Prototype CoolerCase ...a case that cools...

[BaK]

Ok no GPU and no vapour chamber yet, but first results!

This is what I get with 2x Gentle Typhoons 1850 @1200RPM. They are making a little noise at that speed, not dead silent as I like them to be, but I can not lower them more with my fan controller.

I have tested the CPU cooler on OCCT 5.4.1 - 4 threads - AVX2 instruction set, during a 60 min test 100% CPU with 5 min idle before and after.
AVX2 options showed to be the more power hungry scenario.
I did one test with 'Large Data Set' option, another with 'Small Data Set' and a final 'Small Data Set' test with a fixed VCore of 1.05V (vs stock 0.983V @load).

Ambient temp was 21.5°C, here are the records on HWinfo64:

So for a first test, without the help of a vapour chamber to propagate the heat, I am quite satisfied with the result even if I was expecting a little more.
This cooling setup is able to keep an i5 4590 at stock voltage around 80°C at load.
This 84W TDP CPU eats 60W according to HWInfo64 at stock voltage.

With a voltage of 1.05V though, the power used raises up to 70W and the max temp to 95°C. This seems to be the limit of this cooler as it is right now.


I have also monitored the temps on the Stracom HT4 with a temp prob stuck on it, on its side close to the CPU, this is T1.
T2 is a second temp prob on the top of the aluminium cooler.

It's a few points chart only, but I guess it already shows that both the Streacom base and the aluminium heatsink don't follow the temperature raise of the CPU.

I can see three reasons for that:
a) This undelidded CPU is building heat up and do not transfers it properly to the IHS
b) The contact between CPU IHS / Streacom HT4 is not good
c) The contact between Streacom HT4 / Aluminium heatsink is not good

b) and c) seem to be the case according to my pressure tests.
Adding a vapour chamber should improve c).
I will see next if some lapping or delidding is required but I would like to keep both the CPU and the HT4 untouched if possible.


What do you guys think of this? Promising? Not at all? ?

 

[WinterCharm]

I don't actually count on natural convection here, as this is not a passiv system, but on the fans airflow only.

Makes sense.

Any link to the place where you are ordering your cooler?
I'm interested as I first wanted to use heatsink with fins instead of aluminium cooler, but didn't succeed in finding a seller proposing the right dimensions.

I'm in the middle of speaking to design engineers at Celsia, since they do small production runs (0-5000) on heat sinks. Haven't submitted an order with them yet, though, as I'm still in the prototype stage. Not sure if they will make something in the dimensions you require, though. Have you considered a "tiled" design? (small extruded blocks tiled onto a larger plate?)

 

[BaK]

I'm in the middle of speaking to design engineers at Celsia, since they do small production runs (0-5000) on heat sinks. Haven't submitted an order with them yet, though, as I'm still in the prototype stage. Not sure if they will make something in the dimensions you require, though.

Oh I didn't know they were making heatsink themselves, will check that, thanks!

Have you considered a "tiled" design? (small extruded blocks tiled onto a larger plate?)

Not sure to see what this kind of heatsink is, any pic?

 

[BaK]

I finally went with this GPU cooling setup, the previous one was using too much space at the back of the GPU card and thus would have interfered with the PCIe riser or even touched the mobo:

Because of the weight of the copper block, as you can see I played it safe and add a backplate at the back of the card to help supporting the block. A modded CPU bracket did the trick!
The hardest task here was to find the right springs to put in between the backplate and the card, not too big, not too strong but enough of both.
There is a now short 1mm gap above the Direct Power thing, everything fine!

After several tries I have been able to adjust the copper block so that it touches the die properly when attached to the GPU cards. The contact could be improved on on the top left corner of the die, but all in all this time the pressure test looks good to me, unlike the CPU side.

Because of the relatively big size of the copper block, I had in mind to save some of my remaining pressure papers and instead check how the contact was with the thermal paste left on the aluminium heatsink.
However, it turned out that once in contact, there was no way to separate the copper block from the heatsink! Pulling on the heatsink only acted on the backplate springs, I would not have bet on the thermal paste being that sticky!
Anyhow, I concluded that there was a good contact there!



Minus the vapour chambers, the prototype of the CoolerCase mini version is now ready!


I will next check the cooling capacities on the GPU, hopefully getting better results than what I got with the CPU.

 

[BaK]

I have finally been able to test the GPU cooling system efficiency!
OCCT 5.4.1 has been used again, this time with the GPU 3D Shader3 test.
Here under some results:

As you can see, somme thermal throttling occured after less than 10 minutes of load.
With a core voltage of 1.19V, the GTX760 temp rised up till the limit value of 81°C.
To stay under that limit, the core voltage was (automatically) reduced to around 1.1V, which corresponds to 75% of the GPU power.
According to Asus (https://www.asus.com/Graphics-Cards/GTX760DC22GD5/specifications/), the TDP of this GTX760 Direct CU2 is 225W.


So while my early calculations were aiming at cooling a power of 175W, the system right now is able to keep 169W (225W x 0.75) below 80°C.
Quite good I would say, even without the help of an additional vapour chamber! \o/

Temp probes were also attached to the system, but the one on the copper block ended up being defective. The block being not easily reachable, I launched the test without replacing it.
A working temp probe was attached to top of the aluminium heatsink though, showing a max value of 54°C.

That makes a temperature delta of 27°C (81-54) between the GPU core temp and the top of the heatsink.
On the CPU side, the same kind of measurement showed a delta T° of 64°C (95-31), which again seems to point at a bad heat transfer from the CPU die to the heatsink.

Just to check, now that both sides are up and running, I redid the GPU test while the CPU was performing a Large Data Set benchmark on another OCCT. The results were totally similar, which means one side of the case doesn't influence the other side. Good thing!

I am already happy of what have been achieved on the GPU side, but let's see next if all this can be improved with the use of some vapour chambers!