Installation and Performance
The best part about the Celsius S24 using the standard Asetek brackets is that for anyone who has ever installed one before the rest of the installation is normally easy. In the case of our X99 test bench or a new AM4 or Threadripper build, you don’t even have to use a new backplate. You just pick out the correct standoffs, screw them on, then make sure you have the correct bracket and the kit drops on. Then you use the large thumbscrews to hold it in place. The radiator is a little more complicated, you will have to install 8 screws to mount it all. I did all of this but ended up with a problem. I originally thought the new AM4 standoffs with the guards over them were for the 2011 socket and they worked well but they were the wrong height. The instructions didn’t help but I did figure out that the correct standoffs were in the bag with the mainstream Intel standoffs. Labels on the bags would have been huge here, given my experience and I still had an issue you might want to be extra careful not to mix these up. Once swapped out I was completely good to go. If you think you may have them switched, you should be looking out for warmer temps, it did cool for me but the whole setup ran warmer. About 46c-50c for idle temps.
So with everything installed, I started out my testing with idle testing. I always do this just to make sure that we have a good installation before applying load. These numbers don’t matter much other than that. This did help me catch that our installation was wrong as mentioned above.
Once comfortable with the installation I jumped right into temperature testing. I tested with the pump set to PWM for consistency with past testing but I will note that idle temps did go up when I had everything set to auto. The Celsius S24 did a lot better than I expected here. It was up against the CLC 280, another quality kit with a large radiator and fans so coming in within one degree on both tests was impressive, especially in the 100% fan speed test that the S24 came in cooler.
Of course, temperatures are only part of the picture, some people move to water cooling for the noise as well. This tested ended up being interesting because the fan speeds run through the pump. This means I couldn’t control just the fans or just the pump and it also means the RPM readouts are for the pump. In fact, when doing this testing I noticed that at the 25% fan speed setting the two fans were both off. Turning the RPM back up to 1250 RPM turned one fan on and 1300 RPM got both moving again. But it does show that the built in fan splitter might not always be best, especially with motherboards now having both fan and pump fan headers. Anyhow when the fans were working they were your standard Fractal quiet, running much quieter than the loud CLC 280 and at 50% fan speed they were almost as quiet as the Noctua, though to be fair at that setting the fans were actually running closer to 25% fan speed I would imagine given the controller issues. Air cooling with a quality cooler was still quieter though, no matter how you slice it.