Thermal Paste Experiment - So Far So Good

[Markfothebeast]

I've been concocting a thermal paste based on Arctic Silvers MSDS and other chemicals that are good thermal conductors. I didn't have time to complete that project. However, I was just finishing building an air compressor aftercooler from an AC heat exchanger when I realized that the thread sealer I was using was textured, appeared, and felt just like Arctic Silver (expensive as crack) and similar thermal pastes.

The chemicals listed on the tube are good thermal conductors, non-corrosive to metals, and withstand heat very well. No silver (or aluminum oxide) like Arctic Silver, but I have doubts that the less than 5% is going to make a large impact. I figured since I shut off my server because of its loudness (72db @ idle), I'd use it as an experiment to see how the fans may throttle down if it idles cooler.

I cleaned the Arctic Silver from both heat sinks and processors and applied the "Harvey's Pipe Thread Compound". Sounds ridiculous, huh? I started up the loud beast, adjusted the irritating BIOS and such. Idle temps dropped from around 95F - 104F to a now 60F - 68F sitting steady. I don't recall EVER modding any computer cooling system that resulted in such a drastic change. I am happy to report that a day later the temps remain the same at idle.

So, "Harvey's Pipe Thread Compound". Find it at your hardware store for like $5. It's only been running for a day but I don't see this chemical drying out or failing quickly because it's intended to seal pipe fittings under harsh conditions. Edit: It generally remains fluid, it is not glue

 

[bfun_x1]

Interesting. Even the best and worst thermal paste only have a few degrees difference in performance.

 

[Rifter]

Interesting. Even the best and worst thermal paste only have a few degrees difference in performance.

Yeah, must have been something else wrong to get temps so high before. not just thermal compound.

 

[Markfothebeast]

Actually, I removed 2 (out of 4) processor fans and one chassis fan because it sounded like a jet engine. I didn't change anything other than the paste.

Nearly every desktop I've owned idled roughly 30c - 40c, including my current desktops. I had an Intel mobo with the Core2Duo quad processor and that machine idled at 50c-60c. I drilled holes on the mainboard (around circuits) and torqued the heatsink down and it idled at 40ish celcius. What this server is idling at is surprisingly cool in comparison.

 

[XavierMace]

Unless you're running it in a freezer, your sensors aren't reporting properly. Period. Not trying to dissuade you from experimenting, but a temp of 15c/59f inside the OS is basically impossible unless it's being run in a room with below freezing ambient temps. Your original temps of 95f to 105f (34c-40c) is 100% normal for a S5000PAL system.

 

[Pick2]

... I drilled holes on the mainboard (around circuits)

You did What ?? Drilled holes on the motherboard ?? No wonder its staying at room temps

 

[mikeymikec]

OP, what do the other sensor readings say? AFAIK there's usually an internal processor sensor, then a sensor on the board that detects the processor temp, and AFAIK the latter is usually what the board's fan management system bases its decisions on.

I think XavierMace has to be right though. The temp in my computer room right now is probably 18-20C, the SSD is 27C and the 4690K (stock) is 31-32C. As soon as something generates some heat, it's got to be higher than the ambient temp.

I don't have any experience of monitoring xeon temps, but 35C idle is what I'd expect from say a 2010 AM3 processor with a stock cooler. I wonder if the sensors giving you temp readings have always been wrong and that the HSFs weren't installed quite correctly (frankly I dread having to remove an Intel HSF from that era, especially the stock ones, it seems to me to be total pot luck for them to go back on correctly and it's difficult to judge visually so I rely on the temp readings).

Also IMO if "Harvey's Pipe Thread Compound" was cheaper and so much more effective, then surely companies like Arctic Cooling would have copied their formula (or just resold their product) with a higher price tag?

 

[Markfothebeast]

I'm simply stating what the temperatures are reading after my experimental run with pipe thread paste. The ambient room temperature is approximately 58 - 62 fahrenheit day and night. So it is idling slightly above room temp.

Yes, it is a piece of software giving me a reading. I have had false readings on several computers between fan/temp sensor parameters to whatever software was being used. Most of these were BIOS issues that were fixed after a firmware update.

I will run a "stress test" to see how high the temps climb and post all temperature sensor data as well. I have a video of this experiment from the start - which I will share very soon.

 

[BonzaiDuck]

As a hobbyist-enthusiast-DIY-mod-hardware-addicted veteran, I can only know firsthand the expectations of the CPUs I personally buy and use in my computer building projects. But it was pretty much established around the time of the Sandy Bridge cores and two generations before that: there is an error-range for the temperature sensors. Further, some processors -- like the E8600 C2D Wolfdale -- evidenced "crippled" temperature sensors that would peg at something close to 50C for idle temperatures approaching room ambient, but would kick up and measure with reasonable accuracy when the thermal power exceeded that associated with the 50C threshold.

There had been a time when only people who invested heavily in water-cooling parts were more cavalier about going after the "last grain of rice" in cooling improvement or reductions in degrees C stress temperatures. About that time, products like IC Diamond appeared on the market with other formulations and comparative temperature improvements. Then, the liquid metal applications became popular -- like Coolaboratory Liquid Ultra or its predecessors, and Indigo Xtreme metal pads.

The test results would depend on the test-bed and its TDP or peak load thermal power for a given overclock setting that takes the power level beyond TDP. When we went from Arctic Silver to nano-diamond, we could measure an improvement somewhere between 2C and 5C. Similarly, for removing nickel plate from copper surfaces we could get another 2C to 5C. The difference between nano-diamond and something like Indigo Xtreme would be another 3C degrees. And these comparisons and values were in all cases observed for unmodified Intel CPUs in the test-bed. That is -- they all referenced a heatspreader IHS with the nickel-plate unless otherwise stipulated, and comparison tests typically don't involve other improvements as would occur with lapping to bare copper.

I continue to use IC Diamond because it is totally reusable, is non-conductive, doesn't "pump out," and offers a level of improvement over something like Arctic Silver. It is not recommended for bare die applications. The liquid metal formulations may or may not be good for coolers with aluminum in the base construction. They may amalgamate or bond with the surfaces -- more with copper, less with nickel. Indigo Xtreme is troublesome to cure and troublesome to remove.

While TDP specs are high (140W) for hexa-core (or more) Intel processors fabricated with Indium solder, improvements in heatpipe cooler design and the proliferation of AiO water-coolers makes TIM choices less critical as long as the TIM performs to established expectation. But again -- every last grain of rice counts toward overall temperatures of loaded processors.

 

[Markfothebeast]

Also IMO if "Harvey's Pipe Thread Compound" was cheaper and so much more effective, then surely companies like Arctic Cooling would have copied their formula (or just resold their product) with a higher price tag?

Arctic Silver is just one of many different thermal pastes and the chemicals are similar in most, depending on the application.

 

[Markfothebeast]

Here she is. Pasting that bad boy (or girl).
https://youtu.be/W-lusNTEggo