@niclas @lupyuen The T-HEAD cores are open source. I suspect they are odd because they were pulled from their targeted applications.

You can get on GitHub and run a C910 if you have a powerful enough FPGA because they have the Verilog up. It's not going to be the 16 core module that was supposed to be part of a larger system.

At least T-Head released information when alleged competitors hide their information. RISC-V was a good idea but it just lets others blatantly copy and obfuscate the fact that they copied a design. If you want a good example, look at the C910 and other manufacturers who made innovations after it. They outperform it because they took inspiration and made it on a smaller process with some improvements.

If one thing is clear about RISC-V, to me, it's that copycats love open source and T-Head processors show their heritage by being difficult. It's almost as if their processors aren't as trashy as the implementations are but are instead being obtuse because they are not working with hundreds more cores in a system cared for by scientists who know how to make them work properly.

If you have read this far, congratulations. We still don't have an Open FPGA that hasn't been reverse engineered. The problem with RISC-V is that it got controlled by companies and it has little to no integrity. Applaud the company that took a RISC-V processor and paired it to nearly dead GPU technology that still refuses to make it open.

Perhaps I'm too cynical about the open hardware aspect that could have been and I should just embrace the shameless desecration of ideals. Perhaps I'm bitter about past experiences that lead me to believe that a man with an interest in open hardware was misled by a few companies and it rocked their business. Just put legal landmines in the mix and we are at the current state of RISC-V. There's good, there's bad and there's the reality that this is how it's going to be. I don't see RISC-V as a path forward for those who want open source or open hardware.

With all that being said, I could see T-Head destroying competition with what would be a considerably small investment and make a fully open source and open hardware SoC that has its own open GPU that's wrapped up with a non commercial use clause. It might sell poorly but it would be yet another first they could claim. The Party wouldn't like it, their competitors would try to destroy it and the US would be bribed to not accept it. Perhaps it would be successful as it's ideologically pure, modern computation is plagued by inefficiencies which are legal landmines and the fact that no intelligence agencies can trust their own hardware because they had the foresight to embed backdoors in everything.

@AmpBenzScientist

Yes, I read it all... ;-)

I am probably more hopeful that we will see progress (outside copy/paste-china) for all but the most compute intensive usages, i.e. embedded controllers.

With prices for custom asic also dropping quickly, I think there is a future for open source adopted by many non-asic companies, and a decent job market for people with experience in it.

We are now roughly at 1990 in software, long way to go.

Also found; https://opencores.org/projects/neorv32

@lupyuen

@niclas @lupyuen China 2025 is going well. The US is going to have to get off its ass or we'll be left behind. It really shows when China tries something.

@AmpBenzScientist

Yes and No. If you have lived in China, you must have noticed the lack of honorable business practices and precision. So for every pearl of excellence there are a dozen scoundrels trying to ride on that success, and it creates an enormous downward pull on those.

We can already see a higher frequency of microcontroller "bugs" in silicon. GigaDevices GD32 cloning of STM32 has plenty of mistakes in them, much higher than ST. And that is with IP purchased from ARM.

@lupyuen

@niclas @lupyuen That particular company also made a RISC-V MCU that has been called difficult. It's not that the production quality is bad at all but the architecture was not very friendly from what I remember.