Test System
We tested the Kelvin S24 and S36 on our ASUS ROG test bench with the specifications listed below. We always re-baseline the fan profile in the FanXpert 2 software and run PWM mode in the ‘standard’ ASUS pre-set. The MSI GTX980 Gaming 4G is 100% silent when not under load because the fans actually stop, making it an even better choice than the incredibly quiet GTX 760 HAWK that we normally use. We want to ensure that we have as little noise as possible coming from the system so that we can isolate the sound from the CPU cooler itself.
We recorded the delta temperatures for the results – the difference between the reported CPU temperature and the ambient room temperature. In order to record a valid result, the ambient room temperature has to remain the same from the start of the test to the end of the test and must be between 18 and 24 degrees when starting. If the CPU hits 70 degrees over ambient, the test is considered a fail because that delta temperature places the CPU in the 90’s.
To create a consistent loading on the CPU for each run, we ran OCCT for 30 minutes with a set idle period of 4 minutes immediately before and after the run. The maximum temperatures across all 4 cores were taken from the OCCT report and an average generated for the results. This is a synthetic load and the closest activity we have seen that comes close to the test is video encoding. Although many games are CPU intensive, none have punished our CPU in the same way as the OCCT test so you could consider this a worst case scenario.
Although our performance test platform was the ASUS Maximus VI Gene Micro ATX based system, we also installed the S24 on our MSI X99S Gaming 7 board to see how it went on the 2011-3 platform with the different mounting arrangement.
Test Results
Delta Temperature Results
Please note that these are all Delta temperatures where the recorded temperature is the difference between the CPU and the recorded ambient room temperature which was a typically about 20 degrees. So a delta temperature of 35 degrees translates to a real world temperature of about 55 degrees.
PWM Fan Mode Test Results
Let’s start with the PWM tests. PWM is where the motherboard uses the 4 wires to control the fans and adjusts the speed as per the software profile to maintain a temperature. The standard ASUS profile will increase the CPU fan speed as it gets warmer to maintain a temperature under 75 degrees. This means that the temperatures listed in the graph below don’t mean a great deal as all coolers will have different fan speeds to maintain the maximum temperatures listed. However the graph does show coolers that struggle to keep the CPU cool under extended load (high temperature results) and we do use this test to check how hard the CPU has to work at stock settings and when overclocked.
The maximum RPM hit on the “standard” PWM profile was 800 for the S24 and the S36 – both seemed to average out at ~740rpm which is roughly 10% lower than the minimum rated fan speed. This is not unusual for fans to run 10% above or below the rated minimum or maximum speeds on the spec sheet. Both sets of fans remained barely audible and we were unable to get it to clearly register above ambient noise on our decibel meter. Performance wise, it was at the lower (better) end of the pack, just trailing the large heavy duty 140mm heat sink towers. These units were laughing at stock clocks in the PWM profile.
When overclocked, we saw all coolers coming in within 52.1 and 53.45 degrees. The PWM profile had to push the S24 fans harder to 927RPM to maintain the thermal target set by the software. This was still quiet but not as quiet as the S36 – the extra 120mm fan and surface area meant that the fan speeds for the bigger brother stayed at minimum for the whole test, hovering between 750 and 800RPM.
12v Fan Mode Test Results
When we run the fans of our coolers at 12v, we get to see what they are really capable of. It’s a less realistic test in the sense that 90% of people won’t ever run their CPU cooler flat out at 12v for day to day use. This test gives an indication as to how well the heat sink or radiator can transfer heat with the bundled fans.
At 12v under stock CPU conditions (above), the Kelvin S24 and S36 were on par with the other coolers. The stock CPU volts don’t really push the temperatures much and the above results are not as telling as the numbers we see when overclocking to 4.5GHz @ 1.29v (below).
These are great scores. The S36 used it’s massive surface area to beat all of our other coolers with a 43.5 degree delta temperature. The S24 came in on par with the larger 140mm heat sink coolers as expected.
Acoustically, at 12v, the Kelvin S24 fans registered 44dB(A) on our decibel meter at 1m away and we would consider them loud. The S36 also registered 44dB(A) at 12v but seemed louder when they were run side by side. I wouldn’t be running these in anything other than a standard PWM profile because you really don’t need to. The radiator, pump and fans demonstrate an impressive ability to exchange heat and keep our CPU cool. Just remember that this is a synthetic test and a situation that most system builders wouldn’t encounter under normal conditions.
Fractal Design “Kelvin vs Kelvin” Test Results
The above graph shows the thermal results we saw from the different configurations; effectively this graph shows the Kelvin S24 competing against the S36.
Idle Delta Temperatures
The idle temperatures were all consistent and in line with other performance coolers. Measuring idle temps can be finicky due to minor fluctuation among the ‘idle’ processes so we don’t give it much weight. Typically we expect delta temperatures around 8 degrees mark as a standard and the Kelvin S24 and S36 were no different. What this shows is that we had a good ‘mount’ of the water block to the CPU.
Load Delta Temperatures
The Kelvin S24 and S36 easily managed our Haswell i5-4670K at stock settings and overclocked in both the excessive synthetic benchmarks and in extended gaming sessions. The S36 wasn’t really stretched and the fans ran at idle almost all the time in general use, gaming and video encoding – meaning that it is total overkill for a test bench like ours. This is where the expansion feature starts to look attractive and we start to wonder how things would look if we were to extend the loop to something like an ASUS Poseidon hybrid water/air cooled graphics card…
Pump Performance
The other factor not reflected in any of the graphs is the pump speed. For testing purposes, we ran the pump at 12v but our Fan Xpert 2 software from ASUS recognised the pump as a fan and after ‘tuning’ the software to align with the pump, the software determined that the pump could be tuned – the only catch is that the range isn’t that big. The thermal performance differential between the standard 12v and 7v was 2 degrees under stock CPU conditions and 3.5 degrees when overclocking. At 12v/full speed, the pump unit was audible on our S36 model but not annoying. The S24 was a different story and the pump was the loudest thing in the room at 12v but bearable at 7v. This was interesting and unexpected given that both pumps are the same spec and we put it down to a manufacturing quirk of us receiving an early sample of the S24 but a later sample of the S36.
