Another Graphite Thermal Pad Test: Panasonic PGS

[Saccaed]

TLDR; IC Graphite is Panasonic Soft PGS. Both perform identically. Soft PGS can be bought in large sheets and runs ~10% the cost of IC Graphite. Soft PGS(IC Grapite) performs best under the most clamping force as can be safely used for any application(for my laptop, increased clamp force resulted in 4c cooler temps). Cut to fit the clamped surface(extra margin does not perform better). Soft PGS does amazingly well at normalizing temps between cores(2c delta for normal clocks, 4c for extreme overclocking). The material is completely reusable and is great for experimenting; so far I've pulled and replaced the heat sink ~20+ times without any degradation. Do insulate any immediately nearby components as PGS is very electrically conductive.

 

Inspired by the LTT IC Graphite Thermal Pad video I've been researching graphite and graphene products with the goal of overclocking my laptop even more. Panasonic PGS looked interesting due to it's price and claimed properties. I ordered the most conductive sheet listed for immediate order, 0.025mm EYG-S121803, along with a sheet of 0.5mm Arctic MX-2 silicone thermal pad. I ordered the Arctic thermal pad in anticipation of emulating some example use of the PGS material (TUfast racing using the Pyrolytic Graphite Sheet).

 

Before making changes I ran some heating and cooling tests using Prime95 set to run "Small FFTs" mode with 8 threads. I also set the laptop fans to max.

 

The laptop is an EVGA SC17 1070 and for all tests the clock and voltage were kept the same. This first set of tests show performance using IC Graphite in place of the shipped thermal paste(good stuff). The tests I ran to compare IC Graphite to the shipped paste we're not valid for side by side comparison, but a long term test confirmed IC Graphite to perform within 1-2C of the shipped thermal grease.

 

A preface is needed before going into the modifications. Through overclocking and stress testing of the CPU and GPU I found the CPU to be capped by heat and the GPU by power delivery. The GPU peaks roughly 20C below the thermal throttle threshold which began me wondering if I could find a way to wick more heat from the CPU over to the GPU. The cooling solution for the laptop already has a crossover heat pipe that runs from the CPU die, over the GPU die, that passes through both GPU heat syncs. With that in mind the modifications went as follows...

 

Test 1 involved folding up the PGS sheet I received and spanning it between the CPU and GPU dies. After watching a showcase video put out by Digikey I wondered if the sheet alone could improve thermals and found out that it can't. I didn't bother running a more in depth test after discovering that the CPU couldn't handle the previous overclock and would only stabilize below 100C around a 31x multiplier.

Above is how I mounted the folded PGS sheet. The results were lackluster.

 

Test 2 involved skinning one side of Arctic thermal pad with the PGS sheet, then coating the backside of the laptop cooling solution in hopes that cooling performance would improve. This idea came from seeing existing variations of PGS products that already came configured with silicone thermal pad backed with PGS sheet and a short clip from a video demonstrating a real world application of PGS in electric race cars.

This configuration seemed to yield measurable results. So I repeated exactly the tests I performed before attempting modification(prime95, 8 threads, Small FFT, fans set to max).

To consider changes in ambient temperature between the before and after tests, I calculated the ambient temperature for both tests looking up a temperature chart for the day. Testing before PGS application happened at 11am and the calculated ambient is 15.5C. Testing after PGS application happened at 8pm and the calculated ambient temperature is 14.8C. Between tests a calculated 0.7C delta.

 

Analyzing data between the tests, what sticks out for me is the 3C drop in temperature across the CPU and GPU while testing the PGS application. The ambient temperature of the PGS test was 0.7C lower, which to me does not account for all of the observed cooling.

Testing does seem bear out that the PGS + Arctic application did manage to normalize temperature between the CPU and GPU 0.9C; observable, but not dramatic or conclusive. What I failed to test was the Arctic thermal pad without applying the PGS to the back surface. The PGS material is quite fragile and I don't want to attempt separating it from the Arctic thermal pad now that they are stuck together. As a result, I can't know if the arctic pad or PGS played more of a role in the results.

 

Suggestions for improvement are welcome. The rule for this project is no materials that need regular replacing to maintain performance(no thermal grease).

 

Hopefully my testing will be useful for someone.

 

UPDATE 2018 06 13
After looking into Panasonic Soft PGS I found a graph comparing thermal conductivity and mounting pressure that interested me.

 

 

Seeing that my heat syncs are held firm with sprung screws, I wondered if I could shim the springs with washers to increase mounting pressure. I ended up testing 2 configurations each adding 1 washer as a shim. The first test was great with 1 washer shim, the second test was a complete failure as I could not maintain a cool enough temp to even test(I figure that the heat sync began to warp and the contact surface shrink).

 

 

I performed the same Prime95 tests with the same clocks. Absolute CPU temperature dropped from 91C to 87C (4C delta). CPU GPU temperature delta dropped from 24.1C to 17.7C (6.4C delta) while under load. CPU GPU temperature delta normalized back to 19.5C while cooling after load (before PGS 19.7C measured, after PGS 20.1C, after washer shims 19.5C measured).

 

Update 2018 09 02

First, I changed how I built up material around the CPU die by using wider tape then cutting out the die area for fitment; before the tape was built up using many pieces of smaller tape(I didn't have wide tape at the time). I changed the tape application after noticing the uneven imprint left behind on the old piece of Soft PGS. The new application yielded improvement over the old application at the time of testing(4C) but when I compared to the first tests with the previous application there only was a 1C improvement. This has me wondering if there is some change over time that the Soft PGS material is experiencing. I'm fairly confident in the consistency of mounting as I tested many configurations over the course of the day and most yielded no improvement and an exact repeat of the previous tests. Don't know what to make of today's results, going to test again in a couple months with another new piece of Soft PGS to see if I can repeat the observations.

I also tested stacking two layers of the Soft PGS material as a goof. The results were bad, not recommended.

 

Update 2018 09 06

Found improvement by trimming the Soft PGS to fit the die and eliminating the buildup that maximized contact between a larger PGS sheet and the cold plate. I still used a single layer of tape for insulation.

Prime 95 temps dropped from 87C to 84C. My gaming overclock temperature changes were dramatic in comparison, 96C to 81C before and after previous testing.

[Amaranth]

The IC Graphite pad is just a repackaging of Panasonic's Soft PGS that is pre-cut - if you looks at them side-by-side it's obvious that they're the same product (something that is further proved by their performing the same and having the same technical specifications).

[Saccaed]

I wondered if that was the case. Weird that I never ended up finding soft PGS while doing all my research on PGS material. Thanks for the heads up.

[Amaranth]

29 minutes ago, Saccaed said:

Thanks for the heads up.

No problem, if you're interested the part number of that sheet is EYG-S0918ZLX2.

 

 

[Saccaed]

You are a legend. This is exactly what I'm looking for for a janky hard drive project.

[Saccaed]

Updated the initial post. Increased mounting pressure of the heat sync by shimming the sprung screws with washers. Netted 4C cooling improvement under load. A second test involving another washer shim resulted poorly; likely due to heat sync warping which shrunk the contact patch size.

 

Test performed after viewing a mounting pressure vs thermal conductivity graph on Panasonic Soft PGS.

 

 

If I can come up with a way to get more mounting pressure without warping the contact patch of the heat sync I'll see what even more mounting pressure will do for temps. I'm interested in getting another data point to try to figure out where on that graph the laptop likely places: lots more room for mounting pressure improvement, or little room for mounting pressure improvement.

[Hyperactive]

I was in doubt of buying de ic graphite sheet when I found your topic. You did amazing work man! I can buy the same product of the Panasonic sheet as you mentioned. But I can also buy 1950W/mk 10 um thick for less money. Will this be better or worse than your 25 um thick sheet? 

 

I don't know wether I should buy the graphite sheet you listed (which worked the same as the ic graphite) or buy the thinner and better conducting Panasonic graphite sheet. 

 

Here is the link to the product (in dutch but should be fine to understand):

http://nl.farnell.com/panasonic-electronic-components/eyga091201m/thermal-interface-film-pgs-10um/dp/2577414

 

Thanks in advance for your answer.

[PIRATA!]

 

On 6/2/2018 at 5:10 AM, Amaranth said:

No problem, if you're interested the part number of that sheet is EYG-S0918ZLX2.

 

Hi to everyone.

So the IC Graphite is indeed from Panasonic? Is this the exact reference number??

Are there any other Panasonic graphite versions that are better?? ...and so that are better than IC Graphite??

 

Thank you

[Saccaed]

On 8/9/2018 at 9:13 AM, Hyperactive said:

I was in doubt of buying de ic graphite sheet when I found your topic. You did amazing work man! I can buy the same product of the Panasonic sheet as you mentioned. But I can also buy 1950W/mk 10 um thick for less money. Will this be better or worse than your 25 um thick sheet? 

 

I don't know wether I should buy the graphite sheet you listed (which worked the same as the ic graphite) or buy the thinner and better conducting Panasonic graphite sheet. 

 

Here is the link to the product (in dutch but should be fine to understand):

http://nl.farnell.com/panasonic-electronic-components/eyga091201m/thermal-interface-film-pgs-10um/dp/2577414

 

Thanks in advance for your answer.

I'm not sure that thinner will be better. The absolute conductance may be higher, but the amount of heat that can be moved might be less due to the thinness of the material. I did try to source the thinner material for my test but did not find a US supplier at the time. At the very least, in conjunction with a thin thermal pad material the PGS sheet will improve conductance. Sorry for the slow reply. I'm not sure why I didn't see a notification when you posted.

[Saccaed]

On 8/11/2018 at 10:25 AM, PIRATA! said:

 

Hi to everyone.

So the IC Graphite is indeed from Panasonic? Is this the exact reference number??

Are there any other Panasonic graphite versions that are better?? ...and so that are better than IC Graphite??

 

Thank you

Yes, IC Graphite is from Panasonic; Panasonic PGS. The material comes in two distinct forms, one "soft" the other not? The soft material is for use in clamping situations(ie. between a chip and heat sync). The soft material does not seem to conduct as well as the other form, but the other form is useless in clamping situations. The hard? PGS material conducts best and appears to be great at improving the conductance in an area when adhered in some form. In phones and tablets PGS like materials are often used as a heat spreading and routing material.

 

Compared to thermal grease, the soft Panasonic PGS material benefits more from increased clamping force. Finding a cooler that is made for high clamping force without bending the motherboard will pay big dividends. Aside from using a delidded chip, the clamping force will result in the highest performance.

 

EYG-S0918ZLX2 is the reference number for the material identical to IC graphite. I have not explored the other soft pgs materials. From the limited reference materials released by Panasonic and on other forums, I have found soft PGS materials that advertise higher conductance that are thinner, but also have found that apparently there will be a point where material thickness will result in less conductance even though advertised with higher conductance. Others have speculated that there likely will be a butter zone from a given chip and cooler configuration and that the pairing likely has to do with how well the PGS material fills the tiny voids between the cooler and chip. If you are not pressed for cash you might buy a couple types and test.

 

Sorry for slow reply.

[Saccaed]

Did some more tests today and came up with odd results. I managed to gain 4C cooling improvement compared to the day start, but upon comparing to the last collected data the results are almost the same(1C better cooling today). I used a new piece of Panasonic Soft PGS and changed the method of building up around the CPU die. Due to the weird results I'm wondering if the Soft PGS material changes over time.

As can be seen in the image the old tape method left an odd imprint pattern. The pattern got me thinking that maybe mounting pressure evenness could be improved by switching to wider tape could cover the whole die. Thoughts as to why I observed a big improvement in the day results but not when comparing to the previous best results are welcome. First post updated to include today's results.

[Ross Siggers]

A very interesting thread, looking at the graph you posted for the Panasonic PGS, do you know if the IC graphite has a similar reaction when it comes to mounting pressure? I purchased some for my laptop(Lenovo X230), replacing both the graphics chip(thermal pad) and cpu(paste).

But the improvement in temperatures was pretty underwhelming, considering the machine is 6 years old and had never been repasted or cleaned. I consistently dropped around 4-5 degrees, but considering the previous peak was hitting the low 90's, the percentage isn't huge. 

The consolation here is the fact that I was running simultaneous cpu/gpu stress test. I use an eGPU, so it'll never see real-world temperatures that high. Not that it was designed for gaming anyway 

[Saccaed]

Yes IC Graphite performs identically to Panasonic Soft PGS. They are the same material. I can confirm the perform identically as I used both IC Graphite and Soft PGS without any observable change. If your laptop has sprung mounting screws I recommend experimenting with shims to increase the mounting pressure. Likely you will only be able to shim a little before the heat sink warps and performance is effected, but before then I wouldn't be surprised to see ~3-5C cooler peak temps(as long as your laptop doesn't have some heating scheme that allows the temp to reach 90C before full fan spin) and tighter temps across cores.

[Ross Siggers]

@Saccaed Thanks, it does indeed have sprung cooler mounts so I may try that next time it's apart. 

And the idea of a fan curve that isn't at 100% @ 90 degrees is literally nightmare fuel  She's definitely at maxed out at that point, there are two heat pipes that run parallel, and the cpu/gpu are mounted in series.

[Saccaed]

Found some more performance while fiddling. I tested without buildup around the die that maximized contact between PGS and cold plate. I still used a single sheet of tape for insulation just in case. The Soft PGS material was cut to die size.

Prime 95 temps dropped 3C. The overclock I use for gaming showed a 15C drop which blew me away. While gaming I run at 4ghz instead of 3.6ghz I use for prime 95 tests.

Die to cold plate as only contact is by far the best improvement. Previous improvements have directed translated from prime 95 testing to gaming overclock(1 degree less testing prime 95, 1 degree less gaming overclock). This improvement yielded 12C improvement difference between gaming overclock and prime 95 testing.

[T[]RK]

Hello!

 

I also founded sometime ago information about PGS and finally founded some tests!

 

Here is some important links for topic and also materials (PGS and Soft-PGS):

 

Panasonic Industrial official site

https://industrial.panasonic.com/ww/products/thermal-solutions/graphite-sheet-pgs/pgs

 

You may find there data\specs of diffirent materials.

 

On 9/7/2018 at 12:44 PM, Saccaed said:

I tested without buildup around the die that maximized contact between PGS and cold plate. I still used a single sheet of tape for insulation just in case. The Soft PGS material was cut to die size.

What's exact Part No. of Soft-PGS did you used for last test? I looking forward to test it on my laptop - Lenovo Y500.

 

Also, what tape did you used to cover PCB around Die? It's something special (non conductive\heat resistant)?

 

Here is catalog: http://industrial.panasonic.com/cdbs/www-data/pdf/AYA0000/AYA0000C50.pdf

 

NOTICE: 2018-10-02"Soft-PGS" have been renamed to "GraphiteTIM".

[Saccaed]

5 hours ago, T[]RK said:

What's exact Part No. of Soft-PGS did you used for last test? I looking forward to test it on my laptop - Lenovo Y500.

 

Also, what tape did you used to cover PCB around Die? It's something special (non conductive\heat resistant)?

I used EYG-S0918ZLX2. If your feeling adventurous try a slightly thinner Soft PGS. Aside from mounting pressure, thinness will yield better conductivity to a point.

For tape I used blue 3M painters tape(it's what I had on hand that was wide enough to fit over the dies without the need for multiple strips). It's only purpose is to protect the surrounding die components from contact with the sheet(which is highly electrically conductive).

 

[T[]RK]

19 hours ago, Saccaed said:

If your feeling adventurous try a slightly thinner Soft PGS.

Um... no.  I think it's best to test current material and confirm result.

 

Also, i did small comparison of you CPU (Intel Core i7-i7-6820HK) and my (Intel Core i7-3630QM) and diffirence are only:

 

1. Lithography (14 vs 22 nm);

2. Frequency for both: Base and Max Turbo (2.7@3.6GHz vs 2.4@3.4GHz);

3. Cache (8 vs 6 Mb);

4. Bus Speed (8 vs 5 GT/s);

5. Memory Types;

6. Max Memory Bandwidth;

7. TJUNCTION (100°C vs 105°C);

8. Package Size (42x28mm vs 37.5x37.5mm);

9. Die Size (122mm2 vs 160 mm2).

 

But TDP is the same - 45W!

 

So, my die is bigger and with bigger lithography and have got lower frequency, but same TDP. Even with 90x90 pad i should be safe and one pad should be enought. I may even test on dGPU first (GeForce GT 650M), since it's "cold" at start (~50°C). Then run benchmark and look what will happen.

 

19 hours ago, Saccaed said:

I used blue 3M painters tape. It's only purpose is to protect the surrounding die components from contact with the sheet(which is highly electrically conductive).

Yeah it was for a reason, but i would lik to get something proper. Something like this:

 

3M Non-Conductive Heat Activated Cover Tape 2672

https://www.3m.com/3M/en_US/company-us/all-3m-products/~/3M-Non-Conductive-Heat-Activated-Cover-Tape-2672/?N=5002385+3290270507&rt=rud

 

Just need to find PCB size and buy proper size of tape.

[Saccaed]

Funny you should bring up the i7-3630QM, that's whats in my other laptop(HP DV6) and I've already replaced the paste with the same Soft PGS material. Sounds like our laptops are very similar with the difference maybe being that mine has and integrated 650m and yours could have a dedicated 650m. For my i7-3630QM laptop the PGS material has worked well. I improved temps, but the replaced paste was only Arctic Silver Alumina which is hardly a high end paste. When I get some tinker time I plan to shim the springs to see if temps drop further like happened with my EVGA laptop.

 

On wattage, your numbers are right for the stock speeds. My laptop has a full desktop bios and overclocking support which I've been able to push at peak to 79 watts(34 over stock rating, and 16 over the as shipped overclock peak) during synthetic tests(prime95). What's been great is that the switch to Soft PGS brought the core to core temps much closer, and that has largely been what has allowed me to bump my multiplier from x37 to x39. The laptop as it was shipped could have core to core temps vary ~7C, after all the work with Soft PGS and mounting pressure I brought the core to core temps to within ~2C.

[jpmomo]

I am going to be running a few tests comparing the different panasonic pgs graphite thermal sheets with the IC graphite thermal pads vs thermal paste.  one thing that i noticed is that the IC thermal pads state 35(W/m - K) vs 20 for the soft pgs that you referenced.  I am curious how these 2 could be the same part with such different thermal properties.  I am assuming that these are all referring to the Z axis (cpu through the thermal pad up to the heatsink).  I am also not sure how much impact the X-Y axis thermal properties have on keeping the cpu cool.  I will be doing most of these tests on a server cpu (amd epyc 7551 and intel xeon 8160) with an air cooled copper heatsink.  Part of the reason for looking for something other than the IC units is the size of these CPUs.  they are both very large.  the amd is the size of the threadripper and the intel is almost as big.  the 40x40 size of the IC units may not be big enough.  I have tried these IC units and they seem to work ok but I wanted to use a larger size pad to ensure complete coverage.  I am glad that you have performed these tests as it helps with some data points for my project as well. 

thanks,

jp

[Saccaed]

I'm not sure of the reasons for differences in the stated properties. I think it could be as simple as the values being from different sources(maybe xy vs z, maybe someone had to list all the different versions on digikey and ctrl+c ctrl+v copied wrong numbers). Can confirm that they perform the same. Tests first started with IC Graphite then moved to the Soft PGS with no changes in temps. If you don't mind post results. Interested in if the core to core temps remain as tight for the high core count cpus.

[T[]RK]

Saccaed, i forgot to ask... since you already worked with this material how you cut it to the size of die? What did you used for it (regular scissors, razor blade or maybe scalpel)? What do you recommend (maybe you got bad experience)?

 

How you may describe material? It's soft and easy to tear (like paper, maybe foil) or something strong like office file for paper?

 

On today i founded all dimensions of CPU die, GPU die and PCH die (i don't want to disassemble my laptop to just find it's size) and total size about 40.3x11 (all three in a line). So, EYGS0918ZLX2 (180x90mm) is too big (and also expensive!), but i founded in catalog another one EYGS0606ZLSL which is 58x61.5mm. I think it's better option.

[Saccaed]

Scissors. The material offers zero resistance to cutting and is rather delicate. Had some issues dragging a blade during some cuts that resulted in tears. Scissors have worked flawlessly. It's easy enough to handle just be careful to not apply tension to it in any way. Tension and folding causes tearing, any puncture source will pierce through.

 

As for patterning what to cut, I've found that lightly pressing the Soft PGS against what your planning to cover will transfer a light outline to the PGS. 

[Saccaed]

I've not yet run across anyone using the PGS material for anything aside of cpu and gpu cooling in PC applications. Panasonic showed some partners in EV racing using the adhesive backed pgs for cooling of high amp draw components. In some of the promotional video the adhesive backed PGS was being used to expand the dissipation area in devices like cell phones by wrapping the material around.

[T[]RK]

I still collecting proper materials. I ordered Panasonic GraphiteTIM (EYGS0404ZLMP, 38x36mm), i even drawed it on paper and drawed CPU die and GPU die on it just to be sure i may cut it later and don't spent money on bigger one without a reason.

 

Some EYG-S in Panasonic document are designed especially for IGBT modules and they have strange form or have got too many holes. I have to copy it's part No. and google it for exact photo. I planned to buy EYGS0909ZLX2, but it's too big (90x90mm) when max size i will use ~30.5x11mm, also it's expensive (~$50).

Then i selected EYGS0606ZLSL, but after i founded it's photo - i rejected it. It's cut very weird way.

At the end i stopped on EYGS0404ZLMP which have got best size and i will place CPU and GPU around hole and cut it. Its also cost only ~$9.85.

Another question is - some of characteristics which is important for aplication:

 

1. Thermal Resistance (lower is better);

2. Pressure provided.

 

From document we know that more compressed - better thermal resistance (but not more then 40% or 600kPa). I have a problem with understanding how i may provide specific amount of pressure? Looking on the graph i maybe want to get for example 300 kPa... how to do it? It it possible to convert kPa in screws torque? Screwdriver with torque control may help with this.

 

Also i am looking for material to cover zone around CPU and GPU die. I think Polyimide Film will be best, but original material made by Dupont (called Kapton). Problem with this is - it's damn expensive (~$60 for 3M 5413 with size 3/4" x 36 yd) and i don't need THAT much, it's also in a lot sizes (may cover zone with one layer or buy smaller size and cover with two lines). Maybe generic (copy) will be best - i may find it in local store and very cheap (less then $7!).