What’s best – a Corvette or a John Deere? – Part I
A good friend of mine and a board member in Asetek; Henri Richard inspired me to write this article. However instead of asking the car question, Henri was asking whether air cooling or liquid cooling was the best. My reason to pick the car analogy instead, is because the answer here is really simple and clear; it depends on what you need it for…
Recently I read this article at Anandtech, and as you can see their answer is, that unless you pick a very expensive water cooling kit, or cool other components then air cooling is better. Obviously Anandtech is covering this through the glasses of an end user, where I will be covering it from an OEM perspective. However as Asetek’s main activities are based on liquid cooling, the question in itself is of course interesting and highly relevant!
Let us start with a small physics lesson.
In a heat pipe in a CPU cooler you have liquid, typical water, alcohol or a refrigerant. Instead of using a pump there is a wick or a mesh inside the heat pipe, and the capillary forces are used to transport the cooling medium around in the system. When the liquid is heated (from the CPU) it boils and evaporates. This phase change requires energy (CPU heat) and the gas is transported to the cooling fins, where it is cooled (by the fan), turned in to liquid state again and returned to the heat source area. This is a very efficient process that in many ways are similar to liquid cooling – also in the sense that if it leaks, it will not work anymore.
In a pumped liquid cooled system you also have liquid, typically water with an anti freeze and a corrosion inhibitor. A pump is circulating the liquid around in the system. The idea is that you pick up the heat from the CPU and transport the heat to a remote place where you cool the liquid over a large surface area with a slow running fan. As the smart reader might have observed already; provided you use a heat pipe cooler that has the same or greater radiator area than a liquid radiator, it will obviously be just as good or even better if it actually has even larger radiating surface area. A pump will transport the liquid faster; in return a phase change process is more efficient, so there are cons and pros with both technologies.
If you take a look at the Coolermaster Hyper 212 that Anandtech reviews here you will see that it actually outperforms a couple of liquid coolers. I decided to buy the Hyper 212 and have a look at it and compare it to our LCLC. I would like to state that this is a great cooler for the market segment Coolermaster is addressing with it. However as mentioned earlier I am trying to address this from an OEM perspective, where this cooler is not suitable. I could have picked any air cooler, but because of the articles on Anandtech I decided to go for this one.
The first thing I noted, and as you can see on the picture, was that the effective radiator area on the Hyper 212 was at least twice of our LCLC, meaning that from a pure technological point of view, it is not really a fair comparison. Going back to the articles, I would probably have tested with same radiator form factor before judging one technology over the other. Anyway as this is not a performance shoot out I will leave the performance part for now. I have sent the Hyper 212 to our application people, so I might return at a later date with a benchmark, where all things are made equal.
From an enthusiast/DIY end users perspectov, I agree very much to the articles; why spend money on CPU liquid cooling, if you can have an air cooler with the same performance (or better). However things are usually not that black and white. Let us start looking at the system level and look at two extremes:
- An enthusiast grade chassis with a lot of ventilation holes and room for big fans. In such a chassis air cooling will shine. The “cold” air supply to the cooler will be close to ambient, and the “hot” exhaust air from the cooler will easily get out of the perforated chassis and in to the ambient. In this scenario you will need very good liquid cooling with a least the same radiating area to compete.
- Let us imagine another extreme and much more likely scenario (as mentioned before I cover this from an OEM vendor perspective), where we close the chassis completely besides from a small air inlet and one outlet, where we place the radiator.
So these are two extremes, but the truth is in between – always. I think scenario #1 pretty much reflects most DYI end users, and scenario #2 reflects most factory built PC’s. As you can see already now, the answer is still, that you cannot say what is best before determining what it is going to be used for J
Let us leave the DYI end user comparison for now and turn in to why liquid cooling really is superior for OEM integration.
- Form factor: Most factory built PC’s are not big enough to accommodate huge coolers like the Hyper 212 (size of a brick). Gaming market aside, the trend is opposite i.e. PC’s are getting smaller and smaller. Our integrated LCLC can fit in to a very small chassis, and as we have flexible radiator from factors it is still possible to achieve a big radiator area and therefore obtain good performance and low noise. Something that is not possible with air cooling. A good example is this small form factor DTX reference design. With air cooler you are limited in performance and noise is a problem. With air cooling you cannot use more than 65W CPU’s (if you go higher it will over heat). With the implementation of our LCLC we were able to use high power CPU and still maintain a low noise. Another example of form factor could be graphics cards. All high end graphics card today feature 2-slot coolers, meaning that if you want to run a 2-way SLI/Crossfire or even “worse” a 3-way SLI there will be limited to no place available for other add on card. With liquid cooling you can remove all graphics card heat in a 1-slot form factor leaving plenty of space for other add in cards.
- Weight: The Hyper 212 weighs 712 grams (0.7Kg) violating all possible CPU/motherboard standards available. Intel and AMD’s maximum weight specs are about half of that. Because of that no OEM would ever use a cooler like this. The entire PC would be smashed by the cooler during shipping. The cooler will simply ruin the motherboard, CPU socket, chassis and CPU and this is very common knowledge in the industry. The reason why these coolers can be justified is obviously because it is sold retail where these strict requirements for shipping do not apply. Our LCLC is low weight since the main part of the weight is in the radiator that is fastened to the chassis and not the motherboard. Therefore if you need high performance but low weight, liquid cooling is a much better choice.
- Technology: Heat pipes can only be a certain length (and thereby limited in how long they can transport heat) since they work by capillary forces. On top of that, at around 150W a heat pipe will dry out and no longer work. This is a physical limit of the technology. We are not there yet, but getting there rapidly with high end CPU’s. Liquid cooling can be scaled to whatever performance you want (many megawatt Flexibility Liquid cooling can cool several components quietly in one loop. You could argue that heat pipes can cool more components as well like seen on recent enthusiast motherboards (Northbridge+Southbridge+VRM’s), but it will not go further than that (Simply due to the limitation in length and performance). In HP’s Blackbird 002 we designed the liquid cooling system to be able to handle tremendous amounts of heat (2x130W from the AMD FX-74 CPU’s and 2x180W from ATI R600 GPU’s) With air cooling In this case, you would need a 700gr cooler on all your components, meaning that you would need 4 of these coolers inclusive the weight and cost. Obviously this is not feasible, but an interesting conclusion. As I wrote above, the Hyper 212 has an effective radiating area around double of our LCLC. Another interesting thing about the HP Blackbird 002 is that in there our radiator area is around the same as the Hyper 212 even though we remove all this heat.
- Cost: As you might know LCLC means Low Cost Liquid Cooling. It is a given that we cannot match the price of the Hyper 212, but it is getting close! Again looking from a system perspective from the example above our LCLC is highly competitive with air cooling. As soon as you need to cool more than one component, we are very close in cost, or to use Intel’s way of phrasing it, liquid cooling is competitive when you look at cost per Watt of cooling.
Another way of considering cost is the side benefits you could get with liquid cooling. As mentioned before small form factor PC’s, iMac, all in one PC’s etc. seems to be the trend going forward. Many of these feature mobile CPU’s because of the low power. With the low power it is possible to cool them without too much noise in a small form factor (basically on the backside of the LCD screen). Mobile components are more expensive than desktop parts. Let us assume that liquid cooling is efficient enough to enable the usage of regular higher powered desktop CPU’s. The cooling system will be more expensive than the heat pipe, but the cost difference between the mobile and the desktop CPU will more than cover the added cost for the liquid cooling. In this case liquid cooling actually saves the OEM money despite a higher SKU cost!
Summary – Obviously there are many technical aspects in each of the above-mentioned topics that can drag the conclusion in different directions. So what is then the conclusion on all of this? Is the John Deere the better choice? As already stated twice, there is no answer to that question without being more specific of the usage model. From a DIY end user perspective with a good ventilated chassis, no need for shipping/transporting your PC and no particular interest in overall system acoustics, I would easily recommend air cooling.
On the flipside – for an OEM who is looking for high performance, low acoustic emission in a regular or small chassis that can survive cross USA shipping in a truck, I would recommend nothing else than liquid cooling!