Heat management is at the very core of computing. Thanks to those ever-present laws of thermodynamics, the electricity being passed through semiconductors is converted to heat, which then needs to be taken away from the source. This sounds simple, but it isn’t when you are pumping almost 100 Watts of power through a piece of silicon the size of a five cent coin.
Traditionally the solution has been to use a heatsink, which takes the heat generated by a processor and spreads it out across a large surface area in the form of fins. Air can then pass over these fins and heat up, cooling down the fins and effectively taking the heat away. However this kind of air cooling has its downsides. Not only is it highly dependent upon the air temperature within your PC, it is also at the point where traditional techniques struggle with high-end, overclocked systems.
We use the term ‘traditional techniques’ but most modern air coolers are much more than simple fins spreading out from the base. Ten years ago the accepted solution for cooling high speed CPUs was to use a faster fan. The issue was that small fans at high speed produce a tremendous whine, which is an incredibly unpleasant presence in a PC. Rather than persist with this trend, the heatsink manufacturers moved to other techniques designed to avoid small fans.
This resulted in a trend towards ‘tower’ designs. These designs separated the base from the fins of the heatsink, instead using heatpipes to transfer the heat from the processor to the fins. These heatpipes are usually made out of copper and consist of a sealed tube containing a small amount of pressurised liquid. Because the heatpipe is pressurised the liquid boils at a much lower temperature than it otherwise would – what happens is that when the processor heats up the liquid boils and travels up the tube as steam. The heat from the steam then dissipates as the heatsink fins take the heat away from the pipe, at which point the liquid condenses and falls back down the pipe and the cycle begins anew.
Not only does this design mean that a larger cooling fan can be used, it physically lifts the heat up and away from the motherboard. This ensures much better airflow over the entire heatsink, which helps transfer the heat away from the processor more efficiently.
While these kinds of gigantic tower heatsinks have become favourites of overclockers and other high end enthusiasts, they have created their own set of problems. The first is weight. When mounted in a tower configuration the combination of gravity and the weight of the heatsink can stress the motherboard. This is especially an issue for system builders, who have to ship systems with heatsinks already installed. A bit of jostling during shipping and a weighty heatsink can damage the motherboard, especially around where the heatsink mounts.
The other issue is access and clearance around the CPU socket on a motherboard. While manufacturers have made a concerted effort to clear out space around the socket, there are still electrical design requirements that necessitate putting the RAM slots as close as possible (Each RAM slot needs a certain length of tracer between it and the CPU, which keeps everything in balance). Unfortunately, nearly all enthusiast models of memory have large heat spreaders, which rise above the normal height of a RAM stick.
As we have discovered a few times in the labs, RAM heat spreaders and tower heatsinks tend to occupy the same physical space, meaning that often times you can squeeze RAM into the secondary slots, but lose the first one. This is a less than ideal situation, and tends to hit the users who are most likely to want to fill all their RAM slots.
Air cooling and heat pipes aren’t the only solution out there to the cooling issue. The other major means of cooling high end PCs has traditionally involved water cooling. Over the years some marketeers (largely in the laptop space) have called heatpipes liquid cooling, but watercooling is a completely different beast to heatpipes.
A traditional water cooling rig comprises several pieces. First there is the water block. This sits on a processor instead of a heatsink and has channels designed for water to flow through. This connects up to a radiator, which is ideally located away from the heat generating components in the case, and has fans attached to cool the water passing through it. The system also requires a pump and a reservoir, to keep water moving through the clamped on tubing that connects various pieces of the system.
Not only does water cooling provide highly effective cooling without increasing noise levels, it can be tailored to perfectly fit your PC. Choosing the right water block (or blocks if you want to cool GPUs as well as CPUs) and marrying it with the correct radiator and pump is part of the equation, but the major part is the ability to cut tubing lengths and design the system to fit perfectly with your components.
This kind of water cooling is expensive (a quality CPU water block can cost upwards of $150 – and you still have to buy the rest of the parts) and can be fiddly to install. It also introduces multiple points of failure thanks to the clamps and multiple lengths of tubing employed. This alone scares alot of people away from water cooling, and it introduces shipping problems for systembuilders that build custom water cooled rigs (a lot of high end workstations use watercooling to minimise noise and allow overclocked hardware).
The various issues with tower style air coolers and the complexity and expense of water cooling has pushed manufacturers towards a new solution. Known as closed-loop liquid coolers these are effectively mini water cooling systems just for the CPU, which use a similar architecture to water cooling kits but seal the system to minimise risks.
Most of these liquid cooling kits are based on technology developed by Asetek, the company that became famous for its refrigerated ‘Vapochill’ cases a decade ago. Various manufacturers are now making variants of the basic designs in conjunction with Asetek and have been for a couple of years now. They consist of a radiator with one or two 120mmfans attached, which bolts onto the relatively standard rear fan out takeon the chassis. This is connected to a small, circular water block (with inbuilt pump) that is slightly larger than the CPU die itself. This block attaches to the CPU via a special mount, and connects to the radiator via two sealed pipes.
The big reason why these closed loop water coolers are becoming important is that both Intel and AMD have introduced branded closed loop water cooling solutions with their new processors (Intel’s was shown at IDF and will be sold as an option with the upcoming SandyBridge-E Socket 2011 processors).
The reasoning behind this is a combination of the factors mentioned earlier in the article. It is an acknowledgement of the growing inconveniences inherent in tower design air coolers, both in terms of space inside a system and space around the CPU socket. With Intel introducing quad channel memory the constraints placed upon the CPU area are high (there are banks of memory slots each side of the CPU socket), and the relatively small footprint of the closed loop water block fits in perfectly. These closed loop systems should also allow for more overclocking headroom without having to invest in a high-end cooler.
High-end closed loop coolers, and do it yourself kits, aren’t going anywhere though. These new AMD and Intel solutions will likely be comparable to other manufacturer’s base models of closed loop coolers. Those wanting higher overclocks, or the ability to incorporate software fan management and the like, will want to look elsewhere. The good news is that Intel, at least, won’t be shipping any coolers with Sandy Bridge-E. Rather, it will provide branded options that guarantee a base level of performance. If you want more you can get more without wasting money on a stock cooling solution that will go unused.
This article was brought to you by Thermaltake.