Direct-Die water cooling

[Daniel ~]

Gizmo, can you throw in a couple of pictures? Then I'll front page it! ":O}

[Gizmo]

Depends on what you get. There are some other folks around here more qualified than I am to advise you on what prebuilt kit you should buy. You see, my entire rig was built from scratch.....

As for water and electronics; I have dunked my system more times than I care to count. The only, ONLY time I had a problem that wasn't cured by letting the mobo dry out was when I had just made some changes to the system and still had some fine copper shavings in the water. More about that here

Daniel ~, I'll see about getting some pictures this evening.

[Gizmo]

Ok, here's some pictures. I don't have any pictures of it out of the case, but you can always go here.

This is NOT a small heatsink.

First is a picture with the 90 mm Vantec Tornado mounted. A 90 mm fan is a waste on this HSF, IMHO, as about 1/3 of the fan hangs off the edges of the heatsink, and so doesn't do anything for cooling it. On the other hand, it DOES generate a lot of airflow around the CPU and VCore regulator, so I guess it isn't all bad.

Next is a picture from the left side. You can see the caps of my VCore regulator. You can also see the ends of the heat pipes. This HSF doesn't overhang the left side of the CPU socket much at all, so there is plenty of clearance there.

Last is a picture from the right side. Here you can clearly see the heat pipe arrangement. This overhanges the right side of the CPU socket by a good 2-3 cm (about 1").

On my NF7-S, the CPU socket lever is on the right side of the picture. For this orientation, you can mount the HSF with the 'U' of the heat pipe to either the left or the right, depending on how your mobo is arranged. If you have a system where the CPU socket is rotated 90 degrees counter-clockwise to what you see here (CPU lever at the top), then you may have a problem, as the HSF MUST be mounted with the 'U' facing downward in order for the heat pipes to work properly. The HSF base is symmetric on the long axis, so you can rotate it 180 degrees and still mount it, depending on how your mobo is arranged.

The unit comes with mounting screws and a back plate to relieve stress on the mother board. However, you MUST mount this thing using the screws; there is no provision for mounting it using the socket clips (I don't think you could get to the socket clips anyway). This means that if your mobo doesn't have the four mounting holes, you are SOL to use this HSF.

As has been observed elsewhere, the base is not particularly well finished; I could feel slight ridges when I ran my fingernail across it, and you can see the ridges in the finish. ThermalRight claims that the base is flat corner-to-corner to within .03" (a little less than 1 mm). Lapping the thing would probably buy me another 3 or 4 degrees in performance.

All in all, it performs well for a HSF, IMHO. I haven't quite made up my mind if I want to keep it for use on one of my other machines or not. If you don't plan on OCing, this HSF with a quieter fan might just be the ticket to a nice system with reasonable CPU temps.

[Gizmo]

Hopefully, I'll have a Danger Den RBX in my hands in two or three days. I'll take some measurements with it as well.

[Daniel ~]

It's up!

[Gizmo]

I have some pictures of the SP-97 out of the case, now.

On thing I forgot to mention about the SP-97. I believe the unit would have performed better if it had been in a horizontal orientation. Unfortunately, my rig is not very, uh, willing to be turned on its side so that I could test it that way.

The reason I say this is because if you look at the first pic below, you will see that two of the three heat pipes are angled, and one is completely horizontal when the unit is mounted in a vertical orientation. This pipe is not going to be able to transfer a lot of heat, since there isn't a significant convection area. One heat pipe will be able to work fairly well (the one on top), but the bottom heat pipe will not be able to work at all, and in the extreme case, would actually transfer heat from the fins to the base (the exact reverse of the way it should work)!

you can see what I was talking about on the base finish in the second pic.

I have marked the third pic with red arrows to show where the HSF fins are in relation to the fan. You can see that roughly 1/3 of the total fan area is wasted on the HSF, although it might have other benefits (such as circulating more air around the VCore regulator.

[Gizmo]

Got my Danger Den RBX Water block in Friday. Finally got a chance to mount it just so I could do some comparisons.

The test equipment is previously documented here

Configuration: Also as documented above. The only exception is that I no longer have the Vantec mounted because I don't have the HSF installed. In addition, I am now running my water cooling rig. I have two 6 1/2" x 6 1/2" (roughly 165 mm x 165 mm) radiators in series, with a 120 mm 65 CFM fan on each radiator. This is all being pushed by a 700 GPH Danner Mag Drive water pump. All tubing and fittings are 1/2" (12.5 mm), including the RBX.

Edit: I tested using the number 1 accelerator nozzle for the RBX.

Here's what I've got with it.

Idle

CPU vcore - 1.824 as measured by 8rdavcore. 1.82 as measured by MBM5. 1.888 as measured with my DMM on the 2v scale, and 1.89 when measured on the 20v scale. These readings are the same as those previously obtained, and I had no reason to believe they would change.

CPU Temp via Thermistor - 36 C as measured by both 8rdavcore and MBM5

CPU Temp via CPU Diode mod - 35 C as measured by MBM5

System temp - 19 C as measured by MBM5 and 8rdavcore

Ambient - 26 C as measured with IR Thermometer

Load

CPU vcore - 1.792 as measured by 8rdavcore. 1.79 as measured by MBM5. 1.874 as measured with my DMM on the 2v scale, and 1.87 when measured on the 20v scale. Again, these are the same as previously measured.

CPU Temp via Thermistor - 40 C as measured by both 8rdavcore and MBM5

CPU Temp via CPU Diode mod - 53 C as measured by MBM5. (Edit: Uh, that's a typo. It should be 43C.)

System temp - 20 C as measured by MBM5 and 8rdavcore.

Ambient - 26 C as measured with IR Thermometer

The Danger Den RBX appears to be a VERY good Water Block, offering nearly as good performance as my own Direct-Die water block. But it still isn't quite as good as my design. It decreased the temp of my CPU by over 8C under the same load, and is a LOT quieter than the SP-97 that I previously tested.

[Gizmo]

Here's some pics:

Danger Den RBX block, top view, side view, bottom view. The base appears to have a good finish. Not mirror smooth, but good enough that I doubt hand lapping would improve the performance enough to be worth the effort.

First pic is of the top.

Second pic is a side view.

Third pic is a tad out of focus (too close; I gotta get a new camera), but is intended to give an idea of the finish. You can't really see the screw reflection, which is not too surprising, 'cause you can't see it that well in real life, either.

Fourth pic shows the block mounted in my system.

The only real complaint that I have with this block is that, when I got it, I also ordered the extra nozzle kit for it. This comes with 4 additional 'accelerator' plates (which are actually nothing more than baffles), and an allen wrench (hex wrench) for removing the screws that hold the top plate on. I stripped the heads of three of the six screws trying to remove them, and ended up having to buy new bolts. As time permits, I will try to do additional testing using the other nozzles.

[Gizmo]

Well, crap. I appear to have killed my XP3200+.

Fortunately, I have an XP2500+ mobile (IQXA) that I've been wanting to play with. Interestingly, this 2500+ shows temps about 5C lower than my XP3200+ did, under the same loads.

That's IQZFA, not IQXA.

[Gizmo]

Got some interesting results from my two XP2500+ mobiles, running on the RBX WB.

Mobile 1 and Mobile 2 both appear to be 5C cooler than my XP3200+ at the same voltages and loads, according to the diode reading, although the thermistor reading is about the same.

Mobile one tops out at 2.600 Ghz at 1.9 v, no matter what I do, it will not go any higher.

Mobile two tops out at 2.640 Ghz at 2.0 v. I believe it will go higher, but I'm not able to get rid of the heat with this RBX fast enough and the chip gets dangerously hot (approaching 60C according to the diode). When I get my direct-die block back on, I'll take a new set of measurements.

[Gizmo]

WOOT!

I got my replacement SCSI RAID controller in today, AND AMI finally agreed to RMA my other controller!

On top of that, Abit replaced my dead NF7-S, even after I told them exactly what happened to it (and paid them $25)!

They are still looking over my dead AN-7. At least they haven't outright rejected it.

[mookydooky]

Great news Giz! At least you don't have to go get another controller. Those can cost more than mobos...