Comment by LeFantome

> please feel free to downvote

Not sure why anybody would downvote your comment

> RISC-V doesn't make sense to 99% users at this stage

Not sure about the exact percentage but your basic point is valid. Adding "at this stage" makes it hard to argue with you.

> ARM is cheaper for 99% use cases

It may be 100% of use cases today. Facts are facts. You probably need to add "at this stage" again though.

> ARM ... has far more choices on the market

Very much so. Again, today...

That said, it is worth noting that almost all ARM "choices" are licensing the same small number of core designs from ARM. Already there are beginning to be enough RISC-V suppliers that some users may like the RISC-V options better in some niches (see automotive and some edge AI for example).

> ARM ... has ... much better performance

Absolutely. But that may not stay true for long. RISC-V CPUs will appear this year that equal or exceed CPU designs from ARM themselves in performance (eg. Ascalon). And we will see where things go from there. It will be a while before RISC-V beats Apple Silicon of course. And even once RISC-V gets there on performance, ARM may lead on price/performance for a while. That is, until RISC-V volumes start to equal or exceed other ISAs...

> greater software ecosystem and tooling

On the Open Source side at least, this is already a weak point. You can get multiple Linux distributions for RISC-V today, including from critical players like Ubuntu and Red Hat. The Linux kernel has a tonne of dedicated RISC-V support. Even though there are hardly any RISC-V chips with vector extensions in the wild, you already see Open Source packages adding support for these extensions. Both Clang and GCC have great RISC-V support. There are already x86-64 emulation layers for RISC-V. Ecosystems like QEMU support RISC-V. Even niche projects like Haiku OS support RISC-V. And on the hardware side, RISC-V players like Tenstorrent are advancing Open Source tooling and toolkits like crazy. The ecosystem is great now and getting better every day which, given the complete lack of real RISC-V hardware on desktops and servers, shows you how excited the industry is for RISC-V and how much support it is going to get.

Remaining gaps in ecosystem and tooling will close quickly. Starting with the board that we are discussing here, Titan, RISC-V is entering an era of being good enough to actually use. Linux and the universe of software associated with it are going to support RISC-V rather robustly. And while some RISC-V suppliers will follow the ARM path, many RISC-V suppliers are being good about getting support into the Linux mainline.

I expect ecosystem and tooling to be better for RISC-V than for ARM in general (though both will be great).

> For 99% users, the only real "benefit" RISC-V can bring to the table is the _false_ feeling that "I am different"

> _false_

False.

Here we disagree. But again, it may mostly be about the percentage. Because most users just want something that runs their software at the highest speed for the lowest price. And see above for how we agree that it will be a few years yet before people that do not otherwise care about RISC-V will find it the best option based on simple price/performance (though I do think that day will come).

But there are many "real" benefits to RISC-V.

Perhaps the biggest benefit is that it is an ecosystem that cannot be truly dominated by a single player or even by a small few. I wish RISC-V suppliers great success, and many will find niches that make them rich. But the amount of market power they can ever wield is limited by competition. I for one want this to be my future and I cannot wait to get on the train.

This is just my opinion but I think RISC-V is very well designed. I want to build software for the platform. I want to use assembly language on it. It seems much more pleasant than x86-64 and even ARM. This is a big benefit to me.

Similarly, RISC-V as an ISA and an ecosystem will be uniquely scalable. The same basic ISA can be used on the smallest micro-controllers or the most complex AI supercomputers. And it can be used in the support chips every step of the way. The expertise that I acquire using RISC-V will be broadly applicable over space and time.

And, while this is a prediction, RISC-V will have longevity. Suppliers can go out of business. RISC-V is not a supplier. Once it takes hold, it is not going anywhere. Many an ISA has dominated the computing landscape only to be abandoned and forgotten. RISC-V was inspired by MIPS (the ISA) and MIPS workstations used to cost as much as a small house. But now MIPS (the company) is a RISC-V supplier. x86-64 may seem unassailable but ARM has certainly kept it out of many niches and now ARM is starting to be credible on desktops and servers. x86-64 could go away (especially if Intel failed--not impossible). And ARM is very vulnerable to RISC-V (if you ask me). But for RISC-V to go away, there would have to be yet another totally open ISA that the entire world rallied behind. That is not how things generally work. Like Linux, RISC-V is destined to become a natural monopoly in my view and to be with us a very, very, very long time.

And, for now at least, RISC-V is just more interesting. Companies like Tenstorrent are doing really interesting things. Universities are doing interesting work and sharing it with the world. Those two things came together just the other day when RISC-V vector extensions were added to BOOM. It is a fun space to watch and it will be a fun space to be a part of.

And just like every school teaches software in Java, every school is going to teach programming and electronics with RISC-V. It is going to be the default technology in the future. And that means that it will be the go-to for start-ups as well. RISC-V will be the go-to technology for innovation.

Finally, what I will end with is that the real inevitability of RISC-V has nothing to do with users. Companies will choose RISC-V. But not to save a few bucks on ISA licensing like everybody imagines. That may be a benefit but it does that amount of cost is not going to drive most decisions. But the more critical issue with licensing is control.

Take the situation with ARM and Qualcomm. Qualcomm wants to release its own high-performance silicon to compete with the likes of Intel and Apple. They licensed the ARM ISA to do this. And then ARM tried to stop them from releasing this technology over a license dispute. Yes, it was over money (which matters) but the much bigger deal is that ARM (a supplier to Qualcomm) tried to dictate how Qualcomm can run its business. Qualcomm recently bought a prominent high-performance RISC-V designer. I think these facts are related.

If you bet your business on ARM, you better hope that ARM likes your strategy. If they do not, they may try to stop you. Why would Qualcomm want to create a business around ARM if ARM is going to represent a strategic risk like this? Why would Amazon or NVIDIA? Building the same business around RISC-V eliminates that risk. You do not have to ask permission for whatever you do with RISC-V. Nobody can file an injunction on a RISC-V supplier for using RISC-V.

This is why China is so into RISC-V. Even for them, it is not really to save on license costs. It is about reducing legal and geopolitical exposure. US sanctions led to ARM refusing to work with Chinese suppliers like Huawei. The US cannot sanction RISC-V as a technology. Fast forward to today and there are now many, many companies around the world rethinking their exposure to US based technology companies. There is a reason that RISC-V International is based in Switzerland even though RISC-V was invented in California. Europe is investing in home-grown RISC-V solutions. So is India. Tenstorrent came from Canada. Andes is out of Taiwan. It is not just China.

So, no need to downvote. But if you really think RISC-V has no benefits, the next 5 years are really going to confuse you.