Well, where to begin? I guess the beginning.....:shrug:
Anybody remember the glory days of the D-Tek Fuzion? The V1 came out and pretty much rocked everyone's world...it was a great performer with both single die CPUs and dual-die CPUs. The nozzles and washer gave tweakers something to do and test and play with...and ultimately, gave people even better performance. The V2 followed with much of the same design, just modernized in appearance and mounting kit. The performance was a little better and it was fully compatible with the old nozzle kit. With it also came a standard Pro-Mount kit.
The LGA775 Pro-Mount kit was the first great mounting kit for waterblocks. It used 4 captive screws and a backplate...tighten the screws until they bottom out and you're done. You got a consistent, high-pressure mount and you could do it in seconds.
Since then, many, many waterblocks have debuted and the competition has largely caught up and even a few have surpassed the Fuzion V2. Even the great Pro-Mount kit was dethroned by Swiftech's GTZ mounting as best-in-show by using user-friendly thumbscrews.
I tested the Fuzion V2 against the Swiftech GTZ back in early May on a Core 2 Quad, and the much-newer GTZ managed to pull ahead a little bit, though the FV2 still had a strong showing. I haven't tested it on a Core 2 Duo yet (that testbed gets lowest priority due to the fact it's mostly antiquated), but I did test it on LGA1366 with D-Tek's newest Pro-Mount. The results weren't pretty for D-Tek.
While the new Pro-Mount matched the GTZ's awesome mounting kit on all fronts, the performance of the block was really lacking. Or so I thought.
Turns out, through all the the hullabaloo of the LGA1366 Pro-Mount's delayed launch and Fuzion V1 incompatibility, the LGA1366 Pro-Mount was also fatally flawed in basic design. Essentially, it throws like a girl.
I declared the Fuzion V2 as non-competitive on i7 and other reviews didn't show it performed any better. All this time though, the problem hasn't been the block--it's been the mounting system.
I had a hunch it was part of the problem, but at 5 days a test, I wasn't about to engage in any sort of trials to chase a gain that might not exist. Skinnee, in his own tests, thought it might be part of the problem as well. Heck, I'm sure most people who used the Fuzion V2 on LGA1366 probably thought at one point, "this doesn't feel right."
When mounting the Fuzion V2 on LGA1366, there's no resistance when tightening the block down. It's not any sort of WD-40 magic or anything like that...there genuinely isn't any sort of tension when tightening down. When unmounting the block, TIM never had any sort of fractal spread pattern (a sign of a good mount with some TIMs--MX-2 included). Still though, in my mind, the block was just old and showing its age and was no longer worth testing.
Then, recently, D-Tek approached skinneelabs and asked us to try a bow in the block to see if it closed the gap in performance. My new testbed is 98% complete and one of the big advantages of it is that I can completely test a block in a little over a day. If I choose not to do the entire flowrate spectrum of testing, I can get a suite of testing done in half a day. And I had a long weekend....so I took on the task of eeking out a little more performance from the Fuzion V2. Turns out, there's a lot of performance to be gained.
The first thing you test when you want to see what performance gains are available? A control test of course :)
After 5 mounts, I had a good idea how it would perform on my testbed: average core temperatures corrected for 25C water temperature were just under 70C, ringing in at 69.83C.
Then I moved to D-Tek's suggestion: bow the block. I did this by replacing the stock o-ring with a standard #9 o-ring. It's the same overall diameter, but a little thicker. It's a little too thick to fit into the recess of the midplate, so the bow was pretty evident when you held a straightedge to the base. So was the change in performance...just not for the better. Performance got noticeably worse. It was so bad that I stopped after three mounts, I didn't want to waste my time with it any more. Flowrate also went down from 1.77GPM to 1.72GPM. Contact on the CPU went from bad to horrible, with the contact area being smaller than a dime. The average core temperature, corrected for 25C water, was 71.85C. 2C worse, yikes.
Okay, back to the drawing board. Despite the 2C increase in average temperatures, the hottest core only went up a fraction of a degree. Based on that, I figured the bow was something to keep, but that the tiny contact area and general lack of mounting pressure was what needed to be fixed.
I had two possible paths to take to increase mounting pressure: either use a washer (or washers) or swap out the springs. I looked around my 'playroom' a bit and didn't find any washers that fit the bill, so I started to look for an even greater longshot, better springs. Turns out, I got more than lucky...Thermalright did the work for me. When Thermalright released an LGA1366 kit for their vaunted TRUE heatsink, they addressed their mounting pressure issue by using longer and stronger springs. They're 1/8" longer than D-Tek's stock springs and require significantly more force to compress. And they're gunmetal colored too, which is a nice touch.
So I swapped out the springs, and retested...performance got significantly better with the high-pressure mount and the bow compared to both stock and especially the bowed configuration. Flowrate was still 'down' at 1.72GPM, but temperatures also dropped to 68.05C after 5 mounts. Almost a 2C drop over stock and nearly a 4C drop compared to the bowed config, nice.
Contact, visually, still wasn't that great though. Everything but the extreme center of the IHS was pretty thickly clad with TIM. So I thought to remove the bow and just try the stock o-ring in the midplate as well as the high-pressure mount. Results were awesome and made the tests totally worthwhile, with the average core temperatures, corrected for 25C water temps, were a very low 67.46C. That's nearly a 2.5C drop on my testbed...and flowrates were back up to their normal 1.77GPM.
Alright, now that we have all our data, let's put it in a chart:
http://img199.imageshack.us/img199/8263/captureku.png
To put that performance in perspective, I recently tested the famed Heatkiller 3.0 LT and, it's budgeted brother, Heatkiller 3.0 LC on the same exact testbed. If you put 1.77GPM and 67.46C on the 25C water temp chart, you get a datapoint that falls almost exactly on the Heatkiller LC's best-as-tested vertical configuration. It also puts the Fuzion V2 within 2C of the Heatkiller 3.0 LT, regarded by many as the best performing block on the market. Not bad for a block that debuted years ago.
Good enough that I've added it to my Roundup #2 list (being genuinely interested in what it can do with a little bit more tweaking and the nozzles). Good enough that I no longer consider it a block that is a liability of a high-performance loop. Good enough to just be called really good--the old dog still has some life in it.
So how does someone else get this performance themselves? Few easy ways:
1) get some washers and tinker yourself (Home Depot and Lowes should both have washers that work).
2) pick up a Thermalright LGA1366 kit, ~$10. Petra's, P-PCs, Jab-tech, Sidewinder (1, 2), CrazyPC, etc.
3) try some other stuff yourself--turning the bracket upside-down might do the trick as well. :)
For bargain hunters, the Fuzion V2 becomes pretty attractive as well....often you can find them in FS sections of forums for dirt cheap and washers would only be a couple dollars...combine the two and you get a very low pricepoint for a cooler that performs within a couple degrees of the best, not bad. :yepp:
Anyway, thanks for reading another wall of text, I promise real reviews will have more pictures and graphs to digest, but I wanted to keep this one super simple in terms of data :)
