How Aftermarket GPU Heatsinks SHOULD Cool VRM

[angelkiller]

Today's video cards come with VRM that put out some serious power and need to be cooled. It seems like nearly every aftermarket GPU heatsink for these cards always have higher VRM temps than the stock cooler. So much for an upgrade.

I came up with this idea while reading X-Bit Lab's AC Accelero Xtreme 5970 review. The two GPUs are cooled by a large heaptipe/fin array and the VRM gets a little aluminum heatsink. While the AC heatsink does beat the stock cooler in most cases, I've seen other products where this doesn't happen. Anyway.

Instead of attaching a separate heatsink to the VRM, AC should integrate the VRM cooling into the main heatsink. This pic shows the bottom of the AC heatsink. And this is how I would change it. The VRM get its own copper base and a single heatpipe that transfers heat to the main heatsink. Make sense?

This way the main heatsink is cooling the VRM than relying on the heated air from the mean heatsink to cool it. The VRM will get much more surface area as well as better airflow with this solution. There are a couple of downsides though. First, this means that every card needs a unique card. So no more 'one cooler fits all'. But that's not so much of an issue because most high end coolers were only designed for a single card anyway. The second downside is that the GPU will have one less heatpipe. But again, this is a minor reduction in cooling ability. The GPU may lose 10% cooling capacity but the VRM cooling capacity may jump ~40%+. A good tradeoff imo.

Good idea? Bad idea? Great solution? Won't work? Discuss.

 

[newtekie1]

I think if they just included copper VRM heatsinks, instead of the crappy Aluminum ones, it would be fine.

Or even better, if a company started putting out copper Mono-Sinks for the various cards, that would be even better...

 

[JATownes]

I think if they just included copper VRM heatsinks, instead of the crappy Aluminum ones, it would be fine.

Or even better, if a company started putting out copper Mono-Sinks for the various cards, that would be even better...

Definate +1. This is the obvious solution, and I can't believe these are unavailable.

 

[AphexDreamer]

Of course there are better solutions, but they will wait a few years before they do it so they can call it new and charge you more for it.

 

[angelkiller]

I think if they just included copper VRM heatsinks, instead of the crappy Aluminum ones, it would be fine.

Or even better, if a company started putting out copper Mono-Sinks for the various cards, that would be even better...

I don't think making the VRM sink out of copper would make significant improvements. Look at reviews of the Ninja copper and the TRUE copper. Hardly a noticeable improvement.

What do you mean by a copper mono-sink?

 

[Athlon2K15]

a single heatsink to cover all the VRMs

 

[newtekie1]

I don't think making the VRM sink out of copper would make significant improvements. Look at reviews of the Ninja copper and the TRUE copper. Hardly a noticeable improvement.

What do you mean by a copper mono-sink?

Of course the difference with a TRUE Copper over a regular TRUE isn't that great, because the base and heatpipes on a regular TRUE are still copper, changing the fins doesn't really do much. The idea is to get the heat away from the heat producing thing as quickly as possible. With a CPU heatsink, the ideal combination is a copper base with copper heatpipes, leading to aluminum fins. The reason this is ideal is weight, as all copper fins are way to heavy. A small VRM heatsink made out of all copper isn't going to weigh that much, and will be far more efficient at pulling the heat out of the VRMs vs. an aluminum heatsink.

And a mono-sink would be a single plate that covers the entire front of the card, cooling every heat producing object except the GPU core itself.

This is a mono-sink, but an aftermarket mono-sink won't have the fan obviously.

 

[overclocking101]

swiftech does it but only for waterblock, IANDH used to do it with the 8800 cards and 48XX cards, I think if someone with the right machine equipment did this they could be quite successfull especially if the sinks worked with waterblocks and air blocks

 

[angelkiller]

Oh one of those, yeah, that would definitely be a good solution.

I still don't agree with you on whether changing the VRM sink from Al to Cu will get signifant difference. But I don't think I have enough knowledge on the subject of heat transfer to support my claim.

It's not just about getting the heat away from the CPU, it's about getting the heat away from the whole heatsink as a whole. The copper heatpipes take the heat away from the CPU, but at some point, all the heat has to pass through the Al in order to get dissipated. It's like having a pipe that's carrying water.. If you reduce the diameter of the pipe, the flow rate is determined by the amount of water the smaller diameter pipe can carry. The larger pipe is Cu and the smaller pipe is Al. Regular heatsinks with Cu heatpipes and Al fins are the large pipe that shrinks at some point. Heat dissipation is determined by how fast the smaller portion of the pipe can dissipate heat. The TRUE Cu shows that if you made the smaller portion of the pipe the same as the larger portion of the pipe, the heat transfer is essentially the same, even though you technically have more capacity. This supports the following statement, "There are negligible improvements when removing the bottleneck of aluminum in a heatsink." And that statement can be applied to changing an all aluminum heatsink to all copper.

My theory of why this is is that copper has a higher thermal conductivity or whatever, but it doesn't trasnfer heat better to the air, which is the only way to improve temps. Absorbing the heat works up to a point. The only analogy I know of is in a WC loop, increasing the amount of coolant in the res will not improve temps. It just slows down the rate that the coolant heats up. At some point the WC loop will become saturated at heat. Performance depends on the transfer of heat to the air.

That was a huge tangent. My bad.

 

[erocker]

Oh one of those, yeah, that would definitely be a good solution.

I still don't agree with you on whether changing the VRM sink from Al to Cu will get signifant difference. But I don't think I have enough knowledge on the subject of heat transfer to support my claim.

It's not just about getting the heat away from the CPU, it's about getting the heat away from the whole heatsink as a whole. The copper heatpipes take the heat away from the CPU, but at some point, all the heat has to pass through the Al in order to get dissipated. It's like having a pipe that's carrying water.. If you reduce the diameter of the pipe, the flow rate is determined by the amount of water the smaller diameter pipe can carry. The larger pipe is Cu and the smaller pipe is Al. Regular heatsinks with Cu heatpipes and Al fins are the large pipe that shrinks at some point. Heat dissipation is determined by how fast the smaller portion of the pipe can dissipate heat. The TRUE Cu shows that if you made the smaller portion of the pipe the same as the larger portion of the pipe, the heat transfer is essentially the same, even though you technically have more capacity. This supports the following statement, "There are negligible improvements when removing the bottleneck of aluminum in a heatsink." And that statement can be applied to changing an all aluminum heatsink to all copper.

My theory of why this is is that copper has a higher thermal conductivity or whatever, but it doesn't trasnfer heat better to the air, which is the only way to improve temps. Absorbing the heat works up to a point. The only analogy I know of is in a WC loop, increasing the amount of coolant in the res will not improve temps. It just slows down the rate that the coolant heats up. At some point the WC loop will become saturated at heat. Performance depends on the transfer of heat to the air.

That was a huge tangent. My bad.

Well, copper displaces heat better then aluminum. When we are talking about something small like a mosfet cooler, I would think copper would make a good difference. Right now I'm trying to find a good copper heatsink for the mosfets on my 5850's.

 

[newtekie1]

The problem is that aluminum can't disappate what it doesn't absorb, which is why copper is needed to absorb the heat. That is why people use forged copper heatsinks like the enzotech ones over the crappy aluminum ones.

If you want a good example of what copper does in an idential heatsink, when the actual base material use switch to copper, along with the fins, check this out:

http://www.tomshardware.com/reviews/10-mammoth-cpu-coolers,1053-17.html

It is a comparison between the CNPS-7700 CU and ALCU. The ALCU, which has a good 2/3 fo the fins and base made of aluminum instead of copper, but still about 1/3 made of copper, gives noticeable temperature differences for the worse. It is showing 6-7°C difference, and I bet if the whole thing was made out of aluminum it would easily be 10°C or worse.

 

[angelkiller]

Well, copper displaces heat better then aluminum. When we are talking about something small like a mosfet cooler, I would think copper would make a good difference. Right now I'm trying to find a good copper heatsink for the mosfets on my 5850's.

No offense, but my whole post was showing how while copper has a higher thermal conductivity, making a Al heatsink out of copper may not make a huge difference.

But in response to what you said, if copper does displace heat better than Al, why doesn't the TRUE Cu perform leaps and bounds better than the regular TRUE? In order for heat to be dissipated, 100% of the heat must travel through some portion of Al at some point.


And if you'll excuse me, I have to take a final now. I'll be back this evening.

 

[newtekie1]

The TRUE doesn't show a huge difference because of the huge number of aluminum fins, they are capable of disipating a lot of heat because there is a lot of them, they aren't being saturated with heat. That isn't the case with a small VRM heatsink.

I'm not saying the solution of a copper heatpipe leading up to the aluminum fins, like your idea, wouldn't be great, or even that it isn't the best solution. I'm just saying that a copper VRM heatsink would be an easier to implement and cheaper to manufacture solution, that would be good enough to give better VRM temps then the stock cooler, while being more likely to implement(and actually it is already avialable right now).

 

[erocker]

No offense, but my whole post was showing how while copper has a higher thermal conductivity, making a Al heatsink out of copper may not make a huge difference.

But in response to what you said, if copper does displace heat better than Al, why doesn't the TRUE Cu perform leaps and bounds better than the regular TRUE? In order for heat to be dissipated, 100% of the heat must travel through some portion of Al at some point.


And if you'll excuse me, I have to take a final now. I'll be back this evening.

Because both use copper heatpipes and a large CPU heatpipe cooler works differently than a direct contact heatsink. I currently have an aluminum mosfet heatsink on my 5850 and I'm not very happy with it, temps kinda suck. I have some Enzotech's on the way and I'm pretty confident I'm going to see a considerable difference. I know they worked well on a 4850 with very hot VRM's. Copper is denser.. there's more mass to it. This will work better when you are limited with space/size. Good luck with your final.

+1 Newtekies post.

*I found some old Enzotech ramsinks that I'm going to try tonight on the VRM's.

 

[BababooeyHTJ]

I don't think making the VRM sink out of copper would make significant improvements. Look at reviews of the Ninja copper and the TRUE copper. Hardly a noticeable improvement.

What do you mean by a copper mono-sink?

Those little enzotech MOS-c1 sinks cool about as well as the big uni-sink thats on my GTX280 now. Those vrms run notoriously hot too.

You have to remember that the base and heatpipes of both the True and the True copper are both 100% copper except for the plating.

 

[LAN_deRf_HA]

I think integrating it into the cooler like that won't be good business wise. You'd need a different version for every card.

It occurs to me the size of the vrm is small enough to fit machine made diamond blocks. From what I last heard it only costs them $5 to make 1 carat, and that was years ago. It might be better now. Diamond might even sink so well that they don't need to cut it into fins, further reducing cost.

 

[newtekie1]

I think integrating it into the cooler like that won't be good business wise. You'd need a different version for every card.

It occurs to me the size of the vrm is small enough to fit machine made diamond blocks. From what I last heard it only costs them $5 to make 1 carat, and that was years ago. It might be better now. Diamond might even sink so well that they don't need to cut it into fins, further reducing cost.

Well in the case of the cooler in the example, it is only made for a single card anyway.

 

[LAN_deRf_HA]

I noticed in that review the second gpu ran cooler with the after market part, is that because of minor contact differences or proof that they really do use better binned parts for the second gpu?

 

[rangerone766]

i'm just parroting what i've heard. so i don't know the physics of it, but its reasonable.

copper soaks up heat and and can move that heat faster than aluminium. it takes more heat to change copper's temp than al's. but aluminum can radiate that heat to space faster than copper can.

thats why copper heat pipes with alumium fins are so prominant. if all copper really was that much better, more companies would make them. because we would buy them.

 

[largon]

i'm just parroting what i've heard. so i don't know the physics of it, but its reasonable.

*braak!*
*smacks the parrot with a frozen trout*

copper soaks up heat and and can move that heat faster than aluminium. it takes more less heat to change copper's temp than al's. but aluminum can radiate that heat to space faster than copper can.

Fixed.

 

[newtekie1]

Aluminum can't radiate heat faster, copper is better in every way than aluminum in terms of heat absorbtion and radiation. If copper radiated heat worse than aluminum then the TRUE Copper would actually be worse than the regular TRUE, but it isn't worse, it is better. Aluminum is used for two reasons, and two reasons only, to save weight and cut costs.

 

[angelkiller]

Sorry for the late response, the last 48 hours have been extremely busy. And thanks for wishing me luck on my final erocker. I already took my math and chem finals so the one I took on Friday was easy. It was basically an Intro to Computing class. We learned the different types of wireless standards, unix commands, how to make simple webpages, etc. Complete waste of my time. :shadedshu But I placed out of it for the fall, so I guess that's good.

The TRUE doesn't show a huge difference because of the huge number of aluminum fins, they are capable of disipating a lot of heat because there is a lot of them, they aren't being saturated with heat. That isn't the case with a small VRM heatsink.

I'm not saying the solution of a copper heatpipe leading up to the aluminum fins, like your idea, wouldn't be great, or even that it isn't the best solution. I'm just saying that a copper VRM heatsink would be an easier to implement and cheaper to manufacture solution, that would be good enough to give better VRM temps then the stock cooler, while being more likely to implement(and actually it is already avialable right now).

I totally get what you're saying. Especially about the aluminum not being saturated. With a smaller heatsink, I can definitely see how a copper VRM sink would improve temps. I'm not totally convinced, but I can see it happening.

I would agree that a copper VRM sink is a solution that companies could easily implement. But I also think that integrating the VRM into the main heatsink can offer even better results than a copper heatsink. The main heatsink gets much more airflow and you could probably get more surface area for heat dissipation than with a standalone heatsink.

Because both use copper heatpipes and a large CPU heatpipe cooler works differently than a direct contact heatsink. I currently have an aluminum mosfet heatsink on my 5850 and I'm not very happy with it, temps kinda suck. I have some Enzotech's on the way and I'm pretty confident I'm going to see a considerable difference. I know they worked well on a 4850 with very hot VRM's. Copper is denser.. there's more mass to it. This will work better when you are limited with space/size. Good luck with your final.

+1 Newtekies post.

*I found some old Enzotech ramsinks that I'm going to try tonight on the VRM's.

I would be interested in seeing those results. I don't think your experiment will be able to prove/disprove my point, but they will be interesting.

Speaking of an experiment, this would make an interesting one. We need two identical heatsinks made, one out of aluminum and the other out of copper. They need to be attached to a small heat source that can put out ~30W or so. A groove should be cut out for a temp probe at the junction of the heat source and heatsink. The temperature difference should be recorded.

You have to remember that the base and heatpipes of both the True and the True copper are both 100% copper except for the plating.

I understand that that, but I don't exactly get what your point.

I think integrating it into the cooler like that won't be good business wise. You'd need a different version for every card.

It occurs to me the size of the vrm is small enough to fit machine made diamond blocks. From what I last heard it only costs them $5 to make 1 carat, and that was years ago. It might be better now. Diamond might even sink so well that they don't need to cut it into fins, further reducing cost.

I noticed in that review the second gpu ran cooler with the after market part, is that because of minor contact differences or proof that they really do use better binned parts for the second gpu?

I addressed this issue in my first post. My solution does require that every card have their own version of the heatsink. But this isn't a huge issue, becasue most heatsinks for high end cards are already like that.

Diamond for cooling? Never heard of that one.

 

[stevednmc]

seems to me that when you have a surface area issue, ie small, the material of the heatsink WOULD make a greater difference, say in the VRM heatsink area. But, when you get into cpu cooling, most of the dissipation is handle by surface area, and more surface area gets heavy. Therefore aluminum is better, its lighter, and gives the opportunity for more surface area for DISSIPATION of heat. Having the base of the sink in copper allows better conductivity of heat to the larger surface area. Just my opinion and understanding of sciency stuff tho...

 

[angelkiller]

seems to me that when you have a surface area issue, ie small, the material of the heatsink WOULD make a greater difference, say in the VRM heatsink area. But, when you get into cpu cooling, most of the dissipation is handle by surface area, and more surface area gets heavy. Therefore aluminum is better, its lighter, and gives the opportunity for more surface area for DISSIPATION of heat. Having the base of the sink in copper allows better conductivity of heat to the larger surface area. Just my opinion and understanding of sciency stuff tho...

Yeah, I think that's what newtekie1 was getting at. With CPU heatsinks, the ratio of heat to surface area is so large that changing materials doesn't matter much. But when the ratio is much smaller, the material really makes a difference. A CPU heatsink with Al fins may not reach the limits of Al in terms of thermal conductivity, but a small Al heatsink on a really hot VRM certainly may, which is where copper could help. I think you've convinced me on that point now. All things equal a copper VRM sink should offer noticable improvement over an aluminum one.




But the thread was about integrating the VRM cooling into the main heatsink!

 

[stevednmc]

Well, said, that was exactly my point. Thats part of the reason cars went from copper to aluminum in radiators..weight, cost, and not much loss in efficiency,Esp when you can pack more fins together for the same weight for more surface area.