CRS328 does not suck 😁

tangent · August 30, 2022, 8:33pm

There’s too much griping on this board, so in an effort to counter some of that, I offer this:

% iperf3 -c …
Connecting to host …, port 5201
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-1.00   sec  1.09 GBytes  9.37 Gbits/sec                  
[  5]   1.00-2.00   sec  1.09 GBytes  9.41 Gbits/sec                  
[  5]   2.00-3.00   sec  1.09 GBytes  9.40 Gbits/sec                  
[  5]   3.00-4.00   sec  1.10 GBytes  9.41 Gbits/sec                  
[  5]   4.00-5.00   sec  1.09 GBytes  9.40 Gbits/sec                  
[  5]   5.00-6.00   sec  1.10 GBytes  9.41 Gbits/sec                  
[  5]   6.00-7.00   sec  1.09 GBytes  9.40 Gbits/sec                  
[  5]   7.00-8.00   sec  1.09 GBytes  9.41 Gbits/sec                  
[  5]   8.00-9.00   sec  1.10 GBytes  9.41 Gbits/sec                  
[  5]   9.00-10.00  sec  1.09 GBytes  9.40 Gbits/sec                  
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-10.00  sec  10.9 GBytes  9.40 Gbits/sec                  sender
[  5]   0.00-10.00  sec  10.9 GBytes  9.40 Gbits/sec                  receiver

iperf Done.
% iperf3 -s            
-----------------------------------------------------------
Server listening on 5201 (test #1)
-----------------------------------------------------------
Accepted connection from …, port 55203
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-1.00   sec   938 MBytes  7.87 Gbits/sec                  
[  5]   1.00-2.00   sec   939 MBytes  7.88 Gbits/sec                  
[  5]   2.00-3.00   sec   946 MBytes  7.94 Gbits/sec                  
[  5]   3.00-4.00   sec   935 MBytes  7.85 Gbits/sec                  
[  5]   4.00-5.00   sec   943 MBytes  7.91 Gbits/sec                  
[  5]   5.00-6.00   sec   931 MBytes  7.81 Gbits/sec                  
[  5]   6.00-7.00   sec   941 MBytes  7.90 Gbits/sec                  
[  5]   7.00-8.00   sec   948 MBytes  7.95 Gbits/sec                  
[  5]   8.00-9.00   sec   948 MBytes  7.95 Gbits/sec                  
[  5]   9.00-10.00  sec   944 MBytes  7.92 Gbits/sec                  
[  5]  10.00-10.00  sec  3.71 MBytes  7.84 Gbits/sec                  
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-10.00  sec  9.20 GBytes  7.90 Gbits/sec                  receiver
-----------------------------------------------------------
Server listening on 5201 (test #2)
-----------------------------------------------------------

Okay, what’s so great about that? It’s that this is between an iMac and an M1 MacBook Pro with generic SFP+ modules, not between two studly servers with MikroTik branded optics. Both connect to SFP+ ports on the CRS328, one via 10G copper, and the other via LC-LC OM4 MMF over Thunderbolt 3.

The last time I tried this, it would peak at 5-6 Gbit/sec for a while, then fall over, choke, gasp, and retry. None of the hardware has changed in the meantime. As far as I can tell, all of the improvement is in RouterOS software improvements.

This result is what I was expecting when I went searching for new networking gear last year and ended up picking MikroTik. I’d have been happier to get this result out of the box, but now that I finally have it, this result says 10G to the desktop is here, even for home users.

Thank you, MikroTik!

holvoetn · August 31, 2022, 9:36am

It’s human nature to make more noise about complaints then about what works fine

Good post.

tangent · August 31, 2022, 2:51pm

In case anyone’s wondering why there wasn’t a prior post from me titled “CRS328 sucks ,” it’s because I had a whole list of candidates for why I wasn’t getting the advertised speed, and I use this 10G link rarely enough that I had yet to eliminate any of the possibilities:

Host OS weirdness
General Thunderbolt weirdness
Driver problems with this particular Thunderbolt to SFP+ adapter
Generic modules; hadn’t yet broken down and decided to get a set of MikroTik modules to test with
The pauses were due to some host-side stall, and 5-6 G is considered normal for this class of hardware, such that it’s expected to take hardware advancements to get beyond that. (Really? Yes. I’m old enough to remember when filling a 10 megabit pipe with a single host was difficult.)

I only retested it on a whim and found out that it was apparently none of these things.