Comment by ThrowawayR2
21 days ago
Open standards don't mean a thing; you can't execute code on a standard. There are past open ISAs like OpenSPARC, MIPS, and OpenPOWER that never gained any traction.
High performance implementations, i.e. actual chips you can buy, are going to be proprietary and that's not going to change. Engineering hardware is expensive.
This is a bold prediction but I thing “alliances” will form where industry players collaborate (like we are seeing in video codecs). And the basic core could become an Open Source project just like Linux did. Operating Systems and codecs were (and are) expensive too.
But there are different levels of proprietary. Having your entire software ecosystem impossible to lock-in means something. And competition tends to breed openness.
MIPS certainly did gain a lot of traction. It was a real force at one point and the world is awash in them. But of course MIPS (the company) is RISC-V now.
An operating system can be coded on one not particularly powerful computer by one person and it costs a few pennies to compile and test. A lot of other open source projects were also initiated by one or two talented people. Software is absurdly inexpensive to develop relative to its complexity.
A cutting edge processor requires personnel across several disciplines and millions in specialized equipment to both validate the implementation of the architecture and the electrical behavior of the circuits and each time it's "compiled" (a batch of test chips fabbed and QAed), it takes a few weeks to be delivered and costs hundreds of thousands of dollars. The ISA being open and royalty-free doesn't affect any of those massive costs.
To use a famous quote: "The answer to any question starting, 'Why don't they...' is almost always, 'Money'" Nobody is offering up that kind of money without practical guarantees of success and some kind of profit at the end of it.
The idea that a chip takes more "personnel" than an operating system or a codec is wrong. An individual can make toys of either software or hardware. "Real" ones take dozens or hundreds of people. There are 5000 people involved in the Linux kernel. That is design, not production. Production (manufacturing) is what is free in software.
The Linux kernel may be "free" but it represents millions of man hours (or years) of engineering. Creating a viable RISC-V chip would be easier.
Creating the AV2 video codec cost money. I assure you. There is a reason that the Alliance for Open Media is a list of Fortune 500 companies and not a bunch of individual developers.
I have worked in industries dominated by a single chip supplier that made the chips that everybody used. Video surveillance is a good example. It would have been much cheaper for the major players in that industry to fund the collaborative development of chips they could all use and that could maybe be "tweaked" to add differentiated value for the largest players. It would save them money. It would give them more control (even more valuable).
I assume you know what a "chiplet" is. RISC-V is going to change things. In my view, you are focused on the wrong constraints.
We are both saying that money matters. We are simply coming to different conclusions about what that means.