Making your own cooling?

[Tudz]

G'day all,

I've been looking around at coolers for 4870s and have just been thinking, all these coolers are is a plate to get heat off the chipset, copper heatpipes (optional) to get heat away from the card, then aluminium or copper plates to disperse the heat and maybe a bracket for a fan.

Sounds pretty simple to me, has anyone tried making one themselves?

Its possible it may end up more expensive, but I have access to a lathe to make spacers, mill to make any block type things, tap sets for screws, grinders and relatively cheap aluminium (plates at least, haven't checked rods) so was thinking of giving making a cooler a bash. It means I could make something that takes the air sucked in from outside the case by one fan and then exhausted by another case fan, something most coolers can't do (they usually suck hot air from inside the case and often just blow it back into the case again).

Any thoughts?

 

[Kaldskryke]

Sounds pretty simple to me

Well, yes the whole concept behind heatsinks is very simple. You want to transfer heat from the GPU to a (re)movable material such as air or water. The more surface area there is between the GPU and the air/water, the better the heat transfer. That's why fins are so popular, because they have huge surface area for relatively low cost.

You'll just have to be very careful with joining your fins, because any discontinuity in the metal forms a thermal resistance. A mechanical connection might not be good enough, you'll have to solder things. Extruded fins obviously don't have this problem, but you probably will.

I'm interested to see if you can get good results, but I have a feeling that you'll likely incurr a huge cost for a relatively inferior cooler. A [H]ard project, for sure, but if you're interested in the best temps I'd just get a heatpipe based cooler from the store.

 

[Tudz]

Yeah, I'm not sure if I can be bothered attempting it right now, but if I were to do it, I'd use the stock cooler's interface to the chipset (as it'll be hard to machine such a smooth finish). Mill down the old cooler to a plate, mill grooves for the heatpipes. Make heatpipes from either copper or aluminium rods which lead away to the cooling fins. The pipes would pass straight through holes drilled in the fins. The fins would be seperated by aluminium spacers turned up on the lathe.

Getting it all to fit together would be tough, you can make a press fits on the lathe, but its not easy. Was thinking of maybe welding the spacers to the fins (they'll both be aluminium, which is a bitch to weld, I've only welded steel in the past) then using thermal paste on the heatpipes which pass through the spacers. Soldering might transmit heat better, but welds fuse the metal itself, not sure which would be better in the end. Could possibly heat treat the whole thing at the end to get rid of any grainular dislocations caused by work hardening or welding which might reduce thermal conductivity, but I dont know if I have access to any sort of heat treatment facilities.

Thinking about it all like that, yeah, it'd probably cost at least as much to make as it would to buy it, even with my access to cheap materials and machinery.

The main advantage would be to have it positioned in a position to promote better airflow so its sucking cool air from outside the case and exhausting the air again (currently I'm working on some ducting for my Sonata III case to suck air from the front directly on to the video card, which the video card blows out the back).

 

[JimmiG]

I agree, the theory is simple - the workmanship, quality and finish are very important.. This is part of the reason why the 4850 cooler is so ineffective - people who remove it and replace the thermal compound with a thin layer of AS5 or similar get much lower temps because heat transfer is improved. If you're very good with the tools, you can probably make a pretty good cooler - many cheaper coolers that are mass produced aren't that great quality wise.

 

[vengence]

You won't save money by making one your self. Period. If that is your goal, stop right now.

You CAN make a better air cooler than what is on the market. However it will be BIGGER, and cost MORE, than what is on aftermarket cooling market. You'll need to know your way around heat transfer though for fin spacing. To close together and you'll hurt your preformance. Fin spacing is largely based on the speed of the air moving over them.

But like I said, it's not going to be cheaper. It's not going to be smaller. The people who make aftermarket coolers for video cards don't have engineers that are smarter than the ones who make the OEM coolers. They have a bigger budget. Saving 10$ on a cooler isn't a big deal to you, but that is a lot of money on a 200$ graphics card in terms of profit margin.

You could make a cooler that cools off 1000Watts with only a few C difference from ambient. You'd just need a spare bed room for the Heatsink to fit into.

 

[Tudz]

Well I've been meaning to learn 2D computational fluid dynamics, maybe I could try it on the heatsink I've got the software on my computer along with tutorials, just never tried to use it before (anyone here used Fluent before and could give some tips? ).

You could make it cheaper than buying it, but that's not my motivation ($50 is less valuable than all the time I'd spend making it). I could pick up a large sheet of Aluminium for the fins for almost nothing, and I have free access to all the tools I mentioned above (including a TIG welder).

My motivation would be creating an beautifully cooled system that doesn't require water or 5 million fans running at 110dB to cool it (exaggeration of course ). And also to prove that all the fans in modern "power" cases are overkill, but they're simply set up poorly. Sure a fan might be able to move XXX amount of air per minute, but I guarantee most fans in PCs are just cycling 90% of that air around themselves, not actually moving it to where its needed.

 

[JimmiG]

Sure a fan might be able to move XXX amount of air per minute, but I guarantee most fans in PCs are just cycling 90% of that air around themselves, not actually moving it to where its needed.

I think you would have more success if you built a custom case, then. Something with ducts and thermal zones and a clean layout, to get the air to where it's needed using slow rotating, low noise fans.

 

[vengence]

Make heatpipes from either copper or aluminium rods which lead away to the cooling fins

I didn't catch this the first time I read through. You might do some research on heat pipes, cause it sounds like you think they are just a copper rod. They're not. Wiki Link: Heat Pipes

Well I've been meaning to learn 2D computational fluid dynamics, maybe I could try it on the heatsink I've got the software on my computer along with tutorials, just never tried to use it before (anyone here used Fluent before and could give some tips? ).

I've done very little in CFD. Most of my work is on nonlinear diff eq in simulink.

You could make it cheaper than buying it, but that's not my motivation ($50 is less valuable than all the time I'd spend making it). I could pick up a large sheet of Aluminium for the fins for almost nothing, and I have free access to all the tools I mentioned above (including a TIG welder).

Including your time, you'd never be able to

My motivation would be creating an beautifully cooled system that doesn't require water or 5 million fans running at 110dB to cool it (exaggeration of course ). And also to prove that all the fans in modern "power" cases are overkill, but they're simply set up poorly. Sure a fan might be able to move XXX amount of air per minute, but I guarantee most fans in PCs are just cycling 90% of that air around themselves, not actually moving it to where its needed.

If you want something quite, the big thing is going to be increasing the size of the fans. A larger fan can move more air with less noise.

If I were you, I'd design something you can build in CFD, then start tinkering. You might also model the airflow of your current case as a comparison.

 

[Direfox]

If your trying to build a better cooler for a 4870, my suggestion would be to start out with a decent CPU cooler and see what you can do to adapt it to the Video card. Finding a way to link a plate and a CPU cooler base might work for you and would deliver awesome cooling, but you would most likely have to design some kind of support for it as CPU coolers can get pretty darn heavy.

 

[Potatocodone]

I haven't encounter anyone actually make an air-cooled heatsink but on bit-tech forum I see people making water blocks all the time.

 

[NoNRG]

Heres a guy making a heatsink out of silver for its superior thermal conductivity. He plans to use it for air cooling. No update for a couple months though.

http://hardforum.com/showthread.php?t=1292712&highlight=silver

 

[Tudz]

I didn't catch this the first time I read through. You might do some research on heat pipes, cause it sounds like you think they are just a copper rod. They're not.

I didn't realise that's how they were made actually! Though I have done some fluids, so it makes sense that's how they're done. I could make that still (not with a wick, but a gravity one... actually depends what the wick material is and how much it costs). But it'd be much harder, not sure if I could seal it closed at a low pressure.

If you used water or ethanol in the heatpipe, it'd be possible to calculate the pressure required to lower the boiling point to a usable temperature, then figure out the volume change required to create that pressure and use something like a syringe to get that pressure in the copper rod... no idea how you'd seal it off to maintain that pressure though. Doesn't vapour slowly work its way through materials, even metals, to reduce the pressure over time? How to seal it though, mmm, no idea. Could make a jig with a temporary seal to weld it shut, but it'd be hard to know whether the seal breaks under the temperature of welding, not to mention what'd happen to the liquid inside under those temperatures.

Either way it'd be a bitch to make and whilst it'd be fun to try, I dont think I'd be able to make good quality ones and I dont really have the time to waste on it right now (well maybe I could, there's too many things I dont know how to achieve though, would need to talk to someone more experienced, could ask some of the guys I know in mechanical engineering).

The idea of using a CPU cooler and making a mating plate for the GPU actually sounds attractive and very simple. I could probably just machine a big beastly CPU cooler down to the size needed for the GPU, and make a mounting bracket for it.

Some of those passive CPU coolers are massive beasts, wonder how they'd go on a GPU?

 

[flogge]

When your talking about mating a CPU cooler to a vid card I think you need to watch the mass of the cooler closely. A CPU socket with its mounting holes and motherboard screwed to the tray is alot more structurally sound than a video card held by 2 adjacent sides one being the PCIe slot.

More often than not there is a back case fan right beside the video card so if you could use heat pipes to get your fins attached to the exhaust fan I think you would improve your cooling greatly and eliminate the fan attached to the card itself.

 

[vengence]

If your trying to build a better cooler for a 4870, my suggestion would be to start out with a decent CPU cooler and see what you can do to adapt it to the Video card. Finding a way to link a plate and a CPU cooler base might work for you and would deliver awesome cooling, but you would most likely have to design some kind of support for it as CPU coolers can get pretty darn heavy.

Heavy OC'ed CPUs are only putting out 150 watts, Heavy OC'ed GPUs are putting out 300 watts.

But then again, GPUs run much hotter than CPUs for a reason.

I would suspect if you mounted a 120 Ultra on your videocard you could see some nice temps. HOWEVER, structurally supporting that.... would be another story.

 

[vengence]

<snip>Either way it'd be a bitch to make and whilst it'd be fun to try<snip>QUOTE]

Some people have had tried harvesting them out of coolers and rebending them. You have to bend them on a 1/2 circle shaped groove so the pipe doesn't collapse.

Some of those passive CPU coolers are massive beasts, wonder how they'd go on a GPU?

See my post above.

could ask some of the guys I know in mechanical engineering

Your talking about a really speciallized field. I am a ME and wouldn't be able to design one without a LOT of research.

 

[Kaldskryke]

Dude, making heatpipes from scratch is a nearly insurmountable task. If you really want to reduce the noise of your system (which seems to be one of your primary objectives is to reduce fans) you'd probably be better off getting better fans, a nice big retail heatsink, and maybe modifying your case airflow patterns. With the right fans in the right spots, you can get pretty good airflow at 5V.

 

[Tudz]

HOWEVER, structurally supporting that.... would be another story.

That'd be the easy part Making a nice solid mount isn't too hard.

Some people have had tried harvesting them out of coolers and rebending them. You have to bend them on a 1/2 circle shaped groove so the pipe doesn't collapse.

Yeah, I've bent big things before (1" roll cages and such). Never anything copper though, and generally using heat.

Your talking about a really speciallized field. I am a ME and wouldn't be able to design one without a LOT of research.

Yeah I figured you were an engineer of some sort I'm currently 2nd year Aero Eng and helping with research projects at my University, I figured one of the people I work with might have an idea. Could be wrong.

Dude, making heatpipes from scratch is a nearly insurmountable task. If you really want to reduce the noise of your system (which seems to be one of your primary objectives is to reduce fans) you'd probably be better off getting better fans, a nice big retail heatsink, and maybe modifying your case airflow patterns. With the right fans in the right spots, you can get pretty good airflow at 5V.

That's pretty much what I'm doing at the moment, with a bit of work I used a front case fan (Antec Tricool) to drop my GPU temperature 10 degrees on medium, and 6 or 7 degrees on low (stock GPU cooling on an 8800GTS, just making sure it gets cleaner airflow). Thats compared to just attaching the fan where its "supposed" to go. Unfortunately its not super quite because the fan isn't well secured and vibrates too much (its possibly a fault in the fan too, its way louder than my other tricool fan). That was without any sort of analysis, just setting it up with the thinking stuff like "that looks like a better position than stock" and "a duct through here would reduce separation over this plate". With a bit more thinking, analysis and better manufacturing I could improve on that without touching the GPU and CPU coolers. Those case fans even on low are capable of moving a decent amount of air, its just they're usually set up wrong so end up only moving a fraction of what they're capable of doing. I recently built my friend a new PC with an antec 900 case and was amazed how it had such poor airflow around vital areas compared to my modified sonata III case, despite the fact the sonata III has only got 2x120mm case fans with only 3 air inlets/outlets and the 900 has, what, 3x120mm + 1x200mm fans and the whole case is perforated.

Originally this thread was just a "what if?" and to see if anyone had tried it. As things went on I actually starting thinking more about how I'd actually make it. I agree making heatpipes would be pretty much impossible and I probably wont even try... but obviously someone made them to start with, so it surely isn't completely impossible, just thinking how I could do it with the tools at hand. Controlling the pressure and temperature inside while sealing the tube would be totally trial and error, mostly error because there would be no way of knowing if it worked.

 

[AltTabbins]

I've actually done it a few times. I have a friend who works for a company that is in the oil drilling business and we have done a couple custom side jobs back in the 6800gt days.

There are so many factors that you dont even think of. Perfectly flat contact point, fin distance, fan mounting, air circulation etc are all make or break factors. We attempted 6 or 7 total and only one of them worked enough to drop the temp a couple degrees.

Edit: and they weren't even heat pipe solutions. Just copper heat syncs.