Does thermal paste really matter? We applied, reapplied, reapplied some more, and turned up the heat on our testing CPU to find out.
Ask ten geeks about their preferred thermal interface material (TIM) and you’ll get six different answers. Five will go with Arctic Silver 5 and the others will have five totally different favorites. Ask a non-geek about their favorite thermal paste and you might get slapped.
But is Arctic Silver 5 really the best? Is the thermal testing compound we’ve been using in the lab really cutting the mustard? We gathered seventeen premium thermal pastes and an overclocked test bed and set about finding out whether thermal paste really makes a difference, and if so, which one is the best.
A Note on Methodology
Our test bed consists of an Intel Core-i7-975 CPU, overclocked to 3.9GHz. This gives an idle clock speed of 2.1GHz. Our motherboard is an Asus P6X58D Premium with 6GB of RAM and a Thermaltake Frio OCK cooler with its fans set to their highest setting, to maximize airflow. Our test bed is built into a High-Speed PC Top Deck Tech Station (large), and our PSU is an Antec TruePower 850W.
We tested each thermal paste after an hour of idling, then after three hours running Intel’s Nehalem thermal stress-testing utility at 85 percent thermal load. How’d we get that number? We applied our standard thermal testing compound—Arctic Silver Lumiére—and cranked up the utility until the CPU was consistently just under 90C, then used those settings for every future test. CPU temperatures are an average of the four CPU core temperatures as measured by HWMonitor, and we used TMonitor to make sure the processors were running under 100 percent load at their full 3.96GHz during testing.
For this test, we focused on thermal pastes that are nondestructive—none of this liquid metal stuff—and readily available in the United States.
After a few weeks of testing day in and day out, we’re ready to report our results. Here are your contenders!
Arctic Silver’s OEM testing compound is a ceramic-based TIM which we’ve been using in the lab for years because of its consistent performance and lack of burn-in time. It’s not available to the general public, but we included it to see how our baseline TIM really stacks up to the competition.
Not well, it turns out. We’ve been using Lumiére since God was a kid, and it’s not bad at low temperatures, but in our stress test it pulled down temps on par with the worst performers of the bunch—89C, almost hot enough to trigger our CPU’s automatic throttling.
Noctua’s NT-H1 is a thick electrically non-conductive composite material that ships with the company’s CPU coolers and can also be bought separately. Noctua recommends dropping a small amount in the center of the heat spreader and allowing the pressure to spread the paste across the interface. Given the tackiness of the compound, that’s good advice.
NT-H1 is good stuff. At full burn our processor only averaged 81.25C, and at our processor’s idle temps (still far higher than most processors’ burn temps) were just 41C.
Arctic Cooling—not to be confused with Arctic Silver, which also makes thermal pastes—distributes several thermal pastes in addition to its lineup of CPU and GPU coolers and remote-controlled tugboats (yes, really). MX-4 is a non-metallic carbon-based gray compound that doesn’t conduct electricity, and Arctic Cooling says a single application will last up to eight years. It’s a bit pricy at $13 per tube, but it does the job well, with burn temperatures of just 81C and idle temps of 41C.
Tuniq’s TX-4 is gray and tacky and difficult to apply even with the included spreader, which is the size of a credit card. TX-4 is not electrically conductive. Once we did get the paste applied, Tuniq TX-4 performed the best of any thermal paste in our roundup, and was the only TIM with a sub-80C score—though at 79.5C, that’s within the margin of error for our test. It’s quite expensive for the tiny amount you get per tube, so you may be just as well off with some of the less expensive pastes in the roundup, but TX-4 has top-notch performance and earns our recommendation.
ZeroTherm ZT-100 is gray, silicone-based, and easy to spread. It also comes with a finger cot, so when you spread a thin layer of it on your CPU (as per ZeroTherm installation instructions) you don’t have thermal paste on your fingertip for the rest of the day. With burn temps of 81.5C and idle temps of 41C on our test configuration, it earns our solid recommendation.
Prolimatech has a hard-won reputation as a maker of excellent, if massive, CPU coolers. It also makes PK-1, an aluminum-based thermal paste with bits of zinc oxide in it. PK-1 is gray and pretty easy to spread despite its almost claylike consistency, and doesn’t conduct electricity. It’s also one of the best thermal pastes we’ve tested, with burn temperatures under 81C and idles just over 41C. It’s cheaper than a lot of the other premium pastes, too—$13 gets you a whole 5g tube, or you can grab a 1g pouch from NewEgg for $3.50.
Rosewill’s thermal grease is cheap and easy to apply, thanks to its glorified nail-polish bottle and brush applicator, but its performance is strictly middle-of-the-road, with a burn average of 85.5C. Most of its 20g weight is the glass bottle.
Xigmatek’s memorably named PTI-G3606 is a silicon-based electrically non-conductive thermal grease. It’s very easy to apply a thin layer to the CPU with the included plastic spreader. In our tests it performed slightly better than the Rosewill RCX-TC090PRO, with burn temps of 84.5C.
Tuniq’s TX-2 is a light gray thermal paste that’s stickier than most of the ones in our roundup. It’s not conductive and it’s fairly easy to apply. In our tests it was just half a degree Celsius warmer than the TX-4, putting it within the margin of error for the tests and making it one of the best pastes we’ve tested. It’s cheaper than TX-4 and you get more in the tube, so we’d actually recommend TX-2 over TX-4 for the cheapskates.
The gold standard for high-performance thermal pastes, Arctic Silver 5 was our odds-on favorite to win this roundup. Arctic Silver 5 consists of silver particles in a paste with ceramic particles, zinc oxide, aluminum oxide, and more, to maximize thermal contact. Arctic Silver 5 is slightly tacky, but easy enough to spread. We tested it by applying a small dollop of paste to the center of the CPU and spreading it with a spreader, as is our custom for thermal pastes. Unlike most of the pastes in our roundup, Arctic Silver 5 has a recommended 200-hour burn-in period before it’s supposed to reach maximum efficiency, but who has that kind of time?
Even without the 200 hours of burn-in, Arctic Silver 5 did well, reducing the CPU temperature at full load to 82.5C. It’s not the best TIM in our roundup—at least not before the 200-hour burn in period—but it’s good right off the bat.
Alumina is Arctic Silver’s “Premium Ceramic Polysynthetic Compound,” and it looks a lot like the Arctic Silver Lumiére OEM Testing Compound we’ve been using for years in the lab. It’s white, spreads easily, and requires no burn-in time. It performed better than the Lumiére, with burn temperatures around 85.5C—decent, but not high enough to earn an “Approved” rating in this roundup.
Shin-Etsu X23-7783D is justly renowned in the cooling world. This gray silicone thermal paste is thicker than most and doesn’t spread easily, so we used the dot-in-the-center method to apply it to our cooler. Once installed, the X23-7783D enabled CPU temps of just 80.25C average at full burn and 40C at idle. There’s a reason many OEMs pre-apply it to their CPU coolers.
Nano diamonds, you say? We’ll take eight! Biostar Nano Diamond Thermal Compound is only 10 percent diamond particle; the rest is silicone compounds (50 percent), carbon compounds (20 percent) and metal oxides (20 percent). It’s electrically non-conductive and easy to apply with the included plastic spreader, but its performance in our test bed wasn’t impressive—89C at burn and 43C idle temperatures.
Zalman ZM-STG1 is another thermal grease in a tiny bottle with a brush applicator. The bottle is significantly smaller than the one Rosewill’s grease comes in, and the grease is thinner. We applied the ZM-STG1 with the brush as per instructions several times, but met with the same disappointing performance each time—over 89C at full burn and 43C at idle, just like the BioStar Nano Diamond paste.
Xigmatek’s PTI-G4512 is a light gray silicone-based thermal paste that’s easy to apply with the included spreader. It’s non-conductive and non-corrosive. It performs better than the other Xigmatek TIM, PTI-G3606, with burn temperatures just over 81C and great idle temps at just 40C. The PTI-G4512 is pretty cost-effective, too; a 4g tube is just $10—not quite as cheap as Arctic Cooling MX-2, but close.
The predecessor to Arctic Cooling’s MX-4 is slightly lighter and slightly thinner than the newer compound, but is also non-metallic and electrically non-conductive. Its performance was almost indistinguishable from MX-4—well within the margin of error. It’s also a few bucks cheaper, so if you’re especially budget-conscious, go for the older but cheaper compound.
Sound familiar? Cooler Master’s ThermalFusion 400 is a gray, electrically non-conductive thermal paste that spreads quite easily with the included plastic spreader. It cooled our stress-tested CPU to 81.75C at full burn and 41C at idle—good enough to earn it our recommendation. It’s not exactly cheap—MX-2, TX-2, and ZT-100 all give slightly better value for money—but at this point we’re talking a few dollars and a few degrees Celsius.
Geek Approved?: Yes
On an idling overclocked processor or a stock-clocked CPU, the differences between thermal pastes is minimal—we saw a spread of less than 4C between the best and worst thermal pastes in our roundup. At high temperatures—and we should reiterate that we overclocked the processor to 3.9GHz and used a custom thermal-stress utility to put an enormous thermal load on the CPU—we saw a spread of over 12C. Margin of error is plus or minus 2C to allow for ambient air temperature, which ranged from 23.8C to 25.4C throughout the testing procedure.
Of the seventeen thermal pastes in this roundup, Tuniq’s TX-4 scored the highest. Its burn temperature was 3C cooler than Arctic Silver 5’s. Eleven pastes earn our Geek Tested & Approved badge: Tuniq TX-4 and TX-2, Shin-Etsu MicroSI X23-7783D, Prolimatek PK-1, Arctic Cooling MX-4 and MX-2, Noctual NT-H1, Xigmatek PTI-G4512, ZeroTherm ZT-100, Cooler Master ThermalFusion 400, and good old Arctic Silver 5. We’d give pride of place to Tuniq’s TX-2, Arctic Cooling’s MX-2, and Prolimatech’s PK-1, because they’re slightly cheaper than some of the other premiere thermal interface materials.
So does thermal paste matter? Yes—there’s a big difference between thermal pastes when running a CPU at full burn. There’s a big difference between a thermal interface material that’s good for overclocking and those that aren’t, but with eleven great thermal pastes to choose from, you can’t go wrong with one of them.
One final note: The true hero of this story is Arctic Silver’s ArctiClean two-step thermal remover & surface purifier. It’s nontoxic, smells like oranges, and cuts through the toughest thermal interface with ease. We’ve used it in the lab for years and it’s a lifesaver any time we need to remove thermal paste from a CPU or heatsink. We’ve yet to meet a thermal interface material it didn’t work on.
Idle Temp (C)
Burn Temp (C)
Arctic Silver Lumiére (zero-point)
Arctic Cooling MX-4
Rosewill RCX-TC090 Pro
Arctic Silver 5
Arctic Silver Alumina
Shin-Etsu MicroSI X23-7783D
BioStar Nano Diamond
Arctic Cooling MX-2
Cooler Master ThermalFusion 400
Asterisk (*) denotes best score. All tests performed on an overclocked Core i7-975 @ 3.9GHz (burn) and 2.1GHz (idle) on an Asus P6X58D Premium motherboard with 6GB Corsair XMS3 DDR3, a Radeon HD 5850, and 850W Antec TruePower PSU. The CPU cooler used was a Thermaltake Frio OCK with its fans set to maximum speed. Temperatures recorded after 1 hour at idle and after 3 hours of full-burn testing using Intel's internal Nehalem stress-testing utility. We use HWMonitor to determine core temperatures and TMonitor to keep an eye on clock speeds.