Open Source and FPGA Maker Board for Networking

I've had some fun learning how to implement various bits and pieces of networking on FPGAs as a hobby for a while, and while boards like this that focus on gigabit network are fine, the fact is that there are a lot of FPGA boards with gigabit and 100Mbps interfaces. What there are not enough of are low cost boards that can do 2.5Gbps, 5Gbps and 10Gbps. Lattice has some very affordable FPGAs with 5Gbps SERDES, and their newer 10Gbps capable chips remain extremely affordable.

One of the things I would absolutely love to have are a couple of FPGAs boards in SFP and QSFP form factors. Why might you ask? Because it would be seriously useful to have a PPPoE / L2TP data plane to plug into the port of a 100Gbps capable switch for use in the network edge. Modern ethernet switches have plenty of Layer 3 networking capabilities, but most switch vendors fail to expose any functionality for these protocols even though the underlying ASICs often enough have the capability to handle them. Sure, you'll never see these protocols in a cloud data center, but plenty of incumbent telecoms make use of them in their FTTP networks due to the legacy of xDSL deployments and the need to support wholesale access to those networks. Sadly, developing such a board is beyond my hobbyist electronics capabilities, but I'd have no problem bashing a bunch of Verilog / VHDL into shape to make it work in fairly short order... I just hope it uses an FPGA like the Polarfire for which the SERDES are about 100x easier to use than the gawd awful Xilinx 7 series (KC705, I'm glaring at you for eating weeks of my hobbyist life to that bring up).

Thank you for all the great comments and feedback so far. Note that Betsy uses an Altera Cyclone 10 LP (not GX). This is a low cost, general purpose FPGA. The Ethernet PHY interface is RGMII, which utilizes 5 bits of parallel DDR + clock instead of SERDES.

As many of you probably know, SERDES and specialized PCS/CDR blocks will get you well past 1 GigE, but 1 GigE for RGMII is challenging with 125 MHz single-ended traces.

The project compiles super fast with the Quartus tools and Signal Tap enabled with several active configurations. Quartus bundles the Questa simulator, so there is a great environment for simulation.

Regarding Certus-NX, this indeed would also be a great choice. Lattice does a very nice job exposing their I/O primitives, and I believe the RGMII DDR could be instantiated directly in the I/O cells for both input and output (this could definitely be accomplished with the earlier ECP5). We actively design with Certus-NX, and a future Betsy revision using it is very possible.

NetFPGA has been around since 2007, with NetFPGA-1G. They have some very fancy offerings these days, well past the 4 x 1Gbe they started with.

https://netfpga.org/ https://en.wikipedia.org/wiki/NetFPGA

Maybe I'm just jaded and demanding, but:

1) As others mentioned, two GbE interfaces seems really limited for a 2025 project. Modern FPGAs can support 100GbE and up - I don't necessarily expect that on a hobbyist-level project, of course, but 1GbE is well behind the curve.

2) There don't appear to be any hardware design files (e.g. schematics, PCB layouts) in the Git repository. In fact, the only mention of the current FPGA is a single text file stating that "Cyclone 10 GX port in progress"...

3) There's basically zero open source support for Intel/Altera FPGAs. Yes, you can open-source your HDL, but the vendor tools are all closed-source and there's no alternatives.