[Skinnee Labs] Coolant/Fluid - Thermal Performance

[Ccaution]

Very nice roundup - good to know Hydrx was a right choice

Btw, rjkoneill, you've done some major sleeving there mate, pretty preeeety good

[rakesh_sharma23]

Very nice comparision... it will be super if you can add up pure distilled water performance. Since many people use pure H20 so it will be nice for all..

[rge]

hi there

i have nanofluid in my rig and it performs better than distilled, not really sure why your results came up like that.

also, it is slightly more viscous than water with sort of an oily property to it

i just find it really odd how the fluid that has performed the best in all of my testing, is by far and away the worst in this test.

By your comments, I can assume you are using Fluid XP nanofluid, which did perform the worst in skinnee's tests.

I hope you realize there are 2 types of nanofluids on the market, Ice dragon which has zinc oxide in an attempt to increase thermal performance.

Then there is fluid XP nanofluid, which is NOT an attempt at increasing thermal performance, it has ~50% propylene glycol (which accounts for oil like viscosity you noticed) mixed with distilled water, and nanoparticles that scavenge ions in an attempt to decrease the rate at which the fluid in the loop becomes conductive. If you are using Fluid XP nanofluid, it should not be possible for 50% propylene glycol + distilled water + ion scavenging nanoparticles to perform as well as distilled water, since the thermal conductance of propylene glycol (.147 W/m*K) is roughly 1/4 that of water 0.6. If you look at the thermal conductance of say feser one, which is roughly 85% water (0.6W/m*K and 15% ethylene glycol (0.258) the thermal conductance should be around .85x0.6 + .15x.258 or 0.55W/m*K, which would explain why 15% ethylene glycol or feser one is only a degree or less off. However the thermal conductance of fluid XP nano would be .5x.6 + .5x.147 or 0.37W/M*K, so one would expect it to be at the bottom of the pack. And, again their nanofluid particles scavenge ions, not enhance thermal transfer.

Read their web site about chemical properties and the fact their nanofluid was designed for fuel cells, then read this.

EDIT: I just realized that even fluid XP's website lists the specific heat in J/g as 3.23. Water is 4.18. So even their website shows reduced thermal properties.

@rakesh...feser ultra pure is distilled water (double distilled apparently), then he tested tap water and both were same, so any purity of distilled between 2 should be same, not to mention he tested distilled with few drops pt nuke, which is distilled water, again.

[Petra]

Skinnee: A few thoughts...

First off, that test looks to have been a PITA. That said, how much full-load cure time did the MX-2 on the CPU and GPUs have prior to running the tests (thinking back to Vapor's experiments in charting TIM's thermal characteristics)? I didn't see that information in your methodology section (you mentioned no mount changes, but that was it). Also, looking at the averaged data from the CPU and GPU DTS, I'd wager all of the results (aside from the Fluid XP Nanofluid) fall within or pretty close to the margin of error--a heat source and sensor equipment that wasn't live computer components may have been a better choice....then again, it could also complicate things further.

...bah, I'm tired... I think I'll just shut up and go fix my car.

[skinnee]

I don't recall the exact amount of time, but there was well over a week of sensor calibration and pre-testing that occurred prior to data being captured for the logged results. With the sensors deployed for testing, margin of error is +/- 0.125C, so there is some separation but nothing I would consider significant or note-worthy, FluidXP Nano-fluid excluded of course.

On the heat source... that really doesn't work for the intended audience and indeed does complicate matters. With an representative system, the entire audience can relate to each piece of gear, applications used and so on. This same scenario plays out for every test we take on, it is a balancing act and trying to find the proper mix of testing to engage the LC community. I assure you, this is something Vapor and I discuss each and every time we are in the planning stages for each test/review.

[Petra]

I don't recall the exact amount of time, but there was well over a week of sensor calibration and pre-testing that occurred prior to data being captured for the logged results.

Well, so long as it had a chance to cure to a point where the performance wouldn't change significantly over time (digging through my PMs, I recall Eric mentioning 12 hours or so). Either way, it's a detail which should be included in the methodology page.

With the sensors deployed for testing, margin of error is +/- 0.125C

So, the accuracy, precision, and linearity of the DTS is no longer an unknown and/or has been properly cross-calibrated over a wide temperature range against a known, good reference? If that's the case and +/- 0.125C is your calculated margin of error/uncertainty, then it should be stated in your methodology.

On the heat source... that really doesn't work for the intended audience and indeed does complicate matters. With an representative system, the entire audience can relate to each piece of gear, applications used and so on. This same scenario plays out for every test we take on, it is a balancing act and trying to find the proper mix of testing to engage the LC community.

I suppose we're just going to have to remain in disagreement there... pandering to the perceived level of understanding of a given audience is all fine and dandy when it comes to the analysis of data and conclusions; however, such "audience tailoring" should not play a part in the design of a test or in the presentation of data. In the end, though, they're your tests and you can conduct them however you like. Carry on

[ryan92084]

hrm this goes against the the other dragon nanofluid tests, not that is was a huge difference to begin with. one of which was done by yourself or vapor if i recall correctly. any thoughts?

[skinnee]

I wouldn't say it goes against the previous test I performed. Completely different test bed, test procedure, Nanofluid formula's and so on.

[Martinm210]

FWIW,
I think tailoring the test to what the audience both understands and accepts is critical....even more so than the test accuracy.

I don't know how many times I've had someone ask me which direction is better on a pump curve. Is it up or down?

I think we do have to modify methods sometimes..the old accuracy vs acceptance/understanding dilemma..

[Metroid]

This confirms what I thought in theory. I have been using Feser Ultra Pure distilled water since day one. Thanks Skinnee for the test.

[Petra]

FWIW,
I think tailoring the test to what the audience both understands and accepts is critical....even more so than the test accuracy.

I don't know how many times I've had someone ask me which direction is better on a pump curve. Is it up or down?

My point on this particular matter is that I'd rather educate the audience through explanation than dumb-down the test and presentation of data. This way, those who already do understand are presented with what they want and those who don't understand just have to read a bit (yeah, reading... what a horrible thing to expect people to do ). Compromising a test just to make the results nice and easy for the lowest common denominator is damn near a crime in my book (note: I'm speaking in general, not in specific).

BTW, I read that first line of your post to my wife and it sent her off on one of those 10-15 minute "everything that's wrong with society when it comes to basic math/science" rants

[Martinm210]

My point on this particular matter is that I'd rather educate the audience through explanation than dumb-down the test and presentation of data. This way, those who already do understand are presented with what they want and those who don't understand just have to read a bit (yeah, reading... what a horrible thing to expect people to do ). Compromising a test just to make the results nice and easy for the lowest common denominator is damn near a crime in my book (note: I'm speaking in general, not in specific).

BTW, I read that first line of your post to my wife and it sent her off on one of those 10-15 minute "everything that's wrong with society when it comes to basic math/science" rants

Yeah I know...but not everyone is like us and thinks science is fun or even wants to learn more..

I never did a very good job in my testing at articulating the information...but then again, I was doing the testing for myself..out of my own curiosity. I did learn the hard way about presentation and understanding who your audience is. That's an "Art" all in it's own...

[Sadasius]

My point on this particular matter is that I'd rather educate the audience through explanation than dumb-down the test and presentation of data.

HAHAHA I loved your youtube video pouring water and coolant on live motherboards.

[rge]

So, the accuracy, precision, and linearity of the DTS is no longer an unknown and/or has been properly cross-calibrated over a wide temperature range against a known, good reference? If that's the case and +/- 0.125C is your calculated margin of error/uncertainty, then it should be stated in your methodology.

I agree their is no easy extrinsic method for validating DTS accuracy, but you can simply do a paired t test on his data and show sensor error is statistically improbable to scientific standards of 95% (p<0.05) or even 99% if more runs (but not necessary). I had same question in past, and ran 10 tests of same cured tim at different ambients, and no question i7 DTS sensors are capable of .1 to .2C relative accuracy within above tested range and are linear within a few degrees C. Even on his 3 runs here of each, T paired assuming null hypothesis, comparing 3 feser one cpu temps to 3 distilled water cpu temps their is a 98% probability that a difference is significant and not do to random chance, ie same 98% chance not due to random sensor error. Doesnt rule out systematic error (other controls for that), but that alone to a scientific standard eliminates DTS error, and you can show 95% confidence that error DTS is less than 0.2 and i dont feel like redoing skinnees math, but his cpu error is likely around .125.

I guess you were looking at the discrepancy between GPU results and cpu, for example minnesota tap. Some of the gpu sensor data had SD of .4 vs cpu all were near .1, so perhaps GPU data/sensors have higher error causing slight discrepancy. But always some screwed data points in any testing.