on from A to B, so more surface = better. A thick IHS works better than a thin IHS, because the area of heat input remains the same, but the area of the output is used more effectively. Thin IHSs like intel used to use, only server a purpose of protecting the die from mechanical damage, but it provided little spreading ... calling them an IHS was a mistake really.
3) This is an old Athlon XP (-M actually, mobile ones overcloked better). Custom waterblock with a 250w TEC, and a 12mm thick heat spreader (please do forgive the nasty insulation, it was the first attempt on sub zero cooling. later ones were way better). Do keep in mind there was no IHS, and the CPU die was 150mm2 only ... which means that disipating close to 200w the thing had more heat density than a nuclear reactor. In that situation, removing heat from the die, and spreading it to reach the much bigger TEC, required a thick plate. Basically when you need to move in a straight line, thinner works better. But if you need to move sideways, thicker is the way to go.
Also having a buffer to dampen any spikes of heat input never hurts ...

 

[BoggledBeagle]

Vapor chambers are crushed more easily than heatpipes so their use has fallen out of favour with CPU cooling since mounting pressures are much higher than on GPUs.

I thought that vapour chambers have support pillars in them, and they are superior AND MORE EXPENSIVE than standard heatpipes.

The reason why you do not see vapour chambers in CPU coolers may be purely their cost. You only see them in the most expensive GPUs.

 

[Chrispy_]

I thought that vapour chambers have support pillars in them, and they are superior AND MORE EXPENSIVE than standard heatpipes.

The reason why you do not see vapour chambers in CPU coolers may be purely their cost. You only see them in the most expensive GPUs.

You're speculating, but I'm referencing an interview Steve Burke did with either CoolerMaster or Thermaltake on one of the factory tour videos.

Yes, vapor chambers are more expensive, and yes they have support structures of some kind in them because they need to, but the standard round heatpipes, constrained on each side by neighbouring heatpipes or the rest of the CPU block are even stronger - they're an arch in cross-section which is one of the strongest load-bearing shapes mankind has discovered, hence why they're regularly used in architecture and civil engineering projects going back a few thousand years.

 

[BoggledBeagle]

Have you ever seen a direct contact air cooler, where the heatpipes are flattened where they contacts the CPU and also machined away to form a flat surface? These are weakened flat thin copper walls.

 

[Chrispy_]

Have you ever seen a direct contact air cooler, where the heatpipes are flattened where they contacts the CPU and also machined away to form a flat surface? These are weakened flat thin copper walls.

...reinforced on both sides against crushing by solid extruded aluminium arches.

Those arches are a strong geometric structure, but even more importantly they're huge compared to the tiny little pegs in a vapor chamber.

As I said before, I'm not guessing. It's in the GN factory tour videos. Go and watch them yourself rather that arguing with guesswork.

There is definitely an element of cost involved with making vapor chambers, but it's not the only reason, obviously. When RTX 4090's costing $2000 don't use vapor chambers, and manufacturers can clearly see that people will spend $250 on AIOs for CPU cooling, a few extra bucks for a vapor chamber isn't the reason they're not used on CPUs or GPUs with high clamping force.