wafer.space has just opened our first pooled manufacturing run of GF180MCU with the purchase deadline of 28th Nov 2025.
Think of it like OHS Park for silicon!
You provide a 20mm2 design in the open source GF180MCU technology and you get back 1,000 parts. You can used an existing template or build something completely yourself with either open source or proprietary tooling (no required pad ring or management CPU).
The designs are still more in the MCU size, but you have to start somewhere!
The Google open MPW program also had 10 runs with 40 projects published at http://foss-eda-tools.googlesource.com/third_party/shuttle/ -- All the submissions had to be open source and there were 1000+ of those. I did try pitching to multiple Google Research groups that continuing the open MPW funding would grow the available designs which have been manufactured and that was useful for AI training but didn't get any bites.
The now defunct Efabless also ran a number of challenges in this space which got pretty good results, see https://efabless.com/challenges
I afraid I don't recall the status of DSP blocks in the open source toolchains for Xilinx hardware. Even if the basics are supported I'm sure there are plenty of DSP features that are not.
Many configurations of VexRISCV work fine without using the DSP blocks (and has been working for 2+ years), so not sure that is relevant.
I have a Google Doc at <https://j.mp/softcpus-on-fpgas> which includes a bunch of information about what type of performance has been achieved on modern Xilinx FPGAs.
When did you last check? The Yosys+VPR toolchain currently supports a full Linux capable SoC with Ethernet and DDR memory on the Arty A35T (which has a Xilinx Artix 7 part), see the example at https://f4pga-examples.readthedocs.io/en/latest/building-exa...
While the news seems to focuses on ML and RISC-V, there is a huge amount of *other* activity in the silicon space. With [Google working to improve the open source EDA tools + PDKs while also providing free manufacturing](https://opensource.googleblog.com/2022/05/Build%20Open%20Sil...) there are lots of opportunities.
The "new" FPGA companies like GowinSemi and RapidSilicon are also creating some pretty cool parts.
The interchange format includes a computer readable description of the FPGA which tells the tooling what is available inside the FPGA and how they can be connected.
If you are motivated by learning rather than practical usage, FPGAs are a great way to learn more about how things like CPUs or peripherals can be designed. It is also a way to understand how to evolve hardware and software at the same time.
Two examples that I have been loosely involved with;
* Google's CFU playground which is all about profiling and adding small number of new OpCodes to build an accelerator for a specific ML model (http://cfu-playground.rtfd.io/).
* The Fomu workshop (https://workshop.fomu.im) which walks though treating an FPGA like an embedded MCU where you can then modify the MCU!