So, I've frozen CPUs and GPUs quite a bit, but I'm a total noob getting motherboards or memory cold. But, it's time to get my feet wet, and here is my first attempt, somewhat a success, but mostly total fail! :screwy:


Day 1, pre-testing on air cooling. Very good results, but I was scared to give the memory more than 1.75V as these Hypers are expensive and I didn't want to kill them before getting a chance to freeze them.



Day 2 - I started by test fitting my custom block....didn't play with it for long, but I wanted to see how cold I could get it without TEC....you can see on the thermometer the temp outside the window where my radiator is located (-5 Celsius), vs the temp of a probe inside a hole drilled at the end of my aluminum plate (1.6 Celsius).




I made the block from a piece of 1/4" aluminum plate. I countersunk holes for the water block hold down on the bottom (four exposed silver screws).




On the top side, I countersunk 6 holes for the screws which would attach to the DIMMs on the bottom. (the four extra holes were not used for this project).




Next I insulated the memory and motherboard with foam and paper towel. Since I wouldn't be getting too cold, I wasn't too worries about the PCB itself getting cold, so the motherboard doesn't have any protection here.




With the whole assembly mounted, it looks like this. That's a 225W TEC module, and shouldn't have any trouble handling the small heat load generated by the DIMMs, even heavily overvolted.




The CPU and TEC unit are both in a single loop, and the rad is hanging out the window, unfortunately, temps were a little warmer the second night.




When I first powered on, the temps were -2.7 and 19.5 or so. This is a couple seconds after the first power on.




Within about 10 minutes, the temperature settled at about -16 Celsius with 1.75V vDIMM.




The plated had a thin layer of frost within about 20 minutes.




Temps dropped a bit later in the night, this was the lowest I saw with the plate at -21 Celsius, normal operating temps were typically closer to -18 to -19 Celsius.



After about 5 hours or trying in vain to get the memory any faster, this is what the assembly looked like.




But, it was time to disassemble.








Bone dry underneath...I won't be so lucky next time :p Next week, I'll put the memory and the CPU under LN2, so I'll have to be a bit more thorough with the insulation.



A little frosty on the DIMMs.







Why a failure? Well, the first day, pre-testing on air I got the following score:



Going cold didn't help those numbers at all. This does help a little in that I now know that my current limitation is the IMC. I'm planning to get LN2 next week to get the CPU cold, in the process, I should be able to blast past the IMC's current limit and hopefully get some good results. But I've already got a Dominance cooler for the memory as well, so this probably won't be used next week.