Heat Sink Shootout: heatsinkusa's 3.945 -vs- 4.6

[Rahz]

Starting a new thread for title accuracy. Objectives: Determine which heat sink works better passively and actively cooled. Determine how much airflow is necessary to maintain reasonable temperatures.

The 3.945 profile is a 3" piece. It has a similar weight and cost to a 4" piece of 4.6 so these are the two heat sinks I will be testing. Off hand I expect the 3.945 to be better at passively cooling and the 4.6 to be better at actively cooling, though it may also do a fair job passively cooled. Worth mentioning, the 3.945 (to be called 3.9 from here out) has a thicker base plate and taller fins with more space between. The 3.9 will accommodate a much higher airflow (Not necessarily desirable). The 4.6 has a lower profile and serrated fins.

Method: Each sink will have a CXB3070 attached and will be powered @ 1.4A. I will start a timer and observe the amount of time it takes to reach 70C. After those two tests are done the sinks will be placed in high tech LEEs (lamp enclosure emulators) and will be actively cooled with forced air from a 25CFM fan. I will time and monitor this phase as well but unless there is a high temp problem the results will be a temp measurement at the 30 minute mark. The winning heat sink will then be subjected to a reduction in airflow to see how the temperature is affected. I won't be able to know what the CFMs are in that phase but hopefully I will get a rough idea that will be helpful in the future.

If neither of the heat sinks are up to the task of cooling a 50 watt emitter to reasonable temps I have larger versions ready to sub in. If the 25CFM fan isn't enough (I think it will be plenty) I have a 100 CFM fan ready to sub in. Pics and test results will be posted over the next few hours.

 

[medicinehuman]

Sounds like a very cool experiment, I'll be looking to see what happens.

 

[Mechmike]

I'm interested to see the results of your test. My bet is on the 4.6". The serrations increase the surface area of the fins.

 

[churchhaze]

The amount of time it takes for the heat sinks to get to 70C will have more to do with the heat capacity of the sinks (specific heat * mass) than the thermal conductivity.

Heat capacity is basically meaningless once the sink reaches equilibrium.

 

[churchhaze]

What should be measured with passive is what temperature the sinks converge at after 1-2 hours operation with perfectly still air, in a large room with temperature control.

 

[AquariusPanta]

Bang bang!

There is truth in Church's logic.

 

[Rahz]

How much heat can the CXB take though? I'm not at all sure these sinks can cool 50 watts passively without a failure. Time to 70C for the 4.6" was 1 minute 37 seconds...

What should be measured with passive is what temperature the sinks converge at after 1-2 hours operation with perfectly still air, in a large room with temperature control.

 

[Rahz]

Active cooling for the 4.6 results: Temps were checked every 4 minutes and were maxed out at 53C around the 20 minute mark. This is a hair below the typical temp listed on the datasheet. I might put the 100 CFM on the enclosure emulator to see how much further the temps drop. I'll also hook a fan controller to the 25CFM in a few minutes and see how the temps drop with fan speed once the typical has been reached.

Starting temp: 22C
4:00 - 47C
8:00 - 51C
12:00 - 52C
16:00 - 52C
20:00 - 53C
24:00 - 53C
28:00 - 53C

 

[Rahz]

I put a fan controller into the equation. It doesn't seem to be made for these type fans because at full power the fan was producing a lot less air movement. Maybe 10CFM though I have no way of knowing if that estimate is accurate. At this speed the temps hit 70C at 20 minutes. My guess of 25CFM was good since it resulted in typical temps. This last test makes me think lower temps can be achieved with higher volume through the fins. That's why I'll try next...

 

[Rahz]

100 CFM was able to lower temps to 40C. This represents about 7-8 LPW over the typical temperature according to my calculations. Can anyone back me up on that? Being the case, there are some potential gains to be made over typical using this particular sink, though the additional fan wattage necessary to produce the extra lumens isn't impressive at about 40 LPW. Still, when applied to more than one heatsink in parallel the additional LPW from each emitter would be additive, minus the effect of the additional resistance. Anyway, I'll do the same tests I did this evening on the 3.9 sink tomorrow and we'll see which sink wins the shootout. Here are a few pics of the rig.

01- setup for the passive test.
02- passive test running
03- the high tech lamp enclosure emulator ready for action with a 25CFM fan
04- the LEE modified with a 100CFM fan

 

[SupraSPL]

Thanks for setting up this test. This has got me wondering how well an active cooled chunky heatsink like this would compare with a Alpine11 or Rosewill. Do you think the light of the COB could be heating the black tape interfering with the reading?

 

[Rahz]

Thanks for setting up this test. This has got me wondering how well an active cooled chunky heatsink like this would compare with a Alpine11 or Rosewill. Do you think the light of the COB could be heating the black tape interfering with the reading?

Hmm, I'd like to know how it stacks up to CPU coolers as well.

I'd guess that there is a minimal influence from the photons. At the angle the tape is at there's not a lot of light.

 

[Rahz]

I have a couple of the 50W radial heat sinks left from my previous build. Another interesting test might be running the CXB on these heatsinks as well as the radial @ .7A. I know for a fact the radials can handle a 25 watt load no problem, though I haven't got exact temps on them. They are a little more expensive (though with free shipping) at around $12 each. The radial weighs 12.6 ounces and the 3 inch section of 3.9 from heatsinkusa maxes out my scales and is over 17 ounces. The base of the radial is about twice the thickness but there's a lot less surface area. I don't have a spare .7A driver laying around but I could test them at 1.05A.

 

[churchhaze]

I have experience with the 3.95" profile at 4". With a 90mm coolermaster fan at 5V, it performs about as well as an arctic apline 11 at 5V... although I never measured the temperature.

At first I thought the chunky design was worse than the arctic 11, but it also works fine at 40W dissipation without any fan. The arctic alpine would crap out fast without a fan. In general, I think the 3.95" profile "squared" is just nicer than the arctic 11.





That being said, I've since turned to shorter fin profiles (1 inch tall fins) for my newer designs. Taller fins requires wider spacing between fins to work as well [semi-]passively.

 

[Rahz]

I have a 4" piece I'm holding as a substitute. It is a hefty chunk of metal. I also have a 5" piece of the 4.6. Considering the size of these emitters I'm not sure going much wider is advantageous, so maybe the 3.95 has a chance against the 4.6 both actively and passively.Tomorrow we find out.

 

[Rahz]

Taking Supras comment into consideration I modified the rig so the thermocouple was secured on the sink and then bent the tip so it's own pressure held it against the test spot. These are the results.

3.95"
Max temp with 25CFM fan: 37C
Max temp with 100CFM fan: 34C

4.6"
Max temp with 25CFM fan: 37C
Max temp with 100CFM fan: 30C

These temps are much lower. Again, there is a notable difference with the 100CFM fan, especially using the 4.6 profile. As there is no thermal interface (or tape) to insulate the thermocouple I'm not sure how valid they are, though I do believe they are good relative temps for comparison. I haven't done the passive tests and am short on time, so I'll try and get to that tomorrow. I should have just waited since I will have to swap sinks again. Swapping out heatsinks would be much easier with Vero29s and molex connectors... The CXBs thermal junction is the whole backside of the emitter so I felt compelled to cover the whole thing with compound which is kind of a bitch to clean off, and the holders are okay but don't hold the emitter very securely while you're trying to get the screws in there. I can see why Ideal switched to the keyhole design.

I'd really like to know what the temps actually are at some point so it looks like I'm going to have to solder to do that. It will be interesting to see how much difference there is between the proper test method and the one I used for these results. Also worth noting, because the holders partially cover the test spot, it's possible that influenced the results as well. I'm not feeling good about using them for a soldered test. This has been much more trouble than I had anticipated.

 

[beodrone]

@Rahz - are you using thermal grease for these tests and or where the heat sinks sanded to be flat? Just curious as I am on a crusade to eliminate the need for TIM grease when using this kind of setup.

 

[Rahz]

The heatsinkusa surfaces are fairly flat so I didn't do any sanding. I used Arctic Silver 5 because I have a tube. There are a couple other types I want to test eventually. I apply it and make the coverage thin and uniform with a straight razor. I'm not sure it's feasible to do without some kind of TIM but I could be wrong? Thermal adhesive tape might be a good alternative.

 

[beodrone]

"I'm not sure it's feasible to do without some kind of TIM but I could be wrong?"

That's the million dollar question Any chance you will be doing a test comparing with TIM and without? For some reason I have built up hate towards thermal paste.

PS: I have tested without and it seems just fine but its always good to get someone else's opinion/measurement

 

[bicit]

Sanding and polishing seems like a lot more work than simply using a bit of grease. Good sand paper also costs a fair amount.