News Graphene thermal pad for AMD CPUs promises 17X better conductivity than thermal paste, 2X improvement over Thermal Grizzly

[Admin]

Coracer has launched the GPE-01 graphene pads for AMD's AM5 processors.

Graphene thermal pad for AMD CPUs promises 17X better conductivity than thermal paste, 2X improvement over Thermal Grizzly : Read more

 

[thestryker]

Thermal conductivity by itself doesn't mean a whole lot because thickness is equally important. Roman (der8auer) went through how they test TIM at Thermal Grizzly and what all goes into making a good material:

View: https://youtu.be/1BfSx5IG42c?si=_Bz0ZSLLz7gLkRCX

 

[bit_user]

I don't pay too much heed to manufacturers' claims about thermal conductivity. What really matters is test results. This shows Thermal Grizzly's KryoSheet is truly in a different class than pastes like Arctic MX-6:

Source: https://www.techpowerup.com/review/thermal-grizzly-kryosheet-amd-gpu/5.html
​

It's a pity I haven't seen it tested on Toms. Ideally, we'd see a review that tried it on both an AM5 and Intel CPU with an air and water cooler.

P.S. Some people tout the reusability of graphene pads, but Thermal Grizzly themselves advise against this. You can look at that cynically, if you wish. And yes, I know the reviewer at TPU (above) got better performance, the second time around. That's a sample of 1 (would be nice to have more data points... hint, hint).

 

[John Nemesh]

I am completely disinterested in a product that 1) hasn't been thoroughly tested by independent reviewers and 2) is unobtainable through normal channels in the US.

It could be the best thing ever, but until it has been tested and the results PROVEN, AND it's easily ordered through major retailers here in the US, I won't give it a 2nd thought.

 

[SheLikesMeForMyClock]

I don't pay too much heed to manufacturers' claims about thermal conductivity. What really matters is test results. This shows Thermal Grizzly's KryoSheet is truly in a different class than pastes like Arctic MX-6:

​

​

Source: https://www.techpowerup.com/review/thermal-grizzly-kryosheet-amd-gpu/5.html​

​

It's a pity I haven't seen it tested on Toms. Ideally, we'd see a review that tried it on both an AM5 and Intel CPU with an air and water cooler.

P.S. Some people tout the reusability of graphene pads, but Thermal Grizzly themselves advise against this. You can look at that cynically, if you wish. And yes, I know the reviewer at TPU (above) got better performance, the second time around. That's a sample of 1 (would be nice to have more data points... hint, hint).

Agreed. I'm currently using a Kryosheet on my 9950X. Granted I have open loop cooling, but with PBO and DDR5-8000 (I'm still getting my more custom OCs dialed in) my idle temps are ~10°C below average, and my multi-core Cinebench run tops out 18°C below TJMax at 100% load. I'm obviously working now to rid myself of that headroom in place of some extra benchmarking points, but I was pretty blown away that I even had that much headroom on a modern 16C/32T CPU to begin with. Most people would undervolt, but I'm just gonna raise TDP/PPT until it either asks for more cooling or more voltage, and then go from there. I have to think that Kryosheet has at least *something* to do with it.

 

[psyconz]

I am completely disinterested in a product that 1) hasn't been thoroughly tested by independent reviewers and 2) is unobtainable through normal channels in the US.

It could be the best thing ever, but until it has been tested and the results PROVEN, AND it's easily ordered through major retailers here in the US, I won't give it a 2nd thought.

Good thing Tomshardware is available worldwide and not just the USA! ;P

 

[abufrejoval]

P.S. Some people tout the reusability of graphene pads, but Thermal Grizzly themselves advise against this. You can look at that cynically, if you wish. And yes, I know the reviewer at TPU (above) got better performance, the second time around. That's a sample of 1 (would be nice to have more data points... hint, hint).

I went with Roman Hartungs liquid metal on an Erying made G660 board with a mobile Alder Lake i7-12700H, since its original paste caused it to throttle before the die sensor ever reported more than 60°C.

It seemed like perfect conditions, with the board horizontally mounted and the (naked mobile) die not having any resistors or similar on top, which could short out when metal escaped. The die carrier also featured a ring on its outside, much like a tiny wall around the laquered swimming pool or basin the die sat on.

Results were fantastic, too, I got desktop-class results at 90-120 Watts yet the very low idle typical from mobile chips.

But a few months later, throttling returned and inspection found the metal had all escaped into that basin on the sides of the chip on the die carrier, creating a void between the die surface, the cooling shim that effectively replaced the IHS and the cooler.

It hadn't escaped further (until I treid to remove it), but it certainly could no longer cool the chip. You really need a seal around the liquid metal, it won't just stick in place on its own even in a horizontal mounting position.

Nvidia uses a three ringed seal on Blackwell, if I understand correctly...

I also saw visible pitting on the die's top, which I wasn't very happy about.

I replaced it with Kryosheet next, but the results were nowhere near what I had gotten with the liquid metal and when I checked for proper seat and alignment of the Krysheet, it teared and fell apart rather easily, perhaps because also the die and shim surfaces were no longer had a mirror sheen to them.

I then also read up on testing reports from Igor Wallosek, who stressed that proper pressure was rather important for the correct functioning of graphene pads. My low-profile Noctua cooler doesn't offer much pressure control, and I felt it was simply adding another variable I didn't want to care for.

For a moment I even considered mixing the graphene pad and liquid metal to keep it from running out under pressure, but reconsidered just how smart I had been when it came to cooling things (and that evidently nobody else was doing this).

And if the pitting were to be on top of one of the CPU cores, chances are neither sheet nor paste will properly fill it creating a gap. And Alder Lake doesn't do per-core individual turbo limits, yet, that came with Raptor. So any further pitting and deterioration of a single core would thottle or kill the entire chip.

For the moment I've returned to paste, which gives me identical results to the graphene pads, mostly a turbo limit near 45 Watts while reported die temperatures remain around 60°C. To me that indicates rapid throttling action, but at least a system that still works as designed (for mobile TDP) and with all cores.

I've also ordered a phase change pad which I plan to mount next time I take the system apart, just for sake of curiosity and because I'm not about to delid one of my really expensive workstation CPUs: the G660 was cheap enough to be victimized for the sake of cooling science, if it should come to that. Being phase change it might just be liquid enough to fill some pits, yet not creep out like the liquid metal did, at least that's what I aim for.

But yeah, when I reported my disappointment at the graphene pad's endurance to their customer service, Thermal Grizzly confirmed that it wouldn't support multiple applications, something that Linus unfortunately hinted at in one of his videos.

I guess there is quite a few good reasons, why paste is still so popular!

 

[bit_user]

For a moment I even considered mixing the graphene pad and liquid metal to keep it from running out under pressure, but reconsidered just how smart I had been when it came to cooling things (and that evidently nobody else was doing this).

Inspired by a side-comment made by the TechPowerUp reviewer of the graphene KryoSheet pad, I am planning to try this. I already have some Conductonaut Extreme, for the purpose. I am planning on using less than the normal amount of liquid metal, and hoping the porosity of the KryoSheet helps to hold it in place, along with the slight concavity of the loaded LGA1700 IHS (even when used with a contact frame!). The conductivity of the KryoSheet might also help the liquid metal stick to it. I have two KryoSheets, so that I can test without and with liquid metal, each time with a fresh sheet.

And Alder Lake doesn't do per-core individual turbo limits, yet, that came with Raptor.

Mine does. The i9-12900 supports Turbo Boost 3.0. Two of my P-cores can boost up to 5.1 GHz, while the rest report they can only boost up to 5.0.

I've also ordered a phase change pad which I plan to mount next time I take the system apart, just for sake of curiosity and because I'm not about to delid one of my really expensive workstation CPUs:

You might profit from a quick read of my misadventure with it.

Discussion - Thermal Grizzly PhaseSheet Fail

This is a PSA, of sorts. I just tried applying Thermal Grizzly's PhaseSheet PTM and it was a complete disaster. https://www.thermal-grizzly.com/en/phasesheet-ptm/s-tg-ps The instructions suggest you refrigerate (not freeze) the pad, before applying it. I did this, but then I sized and cut...

forums.tomshardware.com

I plan to try again, this weekend. I'll be updating that thread, whether I succeed or suffer another failure.

Being phase change it might just be liquid enough to fill some pits, yet not creep out like the liquid metal did, at least that's what I aim for.

I've read they recommend 10 heating/cooling cycles, before you should expect to see its best performance.

Also, note that they claim it's not as "low maintenance" as the KryoSheet. I suspect that means it's susceptible to some amount of pump-out or drying.

I guess there is quite a few good reasons, why paste is still so popular!

I just used some Artic MX-6 with an old Cooler Master "direct-touch heatpipe" heatsink and found I had to use quite a lot of it! I ended up using an "asterisk" pattern and then spent like 5 minutes working it back & forth (the mounting mechanism limited my range of motion).

I think a PTM pad would be perfect, for that sort of application. Maybe we'll find out!
; )

 

[abufrejoval]

Inspired by a side-comment made by the TechPowerUp reviewer of the graphene KryoSheet pad, I am planning to try this. I already have some Conductonaut Extreme, for the purpose. I am planning on using less than the normal amount of liquid metal, and hoping the porosity of the KryoSheet helps to hold it in place, along with the slight concavity of the loaded LGA1700 IHS (even when used with a contact frame!). The conductivity of the KryoSheet might also help the liquid metal stick to it. I have two KryoSheets, so that I can test without and with liquid metal, each time with a fresh sheet.

That's the thing, the Kryosheet shouldn't really be porous, once the proper pressure is applied: real pores would be air and thus not conduct terribly well.

As far as I recall, the fibres aren't just a horizontal cross-weave, but supposedly angled somewhat upright, because their best conductivity is end-to-end, not from the side. And then they are supposed to bend and pack tight with pressure, with mostly round-to-square gaps left to fill. Carbon fibers aren't really compressible so the slight compressibility or surface adaptability required needs to come from the adjustable weave and filler material.

That filler material isn't discussed by Roman Hartung and only briefly debated by Igor Wallosek and likely occupies the same round-to-square space that the liquid metal could. That may also be the reason, multiple applications don't really work, because the filler material is no longer optimally distributed between the fibers, while their short length causes easy untangling and separation.

I didn't dig much beyond that, Roman Hartung can't go into too many details for IP reasons and I don't have the type of material, tools and know-how that Igor Wallosek has at his disposal.

Nor did I want to create another mess on a hunch, even if I still have all the raw materials left over.

You may want to read Igor's test reports or search a little further, but if you have an IHS on the CPU, the risks are much lower anyway. The corrosive and pitting effects of the liquid metal on both the naked die and the metal surfaces of the IHS-shim and the cooler had me reaffirm that I really don't like liquids in my systems.

I'm really not an overclocking guy, it's only that at 60°C the CPU obviously wasn't even trying. Now I'm just not sure it's really just not measuring fast enough and throttle kicks in long before permanent damage can be done in that tiny hotspot that no longer gets contact cooling in the pit.

Thanks for the link on the phase change experiences, I'll try to keep that in mind and perhaps wait until Summer is over.

 

[bit_user]

The corrosive and pitting effects of the liquid metal on both the naked die and the metal surfaces of the IHS-shim and the cooler had me reaffirm that I really don't like liquids in my systems.

What was the material of the shim? Since most high-end heatsinks are nickle-plated (as is the one I'm planning to use), I'm assuming I won't have much to fear from it. They do explicitly caution against using with aluminum.

I'm really not an overclocking guy, it's only that at 60°C the CPU obviously wasn't even trying.

Me neither, but I like to get the most out of my heatsink and fans, as well as promoting better hardware stability. About the furthest I venture towards the realm of overclocking is raising power limits, which won't void your warranty. Of course, that places more demands on your cooling, which is another reason I'm interested in the subject.

Now I'm just not sure it's really just not measuring fast enough and throttle kicks in long before permanent damage can be done in that tiny hotspot that no longer gets contact cooling in the pit.

Thanks for the link on the phase change experiences, I'll try to keep that in mind and perhaps wait until Summer is over.

Oh, I'd say that's overkill. My plan is simply to peel off the first layer of plastic immediately after removing it from the refridgerator. I will also have a sharp hobby knife handy, in case it needs a little encouragement detaching from one sheet or staying attached to the other. Once that first sheet is removed, I think the rest should go fine.

Like I said, I'll update the thread with the outcome of the second attempt.

 

[Elusive Ruse]

I have been using a Kryosheet for over a year now and I’m quite pleased with its performance over the MX-5 I was previously using. I highly recommend it to anyone who is thinking of making the switch.

 

[YSCCC]

TG is actually impressive company, and I admire their R&D, now all my rigs use either aliexpress PTM7950 or TG phase sheet PTM for long term install and forget CPU TIM, performs great and no worry of pump out is the best part

 

[abufrejoval]

What was the material of the shim? Since most high-end heatsinks are nickle-plated (as is the one I'm planning to use), I'm assuming I won't have much to fear from it. They do explicitly caution against using with aluminum.

Erying doesn't say.

It doesn't look as copperish as it does on some of their 13 Gen boards, but silver. Yet its color seems to differ a bit between the core part and the rest so it could be different alloys. I'd expect nickel plated surfaces to really shine, but none still do, so either the nickel is compromised (Noctua) or was never there (shim) and now with liquid metals and heat new alloys might be in the making...

I read plenty of fine print, ...after the fact 😕

 

[jonaswox]

What we need is a socket capable of taking a ton of mounting pressure per square feet. If I remember correctly at this pressure a soft piece of metal will fill the gaps and conduct at incredible rates. All while being readily available and cheap.

So if the tim area was designed to withstand such pressures, and you can avoid cracking the silicon, then we would have the most insane solution for 0$

 

[jonaswox]

Erying doesn't say.

It doesn't look as copperish as it does on some of their 13 Gen boards, but silver. Yet its color seems to differ a bit between the core part and the rest so it could be different alloys. I'd expect nickel plated surfaces to really shine, but none still do, so either the nickel is compromised (Noctua) or was never there (shim) and now with liquid metals and heat new alloys might be in the making...

I read plenty of fine print, ...after the fact 😕

The nickel is solely used to protect the copper. It oxidizes very quickly on the surface without protection from oxygen

 

[abufrejoval]

What we need is a socket capable of taking a ton of mounting pressure per square feet. If I remember correctly at this pressure a soft piece of metal will fill the gaps and conduct at incredible rates. All while being readily available and cheap.

So if the tim area was designed to withstand such pressures, and you can avoid cracking the silicon, then we would have the most insane solution for 0$

If you want to apply pressure, you need to have counter pressure to avoid bending and even then silicon can't take that much without damage.

Dual side and precision advance materials engineering is the other way to go, but that would mean that the first couple of cooling layers go back to the chip maker to design something magic, where you then only connect your cooling fluid pipes: no longer PC territory, but mainframe or HPC.

I remember ages ago HP designed a mainboard with large chunks of copper inside the mainboard for better chip cooling, must have been during ECL times. And boy did those ECL mainframe look cool in terms of water cooled mechanical engineering, much like colliders or quantum stuff these days. Only to have their noses shoved into the dust by dirt cheap CMOS in ceramic or even plastic! ECL melted at 50MHz, and we're close to 5GHz with 5 Watts!

Who knows what will happen once we get those carbon nanowires really excited!

In this case I only got started on a journey I've carefully avoided for decades because 60°C was just too far from where CPUs should really throttle, had it been 80°C I'd have never been tempted.

But I learned something, no matter what.