Another Graphite Thermal Pad Test: Panasonic PGS

[Saccaed]

Up to you to go for fancy electrical tape. The voltage is so low that you just need a barrier; I recommend just getting some wide cut 3m electrical tape. As for some torque to compression conversion, I think you are out of luck. You can probably get an estimate of torque on the screw head by using the threads per distance(threads per inch/mm) but simple pressure on the screw heads isn't going to directly represent plate to die pressure. At least plate to die kpa shouldn't exceed screw head to plate pressure. I would first worry about crushing the die or board with too much torque. Make sure you have a good backing plate setup to distribute the load behind the die, otherwise the board will likely warp/ crush.

[T[]RK]

On 11/13/2018 at 4:09 PM, James Evens said:

Installed the EYG-T3535A05A silicon pad today:

Where are you using it? What's device?

[Saccaed]

Noting again that mounting pressure is a big deal for PGS. Due to my laptop's assembly, I discovered I could squish a couple earplugs between the processor heat sync and case to apply more pressure to the CPU cooler plate. Prime 95 control tests dropped 4C from 77C to 73C. That's on top of the spring shimming I've also done which now makes for an 8C difference in core temps just on mounting pressure alone.

[T[]RK]

I also received my PGS, but not yet tested it. I still need to find material to cover zone around CPU die (i will get cheap one) and i have plans to install new CPU cooler (original), but mate who must help me with ordering it from china - still didn't do it. I may just run old CPU cooler at full speed by removing PWM wire. It will give consistant RPM for all time instead of dynamic for testing. A bit loud, but it's just for testing.

 

Also, when i saw on photo how thick PGS and read in data sheet that it's 0.2mm (200um) - i hardly believe in it (you know... maybe photo was aat macro mode). When i actually saw it with my eyes - it's actually have noticable thickness. If place it on die and add heatsink on it - there will be no gaps. It just be compressed. Problem is - it have to be compressed a lot (for laptop design).

 

But maybe it's just me.

 

4 hours ago, Saccaed said:

Prime 95 control tests dropped 4C from 77C to 73C. That's on top of the spring shimming I've also done which now makes for an 8C difference in core temps just on mounting pressure alone.

Temp per cores also the same (small delta)?

[Saccaed]

Per core temp spread has remained unchanged at ~2C. All cores fluctuating within 2C throughout synthetic loads. On material thickness, unused Soft PGS material will squish significantly. I've not put calipers to it yet, but the compressed area can be felt with the finger when handling. I don't think you will have to worry much about not having enough mounting pressure. Everything I've tested has performed a few degrees off good thermal paste under mounting pressures meant for paste. For applications where the mounting pressure can be increased, the Soft PGS has matched and improved on good paste.

[T[]RK]

O.k., i installed GraphiteTIM in my Laptop and here is fast results:

 

AIDA64 - System stability test (10 minutes)

 

Coldest core was Core #1 (in AIDA64, in HWiNFO64 it's Core #0). Always ~5-7°C lower then others (don't know why). It always was like that even with thermal paste.

 

CPU & GPU with removed thermal paste and die cleaned with alcohol:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Added layer of Kapton on whole zone around die (not perfect, as you may see air bubbles):

 

 

 

 

 

 

 

 

 

 

 

 

 

 

This is how GraphiteTIM was planned to cut:

This is EYGS-0404ZLMP (38x36mm) with hole in center. Other area will be used for another test-projects (i have got old GPU - Radeon X700).

 

Everything is installed:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Also, i removed PWM wire from CPU cooler plug, so it always worked on maximum speed.

 

So, what's short conclusion? Well... it's working.  System started, loaded winows, played few rounds in CS:GO and runned AIDA64 test after that. No re-boots, BSODs... nothing. Without load cores in zone around 50°C.

[Saccaed]

Too bad you still aren't maintaining maximum turbo. How much change before and after?

[TechyBen]

Nice... but can I build a * heatsink * out of this stuff (I really do want to try in my super small ITX build XD ).

[T[]RK]

1 hour ago, Saccaed said:

Too bad you still aren't maintaining maximum turbo.

It's strange about this CPU. I even installed Intel Turbo Boost Monitor, and sometime i really see 3.3/3.4 GHz, but just few seconds and mostly when i returning to PC and it's switching from 3.4 to 3.2 GHz. When i looking videos on YouTube i see stable 3.2 GHz. Don't know why, maybe limitation of motherboard. I may guess, that 3.4 GHz only when 1 core is working.

 

1 hour ago, Saccaed said:

How much change before and after?

Don't know, really. Without GraphiteTIM i got very high temp on CPU and GPU in CS:GO in Danger Zone. About 90°C on CPU and 80°C on GPU. I think because of thermal paste degradation.

 

Also, CPU clock was 2.4 GHz in game, but i think it was because multi-core option in-game.

[T[]RK]

1 hour ago, TechyBen said:

Nice... but can I build a * heatsink * out of this stuff.

GraphiteTIM is more like optimization of heat transfer. Yeah... something like this. If one side of "thing" heating more then other - it will help to transfer heat from that side and balance overall heat. Also, it may help to transfer heat more effectively just because it's graphite. Originally it was made to use with heat sink, but not instead of it. Also, it's possible to transfer heat really long away from heat source.

 

I got some ideas about transfer heat from one source to two heatsinks with two coolers. But then i founded second low of thermodynamic.  

[TechyBen]

13 hours ago, T[]RK said:

GraphiteTIM is more like optimization of heat transfer. Yeah... something like this. If one side of "thing" heating more then other - it will help to transfer heat from that side and balance overall heat. Also, it may help to transfer heat more effectively just because it's graphite. Originally it was made to use with heat sink, but not instead of it. Also, it's possible to transfer heat really long away from heat source.

 

I got some ideas about transfer heat from one source to two heatsinks with two coolers. But then i founded second low of thermodynamic.  

Yeah, it's about what I can do with a tiny mini itx build... so this stuff looks really interesting.

[Saccaed]

Thought it worth mentioning that with the earplug wedge my CPU temps dropped but GPU went up. Turns out that the improved conductance in turn allowed for the GPU to clock higher. After observing how the clocks drop to maintain a temp target, I still have room to improve on the GPU.

[Saccaed]

Added another earplug of thickness to the wedge and got further improvements. Top to bottom: GPU Clock, CPU and GPU temp, CPU and GPU load, CPU wattage; 3 hour timescale.

Note the bump in GPU clock afterwards. Temps dropped again a couple C for cpu and gpu. Probably at the limit of the wedging as I'm having to use a 25lb weight to help with reassembly. The method is sound though. Every additional earplug of thickness has yielded improved thermals.

[T[]RK]

I downloaded Prime95, which test you are running? And for how long?

 

Also, i saw "shimming" mod my laptop heatsink springs, so i may try it too. But it will be better to test without it first.

[Saccaed]

Small FFT. I suggest restricting your turbo through the power options in windows. If you set "maximum processor state" to 99% turbo will be disabled and the CPU will stop at it's stock frequency. Good for A B tests and limiting the amount of heat. I run as long as needed to reach a stable temp(normally ~5 minutes).

[T[]RK]

O.K., firs test:

 

Prime95 v29.5. Tortue Test: Small FFT (+maximum processor state - 99%). Test duration ~10 minutes. (actually two runs, but at first i lost picture from Print Screen buffer).

 

CPU Clock at all cores was at stock ~2.4 GHz. Max core t° under load was 80°C (core #1), min core t° was 70°C (core #0). Also, you may see maximum temp of core #1 = 90°C. It jump to this t° when i hit start button, but then dropped to lower t°. Don't know why, but it was just a second. Also, you may see that HWiNFO64 was runned 2 hours and 23 minutes. I started it right after Windows was loaded. So, it's monitoring everything since that time.

[Saccaed]

I suggest you find a log viewer for HWINFO or use Open Hardware Monitor to graph the CPU clocks. You HWINFO max observed clock is unchanged from the previous post you made. If you can view the frequency in graph form you may see the CPU turbo at the start of the test(could explain the high initial temp). Setting Maximum Processor State to 99% should prevent CPU turbo, but a graph of clocks will tell if that's the case. Graph'ed CPU clock and temp will be more useful than simple values as you will see a temp curve and any unexpected CPU behavior.

[T[]RK]

Just did tiny test with ThrottleStop v8.70 (CPU multiplier x32 on all cores) - and Prime95 v29.5.

 

Without proper cooling and LM - VERY BAD idea. Temp. goes sky high VERY FAST - i barely stopped test when temp was 99°C on two cores. It was just few seconds from test start.

 

Good news is - i got washers for spring mod for CPU heatsink. I think i will try it tomorrow.

 

P.S. In last der8auer video i ask him about Panasonic TIM (PGS\Soft-PGS) and opinion about this product\test. He told me that he tested it, but wasn't happy about it.

[Saccaed]

Looked at his reply. Unsurprising that he would not like it compared to liquid metal. Liquid metal is better as a conductor but in something like a laptop is unwise. Considering the video you replied to and that you didn't mention the laptop application I'm not surprised that the liquid metal comparison was made. I replied to his reply specifying the laptop application, maybe he'll reply. A more apt comparison would be to ICY Diamond or some other high end grease.

 

A tip for the washer shims, when tightening pay attention to the spring coils. Stock configuration the spring coils won't bind up when at full tightness. The goal with the washers is to get the spring coils to just barely touch each other with the shims in and the cooler mounted. Once the springs bind, they start acting like solid spacers and pressure will increase much quicker than increased spring compression.

[T[]RK]

Here is same test after i did Spring Mod:

 

Springs #1,2,3,4,5 - x1 shimming ring (DIN125, M3, 0.5mm thick);

Spring #6 - x2 shimming rings (DIN125, M3, 0.5mm thick).

 

 

I set processor in 99% but still... few seconds (around 10 maybe?) it's worked at 3.2 GHz and temp was around 80°C, but then frequency dropped to usual 2.4 GHz and temp also go lower. In general temp go lower after mod. Nice thing, that after i removed cooling system and before i installed it back - i don't need to worry about thermal paste. There is no thermal paste! =))

 

Also, i suspect that it may happen because i enabled (long time ago) and installed all drivers for Intel DPTF. It's do something with Turbo Boost i think.

 

For next test i added ThrottleStop v8.70 with multiplier x26 (2600 MHz on all cores).

[T[]RK]

Here is small test in a game - Counter-Strike: Global Offensive (Danger Zone game mode).

 

I finally was able to configure NVIDIA Notebook GeForce Display Driver and GeForce Experience to work properly on unsupported GPU. Also i added MSI Afterburner with RivaTuner Statistic Server. Everything working together. So, that mean that i can now not only run soft aand take screenshots, but also record games with Statistic Server and Capture Desktop.

 

Playing time about few hours, so CPU and GPU was hot, but stable. All game settings on low, resolution is 720p (i suck at 1080p in CS:GO with GT 650M, but don't really want to connect my GTX 1080 Ti).

[Saccaed]

Adding Thermal Grizzly Carbonaut to the "to be tested" list.

 

 

[T[]RK]

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.

[Saccaed]

Funny, I want the carpet option so I can use it for all sorts of projects. Throughout the video they mentioned pre cut sizes for different cpu / gpu applications. Worst case I imagine you get a bit more than you want and have to cut it down. I hope that a thinner version can be found. As mentioned in the video, the W/(m*K) isn't so significant when material thickness is accounted for. Z aligned thermal transfer is though. The graphite pads have lower z transfer than they do xy which could make the difference between the two products quite significant.

[Saccaed]

Just tested another product Panasonic makes, NASBIS. The product is neat, but my initial application idea yielded repeatable but insignificant results. I have some leftover and am open to suggestions on alternative applications.

 

 

The idea for the setup was to interrupt any convection from die to cold plate back to die as well as experiment to see if a larger sheet might normalize peak temps. The CPU surround ended up not working as the NASBIS measured in at 0.4mm and the die height is 0.3mm which caused contact issues. The NASBIS sheet is plenty flexible, resists pressure and tearing, and will not stand up to any flame. Officially the sheet I ordered was advertised at 0.5mm and is backed both sides; the backing comes off easily if the sealed edges are cut. Seeing that the CPU application was not possible I tested using Furmark to load the GPU. GPU temp remained the same(expected) but a repeatable clock difference was observed; averaged 1542.5 with NASBIS, 1540.7 without; 1582.0 MHz peak w/ NASBIS 1556.5 MHz without. Results were obtained by running furmark for 25 minutes and gathering data on the last 5 minutes. The material is neat, but for this application not useful. I only used ~half the sheet, if anyone has other application ideas let me know.