Corsair’s H100i Pro 240-mm closed-loop liquid CPU cooler reviewed

Earlier this year, Corsair announced its Hydro Pro series of closed-loop liquid coolers. Those heatsinks introduced a stylish new pump-head design kitted out with RGB LED lighting, and they tapped Corsair’s exceptional ML-series fans for air-moving power at the radiator end. The first two heatsinks in that series covered two important bases: the H115i Pro for 280-mm radiator mounts, and the behemoth H150i Pro for cases with 360-mm radiator hardpoints.

The bread and butter of the closed-loop liquid-cooler market lies in smaller radiators, though, and 240-mm heat exchangers are the place to be for most builders with mid-towers and mainstream CPU sockets. Today, Corsair is rounding third and headed for home with the 240-mm H100i Pro.

Since the radiator size on this cooler is the only thing about it that’s changed versus its larger cousins, we won’t recap every detail of this new series of heatsinks here. If you want that in-depth look, have a gander at our review of the H115i Pro and H150i Pro now. Instead, we’re going to concern ourselves with noise levels and cooling performance first and foremost.

There is one major change on the H100i Pro apart from its radiator size. The H100i’s pair of Corsair ML120 fans can spin all the way up to 2400 RPM, compared to 1600 RPM from the ones included with the 360-mm heatsink. Both coolers have a 400-RPM minimum fan speed. You can’t tell the low-speed and high-speed fans apart externally, though.

Past that, the H100i maintains the same simple mounting system, integrated fan controller, RGB LED illumination, and most everything else we liked about the first two Hydro Pro-series coolers. Let’s see how it performs.

Our testing methods

Here are the specifications of our test system:

Processor Intel Core i7-6950X
Motherboard Gigabyte X99 Designare EX
Memory 32 GB (4x 8 GB) Crucial Ballistix Elite DDR4-2666
Graphics card Gigabyte GeForce GTX 1050 G1 Gaming
Storage Intel 750 Series 400 GB
Power supply Seasonic Prime Platinum 1000 W
OS Windows 10 Pro

Our thanks to Gigabyte, Intel, and Corsair for helping us to outfit our test systems with some of the finest hardware available. Aerocool provided the P7-L240 that will serve as the foil to the H100i Pro today, as well.

To test heatsink performance, we run the Blender “classroom” benchmark file three times in succession and report the maximum CPU package temperature observed in HWiNFO64. Blender is a demanding real-world workload that can still stress an overclock to failure. Prime95 Small FFTs generates a lot of heat, sure, but the heat and power consumption it generates are both far in excess of any real-world application we’ve ever observed.

Although our recent heatsink reviews have used Intel’s massive Core i9-7980XE as a heat source, the fact that the latest Extreme Edition chip uses thermal interface material rather than solder under its heat spreader has bugged me while testing CPU coolers. We want our performance results to be limited as much as possible by the performance of the cooler under test, not the thermal transfer capability of a variable we can’t control.

With that in mind, I set aside our X299 test rig in favor of the Core i7-6950X and its soldered heat spreader. At 4.4 GHz and 1.38 V, the ten Broadwell cores of the i7-6950X are still a formidable match for any cooler. We’ll see whether the move to solder helps tease out any performance differences from these heatsinks.

The ambient temperature of our testing environment was maintained at 78° F, plus or minus roughly 1° F, over the course of each test by monitoring with a calibrated thermometer throughout.


Overclocked cooling performance

No revelations here, really. The H100i Pro beats out the already-impressive Aerocool P7-L240 by 2° C under these conditions, and that’s good news for the Corsair heatsink. Still, the margin of victory is slim. Let’s see if our noise measurements can put some distance between these coolers.

Noise levels

To measure noise levels, we use the Faber Acoustical SoundMeter application running on an iPhone 6S Plus. Measurements were taken 18″ from the intake side of the radiator on our test bench.

Turns out that the H100i’s performance comes at a price. At idle and at the 1000 RPM speed we’ve picked to represent light spikes in usage, both of these coolers are practically neck and neck if you go by our meter. To deliver its cooling-performance victory above, however, the H100i Pro has to get noticeably louder at full speed.

dBA measurements don’t tell the whole story, of course. At 1000 RPM, Corsair’s fans barely make any noticeable sound at all. At full speed, the Corsair spinners have a mild but noticeable high-pitched tone. Even so, their overall noise character is remarkably broad-spectrum and pleasant, although it’ll likely be difficult to ignore entirely thanks to its high volume. Still, I was shocked to see that 54.3 dBA number on my meter. The ML120s sound like much quieter fans to the ear.

Past Corsair 120-mm fans really let it all hang out when operating at their limits, so the refinement of these high-speed ML120s is welcome. Even though the Aerocool P7-L240’s fans are quieter in absolute terms, their noise character is considerably more complex than that of the Corsair spinners.

The one potential shortfall in the H100i Pro’s noise character comes from its pump head. Like its 280-mm and 360-mm siblings, the pump on the H100i Pro isn’t the smoothest-sounding affair. The pump’s quiet mode lets it fade into the background of our otherwise library-like testing environment, but its balanced and extreme modes reveal a gravelly whine that’s harder to ignore. I normally would be willing to forgive this behavior, but the Aerocool P7-L240’s pump is nearly silent even at full speed and has no such rough edges.

If the H100i Pro could dynamically adjust pump speed in response to coolant temperatures, I would be willing to forgive it this annoyance, but it can’t—switching pump speeds has to be done manually through iCUE every time a builder anticipates a change in load conditions. Changing the cooler’s pump speeds from even the balanced mode to the quiet mode has a major impact on performance, too, so builders can’t just set up quiet mode and forget it on demanding systems.


Corsair markets the H100i as a quiet cooler, and it can certainly be that if its owner wants. That tack sells the performance of this heatsink short, though. Though its margin of victory wasn’t a large one, the H100i proved more capable at cooling our test system than even Aerocool’s P7-L240, a recent 240-mm favorite of ours.

That high performance does come at a cost. The H100i gets quite a bit louder than the P7-L240 under full load, even if the excellent noise character of its bundled ML120 fans blunts the edge of the absolute sound pressure levels at play. The sound quality of the H100i Pro’s pump at full tilt is also a bit intrusive when the rest of a system is idling, and the mass of bundled cables from the H100i’s pump head might annoy some builders, too.

Of course, it’s not necessary to run the H100i Pro all-out all the time. At moderate pump and fan speeds, Corsair’s latest can barely be heard at all, and those settings are plenty capable of keeping stock-clocked CPUs in check. The aluminum-and-RGB-LED-accented pump head on this cooler looks as good as ever, and its mounting system is easy to use and easy to adapt to different CPU sockets.

All told, the H100i has range. It can be as quiet or as powerful as a builder needs it to be, it’s easy to install, and it’s competitively priced for the performance it delivers at $120 list. If the iCUE utility ever learns how to switch pump speeds in response to coolant temperatures, Corsair’s Hydro Pro-series coolers will be nigh unstoppable. For the moment, the H100i comes TR Recommended.

Comments closed
    • GatoRat
    • 3 years ago

    Interesting review, though it would be helpful to add some non-liquid coolers to the comparison.

    • farmpuma
    • 3 years ago

    “The pump’s quiet mode lets it fade into the background of our otherwise library-like testing environment, but its balanced and extreme modes reveal a gravelly whine that’s harder to ignore.”

    I wonder if the pump motor uses ball bearings? I’ve had many ball bearing fans which were fairly loud at first, but quieted down quite a bit after running for only a few days. I suspect the bearing races just needed a little time to break-in and smooth out.

    • 3 years ago

    What would have been the most valuable thing to me is to see how this cooler compares to its bigger brothers the H115i Pro & H150i Pro, then I could make a decision on how much to pay for whatever cooling differences there are.

    Instead your review literally tells me nothing of value.

    • blastdoor
    • 3 years ago

    Do folks have a sense as to the relative benefits of going from air to liquid versus moving the air-cooled PC into a cooler room? Like, move it from the second floor into the basement?

    If I really bite the bullet on TR2, it will be a headless Linux beast sitting in my basement that I remote into from my Macs. Given that it’s in the basement, I figure (1) it will be in a cooler room and (2) I don’t have to care about sound. This makes me wonder if I’ve got a decent chance at hitting the magic 68 degrees C without anything more than a good TR2-specific air cooler.


    edit — thanks for all the replies! Very helpful. Sounds like air is the way to go, combined with a cool room and perhaps a big fan pointed at the case 🙂

      • Usacomp2k3
      • 3 years ago

      A headless system would make me worried if the pump ever does or leaked or malfunctioned. I would certainly stick to the reliability of a traditional HSF if it were me.

      • dragontamer5788
      • 3 years ago

      For TR 1950x, I can say for 100% sure that the Noctua u14s TR4 air-cooler is sufficient, at least when in dual-fan mode (I bought another $20 A15 fan to do a push/pull configuration).

      I run my house relatively warm at 76F (~25C), and my Noctua u14s can keep TR1 below the 68C throttle point in most practical tasks. (It hits 68C on Prime95, but nothing else, and it only hits that after 10-minutes of warming up my room to ~79F).

      The issue is that TR2 is going to be 250W, which is WAY more power than the 180W from TR1. So we’ll just have to wait and see how cooling solutions work as another +70W added onto it.


      What temperature do you expect your basement to be?

      Another consideration: you might be able to do better than me by adding more fans to the case.

      I’ve never seen anyone run the benchmarks, but I’m pretty sure if you point this fan at your motherboard, it’d do better than any Noctua: [url<][/url<]. There's just more that you can do if we're talking about a dedicated room 🙂

        • Srsly_Bro
        • 3 years ago

        There is a tr4 silver arrow coming out. Tpu posted a news link on it a few weeks back.

      • Srsly_Bro
      • 3 years ago

      For an unintended system, get a tower cooler. There will be a tr4 silver arrow and the new tr4 noctuas that will never leave you worrying about leaks.

        • chuckula
        • 3 years ago

        I like the idea of unintended systems.

          • Generic
          • 3 years ago

          There’s just no stopping unintended systems when your attention is elsewhere.

      • Goty
      • 3 years ago

      I can’t actually find any evidence of throttling at 68C, just a reported “max temperature” by AMD, which seems like it may be more like a prolonged operation maximum more than it is analogous to TjMax on Intel processors (the actual throttling temperature.) There are plenty of examples floating around with people running their cores at higher than stock speeds at 80C or above with no throttling.

        • dragontamer5788
        • 3 years ago

        [quote<]There are plenty of examples floating around with people running their cores at higher than stock speeds at 80C or above with no throttling.[/quote<] Note that Threadripper adds 27C to its tCTL value. So if Threadripper reports 95C, the Tjunction is really 68C. ---------- My own tests have never seen Threadripper (on stock settings anyway) go above 68C. I can drop the fan speed down, but Threadripper downclocks at 68C even as the fans stop spinning. I guess overclockers can overcome this limit, but I expect most people who get an Air Cooler to just run stock.

          • Goty
          • 3 years ago

          I would expect overclockers to be running monitoring software that doesn’t report temperatures with the 20C offset.

    • chuckula
    • 3 years ago

    Great review Jeff, although the 6950X is an interesting test candidate for the CPU. I guess as long as it produces the heat that’s really all you need for seeing what the cooler does.

    Incidentally, with Ripper2 coming soon, what is that going to mean for cooler compatibility?

    IIRC there are coolers out there that work fine with both LGA-2011/2066 and the 2-die Ripper1 because half of the IHS really doesn’t need to be covered. You just need a mounting adapter to work with both platforms. However, that won’t work with Ripper2.

      • derFunkenstein
      • 3 years ago

      With four-die Threadripper, I would think you’ll definitely need a bigger plate. The IHS can’t possibly have been intended to carry heat to a smaller block, leaving CPU dies with (warning: stretched-out metaphor incoming) just a bedsheet and not the comforter, too.

        • chuckula
        • 3 years ago

        Yeah, from what I can gather it’s not a problem to have a part of the heatspreader exposed [i<]if[/i<] there's no chip underneath that part of the heatspreader, which is the situation with Threadripper 1. It becomes a problem when all of the sudden there's now a chip (or two) underneath exposed parts of the heatspreader.

      • Jeff Kampman
      • 3 years ago

      I don’t know anything official yet (honestly!) but I can’t imagine AMD will countenance anything but full-coverage air coolers or liquid coolers with XL cold plates for use with TR2.

        • dragontamer5788
        • 3 years ago

        I can only hope that more full-plate liquid-options were available. I was severely disappointed with the Enermax Liqtech TR4.

      • Goty
      • 3 years ago

      [quote<]IIRC there are coolers out there that work fine with both LGA-2011/2066 and the 2-die Ripper1 because half of the IHS really doesn't need to be covered.[/quote<] I believe the active dice on TR are diagonal to one another, so I don't think that's the reason coolers with smaller cold plates worked adequately with TR.

        • dragontamer5788
        • 3 years ago

        Thermal resistance formulas: [url<][/url<] Estimating the effects of heat is relatively simple if you know how to do circuits / resistors / etc. etc. Temperature differences are equivalent to voltage, heat (Watts) is equivalent to current. Thermal Resistance is equivalent to... yes, resistance. The cold-plate of an AIO cooler is rated for a certain level of thermal resistance. Similarly, Threadripper's IHS on top of the dies has its own value of thermal resistance. The main issue is that a larger heatplate would give more surface area and better take advantage of the larger Threadripper IHS. ----------- Threadripper hits you with one major difficulty: thermal throttling at 68C. While Intel chips throttle at 100C (worst-case design of ~75C differential, at 25C ambient), you only have ~43C to work with in Threadripper. This means that any cooling solution needs to work way more efficiently to properly cool Threadripper. As such, you need to work with each advantage as much as possible. A bigger IHS helps... but only if you have a matching large-sized heatplate as well. AMD solders its IHS and provides a larger area to attach a cooler. Its on the system builder to fully take advantage of these advantages however.

        • Chrispy_
        • 3 years ago

        That’s what I was going to say, before reading all the comments.

        [url=<]Delid pic[/url<] TR1 has a fairly chunky heatspreader, and the dies themselves are likely covered by most waterblocks, if not fully, at least mostly. The IHS, soldered will do a half-decent job of transferring heat from any uncovered die area to the waterblock so although not ideal for local die hotspots, that thermal energy is at least getting away from the CPU and into the water loop.

Pin It on Pinterest

Share This