Hello Guys,
Going to buy CRS812-DDQ switch. However, i want to know if the 8x50G ports can be splitted into 16x25G ports. This means 2x25G per 50G port via a patch cable. Can anyone confirm on this? Mikrotik staff are super rigid and neither they mention such details on the specs nor they tell when contacting them and always ask to contact a consultant.
Thanks
On brochure...
- QSFP56/QSFP-DD ports also support breakout modes to 1G/2.5G/5G/10G/25G/50G
Yes, that is mentioned and supported but my question was about 8x 50G (SFP56) ports the switch has.
As its not in the brochure you can take a guess 
But for more of a direct answer:
No
SFP56 is a single Lane. So only one “connection”.
Youd need a switch 
If you didnt know this maybe you should ask a consultant before buying
Also you got that reply already 3 days ago:
Thank you for the clarification on this. What about the 200G and 400G ports? Can they be split into 100G and 200G?
2x200G (switch) = 4x100G (breakout)
2x400G (switch) = 4x200G (breakout)
According to specs, the Q in QSFP stands for "Quad" ... meaning that interface uses 4 electrical data lanes ... which are converted to 4 optical lanes by QSFP module ... And the DD in QSFP-DD stands for "Double Density" which means that number of interface lanes (and consequentially optical lanes) is doubled to 8. The 56 in name is talking about speed over each individual lane so it doesn't contribute to the "split ratio".
So the above means that breakout cables from QSPF can create up to 4 independent connections and breakout cables from QSFP-DD can create up to 8 independent connections. Which means it can't be split into 16 individual connections (regardless the speed).
Speed of individual connection depends on variety (QSFP vs. QSFP56) and module implementation (I guess not all modules can support different speeds of individual lanes ... even though they are used independently).
And, BTW, that's the standardization (and physics) of QSFP ... not much to do with individual vendor implementation.
https://www.fs.com/products/182807.html
I have this cable and a broadcom 57608
(LRES1260PF-2QSFP112) which does physically has 2x200g ports, and supports “aggregate 2 ports to 1” via uefi setting
In this mode the 2x200g to 400g DD cable from fs.com does not work with the 400g DD cable into the CRS812 side. Running 7.22.1 (stable)
This nic works fine in not-aggregate mode (2x 200g independent ports) with 200g AOC cable.
When in this mode the vendor name and other info about the length, power, voltage and temperatue of the QSFP module is not reported. Says initalizing. Which makes me think possibly it’s a firmware bug or a problem detecting the cable itself and there may be hope for this working in 400g “aggregate” mode instead of 2x200g as shown in patrick’s video?
While this should technically work, why not just use the "disable port 2" (400G single Port) Option?
Because this particular nic does not physically accept 400g dd connectors? There's two versions of this nic one with a single qsfp dd connector and one with two qsfp connectors.
The broadcom manual did state otherwise but I'll take your word for it since you have the card 
Other than that open a ticket at help.mikrotik.com
No; I want to use the 400g port at 400g. It is 400g QSFP56 DD but the device I have is QSFP56/112. So the broadcom nic I have is physically incompatible with QSFP56 DD but does work with QSFP112.
I think (?) a cable like: https://www.fs.com/products/352443.html would work where one end is QSFP56 DD and the other end is QSFP112/400g.
However the cable I was originally talking about is a more “ideal” connection here because one should be looking at it by lanes and not data rate. The singular QSFP112 mode has to be stepped down to twice as many PAM lanes (8 vs 4).. hence why that cable is so expensive vs the other cable I linked.
I think the problem is on the mikrotik CRS812 side given it works in 2x200G mode but does not work in aggregate mode to be one physical interface.
And the real-world difference is that the $1500 cable works (because QSFP112 is 4 lanes at a faster rate stepping down to QSFP56 in 8 lanes at a slower rate into the QSFP56 DD on the mikrotik side) whereas with the cheaper cable it’s just 2 sets of 4 lanes, one set from each physical connector on the broadcom nic, into the QSFP56 DD, so it makes sense that option is way less complicated from a signaling perspective.
However it doesn’t work on the mikrotik CRS812.
You are correct. An active cable or transceivers is the only way to connect QSFP-DD400 to QSFP112. You can get a pair of QDD-400G-DR4-S and QSFP-400G-DR4 transceivers and a MPO-12 Female Type B patch cable for $350 - far cheaper than an active cable.
The naming of that adapter doesn’t make sense. How can a QSFP112 port not support 400G? A QSFP112 port has 4x 100G lanes. I have never heard of a switch that supports hardware aggregation across two switch-side ports. Maybe the NIC can do hardware aggregation across two ports for 400G, but that would mean that each port has only 2x 100G lanes wired up, or the ports are actually QSFP56.
What about packet buffer size? How big IS IT? I cannot find any information about the Marvell chipset.
I asked under the STH review. If RouterOS prints the correct info it’s 8MB.
Buffer allocation works somewhat different on the CRS8XX series btw:
