Phase Change Pads Thermal interface Materials are available in a range of thicknesses

Phase Change Material Pads — When thinner isn’t better

When select­ing the right Phase Change Mate­r­i­al Pad thick­ness for your appli­ca­tion, thin­ner isn’t always bet­ter

As any ther­mal design engi­neer knows, all oth­er things being equal, Ther­mal Inter­face Mate­ri­als with high­er ther­mal con­duc­tiv­i­ty and low­er ther­mal imped­ance are bet­ter than those with a low­er con­duc­tiv­i­ty and high­er ther­mal imped­ance (Rth). But is the same true when it comes to mate­r­i­al thick­ness?

A Thermal Design Experiment for CPU and GPU

Our engi­neers set out to test a range of Phase Change Mate­ri­als to dis­cov­er when select­ing pad thick­ness if bet­ter ther­mal prop­er­ties and thin­ner bond lines trans­lat­ed to bet­ter ther­mal cool­ing of CPUs and GPUs.

We start­ed with a selec­tion of Phase Change Mate­ri­als and ulti­mate­ly set

Prod­uct
(Unit)
Ther­mal Imped­ance
(°C‑cm²/W)
Ther­mal con­duc­tiv­i­ty
(W/mK)
Den­si­ty
(g/cc)
Test MethodASTM D5470ASTM D5470
PTM50000.0864.42.3
PTM70000.0586.52.7
PTM79000.04682.8
PTM 79500.048.52.8
Pres­sure vs. Ther­mal Inter­face of Phase Change Mate­ri­als up to 280kPa (40psi)

GPU Manufacturers matter

There are many Graph­ic Pro­cess­ing Unit (GPU) man­u­fac­tur­ers, includ­ing AMD, Asus, and Intel, but by far the largest GPU man­u­fac­tur­er is Nvidia with the lion’s share of the GPU mar­ket so it made sense that we would car­ry out our test on the Nvidia GPU.

Phase Change Pad Thicknesses

Phase Change Mate­ri­als can be sup­plied either as sten­cil-print ver­sions which require the ther­mal inter­face mate­r­i­al to be applied to a sten­cil of a set thick­ness and then print­ed onto the sub­strate.

These sten­cil thick­ness­es typ­i­cal­ly range from 0.2mm to 0.3mm and this is the thick­ness that is then applied. An impor­tant note is that Phase Change Mate­ri­als pro­vid­ed in sten­cil-print for­mat need a dry­ing step (either 20 mins @ 80°C or 24hrs at room tem­per­a­ture) in order to have the sol­vent evap­o­rate and dry to leave behind only the phase change mate­r­i­al itself.

Using sten­cil print­ing, it is impor­tant to note that the height reduc­tion (and there­fore the ulti­mate thick­ness) is reduced on aver­age about 20%, so it is impor­tant to take this into con­sid­er­a­tion when deter­min­ing your sten­cil thick­ness.

Anoth­er option is to but “pre-cured and dried” Phase Change Pads which are the exact same mate­r­i­al, but with the sol­vent removed, leav­ing only the Phase Change Pad. In these cas­es, of course, there is no dry­ing step and the ini­tial thick­ness is also the final thick­ness.

Phase Change Pad Candidates

After con­sid­er­ing the options, we set­tled on three prod­ucts: PTM7000, PTM7900 and PTM7950, and three thick­ness­es: 0.2mm, 0.25mm, and 0.3mm.

Result: Thinner isn’t always better

The results we got were sur­pris­ing. What we expect­ed was that we would get the best ther­mal per­for­mance (ie. the most cool­ing & thus the low­est tem­per­a­ture rise) by using the prod­uct with the best ther­mal prop­er­ties (PTM7950) with the thinnest pos­si­ble bond line (0.2mm).

The num­bers below were not exact but rep­re­sen­ta­tive of the data we observed to indi­cate the trend in tem­per­a­ture change

Prod­uctThick­ness (mm)Tem­per­a­ture Change (°C)
PTM70000.2mm+2°C
PTM70000.25mm+1°C
PTM70000.3mm0°C (base­line tem­per­a­ture)
PTM79000.2mm-4°C
PTM79000.25mm-5°C
PTM79000.3mm-6°C
PTM79500.2mm-5°C
PTM79500.25mm-6°C
PTM79500.3mmNot test­ed (was not avail­able)

Discussion

What we noticed is that gen­er­al­ly there was an improve­ment as we moved from phase change mate­ri­als with a high­er ther­mal imped­ance and low­er ther­mal con­duc­tiv­i­ty to ones with low­er ther­mal imped­ance and high­er ther­mal con­duc­tiv­i­ty (as expect­ed), but sur­pris­ing­ly, there was no improve­ment as we moved from 0.3mm to 0.2mm.

Sur­pris­ing­ly, the PTM7950 at 0.3mm offered sim­i­lar ther­mal per­for­mance to the PTM7900 at 0.25mm, but because the mate­r­i­al was not avail­able at the time of test­ing, we do not know the result of increas­ing the thick­ness to 0.3mm on the PTM7950.

Analy­sis of the parts and the data sug­gest­ed that part of the prob­lem lay in the Nvidia GPUs them­selves. The GPUs test­ed were quite large and it is sus­pect­ed that these larg­er GPUs were not as flat as small­er GPUs mean­ing that clos­er and more inti­mate con­tact was not pos­si­ble with the larg­er GPUs.

Many cus­tomers have per­formed sim­i­lar tests and we see sim­i­lar results emerg­ing from the their data.

Conclusion

We can con­clude with some lev­el of con­fi­dence that for larg­er GPU’s, that thick­ness­es of 0.3mm are bet­ter suit­ed than thick­ness­es of 0.2mm or 0.25mm. We can also con­clude that the PTM7900 is gen­er­al­ly supe­ri­or to the PTM7000 and the PTM7950 also seems to offer bet­ter ther­mal per­for­mance than the PTM7900. Unfor­tu­nate­ly, we can­not (yet) con­clude that the PTM7950 at 0.3mm would not have giv­en bet­ter results. Stay tuned for these results when they become avail­able.

As such, for large GPU’s we would rec­om­mend the 0.3mm ver­sion of PTM7900

Please vis­it our web­site to view our entire range of ther­mal inter­face mate­ri­als includ­ing our phase change prod­ucts includ­ing the three mate­ri­als described in this arti­cle: PTM7000, PTM7900, and PTM7950. Here you will also find all rel­e­vant tech­ni­cal infor­ma­tion includ­ing tech­ni­cal data sheets and safe­ty data sheets. You can also con­tact us if you have any ques­tions about any oth­er Ther­mal Inter­face Mate­ri­als.

About Chris Perabo

Chris is an energetic and enthusiastic engineer and entrepreneur. He is always interested in taking highly technical subjects and distilling these to their essence so that even the layman can understand. He loves to get into the technical details of an issue and then understand how it can be useful for specific customers and applications. Chris is currently the Director of Business Development at CAPLINQ.

Leave a Reply

Your email address will not be published. Required fields are marked *