Home
Ask a Question
Secure Orders
Latest News
Reviews
Projects
Customer Gallery
Guides
FAQ
Live Testing

CM Stacker Makeover 2

Today we are addressing a further evolution of the CM Stacker project with the CM Stacker Makeover Two project. For some time we here at Radiical have worked on this project and the development of the "Patent Pending" Radiical Dual Head Pump Unit©. Finally we are able to present our work into the innovation of dual loop technology.

Dual Loop technology

In this part of the CM Stacker Makeover Two project we will be focusing on dual loop design. We are focusing on this type of cooling system design because of three basic relationships. These relationships are predicated on the following assertions;

• That as CPU development takes place GPU development follows at the same pace.
• That with the introduction of multiple CPU and GPU cores the heat loads in these areas will continue to reach significantly higher levels than at present.
• That to cool these processor and chipset developments the adoption of dual loop cooling systems will be a necessary and logical progression.

The first question to be answered is "why dual loops"......why not just increase the pump capacity within the single loop instead. To be honest the solutions of the past did focus on increasing the pumps capacity to simply push the coolant through the single loop restrictions. On the face of it this is not an unpalatable solution with the exception that bigger pumps with higher head pressures and flow rate values usually increase the pump heat being dumped into the coolant. When you consider that the whole point of a cooling system is to cool this becomes a somewhat sterile debate.

These more powerful pumps are also usually much larger in size occupying more of the limited case space available and invariably much noisier. So not only will there be the coolant heat dump issues mentioned above there will also be a flow on effect of higher internal case temperatures. For those of us trying to reach low noise levels as well as efficient cooling this can effectively rule out this type of pump.

To avoid this heat/noise scenario it has been suggested that smaller pumps, with less heat dump, joined together "in series" may work out to be a better solution.

From our testing results presented below joining two pumps say Laing DDC "in series" simply increases pump head with very little gain in flow rate. Given that it is the combination of flow rate and pump head which make many of the most popular CPU water blocks "work" we are left with something of a conundrum. We can expend most of the head and flow rate on the first restrictive block but we will then reduce the flow rate to the second and third blocks in the loop using "in series" twin pumps. Really a case of "famine versus feast" dependent upon where you live in the cooling loop.

The other alternative is to think outside of the box and look at balancing the flow rate and pump head using a dual loop. By using slightly less powerful pumps (lower pump head not flow rate) we reduce the heat load being dumped into the system by the pumps (results below) and retain very good flow rates and head pressures to each system component. We can use smaller ID tubing making tubing out of the system easier and as a result we can reach those elusive elements like chipset's with ease. With two separated loops dedicated coolant can be directed straight to the area to be cooled. For example GPU blocks in this sort of dual loop configuration are not expected to be cooled with the exhaust heated water from the CPU block. The importance of this innovation is becoming more and more relevant with today's dual GPU cards where heat loads can often mimic or exceed those of the CPU. It is really all about hitting that elusive cooling "sweet spot" and this is where dual loop systems excel.

Dual Loop

As the name implies dual loops are two separated loops each carrying out a specific cooling solution viz cooling the CPU's and GPU's separately. Normally the two loops formed in a dual system contain all of the elements of a single loop. That is pump, tubing, water block/blocks and a radiator. You can imagine how complex this could become inside a modern computer case.

Complexity has really been the issue with lack of mass acceptance and therefore adoption of dual loop systems. For example how do you make these dual loop units compact enough to fit inside the case without disrupting each other. Moreover how do you stop all the extra tubing, pumps and so on from blocking off the case ventilation required by other non liquid cooled components. How do you fit the radiator/s required for dual loops into or for that matter onto smaller cases.

Lets face it to introduce dual loop technology there is going to have to be a change in the basic configuration of the cooling system. The first change, in our view, is the radiator design.

Enter the Radiical Three Port Radiator©

In the FAQ for the Radiical Three Port Radiator© and Bleeder Wick© we alluded to a specific but unmentioned function of the Three Port Radiator©. That function was to act as a "mixer" for what we have called here "dual loop technology". The mixer function balances the differing temperatures of coolant coming from each of the dual loops. To accomplish this we have investigated two methods of flow pre mixing.

First we researched and then developed a Radiical Three Port Radiator© with a single pass configuration which we will introduce formally a little later on. Secondly we investigated and then developed the Radiical TwoInOne© pre mixer block for dual pass radiators. Lets look at the TwoInOne© device first.

Enter the TwoInOne© fitting.

The TwoInOne© block is a pre mixing unit developed so that both sections of the loops can be joined prior to entry into the radiator core area where full coolant mixing will occur. The TwoInOne© device acts to create a laminar flow for both flows. This is important so that the flow rate is not choked at the radiator inlet port.

Joining the Flow

Normally to join two differing sizes of tubing would involve using reducing adaptors. The problem with reducers is that they will only adapt one tube at a time. The TwoInOne© overcomes this problem by adapting the tubing at the TwoInOne© mixer block so that two pipes can merge flow within the body of the TwoInOne© block. Of course we could have used a "Y" piece instead if the tubing sizes were the same but they are simply too bulky. In the illustration given below the TwoInOne© easily fits between the legs of the traditional "Y" piece.

Using the TwoInOne© not only reduces the bulkiness of the fitting but also makes the tubing layout more parallel providing a very valuable space saving inside the case.

The result is an easier placement of tubing. One other obvious advantage of the TwoInOne© fitting is the option of using whatever sized inlet and outlet barb size you wish. In the picture below we have blanked off one of the inlets so that we can use a 3/8" to 1/2" conversion.

Okay so having designed the premixing block its now on to the radiator setup.

The Three Port Radiator©

The first problem after joining the two loops to flow into the radiator is considering how we can bleed the unwanted air out of the system. This is where the Radiical Three Port Radiator© shines. Filling and bleeding are almost automatic using the Bleeder Wick© attached to the top of the radiator. Using the Three Port Radiator©, which has the reservoir incorporated into the radiator body, also removes the need for a secondary reservoir and the tubing lay out required to and from a separate reservoir. This constitutes a large space saving and simplifies the system design. So with the problems of filling, bleeding and mixing the flow pathways out of the way its time to look at pumps.

The Patent Pending Radiical Dual Head Pump Unit©

We have already discussed the lack of any real flow rate performance gains with "in series" linked pumps so it is timely to introduce our latest product the "Patent Pending" Radiical Dual Head Pump© unit.

To answer the obvious question "why use two pumps in parallel?" its simple they are needed to create the twin loops. They are also more efficient in this configuration. There will be more detail about this later in the performance section when the results of our testing are reviewed.

The other reason behind using two pumps in parallel is that we can now have control of the coolant presenting equal flow rates and balanced temperature coolant to both the CPU and GPU. The result is that both get an equal share of the pump head and flow rate from each pump. The only time that these flow pathways will combine is in the short distance of the TwoInOne© mixer and of course the large area of the radiator core. The additional advantage of the Radiical Dual Head Pump Unit© is that we can now use smaller cross sectional tubing such as 3/8" ID tube with another large "in case" space saving.

Construction Time

Tubing Out the Loops

To create the dual loop system we are going to adjust the tubing in the CM Stacker Makeover Part One from single loop configuration to dual loop configuration. To do this we have first pre-installed the TwoInOne© unit in the picture below in the single loop configuration and using a small amount of Velcro secured it to the case floor. The next job will be to change around the tubing from the single loop shown below to a dual loop configuration.

We already had the radiator setup on the case back from the previous part of the CM Makeover project so the natural progression was to use the Three Port Radiator© in dual pass configuration in this part of the project. We also wanted to demonstrate how easy it is to change over to the "dark side" of dual loop configuration.

First its the CPU flow pathway that is altered by removing the CPU to GPU bridge tubing as shown below.

Next the GPU flow pathway is swapped over to the pump head unit and a new length of tubing added to connect the CPU with the second pump outlet on the dual pump.

That is how simple it is to convert.

SLI configuration

For those of you using SLI or its variant the normal daisy chaining of the GPU blocks occurs within the now separated GPU flow pathway. There will need to be a further adjustment of the tubing to create the GPU to GPU bridge between the two cards.

When the tube out to the CPU, GPU/s and Pump is completed the two inlet one outlet of the TwoInOne© is attached and used to recombine the flow pathway before coolant enters the radiator. This permits one radiator to be used to cool both coolant pathways by mixing the coolant within the radiator. In the next cycle the incoming coolant to the pump unit is balanced at the same temperature point. This way the total capacity of the radiator is shared by both loops.

In the final picture here we have added some uv reactive dye to the coolant to assist with identification of the various flow pathways.

As shown above the presentation of dual loop systems can be as neat if not neater than single loop systems particular when the Three Port Radiator© is incorporated. From a larger perspective the total impact on the available case space and more importantly air flow within the case can be seen. With reduced obstructions air flow cooling performance around the non water cooled components is also improved. Speaking of performance its time to dive into the results.

Performance

It is interesting to evaluate the various theories held by people in general about single loop flow rates, heat dump and generally the overall performance of such systems when using "in series" and other pump configurations. If you are interested in looking at an interesting forum item on series versus parallel configurations etc you might want to look at this forum discussion here;

http://www.xtremesystems.org/forums/showthread.php?t=119271

In our testing conducted here on the CM Stacker and on our test facility using the die simulator the results are very interesting.

First lets just look at some of the theories.

Theory One

For quite some time there has been the widely held belief that a high rate of pump head will outperform a lower rate of pump head at the same flow rate.

Lets face it if the block design being used does not have a high restriction value then additional pump head is simply wasted.

Theory Two

That higher flow rates increases overall cooling efficiency.

Generally this is true until flow rates of over 1.5GPH (6LPH) are reached. After this value the results of our testing show no increase in cooling efficiency in fact the cooling efficiency plateaus out above this value.

Theory Three

Two modified head pumps "in series" will outperform a single pump using the same modified pump head.

This statement was found to be incorrect when tested with the components listed below.

Radiator
A Thermochill triple radiator fitted with three Tricod fans.

Water Block
A Storm water block fitted with the supplied three eights barbs.

Chipset
A Radiical Chipset block was used as balance point to simulate an additional water block restriction.

Pumps
We used two forms of dual pump configuration. We were sent a sample of an "in series" dual head pump unit and we used our own Radiical Dual Head Pump Unit© in a parallel dual pump configuration. The pumps used were two Laing DDC+1 18 Watt pumps. The same two pumps were used by both configurations being swapped in and out for each test series. For the single versus dual "in series" tests we used an Alphacool modified pump head.

Tubing
Three eights inch ID tubing was used in all system tests so as to favor higher head pumps. Inlet to the pump unit/s were maximized using 7/16 inch ID tubing between the radiator and pump inlet and from the GPI flow meter back to the radiator .

We used a GPI flow Meter to assess the flow characteristics and an Ashcroft pressure gauge to measure in line head pressures. Temperature measurements were taken continuously during testing using a series of PT100 RTD's to measure air ambient, coolant temperature in and out of the radiator, pump body temperature and die simulator temperature variations. Comparisons of air intake to coolant return temperatures were used to give a "K" value for observed temperatures.

Pump Heat Dump
The pump casings were monitored during the test series and the temperature records indicate that "in series" the pumps ran hotter.

In Series
Pump One at 40.60 degrees
and
Pump Two at 41.70 degrees

In Parallel
Pump One at 36.5 degrees
and
Pump Two at 36.4 degrees

The consequence of this additional heat generation being dumped into the coolant flow is expressed within the system temperature performance figures. No attempt has been made to assay the heat dump value.

Pump Pressure

We were able to establish that the dual head "in series" pump was able to develop 15 psi at full occlusion. The pressure rating during flow testing was 2 psi.

The Radiical Dual Head Pump Unit was able to develop 6.5 psi at full occlusion. The pressure rating during flow testing was also 2 psi.

Flow rate

The flow rating for the "in series" pump during testing in the system was a very healthy 6.2 L/Min. By comparison the parallel version of the dual head pump was able to develop 4.2 L/Min using the same components described above.

Performance Results

Okay so down to the final stages of the performance of "in series" versus the parallel split flow dual head units. It would be relevant here to point out that it is not our intention to divert the focus of this project onto the wisdom of using "in series" versus "parallel" flow. Neither is it our purpose here to present extensive and complex performance results by way of a de facto review of other products. It is our purpose though to rule out the perceived deficiencies and to change some widely held beliefs about head pressure and flow rate proportions.

After running repeated test series at fan speeds of five volts and 12 volts on the die simulator and despite all its pump head pressure the "in series" dual head pump faired the worst in the results obtained. That is to say that it did not achieve even "equal" performance with the split flow parallel dual head configuration used. It was close at 1 degree difference between the two but close is not good enough. In an area where 1 degree performance loss can be considered significant the dual head "in series" pump was something of a disappointment. Not surprisingly an 18 Watt pump using a single Alphacool head modification was able to beat out the dual head "in series" versions performance results. To ascertain if the results flowed across to other radiators and water blocks we repeated the testing on the die simulator using our Radiical Three Port Radiator© and our less restrictive CWX water block with the same shift towards the parallel dual head or solitary pump head configuration.

Noise

Noise is so subjective that the only comment that would be suitable here is that we felt the "in-series" pump head was unduly noisy for our own use.

Real World Testing

To demonstrate the benefits of splitting the flow and using a dual system like the one we have used here we went onto the test computer in the CM Stacker for some real world testing. The object here was "not" to run both systems that is "in-series" and parallel configurations on the test computer lest this was interpreted as a mini review. What we were doing was establishing the accuracy between the software used to check our previous single loop testing results on the CM Stacker and verifying the ongoing accuracy of our temperature measurement equipment. We did this so that we could compare our previous single loop testing results with a series of dual loop results on the same machine.To do this we used our WHX water block with the RTD machined into the base of the block. This water block "reads" the interface between the IHS and water block base. This process places a RTD of known standardization up against the K8 on chip thermal measurement as a constant comparison. We verified from repeated testing that we were within 0.5 degrees of each other in our temperature reporting results.

We expected the results were hardly going to vary from the original single loop configuration as little had changed with the exception of the inclusion of the restriction from the TwoInOne©. Similar flow rates should be occurring within the system and as a result similar performance figures should emerge from our comparison. We tested both loops line pressures and as was expected even with the TwoInOne© restriction the system stabilized to near balanced pressures between the CPU and GPU blocks. The marginally lower line value was reflected in the psi rating of the restrictive CPU block and did not appear to be significant.

Next we dropped back even further the pump outputs from 18watts to 10watts by using two 10 watt DDC pump bodies in the Radiical Dual Head Pump Unit©. We observed very similar performance results. We intend to look into higher rated pump head designs at a later date as we increase the complexity of the components being cooled by the dual loops.

Summary

So there is the completed project. The increased performance capacity of the cooling system. The move to smaller tubing with the consequent in case space saving and the inclusion of the ability to direct flow have been the major benefits of this move to dual loop cooling. The real gains of this dual loop systems should become more apparent when it is matched against more worthy opponents such as dual cored CPU's and GPU's in the future. Perhaps that may be the next evolution of the CM Stacker.