SFF WC-project: remarks & opinions

[ZaVkE]

The subject:
a shuttle SN45G
2500+ Barton @ 2750MHz & beyond
ATI 9500 @ 9700pro & beyond

The torture:
Fitting all of this inside the case so on the outside (execpt from some missing grills it looks basic):

• #rotor design cpu-block 3/8-intake-fitting
• vga custommade block 3/8 fittings
• chipset custommade block 3/8 fittings
• Black Ice Micro 1/2 fittings
• 2 Hydor S20II pumps (480l/h or 2gpm & rotatable outlets)
• TT UFO fan (+80cfm)

The plan:
Making 2 seperate circuits which share the same rad.

How & Why?
The rad will have 2 1/2 fittings to make it a less flow restrictive as possible. Otherwise the small fittings could put a huge strain on the flow.
Using 2 Y-connectors both circuits (on 3/8-tubing) will enter & exit the rad, making them completely seperated from eachother (execpt for the rad offcourse).

Circuit1:
From the rad the a pump seated (looking @ the front of the shuttle) left of the drive-carriage (perfect fit btw, 1 mm space left ) to the cpu-block (going to use shorter fittings to make more space for larger tube-corner). 2 outlets come together in a Y-piece which goes to the Y-piece connected to the rad. Nothing unusual here so far

Circuit2:
From the rad to the pump on the right side over to the chipset-block, to the vga-block and back to the rad. But because most chipset- & vgablocks are either with fittings on top or with fittings that make a 90°corner, i would be forced to make extremely sharp bends and lose a lot of flow.
To counter this i'm planning on making my own blocks for vga & chipset (both on the same principle) with in & outlets on the side (maybe a bit elevated so that the tubing doesn't hit something). That way i will have almost no bends between pump & chipset and chipset & vga, together with a much less restrictive design of the blocks should result in a better flow in circuit 2.
You can best compare the idea to the zalman vga-block.


The idea for my blocks is to have a lot of small fins placed along with the flow over the length of the block with a entry & exit @ the beginning & end. Kinda like a cascade design without the center inlet.

I'm planning to start with either of these 2 blocks:



Still in doubt, but tending towards the copper one.
Then i'll orient them in the way i need them, drill the holes to mount them & remove all the metal that i don't need (excess fins & base). Afterwards i'm planning on putting a plexi top over it all. For the shap of the fins i'm opting for a V-shaped entrance so that there is a more even distribution along the fins.

I think that this design should work nicely for what is is supposed to do and that the overal performance of the system will be nice (altough it's looking into a crystal sphere for the moment).
My biggest doubts are the fan, but i can't seems to find a decent 80mm (or even 92mm) with a lot of cfm and a reasonable dba-level.

Any remarks, questions or opinions are welcome and appreciated

[a5h]

you don't seriously plan to cool ....a CPU block, GPU AND Chipset block with 1 tiny lil Black ice micro?

sounds like u need a dual heatercore.

[TyphooN]

Don't cool the chipset block at least.. to keep some heat off of the gpu/cpu.. is there any way you can get a single 120 heatercore??

[ZaVkE]

Hmmm,

A BIM should be able to handle 275W
Cpu produces max 100W
vga should be max 35W
& chipset should be max 25W

So with 160W i stay well below the 275W DD states.

My main reason to wc is to reduce the temp-fluctuations between load-idle & the case-temp in the shuttle.
I don't need ultra-low temps, just a bit more peace & quiet and not so high temps and i'm happy

Bigger might be better, but is not always needed
And i'm not going to make anything visible from the outside, no matter what. The only thing that will be visible is the lack of fangrills, that will be all.

[Lithan]

1 BIM wont handle 275W, except MAYBE with dual tornados at full... even then I'd doubt it.


And you forgot pump.

[a5h]

a BIM dissipating 275w is a load of crap! Trust me, the BIM won't do. a single 120mm rad. wud be the minimum required.

Good air wud even beat it. Do it right or stick with air ...this setup may give u silence but it won't give u "not so high temps".

[pauldenton]

Originally Posted by a5h

a BIM dissipating 275w is a load of crap! Trust me, the BIM won't do. a single 120mm rad. wud be the minimum required.

Good air wud even beat it. Do it right or stick with air ...this setup may give u silence but it won't give u "not so high temps".

iirc the black ice ratings are at 1gpm and a 40C temp difference between the coolant and the air...... which i doubt any of us who don't live in sub-zero temperatures would be willing to run with....

[Gokou]

the 275w would be the max theoretical.... so ignore that for a start. so lets say the practical limit is 200w.... but also remember thats when the rad is at its best.... and to be at its best it has to be hot.

So er no.. one rad isnt going to do it. im cooling my gpu/cpu/chipset with one 120.1 thermochill in a full case and its just about doing the job with a half decent overclock. in a SFF case ur going to have a lot more resistance and ur flow will be less with 2 circuits.

hmm!

[saaya]

i have a very similar looking heatsink from thermaltake. it has even more fins and they are machined out of the copper base and not soldered onto it.

def go for the copper one!

and if you use 2 loops you will use 2 pumps as well, right?