Super Silent Supermicro SuperServer E300 8D

Page content

Silence In 2017 I wrote about my new My HomeLab purchase. If you want to take a look on my HomeLab details use link Supermicro SuperServer E300-8D. It costs me a lot because I upgraded it to maximum capacity in terms of RAM and I was very happy about it. I installed my whole VMware environment and I was kind of using it.

Unfortunately, I didn’t use it as much as I wanted. Under heavy load it was extremely loud. You cannot use it 24/7/365 at Home in normal environment.

Finally, I started to read in the Internet on different forums that some folks successfully replaced the fans on other different Supermicro Motherboards marked as X10SDV-TP8F. Following that lead, I purchased three Noctua NF-A4x20 PWM (4 PIN) fans and I replaced the stock ones.

Stock Fan replacement on SYS-E300-8D

In first phase I tried to replace the fans and mount them using rubber Noctua mounter.

Supermicro SYS-E300-8D-1

Supermicro SYS-E300-8D-2

Supermicro SYS-E300-8D-3

Supermicro SYS-E300-8D-4

Supermicro SYS-E300-8D-5

As you see on this photo they look very nice.

Supermicro SYS-E300-8D-6

Mounted to the chassis they didn’t fit. I had to revert to screws.

Supermicro SYS-E300-8D-7

Noctua use some kind of weird screws so bear in mind that you need some angle when using screwdriver.

Supermicro SYS-E300-8D-8

Temperature results

CPU Temperature Fan 1 (RPM) Fan 2 (RPM) Fan 3 (RPM) CPU Fan (RPM) Noise (dB)
Stock Fans Boot 48 4700 5000 4500 N/A 69
Idle 44 5200 5100 5600 N/A 53
Full 3 min 60 6300 6200 6700 N/A 61
Full 10 min 61 6300 6200 6700 N/A 61
Noctua NF-A4x20 Boot 56 1100 1200 1300 N/A 35
Idle 61 2000 2100 2000 N/A 36
Full 3 min 84 4300 4400 4300 N/A 41
Full 10 min 88 4700 4800 4700 N/A 45

Conclusion

Replacing stock fans really pays of! If you have X10SDV-TP8F based Supermicro server do it. It is totally worth it. Although you have to accept much higher temperatures.

Replacing Stock SYS-E300-8D Heatsink

I didn’t stop there and I wanted to try out if replacing stock heatsink will help lower down temperatures. I purchased SNK-C0057A4L without knowing if it fits 100% on my server. Although Paul Braren from TinkerTry tested many Supermicro servers, he didn’t test Supermicro E300-8D with active CPU cooler SNK-C0057A4L. Without much to loose (only return shipment to the store if it wouldn’t work :) ) I replaced stock CPU heatsink with new active CPU cooler.

SNK-C0057A4L

Heatsink removal

I suggest to heat a CPU a bit to make it easier for you to remove old heatsink. You can do it with hairdresser or as I did. I started my HomeLab for a few minutes, quick shutdown and after removing screws I was able to remove it without issues.

In package there is as well backplate which was supposed to be installed on the motherboard. In my case I skipped it.

Supermicro SYS-E300-8D-9

Supermicro SYS-E300-8D-10

Supermicro SYS-E300-8D-11

I cleaned up the processor and without any issues I installed new CPU cooler.

Supermicro SYS-E300-8D-12

Temperature Results

Noctua NF-A4x20 & SNK-C0057A4L Boot 59 2200 2100 2200 4100 43
Idle 56 2400 2400 2400 4300 39
Full 3 min 73 3400 3400 3500 5400 44
Full 10 min 75 4000 3900 6000 3900 48

As you see, processor temperature decreased significantly, especially under full load from 88° to 75°.

BIOS settings

Almost at the end of the test I forgot to check what kind of settings I configured in BIOS 🤦‍♂️. Generally speaking all test were conducted on Set Fan to Optimal Speed settings. I realized that if I change the BIOS fan setting to Set Fan to PUE2(Power Utilization Effectiveness) Speed.

Power Utilization Effectiveness BIOS settings

Power Effectiveness Boot 56 1300 1200 1200 3200 38
Idle 59 2200 2100 2100 4100 38
Full 3 min 76 3200 3100 3100 4800 42
Full 10 min 78 3400 3200 3200 5200 44

With that setting in BIOS configured I finally found the sweet spot between noise and decibel level. 44 dB is ideal for my living room and it generates less noise than my Ubiqiti switch.