I have quite a few Sipeed devices for FPGA and RISC5. I think I have 7 or 8 different boards. I wanted them to be good, I really did. However, the tool chain for both RISC and FPGA are abhorrent and the chips themselves are not really that great tbh. Granted this was like 6 years ago for RISC stuff and less than 2 years ago for FPGA so many things have likely changed. When I used them there was effectively zero usable documentation and almost nobody using them and very sketchy downloads (browser claimed malware on a number of downloads and after a Chinese VPN I could finally access a Chinese IP only page that allowed for clean downloads). After a few months of banging my head against the wall there was "leaked" Chinese documentation on that same host, that was not very helpful since translation didn't do a great job, but did include a newer toolchain that helped fix some of the issues I was having at the time. (USB connection dropout during uploading was a big one i remember) Hopefully there's better documentation, a better toolchain, and resources accessible without a VPN now. After much struggling and a bit of reverse engineering, I did manage to get them all running for what I wanted them for.

All that said, I wouldn't recommend them to my worst enemy. I ended up swapping all the Sipeed stuff for more well known alternatives. If you're using an already known working example from Sipeed or a known working user repo that supports the specific board you want, and not developing something from scratch, it's probably fine but I wouldn't expect it to just work without understanding the smallest amount of FPGA dev (compiling and bitstream generation and uploading). If you're going to actually develop something from scratch, don't bother and save yourself time and money.

If your goal is a DIY Amiga of sorts, there are plenty of older/reclaimed Intel/Xilinx FPGAs mounted on new PCBs that Chinese sellers sell for relatively cheap and I've had great results with those. Though if you're trying to just use someone's repos without modification, then you'll need the exact supported hardware as pins and stuff are different unless you want to modify the files yourself. Different boards will have different FPGA packages, pin assignments, clocks, memory arrangements, peripherals, etc. So it can be a bit of work to get code working, and more work to get it working the same. Though sometimes it can be as easy as just modifying the pin assignments, depending on the chip and board you select.

I have no clue about the state of the dev resources today, but when I used them I wanted to smash them all with a sledgehammer... I did do a precursory search and the situation seems to have changed substantially since my experience. Sipeed now has a reasonably substantial English Tang documentation tree and examples, Gowin provides its EDA software and programmer directly, and there's even an open source Gowin ecosystem through Yosys, nextpnr, Project Apicula. I can't say how good these are but it's already a massive step above what I had access to.

If you do end up trying them, I'd be interested in your experience, because like I said, I'd love for them to be good. Competition for AMD/Intel is a good thing.