Decibel level measurements

I made sound level measurements on 28-Feb-2026. The purpose was to get an idea of how the noise level of Mark Ross-Lonergan's new server, mannanan.nevis.columbia.edu, at full load might impact the working environment.

I made the measurements with two devices:

• An app called Decibel Meter running on my iPhone 16.
• A cheap decibel meter, TopTes TS-501B, I happened to have for measuring the volume levels of my home stereo.

These are just relative results; see the notes below. Also, there are John's concerns that, since mannanan is plugged into a single 120V source when it's supposed to be plugged into two 208V/240V outlets, the sound I measured may be lower than when the correctly-powered machine actually runs at full 2500W load. At best they're a guide. They're certainly not a professional acoustic measurement.

Units are decibels (dB)

Room	iPhone	meter
116 (Bill's office)	38	31
104 (Mike's office)	39	35
103 (John's office)	41	36
119C (SW storage room)	43	40
120 (Neutrino student office)	45	41
120 (Neutrino student office)	46	40
119 central area	53	49
119C (SW storage room)	53	49
119 central area	62	58
119K (computer room)	73	72
119B (network room)	76	70
119K (computer room)	83	80
Ambient noise levels
During mannanan boot

General comparison, from the web page I referenced above:

40dB	Comfortable background noise.
50dB	Normal household activity.
60dB	Noticeable but not disruptive.
70dB	Loud; sustained exposure can be stressful.
80dB	Hearing damage possible after 2–4 hours.

Notes:

• The iPhone app had a systematically higher reading than the meter. The small size of an iPhone microphone might have something to do with this.

• For the ambient sound levels, I left the app and meter running for about a minute and used their built-in averaging tools for the number in the table.

• I took readings in our offices to give us a "feel" of the normal background noise. John's office is a bit louder because of that fan unit (A/C?) outside his window.

• Room 120 is louder because of the fans in the network room (119B). By far the loudest fans are those in the firewall. Perhaps the new firewalls will be quieter.

• I measured the sound in room 119C (where Amy stores supplies for the Science-on-Hudson talks) to represent the sound level that the SEAS faculty might be exposed to.

• A ~20dB difference in noise level between inside the computer room enclosure and the central 119 area shows that there's already some level of sound isolation between the two.

• The loud sound during mannanan's boot only lasts a couple of minutes. I did the best I could to average the readings over a 10-15 second interval.

• Within the general 119 area, mannanan's boot sound adds ~10dB to the ambient sound. Sound decibels aren't additive, but that's the effect on the values I measured.

• Once you go outside room 119, you can certainly hear the sound of mannanan's fans (all the way to my office!), but it becomes part of the general background sound. In room 120, the sound from the network room dominates the sound from mannanan; though the boot sound was audible, it didn't add significantly to the ambient noise.

• I tried to look up the power supplies' specs. Their model number is R-41230782 and they're made in China. A web search gave me some electrical specifications, but they did not include noise levels.

• Most of the system's noise is coming from cooling fans that I assume direct air over the GPUs. Those fans' model number is PIE080Q12H; they're also made in China. I found some certifications for these fans (published by Foxconn, located in Taiwan) but there was no information on noise levels I could find.