Benchmark: 1 Gigabit Ethernet
The Gigabit Ethernet performance is tested between Intel I217-V and I219-V network interfaces. Switches without 1Gbps ports will use the faster 10Gbps ports. However, when 1Gbps ports are present, those are used.
TCP Performance
To test the maximum throughput ability, I send a fixed block size of 16384 bytes. The test runs for a period of five minutes. With a higher block size, we are sure to get as high a result as possible. A similar scenario you’re familiar with, is with files. When you copy a single large file, it is faster than copying a lot of small files. Even though the two setups have the same total size.
In a real-world situation, the blocks of data that pass through a network adaptor are not of the same size each time. To give a more realistic impression of how an adaptor performs, the adaptor is once again tested for five minutes. However, this time the block size will vary from 32 Bytes up to 16384 Bytes in increasing steps of 148.7 Bytes each time.
UDP Performance
To test the maximum throughput ability, I send a fixed block size of 16384 bytes. The test runs for a period of five minutes. With a higher block size, we are sure to get as high a result as possible. A similar scenario you’re familiar with, is with files. When you copy a single large file, it is faster than copying a lot of small files. Even though the two setups have the same total size.
In a real-world situation, the blocks of data that pass through a network adaptor are not of the same size each time. To give a more realistic impression of how an adaptor performs, the adaptor is once again tested for five minutes. However, this time the block size will vary from 32 Bytes up to 16384 Bytes in increasing steps of 148.7 Bytes each time.
Product Comparison
For the product comparison, I’ll be using the average performance as primary sorting.
“These kind of fans aren’t silent, but they aren’t as loud as other’s I’ve heard. Still, it might be worth upgrading them with something like Noctua’s NF-A4x20.”
I can tell you for a FACT that the Noctuas do NOT work in this switch, so don’t waste $30 on a pair of them. After the switch is up and running, it drops the fan header voltage back to only 7v (down from 12v during start-up), and the Noctuas won’t even try to spin at that voltage. Just for fun, I tried out some 80mm fans after modifying the 3 pin header to work in the BS-MP2008. Turns out that if the fan isn’t spinning fast enough, the DIAG light on the front of the switch flashes all the time. So picking an aftermarket replacement isn’t as easy as the author of this article makes it out to be.
Regarding the original fans, the BS-MP2008 uses a pair of Deltas that are horribly out of balance, which gives them a weird sound when running, even at the lower speed caused by running them at 7v. If you fire one up loose in your hand, it literally bounces around, that’s how bad they are. I’m sure there is a quiet, smooth running Sunon MagLev fan out there that will work in this unit, but it’s proving elusive to find.
Nice article. It would be nice to know what “silent” fans do work well.
You can off-course also replace the fan with a higher quality, less noise one, using another 12-volt connection. It should be fairly straightforward, if you like modding stuff.
I also replaced the fans in my noisy Ubiquity 48-PoE switch with the Noctuas. But fortunately you can use the existing 12v connections in the Ubiquity, no modding required.