TP-Link TL-SH1832 Internal Overview
Inside the switch, there is a power supply as well as a switch PCB. One will immediately notice that there are no fans inside the switch.
The power supply is a single internal unit. This is a step down from MikroTik switches in this price range, like the MikroTik CRS504-4XQ-IN.
Here is a view of the PSU side of the switch.
Here is a look at the switch chip and SFP+ cage side of the switch.
Something that immediately sticks out is the massive thermal pads from the SFP+ cages to the top of the chassis. We have not seen a lot of these on SFP+/ SFP28 cages in the switches we have reviewed.
Next, let us get to the performance.
TP-Link TL-SH1832 Performance
Since this is an unmanaged switch, it is a plug-and-go experience. We hooked several of the 2.5GbE fanless units and other systems we had to the ports and were expecting terrible performance. Here is what we saw:
That is actually not too bad from a performance perspective. We were expecting something abysmal, like 60Gbps on this switch, from some blocking design. Instead, this appears to be a non-blocking switch.
Next, let us get to the power consumption and noise.
This is quite an interesting switch for the right application. Maybe Video Workstations with shared storage. The fact its a “dumb” switch really limits its potential for lots of home lab configurations. Still it nice to see 2.5G/10Gb mixed ports. Seems like 1Gb has been the default for so long now. Easily 15 years now.
Suddenly I wonder how the performance graph is measured. Is there a place where the testing methodology is explained?
What happens if all 2.5Gbit ports try to send data to one of the 10Gbit ports simultaneously? Do they each get a fair share of bandwidth? Are packets dropped? What about average and tail latency? What is the longest round trip when the switch is loaded?
What happens if each 2.5Gbit port is paired with another 2.5Gbit and each 10 Gbit port is paired with a 10Gbit port and bidirectional communication bandwidth measured between the pairs? What about tail latency in this case?
What happens if each port as simultaneously as possible sends data in turn to all of the others? Is this much different than the case when the data flows only between pairs of ports?
Given the equipment and switches available, it would be great to develop some tests that might characterize performance between switches in a more distinctive way.
I wonder if, like many other cheap unmanaged switches, if the switch ASIC actually has the capability but there’s just no management card installed. I couldn’t get a good look at the PCB to see if that might be the case.
The lack of fans jumped right out at me from the first edge-on photo. I’d screw this vertically to a backboard in a closet and trust it to convect just fine, but laying flat? Nah. That’s too much heat for comfort.
It looks like it’d be trivial to stick some fans in the case, though. And since it has massive heatsinks, they could be slow quiet fans indeed, and spaced back from the grille to reduce noise, while still getting the job done.
Interesting! Thank you for the article, I would love to see this in greater depth. Also, can you reach out to TP-Link for comment about the market segmentation?
10GbE switches haven’t come to consumers because manufacturers are colluding. These cost no more than 1GbE switches. It’s all a scam to extract more profits.
Is there a cheap “managed” version of this switch available in China?
Why is 2.5GbE and multi-GbE so expensive in the U.S.? Price fixing/colluding between networking companies.
This is the same nonsense pulled by the memory companies on multiple occasions in the U.S. over the last several decades, and it’s being pulled by GPU manufacturers at the moment as well.
Why sacrifice profits when you and your 1-2 competitors can all agree to keep prices artificially high?
i wish Metgear made a layer 2+ managed of this switch with 300W of PoE with 4 SFP+ cages
@Glen : For 10GbE it remains this patent : https://patents.google.com/patent/US7164692B2/en which expire this year by july so after that the cost by port for a switch with 10GbE ports will be more affordable
MikroTik please make this managers with PoE60
@flo With the way Google has been acting and performing, I wouldn’t be surprised if they draw some aspirin tablets onto the diagrams, claim it’s actually a pharmaceutical product and argue they should be granted an extended period of market exclusivity.
C’mon guys, do a proper review.
What is the Network SoC?
Is this really unmanaged or does it have ‘lite’ management?
Proper throughput testing?
They didn’t name the network chip but if you’re being fair, they’ve said they didn’t see anything on Wireshark, and a management interface would be easy to see. They also did throughput testing to the level anyone using an unmanaged switch would care about since they’ve shown it’s at least non-blocking. STH isn’t perfect, but you’ve gotta be fair.
uzman77 – the fact that this has 2.5GbE ports implies it is aimed at the consumer/prosumer market. There’s loads of problems with 2.5GbE NICs. Would be nice to do throughput testing with NICs, SFP+ transceivers etc
STH has done this in the past – for example with SFP+ transceivers where – to paraphrase you, it did as you say ‘throughput testing to the level where anyone wanting basic NBase-T would care’. (https://www.servethehome.com/sfp-to-10gbase-t-adapter-module-buyers-guide/) People who purchased on the basis of that got burned when the SFP+ transceivers turned out to be highly directional in throughput
@WB the patent is owned by Intellectual Ventures, not by Google.
I hope prices drop quickly after July 18th 2023.
Curious as to the Network SoC and other ICs on board
Now…
Make it managed
Have POE++ on at least 4 ports
POE+ on the rest and We’ve got a deal.