It would be really cool if there was a version of the Ten64 that supported the LS2088 so that you could use DPSW features in the DPAA2 architecture.
I’m mildly curious how come the LS1088 was used, and how hard would it be to get a version with the LS2088? Is it a drop-in replacement that a batch could be made for? I’d only buy one… but I bet other people would find it super interesting since it would allow you to hardware accelerate network switching and would make for a super handy network switch built-in for home router applications. Right now, I have to use a software bridge (which works alright, I guess).
The next two SoCs in the family, LS2088 and LX2160 are basically 2x-4x bigger in every dimension: cores, number of pins, physical size, power/thermal consumption and most importantly, $/part cost.
Before the LS1088 we had a LS1043 design (about 0.5x of the LS1088 in most dimensions) and various smaller SoCs (like the Geode/x86), so it’s been a nice progression for us.
They (LS2/LX2) are conceptually similar in how the “chassis” integrates into the rest of the system, so I could conceivably start with the Ten64 schematic and swap to an LS2088, but the board layout would need to be completely redone to accommodate the bigger part.
It’s likely the next generation designs will have a switch (either external or integrated) as the design trend with newer network SoCs is towards fewer inbuilt MACs (or if built in, only a small set capable of 2.5 or 10G)
Thanks for the thoughtful response. Too bad it isn’t drop in because that would make it more feasible for a small run. The DPAA2 DPSW module looks super attractive but there’s basically zero prosumer hardware that even uses it.
In general, some of the things DPAA2 offers are pretty revolutionary and it would be incredible to see more applications of it in the networking space. Being able to move certain parts of applications such as TLS termination into flexible hardware is compelling.