Nanofluid: Dual Core Tests and Review

[skinnee]

The paper is quite good and gives a lot of background information on nanoparticles and the affects on heat transfer. The things that are difficult to calculate are exactly what the Reynolds number for a typical CPU loop, CPU, NB loop or CPU, NB, GPU loop. My estimated calculations put it somewhere around 6000-7000. If you follow the curve of the test data provided in the paper, we would see a potential 1-3% performance gain using nanofluid. This actually lines up with the test data I collected.

It will be interesting to see if the Quad tests result in roughly the same performance percentage gain.

[Kayin]

Anywhere I can get this from now? I'd love to run it in Mithril...

Black, white and G48 blue would be so badass...

[123bob]

First off, to skinnee, nice post. I'll have to graph now, in order to match your data format....


Whoa ! , to the above posts, where I'm mentioned folks....

Like a few of you, I approached this with some healthy skepticism myself. However, how can we march forward without trying new stuff? I would put it to the nay-sayer's and the overly enthusiastic folks on this thread, wait and watch. Let the science prove it to be valid, or not.

I am doing some very detailed testing since I'm a 100% dedicated cruncher. What does that mean? I run full bore, 24/7/365 with the 16 quads on my farm. 9 of those Quads are water cooled. One of them is my main rig, a QX6700 B3, which I'm testing on. This test protocol is going to take a few days since I'm allowing plenty of time for thermal equilibrium to occur before taking measurements. (minimum of 2 hours each.)

I am aware of the whole fluid temp vs ambient temp vs heat dump of a rad, vs, etc, etc. I think my resulting test protocol will prove that. I have the thing instrumented as far as I can go, with some rather high tier gear.

It is of great interest to me if I can achieve higher average clock using nanofluid rather than distilled. That is the ultimate claim of this product. That it will allow higher stable clocks than just water. We'll see....The test protocol is designed for this particular aspect of the potential product.

Yes, I realize the end price of the product needs to be justified, rather than adding another quad worth of hardware to my farm, but so far, my initial tests dovetail Skinnee's results. We don't know the price of this stuff yet, and I only fully formulated a test plan a couple of days ago, so, let's please not get ahead of ourselves.....

My setup and test results will be detailed in the WCG section, which is my home part of the forum. I will add those results to this thread here, so we all have it.

Under heavy heat load, (and I mean really heavy 24/7 stuff) this is showing some improvement. What I want is more overall clock for that same 24/7 requirement.

(Shameless advertisement...) As a side note, a lot has been given to how you load. I would put it to all of you here that crunching BOINC with us at WCG is the max stability test you will ever see. Put that cruncher on, helping the team out, and watch it run for weeks.... This ain't benching, which is also a sport I respect. Why? Because some of our best clocked machines on WCG are owned by good benchers. Come on over, hang out, and perhaps join the team. Link below.

@ Naekuh, I did the "stick the meter in the fluid" test to measure it's relative conductivity to pure straight distilled water. I find it to be no more (or less) conductive than the distilled. I also detailed a drying test I did on the WCG post I made. I was concerned that this might be like latex paint. It was not. I dried a puddle in the bottom of an empty beer can which dried looking a whole lot like diluted latex. I took a damp rag to it and it wiped right out. So, no permanent issue noted in screwing up blocks or reservoirs.

The smell, that's got me puzzled. As noted by skinnee, it is very familiar. He suggested plaster, I suggested latex paint mixed with some sort of perfume. Of course, I could put some through chemical analysis, but that would not be in keeping with the proprietary nature of the product.....

@reltemp, I did ask this before on the composition. I can dump it down the drain without killing all the fish in the river, right? Also, if my results on the single machine prove out well, I will need another 5 liters to fill my mass cooler and get the other 8 quads. Test results on that to follow too...

Regards,
Bob

[123bob]

Skinee, thank you for taking the time to do this.
Excellent work.
Do you have to use gloves with the nanofluid?

Double post, sue me. This is an important point.

I did not use gloves of any kind. It does feel somewhat sticky as it dries, hence my concern and the execution of the drying test detailed in my prior post.

However, I would not taste test this since I don't quite know what's in it. Perhaps with some Brandy, it might not be bad....

Regards,
Bob

[123bob]

My response to the hope that 123bob's testing would show a significant difference where the tests I peformed do not show a large margin one way or the other. I do not see how that is really possible unless there was a recipe change to the fluid that 123bob recieved from what I did. His test numbers will be very close to mine, he can even use the same test steps I did. That is why I document and present that information, so others can repeat the same tests and compare data. This is the science part of testing, repeatability.

OK probably triple posting now. Old age, you know....

Amen brother. Science counts. I'll let my ultimate results speak for itself. Opinions, well we all have them.... So far, our results are in lock step. It is encouraging that you can repeat my result, and I can repeat yours....I think we have the same formulation of the stuff, only reltemp can confirm that.

Regards,
Bob

[skinnee]

OK probably triple posting now. Old age, you know....

Amen brother. Science counts. I'll let my ultimate results speak for itself. Opinions, well we all have them.... So far, our results are in lock step. It is encouraging that you can repeat my result, and I can repeat yours....I think we have the same formulation of the stuff, only reltemp can confirm that.

Regards,
Bob

Following your posts in the WCG section, we're spot on. Slightly different test plans, but we're achieving near the same results. Its a good check and balance for both of us.

[iandh]

Regardless of the final breakdown of all the details, any fluid that drops max load temps (higher ability to absorb instantaneous temp spikes) could have value to OC'ers, regardless of whatever the average temps work out to be.

Reasoning is, that I've had chips where a 1-2C drop in load means the difference between stable and doesn't even boot. I remember I had an X2 3800+ that would run NOT run stable if load temp exceeded 55C, regardless of what the speed/voltage settings were.

[relttem]

Ah...the Reynolds Number.

Skinnee sent me a PM about this, and I thought it would be a good thing for all of you to know..for those of you that don't that is. The Reynolds Number, named after Osborne Reynolds, is a non-dimensional number that is very important in fluid dynamics. I am sure Wiki has a good write up about it, so I won't go into the details of it. In heat transfer you have a few non-dimensional numbers that you commonly use with the Nusselt number probably being the most common. The Nusselt number, Nu, is related to the Reynolds Number, Re by the equation:
Nu = X*Re^y*Pr^z

Pr = prandtl number
X, y, z = constants depending on fluid and flow.

In the paper I have been sending out, we were using the LOCAL Reynolds number. This is a measure of the Re at a specific location. The equation for Re is:

Re = (rho*V*D)/mu

rho - density
V - velocity
D - diameter
mu - viscosity

also,

Re = (4*mdot)/(mu*pi*D)
mdot - mass flow rate

You have to make sure you are using the same system of units - all your units should cancel out. Thus, the non-dimensional. There is a good website, onlineconversion.com that will let you change gpm to kg/s etc.

local Re is

Rex = (rho*V*X)/mu

X is the distance you are measuring the Re

So, if you wanted to compute the Re of your system you need to know a few things, and then crunch through some equations. First you have to know the gpm your pump puts out. Then, go to onlineconversion.com and convert that to kg/s. Next you have to know the diameter of your tube..preferably in meters...you can use the same website to convert inches to meters if you are unsure of how to do it. Now, you can get your Re using the equation above. You'll have to find the dynamic viscosity and density for water at your systems temp, but you can get that online. With those numbers you can get a very close estimate of the fluid velocity using

mdot = rho*V*A
A - area (pi*D^2)/4

solve that for V

Those are the V and Re for your Tube. Now, you have to move to your water block. You would have to know the dimensions of the flow channels in the water block, and if there is branching flow etc to accurately calculate the Re. It gets a little tricky, but you can make some general assumptions and get a reasonable estimate of your Re in the block. Knowing the Re you can use the graphs to get an estimate of the convective coefficient, h. Then, you can determine how much heat is being removed using

Q=h*A*(T2-T1)

That will give you a ballpark number since there is a lot more going on that just that.

thanks...I will be here all week.

[NaeKuh]

okey so who do i need to bribe to get a sample of this stuff to try on my gpu loop?

[TI66ER]

For anyone interested,i got this



From this pdf file

http://www.osti.gov/bridge/servlets/...able/11048.pdf

It is pretty interesting.

[relttem]

If you look at the results from Skinnee, you can see that at the slower pump speed is when the temp difference was most pronounced. Something that that article points to also is that you can also use a smaller radiator due to the increased thermal characteristics of the nanofluid (less fans blowing), which makes the whole system smaller and quieter.

[NaeKuh]

so increase pumping power and radiator capacity allows you to unlock the fluid even more?

or it just works with lower pump capacity?

And the risk of corrosion lets say if you did mix metals or is it safe in an all alu loop?

[Kayin]

Aside from the increased performance, it's easier to pump too?

Seriously, I need some of this stuff.

[relttem]

the density change is negligible, so the pumping power doesn't change. We looked into this. It will work fine with an all aluminum set-up. I think I mentioned, maybe in the WCG section, that your heat transfer at the block is based on this:

Q = h*A*(T2-T1)

Q is the power input..this will vary depending on what you are doing on the computer.
A is the area of the block..this is constant.
T2-T1 is the fluid temp difference between the exit of the block and the entrance.
h is the convective coefficient, which is higher for nanofluid than it is for water - that is what we found in our research, and other research projects have found the same thing.
So, straight up, based on the Q-equation, if h increases and A and (T2-T1) are constant. Then Q has to increase to balance the equation, which means you can dump more heat into the nanofluid.
If you look at the equations above, you have Nu = XRe^aPr^b, Pr is going to be pretty constant. If you increase Re your Nu will go up. Nu= (h*D)/k. If Nu goes up then h has to go up, because D(diameter) is constant, k(thermal conductivity of nanofluid) is constant - h has to change to balance. That is what that paper/quote above about pump power is talking about. But, your pressure drop is going to increase the faster you pump..so much that you might have leaks and other issues. It is a study. You have to find where everything can reach its highest point without any failures. And, it is a trade-off..bigger pumps, larger radiators, more fans = $$ and noise.

[NaeKuh]

how does that have anything to do with corrosion?

if the nano particles are a form of Cu+ suspended, it would eat into the ALU and hence the coolant would cause corrosion.

Im trying to work out all the fine points on your coolant b4 i give it a test. So sorry if i sound a bit curious.

What got me thinkn this is how it reacts to copper, it seems the solution is safe and has no effect on copper, however how would the solution handle alu? Or even better, how would it handle a mixed loop scenario since we still cant get people to NOT USE ALU.

Once again sorry if i sound a bit curious, im really interested in your solution.

[TI66ER]

On google i found a site that say the nanofluid they use/have/make is copper particles in the fluid which is fine if you have an all copper loop.

[relttem]

During our tests we were flowing the nanofluid over a heated copper plate with no adverse effects. It will not harm aluminum either.

[rosco]

The nanofluid dispersion was purchased to Nanophase according a previous post, and according to their products and the white color, this stuff should be composed of aluminum oxide, the most classical nanofluid, or zinc oxyde (you talk about it in your paper, you probably forgot to erase some info...). They have tin oxyde too, that gives a white color, but I doubt about its use here. Hence, no problem of corrosion because nanoparticles are already in an oxidized form.

[Hondacity]

is this rohs approved?

[NaeKuh]

Hence, no problem of corrosion because nanoparticles are already in an oxidized form.

ahhh that is a very good explaination.

thanks.

[rosco]

Only applies if they're made of a metal oxyde (can't oxidize more given the conditions), not a pure metal obviously. A nanofluid made with pure metal nanoparticles needs special methods to avoid oxidization during the process using a single step method especially.

[Boogerlad]

will it be possible for it to be in different colours? seems promising. Better performance, bling, and no need for a silver bar to prevent algae! Does it have any anticorrosive properties?

[relttem]

we are going to leave the coloring up to the people for now - it would be tough to provide certain colors for everyone. We will probably get some of the dye you can buy and add it to it and see what happens - we don't think anything negative will happen. We originally were going to make it fluorescent blue.

Anti-corrosive? does it prevent corrosion? it doesn't corrode plastic or your pump, block etc. Over a long LONG period of time, like 10 years you might see something, but that could be due to all kinds of things outside of the nanofluid.

[TNTSunshine]

For these conditions, is the 1 to 2 degrees difference statistically significant? If so, with what confidence interval?

And you still wouldn't want this stuff in a mix Al/Cu loop. You'd still get corrosion, unless you had an all Al loop, but then that wouldn't be mixed.

[Jimmer411]

Distilled H20 + PT Nuke $6 for a whole loop

How much would this run?

Tapwater and prestone, totals about 25 cents per loop tops.