SUB-ZERO ExodusFlow Altauna CPU Water Block Review

Water Cooling by KeithSuppe @ 2007-07-08

Watercooling.de has had design and manufacturing input in several water blocks they\'ve sold online over the past five years. My first introduction to German engineered water blocks was through this store when I tested the original CF1 Microstructure block and later with the “upgrade kit” compared it to popular USA system Dangr Den TDX vs. CF1 Rev.2. Watercooling.de\'s recent arrival has already received high praise from other testers, so today I test the new SUB-ZERO ExodusFlow Altauna, will it earn my “Brass O-Ring”?

Introduction and Differences between US <> EU WC

Watercooling.de

This online shop has been providing Europe with some of the very best German engineered water blocks for over five years. My first introduction to German engineered water blocks was through this store when I reviewed the original CF1 Microstructure block and later with the "upgrade" kit compared it to popular USA system Danger Den TDX vs. CF1 Rev.2. In addition to all the water cooling kit the store offers a wide selection of PC-hardware.

Let’s take a few moments to examine the differences between Euro and USA H20-cooling.

Madshrimps (c)


I realized rather quickly a water-cooling system based on 1/4" or 10mm tubing offered many advantages. These attributes went beyond ergonomic appeal such as ease of installation and minimal clutter. With 1/2" systems tubing is cumbersome, difficult to manage and based on a poorly conceived connection system, a propensity for leaks. Unfortunately 1/2” systems utilize barbs on each component over which the tubing is stretched and then fastened with plastic fasteners. Once pressurized 1/2" tubes become very heavy and in some cases the slightest twist or tension can dislodge the tube at the connector. If you were to query more experienced water-cooling Enthusiasts, you'll discover most eschew included hardware and they replace it with stainless steel "worm drive" hose clamps as shown below:

Madshrimps (c)


In contrast 10mm lines insert into their perspective connectors either locking into an O-ring or sliding over a small barb over which a metal collar is tightened down. This design is practically leak-proof and can withstand much more wear and tear than its 1/2" counterpart. At the very least water-cooling experts should be including a version of the stainless hose clamps with their kits. Many makers such as Swiftech are now using slightly thinner walled tubing which has the flow rate of 1/2" without the bulk. Die hard supporters of 1/2" system rebuttal remains; "larger tubes = greater flow = lower temps." This is a generalization which needs qualifying and we won't debate that here, suffice to say the CF1 out-performed several top performing 1/2-ID based blocks.

Madshrimps (c)


The original CF1 much like the new Sub-Zero begins with a 0.5mm thick copper base; however this is where the two blocks differ as the CF1 utilizes an extruded pin design. Sub-Zero improves on this design replacing the micro-pin with a cross flow channel design. With micro-pins flow-rate is essential and I've tested some blocks where an increase in flow rate actually resulted in higher temps. In the Sub-Zero channels with sharp angles are machines into the base plate which effectively creates a "cross-flow" design. Such a design can benefit from higher flow-rates. This is where base-plate thickness plays an important role. Unlike air-cooling in which the copper mass needs to be large enough to accommodate more surface area, within a water block the coolant is the catalyst attracting heat through that base. A thicker base equates to mass which is counterproductive, the best analogy might be to think of Direct-Die water-cooling.

Besides thickness the CF1 base plate was unique because of its minimal footprint, something I came to appreciate for its thermodynamic benefits, the simplest being the coolant is concentrated right where it needs to be. The block has a Derlin (PTFE) core containing threads for inlet/outlet connectors and below the inlet a copper plat is seated. The plate has any small holes drilled through it to force the water into mini-jets, seen below.

Madshrimps (c)


The CF1 remains a classic although relatively unknown in the USA. Shortly after the CF1 was cooling Socket-A and 478 CPUs the world over, an "improvement" was released. This introduced the CF1 Rev.2 which I reviewed in my article Danger Den TDX CF1 V.2 comparing both blocks using the same pump, 1/2-ID tubing and radiator. I used a 1/2-ID to 10mm adapter for the CF1, which doesn't replicate flow through the block as it was intended, however; it's the closest one can get to eliminating variables. Below we compare the major change made from CF1 to Rev.2

Madshrimps (c)


The CF1 Rev2 wasn't my favorite because it eschewed the copper inlet plate for a Lucite center with threading for connectors and then small holes drilled to replicate the mini-jet effect the copper plate once performed. It was an obvious cost cutting method as the block became more popular; Lucite is less expensive then copper and Derlin. The CF1 Rev.2 changes served the maker rather then the end-user although they did try to improve ergonomics drilling a receptacle in the Lucite to accept a LED.

Madshrimps (c)


The Sub-Zero is a different breed of 10mm animal ->

Sub-Zero ALTAUNA ExodusFlow Close Up

Sub-Zero ALTAUNA ExodusFlow

- Product Page

My initial tests on the Sub-Zero began with the same E6400 mounted on Asus P5W DH Deluxe. I was pleasantly surprised this CPU cooler performed on par with some popular high-performance 1/2-ID based USA kits and matching water blocks. Watercooling.de's description (German) of the Sub-Zero (English) can be summarized quoting two terms "wide microstructure" and "high flow" which we'll describe in detail shortly.

Madshrimps (c)


As seen above and below the block's surfaces which include a brass top and copper base are finely polished. It's size belies it's cooling ability and from an ergonomic standpoint it is the antithesis of today's towering heat-pipe coolers. These air-cooled monsters not only stress the socket, for all their weight and mass they can't touch the performance of quality water-cooling.

Madshrimps (c)


The Sub-Zero measures 13.5 mm x 50mm x 50mm (H x W x L) and weighs in at a feathery 250-grams. The block arrived with several mounting kits and two 10m/8mm (external/internal) connectors. The block has a small footprint and just covers the IHS it sits upon. As I stated earlier this design requires very little copper leaving a minimal amount of material between the heat-source and coolant. The finish on block is top notch and this includes all surfaces.

Madshrimps (c)


The finish on the base plate was absolutely perfect and about as flat as can be determined by placing a drop of water on the center of the base plate and then placing a sheet of glass on top. If the water spreads evenly to the edges and you can literally pick-up the heatsink with the glass it's flat.

Madshrimps (c)


As I stated in the introduction, Sub-Zero departs from the micro-pin design of the CF1 Rev.2 as both the original and Rev.2 versions of that block used a small channel under the inlet which re-routed water flow directly over the micro-pins. Such a channel tends to restrict flow and is usually an ad hoc alternative to moving the inlet directly over the center of the water block. Water re-routed to flow directly down onto the center of the base plate is known as impingement. There are many variations on this and Danger Den's RBX reviewed here was such a design. The RBX was based on 1/2" and benefited from a high delivery height /flow rate pump. The Sub-Zero (as with most 10mm designs) is very efficient and its cross-flow design is the least restrictive. As seen below Sub-Zero utilized deep channels with severe angles to generate turbulence and slow flow-rate just enough for thorough heat absorption. Turbulence ensures more water molecules absorb more heat. This is one reason why increased flow can improve performance in this type of design.

Madshrimps (c)


Another user-friendly attribute of the Sub-Zero are its many mounting options. Simply loosening four bolts disassembles the block for quick mounting-bracket replacement. Below we see hardware supporting desktop Socket versions 775, AM2, 754, 939, and 940. Also available are 478 and Socket-A.

Madshrimps (c)


Onto the test-systems ->

Correct WC Components Are Vital to Cooling

Matching components:

I originally tested the Sub-Zero in the NZXT ZERO mounted on the Asus P5W DH Deluxe where it played a crucial role in the article entitled Asus P5W DH Overclocking. The test system was 100% benchmark stable to 445FSB and stable in all but a few benchmarks at 460FSB.

Madshrimps (c)


For this review I re-mounted the Sub-Zero onto the Gigabyte GA-P965-DS3 motherboard in hopes a higher overclock could be reached. In addition to the P965-DS3 I've mounted the entire system into my new test-case, Thermaltake's Mozart TX.

Madshrimps (c)


One of many advantages to this case is the ability to remove both side-panel "doors" resulting in a work-station like tower. This effectively eliminates any temperature differential which exits between ambient room temp and a sealed case with a higher internal temp. The Mozart TX will be reviewed here soon. From the rear you can see four 120-fan vents allowing Mozart TX to support some of the larger radiators out there 2x2 stacked.

Madshrimps (c)


The radiator used in this review is Alphacool's NexXxos PRO-II. While it’s difficult to argue in a cohesive system which component is most important, I've found the radiator (heat exchanger) / fan combo, has greatest affect on temperature. The semantics behind the name heat exchanger say it all; this is the location where H20 carrying energy from the processor will dissipate that energy (heat) into the environment.

Madshrimps (c)


Matching fans are equally important and choosing your fan is where you can have a direct impact on water-temps with the least cost. If you’re looking for silent operation then you’re going to sacrifice a slightly higher temp for peace. The obvious answer is to find very powerful (high CFM) fans and use a rheostat or saving more money simply plug the fans into your mainboard headers. Just about every motherboard on the market has a BIOS fan controller and most offer software giving you fan speed control through the Operating System. The only way I know of combining ultimate silence with temps close to ambient would be a large (usually very large) passive radiator. I tested Alphacool's CORA 662 and found it to be competitive with dual and even triple radiator systems depending on the fans. I chose two Papst 4412 2/GL's which move 74CFM each using 12V with a noise rating of 26dBA (ISO standard measured, except 40+dBA ~ @ 50cm in real life).

Madshrimps (c)


Finally the pump which is another source of contention as to which component is most critical as part of an H20 system. This argument is easily answered when you compare an increase (or decrease) in flow-rate and/or height delivery compared to installing higher CFM fans. Of course as with most technical subjects the argument becomes more complicated when you throw in a water-block based on 1/2-ID tubing and a mini-jet design. For 10mm system testing I've been using the Alphacool AP1510 centrifugal pump with external DC adjuster from 12V ~ 24V in 4V increments. The pump is made specifically for Alphacool by OASE.

Madshrimps (c)


This pump produces 1500l/h with a 6m max delivery height, while consuming 17W. I did increase the voltage and found there was some improvement in temps at higher flow rates. I've eschewed that data since many End-users cannot afford the 1510 and DC voltage converter. Therefore the DC adjustment hence delivery was set to default pump voltage at 12V.

Onto test methods and results ->

Test Setup, Results & Conclusive Thoughts

Test Setup and Methodology

Using the CPU stress-test utility S&M v1.90 the processor was run under 100% LOAD. Temperatures were monitored using both Core Temp 0.95 and Intel's TAT or Thermal Analysis Tool (zipfile). The processor was tested at it's default speed 8x266FSB = 2130MHz, then at the highest overclock attained at default Vcore; 8x400FSB = 3200MHz and finally the highest stable overclock under S&M, which was 8x450FSB = 3600MHz at 1.40Vcore. Thermal Throttling, Speed Step and EIST were disabled in the BIOS. These exact conditions were then repeated with the Alphaool NexXxos XP WB and the stock air-cooler provided by Intel with the Retail E6400

Intel Socket-T Test System
Madshrimps (c)
CPU Intel Conroe 6400 SL9S9 L624A851
Mainboard 1.) Asus P5W DH Deluxe/WiFi (BIOS 1901)
2.) Gigabyte P965-DS3 (BIOS F12)
Memory 1.) Crucial Ballistix Tracer PC2-5300 (2x1GB)
2.) Mushkin X8500 DDR2 (2x1GB)
Graphics Leadtek 7950GX2
Power Supply PCPower&Cooling 1KW
Cooling 1.) Sub Zero CPU waterblock
2.) Alphacool NexXxos XP
3.) Stock Air cooler
Alphacool NexXxos Pro II Radiator
2x PABST 4412 FG/4L fans 73CFM
OASE 1500 12V ~ 24V DC-pump with Adjustable DC box
10mm ID tubing
Storage 2x Seagate Barracuda 80GB SATA Perpendicular
Optical Plextor PX755-SA DVD/RW SATA
Video Monitor Samsung 930b 19"
Operating System Windows XP


Clicking on the thumbnails below will show screenshots taken to record processor temp, speed and voltage during LOAD tests with the Sub-Zero installed.

8x266FSB 1.320Vcore | 8x400FSB 1.320Vcore | 8x450FSB 1.40Vcore

Madshrimps (c) Madshrimps (c) Madshrimps (c)


In the charts below IDLE essentially describes the system with very few active programs running. Although just snapping the screenshot with Mirek's freeware utility MWSnap will load the CPU slightly. I had to repeat the step five or six times to capture the lowest temp. Again this is indicative of the accuracy of internal thermal diodes. While many criticize their accuracy and the accuracy of the software reporting temps, as chip-makers rely on such devices for TM2 and other features their reliability is critical.

Madshrimps (c)


LOAD is arguably a difficult concept to measure since no-one can really say what a full-load should look like to the processor. Stressing the processor with Orthos and similar programs does raise the temp although in most cases it exceeds what your processor will ever see. In that respect such programs are similar to synthetic benchmarks since very few people would continuously run that program on their PC. Where crashes normally appear in the "real world" usually occurs during heavy 3D applications/games. I used to run Seti@Home and S&M to mix things up, however; I've found S&M which targets specific processor areas is the most rounded and certainly able to raise temps as high as any other software. Except for the 8x450FSB level of testing where 87% was the highest the processor could endure without crashing, all other tests were run at 100%.

Madshrimps (c)


Conclusive Thoughts

Sub-Zero represents a new level of excellence among Watercooling.de’s catalog. The block competed with and even out-performed the NexXxos XP which is considered to be one of, if not the best 10mm based water block on the market. Watercooling.de will integrate the Sub-Zero into a single 120, dual 120 or triple 120-radiator based ExodusFlow kits with a LAING DDC-pump. At this point the only other water block I might recommend to compare against the Sub-Zero would be the Masterfrezzer M-II. Today, the Sub-Zero takes the "Brass O-Ring."

PRO
Excellent performance.
Great Ergonomics.
Design improves slightly from high flow.
Works with all current and older platforms.

CON
Not available in the USA


Sub-Zero can be found at the Sub-Zero Shop Watercooling.de

I would like to thank Florian J. for the sample and his patience.

Madshrimps (c)
  翻译: