Cooling A better MXM cooler?

[Kmpkt]

Hey Guys,

Wondering if anyone with decent modelling skills would be keen to partner up to make a better heatsink for MXM GPUs. I would ideally like to make something longer and wider than the stock Micro STX heatsink that would utilize a larger GPU style fan (92mm) and as a result run quieter. Furthermore I would like it to be slightly lower profile if possible.

I have a Nomad 883 CNC in my garage to make these on and will gladly make one for whomever cares to help me. If the heatsinks are a success then I would likely make and sell them here. While my eventual goal would be for the heatsink I am developing to fill this role, production is likely a ways off and I would like something better now.

I am happy to ship a spare DeskMini heatsink to the designer in order to allow for accurate measurements/modelling/etc. The CNC machine I would be using is here:

http://carbide3d.com/nomad/

Cheers,

KMPKT

 

[confusis]

FYI to prospective designers,

 

[confusis]

Should have streamed this.. Forgot how much I love working in Sketchup

 

[blubblob]

Have you tried milling aluminium with a Nomad 883?

While I have no personal experience with the machine it seems unfit to properly deal with it. Low spindel speed, questionable rigidity and no cooling - I would be surprised if it could mill aluminium in anything but a snails pace at shallow depth, probably ruining one mill after another because it's more rubbing than milling the material.

Which alloy do you plan on using?

 

[GuilleAcoustic]

Should have streamed this.. Forgot how much I love working in Sketchup

I so know the feeling. I miss designing stuff I'll never build . Just can't find enough time for that.

 

[smitty2k1]

Should have streamed this.. Forgot how much I love working in Sketchup

Do you have any good resources on getting started in Sketchup? I did a lot of Autocad back in high school, and then Pro/E in college, but since then I've been out of the drafting/modeling game. I downloaded and opened up Sketchup and am... lost.

 

[thewizzard1]

I know I've mentioned it before someplace, but I think an adapter plate would be a far better development for MXM cooling.

Making a heatsink from scratch is horribly difficult, considering the energy density of the cards we are trying to work with.

Making an adapter plate, with hand-polished top and bottom surfaces, could be very easy. Your CNC machine can accurately and quickly make holes, and the perimeter cut. Fins, on the other hand, are nearly impossibly hard to make to approximate the level of extruded heatsinks, which is still baseline for performance in design of heatsinks.

I'd be happy to make up a model and send it your way, once I get time.

 

[Kmpkt]

I definitely like the idea of the adapter plate. If you wouldn't mind sending something along that would be great. Something L9i compatible seems like the best solution. Also if you needed a significant offset I am not averse to soldering the plate to the bottom of an L9i to eliminate one of the thermal barriers

@confusis that looks pretty awesome. Will chat more in PM.

 

[Kmpkt]

Have you tried milling aluminium with a Nomad 883?

While I have no personal experience with the machine it seems unfit to properly deal with it. Low spindel speed, questionable rigidity and no cooling - I would be surprised if it could mill aluminium in anything but a snails pace at shallow depth, probably ruining one mill after another because it's more rubbing than milling the material.

Which alloy do you plan on using?

I haven't, but apparently it's capable. 1050 would make sense to me for the purposes of this heatsink as it is softer than 6061 (more easily millable I believe) and has one of the highest conductivities at 229 W/m. K. Anyone who cares to put in a more experienced two cents on this, I am more than willing to take advice.

 

[blubblob]

The properties of 1050 make it poor for machining - it's a great alloy for bending and anodizing though. "Soft" is a quality you want to avoid in aluminium if you plan on milling it as it will result in the material rubbing away instead of being cut. (I'm oversimplifying here, but it's still mostly true.)

To mill a heatsink you obviously need cuts that are both very thin and deep. The ability of the machine to cut a 3mm sheet with a 3mm endmill does not tell us much in that regard - aside from the fact that it does need coolant (which will result in an incredible mess).
Assuming you want to produce the heatsink on that machine you might want to make initial tests on how thin a mill you can make work with the desired length(=depth) without breaking a mill every few minutes - that would then immensely inform the design of the heatsink.

 

[Windfall]

Hey KMPKT-

I use both fusion 360 and autodesk inventor regularly. If confusis doesn't meet your needs (he probably will) then I'd be up to it!

Cheers, windfall

 

[Necere]

Yeah, I don't think trying to mill a heatsink is going to go well. The channels are too narrow and deep, and you're likely to be breaking bits constantly.

Your best bet is probably going to be to stick with one of the established methods of heatsink construction if you want any hope of it being "better" than existing coolers. For a one off you might be able to adapt an existing 1U heatsink or such. For low volume, aluminum extrusion could work, though I wouldn't expect performance to be great as the fins/fin spacing is larger, making surface area smaller.

 

[rook]

Alternate manufacturing methods for the kind of design you have are wire EDM or even 3D printed metals... might not be cheap but they're worth a look.