Ideal Block Design for i7

[zlojack]

I've been looking around and there seem to be a lot of different opinions about what would be an ideal block design for Core i7 chips.

Given the compact natue of the die and the heat they generate, this kind of discussion could lead to some interesting new designs, or at the very least lead to some good experimentation.

It appears that the champions on Core2 chips just aren't as effective on i7.

Coolmiester has come up with an interesting proposal of using the Danger Den MC chipset block as it would concentrate the cooling in the small area over the die of the i7 chip. Of course we have the impending arrival of the Heatkiller V3 which is supposed to be very effective for i7, and is a completely different design.

What are your thoughts/ideas/concepts for this new platform and the waterblock requirements?

[RedRaider]

The HeatKilker is not a "revolutinary" design, rather a different take on the EK Supreme.

I'm anxious to see if it prefprms well, as I like laminar flow designs...:)

[Septim]

first question, what size is the i7 die area?

for now we know its still single die, but in the future, we don't know how many dies intel crams in under that IHS...

[RedRaider]

I applaud Coolmiester for his innovative idea.:D

[Kayin]

I got one for ya...

http://img.photobucket.com/albums/v310/KayinStorm/Waterblock21.jpg

http://img.photobucket.com/albums/v310/KayinStorm/Waterblock2-1.jpg

http://img.photobucket.com/albums/v310/KayinStorm/Waterblock2top.jpg

Can't find anyone able to make it. However, it was designed to handle both single and dual dies by utilizing convertible mounts and inlet/outlet configs.

[zlojack]

The HeatKilker is not a "revolutinary" design, rather a different take on the EK Supreme.

I'm anxious to see if it prefprms well, as I like laminar flow designs...:)

I didn't mean it was revolutionary, I meant it was a totally different design to what Coolmiester had mocked up.

I know you're not a proponent of DD, but I think the concept is interesting. Concentrate on the heart of the heat source as opposed to trying to cool the whole IHS.

Septim, I am, of course, speaking about current i7 chips and their single-die design.

Here's a shot of what the core layout is like (source (http://www.sharkyextreme.com/hardware/cpu/article.php/3782516)):

http://www.sharkyextreme.com/img/2008/11/core_i7/chip_2.jpg

[Kayin]

My block was designed for Phenom (also monolith die) but you can see how it would work on i7 very easily. It is also configurable for a center in, dual out that utilizes the center as impingement and blocks the open areas to divert flow into the fins to continue to pick up heat.

However, I'm looking at having to cast it to do anything...

[Septim]

how about adding some micro-channels to a storm2 base?

[Kayin]

I need to post my proposed replacement base for Storm. It's a lot more surface area, since it's based on a different geometric principle.

[Hondacity]

i've revied/tested some waterblocks on my i7 920, ek supreme sorta holds its ground..but k350 sorta kicks its ass...i've got the fusionv2 next..and hopefully the pornstar hk3 comes my way soon...

RRR

when's the eta for the hk3 cu?

[zlojack]

So you've seen some good results from those two blocks.

Do you think the impingement design is still the best? Is there anything about either of those blocks that you think could be better for i7?

[Fitseries3]

http://chip-architect.com/news/Shanghai_Nehalem.jpg



looks like i've roweled up a good bit of talk.

[Hondacity]

So you've seen some good results from those two blocks.

Do you think the impingement design is still the best? Is there anything about either of those blocks that you think could be better for i7?

its only a matter of 1-3c which won't really help much...but only reason i see why impingement designs are better is that theres probably more turbulence in them...vs say a fusionv2....where water just flow...

[zlojack]

its only a matter of 1-3c which won't really help much...but only reason i see why impingement designs are better is that theres probably more turbulence in them...vs say a fusionv2....where water just flow...
The thing is, 1-3C gain is significant.

I look at it this way: Every little thing you can do in your loop in combination that gains you 1-2C here and 0.5-1C there will eventually add up to a solid 5-10C drop in your temps overall.

Say you do every little trick to shave 1C off your temps and you have 8 tricks. There's 8 degrees. That's significant. So if a waterblock gets me 3, a pump gets me 1, a rad gets me 1.5 and my tube routing or tube size gets me 0.5, then all put together, I've gained 6C off my temps.

[Septim]

give a radiator size of 2x120, that small increments of C reduction adds up nicely...

[RedRaider]

When i7 matures to split dies, then I feel like the champions of the 775 quad cores will shine again. At the moment, one of the manufacturers should pony up and design a block that is specifically made to tame the i7.

I hope it is one with laminar flow and high restristion. The sand blasting effect of the Supreme is pure ecstasy...:D

[zlojack]

When i7 matures to split dies, then I feel like the champions of the 775 quad cores will shine again. At the moment, one of the manufacturers should pony up and design a block that is specifically made to tame the i7.

I hope it is one with laminar flow and high restristion. The sand blasting effect of the Supreme is pure ecstasy...:D
Will they go back to split dice?

I imagine that would be insanely hot! Two of these beasts on one chip!

[grumpy]

Probably wont do a dual die until the shrink to 32nm.

[cheapskate]

@Kayin - The mounting holes on your model are inside the O-ring...

[Kayin]

Being fixed.

Talks are in the works about a test run.