Nanofluid: Dual Core Tests/Review

skinnee

n00b
Joined
Oct 5, 2008
Messages
47
To start, I would like the say thanks to the XtremeSystems World Community Grid Team (specifically, Hicks and litteowl) for getting me involved in this test. In addition, a big thank you to relttem and his business partner for supplying me with the fluid to test, without the fluid...no test!

Intro
There is not a lot know about this mysterious white fluid known as Nanofluid, what I do know is that there are nano sized particles in the fluid that give this fluid the ability to gather and store heat with the ability to release that heat in a highly efficient manner.

I can dispel some rumors I saw posted though. The mysterious particles are not aluminum. So no, the fluid will not cause corrosion or presents a corroding substance into your loop.

With that as a precursor to the testing and review, here is the eye candy you all so eagerly desire.

Caution, this is not Hydrogen Peroxide, just the vessel that I received the fluid in through USPS.


The fluid has an oddly familiar smell to it, reminded me of the smell of plaster. Here is how the fluid looks poured into small dish.


No UV reactive properties, I was surprised by that, I figured it would be like white lettering on a shirt under UV. Nevertheless, absolutely no UV reaction.

Here is the test bench setup for this test with the loop filled with Nanofluid.


Test Setup and Methodology
Test Specifications are detailed with each of the data tables and graphs. Each and all tests was performed using Prime95 version 25.7 using In-place large FFTs for a 15 minute warm-up period with a logged test cycle of 60 minutes following the warm-up period. Ambient temperatures are averaged between two different sensors. After each test, a 30-minute idle period was observed before starting the next test.

High Flow: Very close, averaged temps are 0.35ºC and 0.33ºC difference for a 60 minute load cycle with Nanofluid having a 1ºC advantage on max core temps for the test cycle.
Nano_dualcore_graph-highflow.png


Medium Flow: Medium flow produced strikingly similar results to High flow, Nanofluid has a lower Max Core Temp by 1ºC, and the difference between average temps shrinks to 0.24ºC and 0.25ºC. At this point, it is looking like a real duel.
Nano_dualcore_graph-mediumflow.png


Low Flow: High heat loads is where the Nanofluid really starts to outpace Distilled. Nanofluid still grabbed the lower Max temp by 1ºC, but the average temps were what really stood out, Nanofluid has a 1.39ºC and 1.42ºC advantage in average temp and a 2ºC minimum temp advantage.
Nano_dualcore_graph-lowflow.png


Closing thoughts
The low flow tests really makes me eager to put the fluid through a much longer and more rigorous set of test runs with a Quad due to the heat load being higher from a Quad versus that of a Dual, I have a feeling this is where Nanofluid will pull away from Distilled and show a bigger margin. I also plan to do multiple runs of the same test settings with a quad, this will allow me to see if there is any major deviation from run to run for the same test. This is speculation at this point, but a somewhat educated speculation.

For those of you interested in purchasing Nanofluid, I cannot help you there, as to my knowledge this is only a prototype and is not ready for sale to the masses.

In closing, I would again like the thank relttem and his business partner for the opportunity to test and report on their creation. Thanks for reviewing the Nanofluid Dual Core test results and review.
 
I was following this thread over at XS and I have to say that I'm really interested in what this fluid contains. a 2c load difference on the same equipment is nothing to sneeze at. I'd really like to see these loaded up on a Quad core using the same D-Tek block both with and without quad nozzles for comparison's sake to your chart above.
 
Yes, there are a couple threads at XS that go into a lot more detail. I did not know the policy here on linking to another forum so I did not provide a link to the thread over there.

BTW, what is the policy on linking to another forum for discussion purposes?

The quad tests are coming (have some D5 tests happening first) and the testing will follow the exact same methodology, but there will be a few differences. For the quad tests a GTX480 instead of an MCR320 and each individual test will be run 3 times versus the single run. I need to be able to show a standard deviation for the tests, basically the amount of swing or variance in the test groups.
 
I haven't seen any mod have a problem with linkage, might be just my luck :D
 
I wonder if performance would degrade over time from simple time elapsed degradation or accelerated by heat or pump action.
 
Would this be the D5 aftermarket top results? Woot! You should link them here too when they are ready.

I am really looking forward to those tests, I am a D5 over DDC3.2 user. And now that I am finally a member here, all the tests/reviews will be posted here as well. :D

I wonder if performance would degrade over time from simple time elapsed degradation or accelerated by heat or pump action.

The test bench crunches for WCG full time when not performing tests, but I still constantly monitor and log all data. I can say that over the last two weeks when the initial testing completed I have not seen a single degree rise in load temps. However, this is certainly not enough time to give a definite answer for performance or fluid degradation. I will continue to watch things especially before starting the quad core test runs.
 
Okay then...Admins/Mods, please let me know if this is not allowed.

XS Nanofluid Discussion

np, its the direct linking of images etc. from a site you do not host that is a problem. This is fine.

This is a very fine workup and done very well and is very interesting.

I hate to have to say this and it is provided as information only as the results are there and it does seem the fluid has some positive effect. The thing I wish to mention is that the temp sensor is accurate to +/- .5C , because of this it is impossible to place any credence to the 100ths of a deg C digits in the results as the accuracy of the values (significant digits) of measured data cannot be more accurate than the valid significant digits of the least accurate device used in testing. This may also put the tenths digit into question but honestly I cant remember. To be fair this is a bit of nitpicking as the same sensors where most likely used in all the testing and so the absolute error is probably smaller than the overall 1C range of accuracy of the temp sensors and as the temps in all instances did not vary greatly (keeping the temp sensor near the same operating point and thus the error fairly consistent - one would hope). This would depend on the "drift" of the sensors error and that is something I did not find in the documentation for the DS18B20 but I did not look at the full data sheet which can be found here: http://www.maxim-ic.com/quick_view2.cfm?qv_pk=2812

Again very will done and outside of using 10's of thousands of dollars of equipment in a lab I do not think it could be done much better, certainly not be me and the presentation is excellent. Kudo's.
 
As Bill stated, a 1-2c difference in temperatures is quite close to a 4% difference, which is well within most margins of error...

But, Considering the Distilled water was replaced directly with the nanofluid, there would be very little margin for human error... The difference may very well be smaller than 2c (inflated through rounding) but it's probably a real difference... Rather than a statistical or mechanical error...

But, 1-2c just isn't worth it IMO, this stuff can't be cheap... Is it?

Nice to have you on the board though Skinnee, We need more knowledge powerhouses around here :)
 
np, its the direct linking of images etc. from a site you do not host that is a problem. This is fine.

This is a very fine workup and done very well and is very interesting.

I hate to have to say this and it is provided as information only as the results are there and it does seem the fluid has some positive effect. The thing I wish to mention is that the temp sensor is accurate to +/- .5C , because of this it is impossible to place any credence to the 100ths of a deg C digits in the results as the accuracy of the values (significant digits) of measured data cannot be more accurate than the valid significant digits of the least accurate device used in testing. This may also put the tenths digit into question but honestly I cant remember. To be fair this is a bit of nitpicking as the same sensors where most likely used in all the testing and so the absolute error is probably smaller than the overall 1C range of accuracy of the temp sensors and as the temps in all instances did not vary greatly (keeping the temp sensor near the same operating point and thus the error fairly consistent - one would hope). This would depend on the "drift" of the sensors error and that is something I did not find in the documentation for the DS18B20 but I did not look at the full data sheet which can be found here: http://www.maxim-ic.com/quick_view2.cfm?qv_pk=2812

Again very will done and outside of using 10's of thousands of dollars of equipment in a lab I do not think it could be done much better, certainly not be me and the presentation is excellent. Kudo's.

The tenths of degrees in the charts are purely from number crunching in Excel, but I do agree with you that the Maxim sensors have a accuracy of +/- 0.5ºC so the tenths of degrees could easily represent a margin of error on all temps logged through them. You're not the only one who has noticed and brought this issue up. I will definitely take your advice and implement the multiple runs of the same test in the quad runs and future tests, thanks! Moving forward, I will run the same test a minimum of three times in order to calculate standard deviation and confidence factor.

You are correct, sensors were not adjust or moved in any way, the only thing that changed was the fluid and the pump speed (flow and pressure) during the tests.

Thank you for your kind words and feedback, I truly appreciate it.

As Bill stated, a 1-2c difference in temperatures is quite close to a 4% difference, which is well within most margins of error...

But, Considering the Distilled water was replaced directly with the nanofluid, there would be very little margin for human error... The difference may very well be smaller than 2c (inflated through rounding) but it's probably a real difference... Rather than a statistical or mechanical error...

But, 1-2c just isn't worth it IMO, this stuff can't be cheap... Is it?

Nice to have you on the board though Skinnee, We need more knowledge powerhouses around here :)

I am really trying to take the scientific approach to testing, where every step and piece of information is captured, logged and noted. This way I have every detail possible given for repeatability and want to be sure that I keep any and all bias out of my testing and reviews.

I honestly do not know the price and would not even begin to speculate, seems I am always wrong when I speculate. ;)

Thanks, but I wouldn't consider myself a powerhouse by any means. I have far to much to learn before even being able to look at a powerhouse. :D

Again, thanks for your feedback and kind words!
 
Back
Top