L3+ Cheap and Effective noise reduction

Here's what $16.88 at Walmart will get you!
Here’s what $16 at Walmart or Amazon will get you…

When I bought my first L3+ miner in early 2021 I was excited to plug it in, set it up, and watch the money pour into my account. I had planned for everything I needed to hit the ground running once it showed up. I had installed a 240V/20A breaker so I could run 3 off the circuit, got some insulated ducting to run the hot air outside, a network router and cables, and of course the 250V/15A rated NEMA cords.

Why a 240V/20A for 3 you may ask? Well, I’m an EE by trade (haven’t done that in a while) so I put my old knowledge to work and did the quick calculation. The rough numbers, 240V x 20A = 4800W. Taking into account electrical 80 rule and not completely trusting what the continuous loads on the L3+ are, I don’t want to load the circuit anymore than 70% (80% is the 80 rule, I’m just being more conservative since it’s running in my house.) So 4800W x 0.70 = 3360W. If we did follow the 80 rule, we are at 3840W, just about what we need to run 4 units.

When I installed my first unit my first thought was, “Holy crap this is loud, my wife is going to be pissed.” Well, let’s say not pissed, but she kindly encouraged me to find a solution to the noise or find a new hobby. That led to quite a bit of research (hence why I started this website in the first place) of ways to quiet the units down without compromising them on heat. I saw various oil based systems (too messy for home), specialized server cabinets (too big and expensive for me), walling in and insulating part of your house to damped noise (I’m not that dedicated yet), then I saw someone post about a small cooler. I wish I could find the original post I read, it really got me thinking. Most of the comments annihilated the person for doing this, “you’ll overheat”, “you’ll melt the cooler”, “you’ll burn out the power supply”, and “where will you put your beer after you cut that all apart?”

Being new to this I took it all in, pretty much all valid points. I decided to move forward with utilizing a cooler, in fact, I tried with two different sizes (a 28 quart and a 48 quart), thank you Amazon! The purpose behind the two sizes was to try one with the power supply outside and one with it inside.

48 Quart (with power supply inside)

Realizing these 4″ fans push a massive CFM for their size, and regularly run 4-6K RPM, I opted to make a much larger intake opening (6″) in the cooler and not restricting it to a 4″ duct. On the exhaust I did install a 4″ duct adapter (4″ ducting end cap from Coolerguys.com) and ran this directly outside. I put a 6″ insulated duct hose on the input that just ran to the cooler and a 4″ insulated ducted hose to the exhaust end cap directly attached to the L3+. This would insure all the hot air would blow directly outside and not into the cooler and plenty of cool air would make it in to cool the L3+ and the power supply. I did some quick measurements of the db levels before and after for reference:

  • 10 ft from L3+ in open air – 75db
  • 2 ft from L3+ in open air – 90db (Bitmain says 60db but I don’t believe that includes the power supply)
  • 10 ft from L3+ in cooler – 61db
  • 2 ft from L3+ in cooler – 71db

But the most important number is that I was under 40db with the door closed, whereas before it was over 60db. My wife can now walk by the room to a light white noise instead of a loud APU from a 747.

I also checked the temperature on each of the hash boards. I was running the system on Hive OS, auto-tuned at the 610 MH/s at ~890W, and before running in the cooler could maintain my hottest chip temp around 67. After putting it in the cooler I only jumped up to 69, essentially no change at all. I credit this to two things, first is the amount of cool air I allow in by using a 6″ opening which more than doubles the area of the exhaust. The second is the variability in the fan speeds. I did worry my fans might be running too hard after this however they only run a few hundred RPM faster and hover around 5000-5200 RPM.


28 Quart (with power supply inside)

I followed the same setup with respect to ducting however the power supply was placed externally due to room. Since I didn’t have to worry about overheating a power supply I opted to do another slight modification, I added in 1″ acoustic sound-proof foam to see if I could knock the noise down even more. What I quickly realized was that this foam only helped raise the chip temp to the mid-70’s. I took my db readings and realized that our noise problem wasn’t just the L3+, it’s also the power supply. While not nearly as loud as the fans on the L3+, my db readings were almost exactly the same.

Not quite the same success…


I would say if you have the room and $20 extra that going with the 48 quart cooler option is nice. However if you don’t need it, no reason to do it. Here’s a quick video of my first iteration that includes the sound to get a real feel of how the cooler knocks down a significant amount of sound.

Update 7/1/21

If you’ve wondered why I did insulated ducting over traditional dryer duct, I did it for noise and not for any ambient heating purpose. However I have found that switching to 4″ flexible metal dryer ducting works just as well and is half the price. The temperature never gets high enough from the exhaust to warrant one over the other, nor does the noise, but in the end cost rules the day. 25′ of 4″ insulated ducting is about $40, whereas 12′ of 4″ dryer ducting is about $9.

Update 12/4/21

I’ve found, after buying a few different power supplies and experimenting with fan speeds, that I can run the units just as quiet by using a better power supply (APW7 or Gold/Platinum ATK) and setting the fan speed manually. I’ve also found that usually off-brand cheap power supplies from China are inherently loud and unreliable. Also stay away from server power supplies (like the HP or Liteon rack mount) as these are also louder than the L3+ themselves.

One thought on “L3+ Cheap and Effective noise reduction”

Leave a Reply

Your email address will not be published. Required fields are marked *