Hi All,
Just discovered these forums. Many names here familiar to me. I thought I'd share this water-block that I've been developing here with you guys.
A while ago I developed a concept prototype block that focused on merging the concepts of jet-impingement, micro-channels, and ultra-thin bases, that focussed purely on cooling CPU dies of the P4/AMD size, and making no concessions for peltier cooling performance. That is, I wanted to see if water-block performance had reached a peak since all the current designs are performing roughly the same, or if there was more to be gained.
The "journey" is detailed in a monster thread over at Overclocker's Australia here.
After much research, and development with a very friendly batch of machinists, the near-final blocks look like this:
Note that in this picture the nozzle in the central plate differs slightly from the final design. We've just moved house and I've packed up my digital camera so a more up to date picture will have to wait.
The top plate uses 3 x 1/2" barbs (with 3/8" BSP/NPT threads - meaning that if you wanted to fit 5/8" barbs this would not be a problem), 1 inlet in the middle, through the nozzle in the middle plate jetting down and impinging directly over the CPU core area, then splitting out to the two outlets. The two outlets are merged back together again by an external Y-piece.
The total block dimensions are 80mm x 50mm x 12.5mm (not including the barbs) and will fit a Socket A motherboard through the 4 mounting holes. An additional aluminium mounting plate with longer clamping screws is an optional extra for Intel P4 Socket 478 mounting.
So how does it perform?
Performance of the final design is slightly better than the prototype that I demonstated in the OCAU thread mentioned above. For an AthlonXP @ 1925MHz/2.15v I'm seeing >5C better full load on-die temperatures on my motherboard than with either a properly lapped and mounted Silverprop Cyclone 5, or a properly lapped and mounted Maze 3 (both of which perform about the same being within 0.5C of each other). For a Duron @ 1200MHz/2.15v the difference is 3C, for a P4 @ 2.6GHz/1.8v the difference is also 3C. For me, a better overclock of the CPU was achieved in all cases.
People who have gotten a block off me have reported performance increases ranging from 4-6C over their old waterblocks, which have included Maze 3, Silverprop Cyclone 5, DTek TC-4, Swiftech MCW462-UH for Athlons, and a 3-5C gain vs the Maze 3 on P4 systems.
I've since had a batch of blocks made up and a few people have acquired some from me. There's a results comparison thread over here which seems to back up my own research:
http://forums.overclockers.com.au/sh...hreadid=109948
Basically the block was designed on a thermal simulator I wrote, and I then tweaked it through several revisions via a very tolerant group of excellent machinists who made the block up for me. The size of the cooling performance really surprised me. I had begun to think there for a while that we really had reached a peak in block performance.
Cheers.
Reply With Quote
Bookmarks