T[]RK

This thread is following more "research" path then actual use of PGS. It's very interesting product, especially on paper, but make it actually work as we need (replace completely thermal grease) is a big challenge. Especially with HIGH TDP products. I think it will work and keep CPU "in spec", but this way (how to do it exactly) isn't discovered yet. It's easy with LM: get it, apply it, assemble, run, enjoy temp drop. Not interesting because it already discovered. Mix PGS and LM - now this may be interesting! Add PGS under CPU IHS - this is interesing (especially under Ryzen 3000 with it's small chiplets design).

No, i would like to keep data consistent AND keep it in one thread. So, Panasonic have got two new products: 1. Multilayer PGS (ML-PGS?); https://eu.industrial.panasonic.com/products/circuit-thermal-protection/thermal-protection/pyrolytic-graphite-sheet-pgs/Multilayer-PGS#overlay-context=products/circuit-thermal-protection/thermal-protection/pyrolytic-graphite-sheet-pgs/Multilayer-PGS 2. Thermal Storage Sheet (TSS). https://eu.industrial.panasonic.com/products/circuit-thermal-protection/thermal-protection/pyrolytic-graphite-sheet-pgs/thermal-storage-sheet# ML-PGS is basicly few layers of 0.025mm PGS with 0.05mm adhesive layer. This wasn't discovered before, but it's perform interestingly since it decrease hot spot size. After some time use PGS in laptop i no longer sure that PGS should be "cut to die size". Heat from small die is concentrated on same small area. I think bigger PGS with adhesive layer which stick to heatsink may perform better (spread heat from small die to heatsink). ML-PGS is good opportunity to test this idea.

Just for the record... i upgraded my CPU to i7-3820QM and runned same test like before - AIDA64 CPU Stress Test for 10 minutes. Here is results: Room temp is 29°C (HOT!) I didn't apply Kapton Tape around CPU die this time. In Cinebench R15 i got 542 points, which is lower then i got with i7-3630QM (563) with ThrottleStop, but with i7-3820QM (and no Throttlestop) i hit almost 45W (CPU Package Power) and there was throttle. Maybe because dGPU hit 58°C... Can i exceed TDP limit? Update: Yeah i can... another run. Max Power: 45.5W dGPU was 66°C.

One more thought to lower temp on CPU is to focus on specific die zone. Here is die scheme cores image i founded in web: Ivy Bridge die. In my Lenovo Y500 iGPU (processor graphics) is disabled, so i need more pressure at cores zone. I don't know from what's side to count cores (left to right or right to left) to locate my hottest core (core #1). And here is picture of CPU itself with die scheme (not made by me).

Incorrect! EYG - Pyrolytic Graphite Sheet (PGS Only); EYGA - Taping PGS (with adhesive or Insulation + adhesive or high heat resistance); EYGE - Thermally conductive elastomer (semi-sealing material - SSM); EYGS - soft-PGS (Compressible Type) later re-named in "GraphiteTIM"; EYGT - Graphite-PAD. For our use best fit only EYG (PGS Only) and EYGS (soft-PGS\GraphiteTIM), since they offer best (of all products) thermal conductivity. Unfortuanly, it can't fill micro gaps like usual thermal gease do it. I think problem with PGS\soft-PGS density. It have good thermal conductivity (26 W/m*K for 100um PGS-Only), but low density (0,85g/cm3). If we look Thermal Grizzly Kryonaut there is 12,5 W/m*K and 3,7 g/cm3 OR Conductonaut - 73 W/m*K and 6,24 g/cm3. If you look results from Carbonaut review - there interesting thing: less density - higher temp (mostly). So, we need something that have both: very high thermal conductivity (W/m*K) and very high density (g/cm3). First will help to fast trasfer heat, another - fill smallest micro gaps between CPU\GPU die and heatsink. I even starting to think, that *CNT array is a bit extreme variant, need to step back. Close to current best solutions - LMs, maybe start from adding liquid *CNT in them to boost thermal conductivity. I saw interesting Foxconn patent where they combined LM and *CNT Array. LM was filled inside array to fill all gaps, but it requied special lab equipment and methods (PVD or CVD). Regular user can't do it by himself, also he can't buy it - very expensive and will be made for order with waiting period. That way solution may sound crazy - custom made LM (DIY Liquid Metal) + test addition or complete new formula... maybe same Galistan (In + Ga +Sn), but made of nano particles to boost density.

This is how package look's like. Bag and flat box inside with instruction and certificate of origin. Finally some actual tests of Thermal Grizzly Carbonaut: Source: https://vonguru.fr/2019/05/13/test-carbonaut-thermal-grizzly-flop-ou-bonne-surprise-partie-1/ Very similar to IC Graphite. So, look's like our next stop will be... CNT array - yes, carbon nanotubes arrays. Single walls, Multi walls (whatever). I saw their price... not in next 10 years ($450-900 for small piece). VERY expensive.

Here is another test, but with small mod to cooling system of my laptop (i called it Kapton Mod): Heatpipe of GPU go throught CPU heatsink base and there is possible chance that GPU heatpipe MAY transfer heat to CPU heatsink base OR CPU heatpipe. So i isolated (wrapped) GPU heatpipe in one layer of Kapton tape to block any heat transfer at this area. Conditions: 1. Cold start of laptop; 2. Running ThrottleStop with BD Prochot enabled and set multiplier x32 on all cores (if GPU temp will be more then 52°C - CPU will NOT throttle); 3. Running O.C.C.T Test; 4. Room temp.: 27°C. Max core #1 temp was 97°C (so, lower on 2°C AND higher frequency on 200 MHz). Lowest core temp was 89°C (core #0 and core #3). Again, lower on 1°C AND higher frequency on 200 MHz. GPU temp also lower on 2°C. I runned test when i did mod (few days ago), but got Error on CPU #1. Only today's attempt everything was fine, i don't know why.

O.C.C.T save results in nice graphs, so i decided to share my results too. Core #1 is hottest of all cores. Test duration was 5 minutes with 1 minute of idle period at beggining. I set in settings max cores temp 100°C. Also, this test was run with ThrottleStop with enabled BD Prochot option. It will not allow hot GPU (more then 52°C) to force CPU throttle and fixed CPU frequency at 3.2 GHz for all cores. Obviously such frequency is too much. If i lower CPU frequency on 200 MHz then i may complete 5 minute test. Temp of this core is high anyway - 99°C. Lowest core temp was on core #0 - 90°C. It's maybe silicone lottery OR bad contact with soft-PGS.

Yes. And more then that - we have proofs!

If Thermal Grizzly do it (rebrand Panasonic Soft-PGS) - if will be simple scam - soft-PGS - is Graphite and Carbonaut is Carbon. This diffirence is a key. Back to topic: I have idea\question, but it's really difficult for me to fully understand such technology. What if we try to use X-Y heat transfer of soft-PGS(or standard PGS) instead of Z heat transfer? How? Usual use of this "pads\sheets" is just like any regular TIM - place it on IHS, add heatsink on top. Heat from bottom (IHS) go through TIM and finally to heatsink. What if we place "special" soft-PGS\PGS with shape like "L" (technology allow to manufacture it such way)? Bottom part of "L" contact with IHS and transfer heat to another side (not outer side, where heatsink is, but inner side) and on that side add heatsink. Probably still need to add regulat heatsink and fan, just to don't let CPU to fry. And yeah, that way cooling system will have two heatsinks and two CPU fans. Something similar to Prolimatech Genesis. I can make a picture to better explanation of idea.

In case you didn't saw yet. Science Studio got Carbonaut for review. I will wait a review from der8auer anyway. I hope he will test not only on 95W i9-9900K, but also on something like 165W SkyLake X i9.

Here it is! https://www.caseking.de/en/search?sSearch=Thermal+Grizzly+Carbonaut Not in stock now, but there are some pictures of a box and material itself. 25 x 25 mm (GeForce RTX 2080) — €9,90; 32 x 32 mm (Intel LGA115x) — €9,90; 31 x 25 mm (GeForce RTX 2080 Ti) — €10,90; 38 x 38 mm (Intel LGA20xx и AMD AM4) — €12,90; 51 x 68 mm (AMD TR4) — €25,90.

Back to topic. I talk with Thermal Grizzly Support about Carbonaut

New Graphite sheet is discovered. It's made by Hitachi. It was used in new AMD Radeon VII (TIM model is TC-HM03). Here is it's data sheets: http://www.hitachi-chem.co.jp/english/products/cc/026.html

Yes, this is very interesting product according to spec - 60+ W/(m*K). I am sure it also depends of compression, but specs is so high... Hope there will be small size - just to cover CPU die. I don't want to buy a carpet.