Okay - I have a mini-itx project going right now and I wanted to make custom copper heat block for it. Its not for water cooling or anything but it will be attached to the outside of the case with a heat sink - I know it doesn't sound good - but trust me my question is that my plan is to use a series of copper plates. I was going to 'weld' these togther with some high-silver soldier. Any thoughts on whether the soldier is going to prevent the heat transfering from one end of the block to the other? Also - any suggestions on soldier? Cheers
Normal solder will do - or flux iirc which u smear on and use a heatgun or flametorch to weld them together.
thanks do you think its going to limit the performance of the block at all to have a couple thin layers of soldier there?
For one it depends on how many blocks you are adding together to form a single block, and another thing to consider is how much of the actual solder you use. Think of the solder as thermal compound between a HS and CPU, the finer it is, the better (also the more contact between both surfaces the better) Some solder changes from liquid to tacky or hard in a very short time so be careful if yo do decided to continue with this plan.
Okay - I've done a quick sketch of what I was thinking. The solder is shown in grey and the thermal goop is shown in white. The idea is for it to transfer the heat to the case which is against the vertial part (to the right) is the one thin layer going to hamper this? I was going to use some heatpipes but space is kinda limited
There is a way to use thermal goop instead of solder for the vertical part: Drill some holes in the side of the horizontal part & in the surface of the vertical part, tap the holes, apply the goop and screw them together.
I thought with a good lead free solder that there would be better conductivity between the two with the solder then the goop - any thoughts?
Goop/solder, i seriously doubt youll see that much difference as long as the contact is good with out airgaps. The advantage of goop is that you can take it apart easily and itll squidge out, whereas solder needs to be well covered and compressed when heated id imagine.
Your not going to get good performance atall from that. It is far too long for the heat to travel. it will be like this: |||| |||||||| |||||||||||||||| |||||||||||||||||||||||||||||||| |||||||||||||||| |||||||| |||| Where the amount of | = heat transfer, in relation to distance. you definatly need heatpipes for something like that, well a heat pump of somekind - a TEC would probably do... although heatpipes are a better alternative. Cant remeber whre you can buy heatpipes, or the cost of them, but theres quite a few guides to making them yourself floating around on google...
I've actually shortened it to cover only the cpu. Now its about 40mm long so I think it will be okay. I do have a line on some heatpipe material but wanted to see if this works first
Update and feedback please Okay - so I got the copper block made and installed. I cut it a bit shorter so that it only covered the cpu. It seems to be working a treat but I wanted to run some temps by you guys and get some general feedback to see how you think its working. its in an all aluminum enclosure. With no openings (completely sealed) I get about case: 55deg cpu: 42deg on no load - on load its about case: 68deg and cpu: 55 If I open the pci plate on the back I get case: 55deg and cpu: 40 deg on no load - with load its about case: 60 and cpu: 45 The biggest thing is that with no openings the temp stays up there - for quite a while This is in a moderate room with a load time of about 8 hours (thats 100% cpu for about 8 hours). The only thing that slighly concerns me is the constant temp although it doesn't seem to cause any probs. Any thoughts?
Whats the heatsink on the outside of the case look like? Is it one designed for passive cooling? Those temps are a bit high but so long as your running with no stablity issues then its fine. I once had a CPU running at 91C after the heatsink became unseated and its still going strong 2 years later Its a bit strange that your case temps are higher than your CPU temp though.
I know - the case temps are a bit wacked Its because with all the kit inside - the ambient temp increases to a certain threshold. When the case is sealed then it sits at this threshold because it can't dissapate the heat until the temp rises past that threshold and the aluminum in the case acts as a conductor. At least thast my understanding of it - I may be wrong The heatsink is copper so it abosrbs much of the cpu heat and transfers it directly to the case - thats why its cooler. The case isn't really designed as a passive heatsink (primarily because the ribs on the side are horizontal not vertical. But it does have quite a bit of mass and surface area which I think is helping. So - 50/60/70 deg is fine for inside the case? (and just to confirm - its actually the motherboard temp not the case temp)
So long as the computer isnt crashing its fine. When you get stability issues you know its too hot. I know thats not a very scientific answer but it works and all your components have different tolerances too heat so its very hard to say at what point things will go wrong.
The case temp cannnot be hotter than the cpu temp if the cpu is the only thing heating it. In thermal transfer energy will only move from hot thing to cold thing. If the case is hotter than the cpu, then it will be HEATING the cpu. So unless you have some kind of heater in there, your wrong. You either have your sensors mixed up, or the cpu sensor is not atall accurate (both are very likely...). If its the sockets thermistor your using for cpu make sure its bent up so it will push against the cpu and putting some thermal paste on it also helps. Idle temps mean nothing just so you know. To work out how well this is performing we take the full load temps, measured with the cpu under 100% load, with one of the optimised heat output programs (burnk7 I think its called). This will give about 85% of the total max possible heat output, as detailed in the amd or intel datasheets. For a 1700+ for example, rated at 49W, your temps would from 41.65W of heat because cpu's in normal operation cannot have all the transistors switching at the same time. If you wanted to test with more accuracy you could use a cpu simulator - a device that makes ~50w of heat in a small area, with the rest of the deivce insulated. Taking your max load temp there, of 68c. Take away room temp, which im guessing 25c = 43c dT Divide the dT by the wattage and that gives you a c/w rating off just over 1C/W, which isnt particually good. To improve this you really need heatpipes - if moving heat over a distance, of which in this distance no disipation will be happening atall. The heatpipes would add to the thermal resistance so the cooler on the end of the heatpipes would need to be slightly better (more [effective] surface area) than if it was sitting diretly on the cpu, in order to get the same performance. If your not bothered that much about improving the performance (your cpu isnt going to die at 68c... without vmods anyway) then just leave the cooler as is. I would however do more accurate measureing of the load temp just to make sure it really is 68c.
