The test platform used the following components:
- ASUS Maximus X Z370
- GSkill Trident Z 16GB CL14 3200MHz
- NZXT Kraken X62 at max fan & pump RPMs
- EVGA GTX 1080 Ti FTW3
- NZXT Hale90 V2 PSU
For testing, we’re using in-house test automation and 30-minute test runs, averaging CPU core thermal data after steady state is achieved. We are also logging liquid temperature, ambient temperature (second-to-second), and current at the EPS12V cables (via clamp). The ambient temperature logging is used to create a delta T over ambient value. Ambient temperature was typically 25C, +/-1C, which is accounted for in test data. There is no significant change in the thermal properties of the air with a +/-1C resolution.
Prime95 with a custom configuration is used for thermal torture. We have tuned our P95 pass to simulate a Blender rendering pass, which is a realistic AVX workload for high-end use cases.
Testing was conducted with Thermal Grizzly Conductonaut (Amazon) between the IHS and die (in both instances), and thermal paste atop the IHS for each test. A graduated syringe was used for thermal compound each time, ensuring the same amount of compound used. Compound was manually spread over the surface of the heat spreader.
Given the potential for human error in a sensitive test such as this, where result resolutions are expected to be low, we conduct each test with three applications. That means re-applying the liquid metal and thermal paste and repeating the thermal torture for three iterations. These test passes are then averaged, with an eye out for outliers.
We do not re-seal the IHS. We have found this to be a performance killer.
Custom Intel IHS vs. Stock – Thermal Results at Steady State
This chart is what we ultimately ended up with. At 5GHz and 1.42V, we end up generating between 200W and 220W at the EPS12V cables, depending on test. Give or take a couple watts for the two tested voltages here. Stock temperature monitoring showed 49.5 degrees Celsius over ambient for baseline, averaged after the 30-minute burn-in mark on our X62 (with max fan speeds). The peak 10-second high was 51.6C, with liquid temperature at 13.8C.
For the Rockit Cool IHS, we measured 44.5 degrees for the same test, or 48.2C across the peak high. Liquid was roughly equivalent and within margin of error. These differences are just barely outside of our error margins, and only exit margins because we ran the test 3 times on each IHS, with each one using a new liquid metal and paste application. This helps ensure we’re using the same application technique each time.
The result is a slight temperature improvement, roughly in the range of 3-5 degrees Celsius. Not bad for a copper brick.
At 1.44V and 5.1GHz, we saw nearly identical deltas: The stock IHS reported 56.8C, or 60C peak high. The Rockit Cool IHS resulted in a 5.18C AVG, or 56.5C 10-second high. The improvement, again, is in a range of about 3-5C, depending on error margins. Liquid temperatures are about the same.
As far as temperature over time, the IHS change doesn’t meaningfully impact or elongate that heating period in our test setup. The Rockit Cool kit is routinely below the stock IHS by a few degrees, but there is no real impact in ramp-up or ramp-down time; at least, not one that exits error margins.
As stated in the video, we see this as a fun Saturday project for an enthusiast. It’s $14 to $20, which makes for a low-risk spend (although the process of delidding and applying liquid metal carries inherent risk), and it’s also accessible. Don’t expect great things out of the Rockit Cool IHS, but it does decently. Certainly well enough to be worth buying for someone who already has a thermal objective in mind: Maybe you want to reduce fan speed by 100-300RPM (thereby reducing noise), or maybe you’ve got some numerical OCD about thermal numbers, or you just think it’d be fun to play with computer parts for an hour. These are all good reasons to make the purchase. Pushing for higher overclocks or exceptionally lower temperatures would not be reasons to buy the IHS; if that’s what you’re after, you’d be better off buying a higher-end cooling system. Any bit helps when it comes to thermals, but other components can help more.
We liked the product for its affordability, machining quality, and value as a time-passer. Again, this isn’t something that’ll get you more frequency or lower voltage (meaningfully), but it is something that’s kind of fun to work with.
Editorial, Testing: Steve Burke
Video: Andrew Coleman