Thermal Grizzly KryoSheet

[lexluthermiester]

Would love to see a PTM7950 comparison as well.

Why?

All cool to work with water blocks? Thanks.

Can't see any reason why not.

 

[BoggledBeagle]

Anothers review with a @ 14900K & 7950X would be amazing too like a part #2. Testing with a watercooling setup on CPU and pushing high load, for see how the sheet is good versus like KPX paste and LM

7950X is not a great CPU for testing thermal interface because it has high thermal resistance due to the thick heatspreader.

14900K bends slightly in the socket with a stock mounting mechanism which may be a problem for the sheet.

The ideal test setup would be 14900K with a mounting frame, it can deliver a lot of heat too...

 

[Assimilator]

I too would like to see PTM7950/PTM7958 tested. I'd also be very interested in seeing long-term tests of thermal pastes WRT pump-out, although I appreciate this is rather more difficult to do.

 

[Philipp_Counter]

What in the world is „Thermal Grizzly's own Conductonaut paste“?

I was so confused by the low performance of liquid metal until I read this. Now I‘m assuming they mixed up kryonaut and conductonaut, bringing those temps much more in line with my own experience.

(btw. from my testing hotspot is almost 10C better on LM while other pastes and the Kryosheet was very similar)

 

[lexluthermiester]

7950X is not a great CPU for testing thermal interface because it has high thermal resistance due to the thick heatspreader.

Nonsense. The IHS on the AM5 line-up has not shown any attributes that would indicated inhibited performance.

14900K bends slightly in the socket which may be a problem for the sheet.

How so? The sheet is flexible and thus would flex right along with any "bend" from the CPU.

The ideal test setup would be 14900K with a mounting frame, it can deliver a lot of heat too...

No, an ideal CPU based test would be an OC'd Threadripper, full stop. However, the direct-die-contact made by the Radeon GPU W1zzard used provides the most ideal testing scenario one can get regardless of the type of heat-source. A better testing situation simply doesn't exist.

 

[BoggledBeagle]

Nonsense. The IHS on the AM5 line-up has not shown any attributes that would indicated inhibited performance.

Incorrect. AM5 CPUs thermal throttle with far lower thermal output than LGA1700 CPUs. You can pump 300W from LGA1700 CPU without hitting thermal limit, you simply cannot do that with AM5 CPU without delidding it.

How so? The sheet is flexible and thus would flex right along with any "bend" from the CPU.

The sheet is flexible but the base of the cooler is not.

No, an ideal CPU based test would be an OC'd Threadripper, full stop. However, the direct-die-contact made by the Radeon GPU W1zzard used provides the most ideal testing scenario one can get regardless of the type of heat-source. A better testing situation simply doesn't exist.

I actually do not think that the sheet is perfect for direct die cooling, because it does not seem to be perfectly homogeneous, unlike a layer of metal TIM under a heatspreader (provided there are no bubbles/voids in it).

 

[TheQuentincc]

Is there a typo between the "conductonaut"/"konductonaut" liquid metal and the "kryonaut" thermal paste ?

 

[lexluthermiester]

Incorrect. AM5 CPUs thermal throttle with far lower thermal output than LGA1700 CPUs. You can pump 300W from LGA1700 CPU without hitting thermal limit, you simply cannot do that with AM5 CPU without delidding it.

Citation?

The sheet is flexible but the base of the cooler is not.

Sit back and think that over for a moment or two..

I actually do not think that the sheet is perfect for direct die cooling

You're welcome to your opinion, but the laws of physics would argue(successfully) that direct-to-die testing is optimal.

because it does not seem to be perfectly homogeneous, unlike a layer of metal TIM under a heatspreader (provided there are no bubbles/voids in it).

You seem to be missing the context of both the testing parameters and the point of direct-to-die testing as a methodology.

 

[Veseleil]

Certainly a nice alternative to paste, especially with GPUs. I will try my best to get this on each of my GPU dies, the next time a disassembly is about for any reason.
Now we need someone to make these in 0,5; 1; 1,5; 2; 2,5; 3mm so I can deshroud, clean & apply and forget about removing a heatsink in a lifetime!?

 

[BoggledBeagle]

Citation?

Common knowledge.

Sit back and think that over for a moment or two.

"flexible" that means "able to flex" does not mean, that it can move the material from the area with high pressure to the area with low/no pressure unlike the normal thermal paste.

You're welcome to your opinion, but the laws of physics would argue(successfully) that direct-to-die testing is optimal.

It is not. Optimal for testing coolers and TIM is a heatplate with configurable and exact heat output. Using CPU or GPU as a heatsource is a lazy amateur approach.

 

[Colddecked]

Why?


Can't see any reason why not.

Well, to see how it compares? I'm looking at purchasing some to put in my laptop.

 

[Lomskij]

Any plans to test indium foil next?

 

[Chrispy_]

It seems pointless for anything under about 250W, as it's 50x the cost of paste and performs worse.

it's only at above 250W where the Kryosheet starts to outperform the pastes, and that's basically three GPUs and one CPU exclusively.

 

[Lewzke]

PTM7950 is the holy grail of thermal interfaces * especially for portable devices or GPU's (direct die contact), it should be on the comparison. The overall behavior (longetivity + performance + non-conductiveness) is hard to beat. The only negative is that it's difficult to apply. It is just slightly worse than liquid metal, but it's safe to use. I think it is better choice than this conductive kryosheet. The Kryosheet is exclusive choice for high TDP desktop components.

 

[AMD718]

Have had excellent results on my XFX 7900 XTX at 460w. Allows me to run with minimum fan setting and only hit 90c hotspot. Delta is <=20c. With moderate fans delta shrinks to <=12c and hotspot doesn't exceed 80c.

 

[THU31]

I was thinking of getting my Xbox 360 Slim cleaned and re-pasted. I might actually get a thermal pad so I don't have to do it again.

 

[W1zzard]

Also you can hook up thermal shrink tube to any tweezers and make it as a gentle pick up tool. Be creative.

Pro tip is to use a bit larger and put some small thermal paste dots in the substrate corners, so it sticks while you align the heatsink on.

Great suggestions

 

[evernessince]

Citation?

I don't know why he couldn't just link you a source but here is what you are looking for:

According to Derbauer the IHS on Ryzen 7000 series CPUs is both thicker and has less surface area.

Common knowledge.

If you are going to make an argument a provide a source. You aren't going to win anyone over to your side with a trust me bro angle.

 

[rusTORK]

It's REALLY cool to see review of KryoSheet on TPU!

I hope it help boost (a bit) development and wide spread of such products (and more research in this direction).

I personally tested Panasonic PGS and soft-PGS sheets. Latest i tested was Thermal Grizzly Carbonaut on RTX 3080 Ti FTW3 Ultra Gaming. There is small catch with this things:

0.2 mm thickness

On my GPU i got 2 mm (instead of 2.25mm by default) thermal pads (GP-Extreme from GELID) on VRAM. This give me memory overheat. Even if it wasn't part of a testing - it show problem. Of course when i removed Carbonaut and apply thermal paste again - VRAM temp back to normal.

So, need to pay more attention not only to dimensions (cut to die), but also thickness of sheet.

Last experiment was to apply GC4 from GELID, but VERY thin layer (0,05mm - thickness of Kapton tape) and test it. I got doubts it's even work, but decided to test anyway. It work awesome! After that i decided to apply same layer on my CPU - 5800X3D and got decent temperatures again!

 

[Deleted member 237269]

It's REALLY cool to see review of KryoSheet on TPU!

I hope it help boost (a bit) development and wide spread of such products (and more research in this direction).

I personally tested Panasonic PGS and soft-PGS sheets. Latest i tested was Thermal Grizzly Carbonaut on RTX 3080 Ti FTW3 Ultra Gaming. There is small catch with this things:

0.2 mm thickness

On my GPU i got 2 mm (instead of 2.25mm by default) thermal pads (GP-Extreme from GELID) on VRAM. This give me memory overheat. Even if it wasn't part of a testing - it show problem. Of course when i removed Carbonaut and apply thermal paste again - VRAM temp back to normal.

So, need to pay more attention not only to dimensions (cut to die), but also thickness of sheet.

Last experiment was to apply GC4 from GELID, but VERY thin layer (0,05mm - thickness of Kapton tape) and test it. I got doubts it's even work, but decided to test anyway. It work awesome! After that i decided to apply same layer on my CPU - 5800X3D and got decent temperatures again!

Try to use Putty instead of pads. You will not have to worry about thickness.

Arctic mx-6 and upsiren u6 pro on my laptop solved my issue.

For your GPU i recommend Grizzly Kryosheet / Honeywell PTM7950 / Gelid Solution HeatPhase Ultra / Arctic MX-6 and for the VRAM try on this list. I got myself a 50g Upsiren U6 Pro qnd UX Pro from this store.

Hope this help you.

 

[wolf]

Awesome review @W1zzard and what a nifty product, I have seen other publications testing with it, and I think curiosity has gotten me, I'll be trying this on future builds/cards for sure.

 

[watzupken]

I feel graphene sheets may not work that well for a chiplet design as the height of each chip may be differ. Probably if this was tested on say an Nvidia GPU die, it may perform better. The beauty of it is that it will not dry, get pumped out, and conductivity will always remain consistent.

 

[lepudruk]

I was never about to find out how it performs because the fabric litteraly fell apart during cutting. It's very fragile. It looks like many linen fibers combined into fabric with some gray grease and pressed into material.
I followed the instruction, used sharp scissors and was very careful. I don't know, perhaps my piece was old, faulty or something but in the end I was left with few shreads of the fabric I couldn't do anything with. Expensive and tricky to play I would say.

 

[lexluthermiester]

Common knowledge.

So you don't have any, that's all you had to say.

It is not. Optimal for testing coolers and TIM is a heatplate with configurable and exact heat output.

In the context of testing absolute heat transfer in laboratory conditions, maybe. In the context of testing with ACTUAL electronics intended for use? No. The GPU test used in this review was an excellent choice. The only other choice would have been to use stock CPU setup, like a 14900K, 7950X or Threadripper model. When it comes to testing thermal interface materials, you don't use lab conditions, you use contextual testing, IE, you test with examples of equipment that the TIM is intended to be used for/on/with. Only then can you get a proper understanding of ACTUAL, real world performance.

Using CPU or GPU as a heatsource is a lazy amateur approach.

Says the person NOT doing the professional work? You're not helping your argument by taking silly jabs at known credible methodologies and people who are known for be very through in their work. You think you can do better hotshot? Prove up or shut-up. Do your own tests with your own equipment and let's see the results.

I don't know why he couldn't just link you a source but here is what you are looking for:

According to Derbauer the IHS on Ryzen 7000 series CPUs is both thicker and has less surface area.

Ah, right, I've seen that. Here's the thing, DB's conclusion is based on a specific set of metrics and school of thought. Even he stated those were just his thoughts. In practicality, whatever limitations exist are not causing enough thermal transfer bottlenecking to be a problem even at full load for the 7000 series CPUs. So the argument of the thicker IHS, while worthy of momentary consideration, is not something to be concerned about.

Well, to see how it compares? I'm looking at purchasing some to put in my laptop.

The short and easy answer, it will do fine. Most general use TIMs perform within 5% of each other these days. Unless you're doing extreme OCing or you have a very high-end part, go with one that doesn't dry out or need replacing every year and you're good. Laptop you say? Buy what you want, smear it on. You're good to go.

 

[Icon Charlie]

Thank you w1zzard for the article. These are the articles that I enjoy the most as they are informative in nature.