Do we know if there will be MPLS hardware forwarding on the new switches?
mikrotik is heavily investing into Marvell's PRESTERA DX series SOCs. the crs326, the crs328 from the existing range use the 98DX3236A1, the beefier crs317 based on 98DX8216B0.
the just announced 20SFP version of crs328 has the very same cpu as the old 328.
https://origin-www.marvell.com/switchin ... 336_pb.pdf
https://origin-www.marvell.com/switchin ... _brief.pdf
just guessing, but i don't think it is a coincidence that all new switches are from the CRS3xx range, and that could mean all of those are Marvell SOC based.
for example we have CRS332-32S+RM, which may be pretty much powered by the 98DX83xx SOC, at least the product brief of the SOC is just a perfect match for the 32S+ device.
and now the "hard to swallow pills":
- 98DX3236, codenamed as Alleycat is not packing no serious L3 / carrier features. this means, no MPLS for the CRS326/328
98DX3336, called PONcat does have two A7 cores and does support MPLS/VPLS, but there are no signs that this is utilised by any of the new boxes. otherwise they are very similar: 24xGbE+4x10GbE switches
98DX4251, called bobcat2 is again an MPLS enabled, but i don't really see this around here
98DX8212 can be be heart of CRS312-4C+8XG, they even have a very similar reference design that uses 3x 88X3240s to have 10GBaseT and SFP+ ports. this can't be unseen.
98DX8216 is what powers the current CRS317 is somewhat similar to 8212. the product brief did not explicitly mention MPLS but it says something about the its purpose: "primary application for these chips is pizza-box switches for medium-size companies, which must connect new 10GbE-enabled servers with 10GbE uplink ports on LAN-access switches"
with exception of the bold ones, no docs state HW MPLS capability. but i did not find them in any of the current Mikrotik range. and until now i wasn't able to figure out what the new 40G equipped boxes are made of.
and i expect some untold tales to hide up their sleeves as i checked out the Prestera EX series SOC docs:
this might be a root of a new breed of CRS4xx series with 100GE QSFP28s.