Intel’s liquid CPU cooler: Is water worth the cost?

When Intel launched its high-end 3960X CPU last month, it also debuted a new liquid cooling solution. Typical boxed Intel CPUs ship with a heatsink+fan, but the company noted that the majority of its enthusiast customers bought third-party solutions. The new liquid cooler — it goes by the inspiring moniker of RTS2011LC — is meant to offer enthusiasts and OEMs an Intel-brand solution they’ll actually use.

The cooler is actually manufactured by Asetek, which also builds solutions for AMD, Antec, and Corsair. These types of closed-loop liquid coolers have become increasingly popular in recent years as manufacturers have experimented with maintenance-free systems that don’t require cleaning or occasional topping up.

One important thing to remember is that just because Intel and AMD have started offering liquid coolers doesn’t make them necessary. Intel’s six-core 3960X CPU carries a TDP of 130W, just like the first generation of quad-core “Extreme Edition” Nehalem CPUs. Any heatsink capable of dissipating 130W or more will have no trouble handling the 3960X, provided you have access to the appropriate mounting hardware.

Intel unintentionally underlines this point in its own product literature when it notes that it intends to launch a sub-$20 air-cooled solution for current and future Sandy Bridge-E and Ivy Bridge-E parts. Granted, such coolers probably won’t be the quietest things in the world — but a 4x price premium is a hefty one.

Performance

We compared Intel’s RTS2011LC (estimated price tag between $85 – $100) against Corsair’s high-end H100 Extreme. The H100 is ~$20 more expensive than Intel’s solution and uses a large rectangular radiator that’s designed to be mounted at the top of a case. It also ships with a pair of 120mm fans while the Intel cooler makes do with one.

One point of difference between the two coolers that we need to address up front is their fan speeds. Intel’s liquid cooler automatically increases or decreases based on CPU temperature from ~850 RPM at idle to ~2400 RPM when overclocked at full load.

Corsair’s H100 does a certain amount of dynamic adjustment, but the button on top of the pump is used to select one of three preset ranges. There’s therefore more potential variation in noise and performance. Our first two tests were conducted with the H100 set to “Quiet” mode while our final overclocking test used the “Performance” preset.

Given the difference in price and configuration, we expect the H100 to outperform the 2011LC — but by how much?

Testing

Idle temperatures were measured by letting the system sit at desktop after booting. Load temperatures were measured using the “Stress Test” option that’s included as part of Intel’s Extreme Tuning Utility. Since the XTU isn’t as widely available or well known as other overclocking stress tests, we also measured peak power consumption values using Prime95. Tests were looped for an hour before results were measured.

Both coolers were mounted in a Corsair Obsidian 800D.

At first glance, the H100 decisively outperforms Intel’s own solution — but the temperatures don’t quite tell the whole story here. We couldn’t get an accurate measurement of the H100′s fan speeds — the Intel tuning utility reports 2100 RPM regardless — but we were able to measure the RTS2011LC’s fan speeds at each of the three clock speeds we tested.

Intel has clearly tuned the X79SI’s BIOS to emphasi e silent running over low temperatures. These settings can be changed in BIOS, if the user wishes — but the gap between the H100 and RTS2011LC at stock speed is wider than it has to be.

Once we start overclocking, the situation changes.

It’s impossible to predict exactly what CPU voltage will allow a chip to hit a given clock speed; as a result, overclocking temperatures are always approximate. In our first overclocking test we deactivated Turbo Mode, set the chip for a 1.25x multiplier, and ran the 3960X at 4.13GH at 1.35v.

1.35v is nearly at the top of Intel’s maximum recommendation for the CPU, so it’s not surprising to see the RTS2011LC panting to keep up.

We were forced to increase CPU voltage to 1.37v to reach 4.5GH , and the CPU only remained stable if we set the fans for Corsair’s “Performance” preset. Intel’s RTS2011LC proved unable to keep up with the increased voltage and clock speed; the chip remained stable during Intel’s included stress test but failed in Prime95 within minutes of starting the test.

A check of the CPU’s power consumption at our three clock speeds is enough to illustrate why Intel’s cooler has trouble.

At 4.5GH , our 3960X system is drawing 125W or 53 percent more power than it drew at 3.6GH , despite a clock speed that’s just 25 percent higher. Overclockers on the prowl for absolute top-end single-threaded performance may have better luck pushing a quad-core CPU.

The H100′s larger radiator and dual fans were able to keep the 3960X stable at 4.5GH , but the chip was very nearly topped out. At 4.67GH and above the CPU would lock almost immediately after booting to OS. We’ve seen better results from certain websites, but the six-core Sandy Bridge CPUs don’t appear to push much higher than ~4.6GH without sub ero cooling.

Is water worth it?

Intel’s RTS2011LC does a fine job of keeping the CPU cool at stock speeds, particularly considering that it’s possible to tweak BIOS settings to bring temps down a bit without turning one’s chassis into a howling tornado. There are any number of modestly positive adjectives that one could apply.

Is it worth $85? Not really. While there’s nothing wrong with the rebranded Asetek cooler, there’s also no empirical evidence that it offers a particular benefit over the upper echelon of air coolers in the $50-$65 range. Intel and Asetek claim that superior cooling technology allows them to spin fans more slowly, but the pump on Intel’s cooler is audible, if only slightly.

The fact that Intel believes it can manufacture and sell an air cooler capable of handling SNB-E and future Ivy Bridge-E chips for under $20 underscores the fact that it also believes enthusiasts are willing to pay 3-4x as much more a cooler that doesn’t deliver anything like 3-4x better performance.

If you’ve simply got to have a liquid cooling kit, Intel’s is fine. If an OEM is willing to toss one in a box for a pittance, go for it. Otherwise, we’d look for something cheaper.

Intel’s liquid CPU cooler: Is water worth the cost?