Thats great news.
I hope RISC-V catches on sooner rather than later because that would slow down ARM's dominance in the mobile space. Remember that ARM is proprietary and is in the process of being bought by Nvidia, which has historically been hostile to open source drivers.
And we'll have three competitors, x86_64/AMD64, ARM and RISC-V in the desktop/server space.
Perhaps a dumb question, but is there good hope that it will catch on at all? From where I'm standing it looks like both x86_64 and ARM have so much more funding than RISC-V that it does not seem likely that RISC-V will ever catch up in capability to the bigger players.
I am pretty sure every company that makes microcontrollers and embedded processors are at least evaluating RISC-V. I think NXP even announced they will make a chip.
Doesn't free guarantee a lack of fat profit margin, unless you can add some sort of proprietary value on top of a reference design? It's hard to imagine who would make a big investment other than to undercut on price. You'd need a vertical technology stack or a captive audience.
Yes, but on the other hand, when the core is free, all your costs go into integration (which you’d have to do anyway) and those precise value-adds. That means you spend money on the differentiator alone.
It's all about adding some sort of proprietary value.
Think of Google Tensor: some boring ARM standard reference design cores, but an exciting new tensor unit.
Even with Apple Silicon, where the CPU cores definitely look exciting, they spent a lot of time in the presentation on some obscure video codec cores - which enable the insane rendering performance.
Microcontrollers are another example: you mostly don't buy them for their computational performance, but for their peripherals. Their USP are the right amount of the right interfaces that you happen to need, a crypto unit, the right number and quality of ADCs or DACs, and so on.
I think RISC-V will lead to lots of new chips, with the same old boring CPU core, but exciting new co-cores for special use cases.
At least that's what I'm pitching to my boss so he will let me start an evaluation project :)
If semiconductor technology reaches a point where it stops improving then maybe RISC-V will eventually get close enough where for at least some applications it's worth it if it's cheaper/easier to customize?
Maybe chinese tech companies are working to provide alternatives (even for freeish) to break the capital inflow of western rivals and hinder their ability to outspend them.
There is a missed oportunity for synergy between opensource movement the fact that in ex-Communism cultures, Intellectual Property is a new concept and an average person simply doesn't get it.
The typical narrative of "China/Russia stealing our tech" owes a lot to the fact that average Ivan does not know what IP means, but has been told that "Micorosoft owns Windows" and "Apple owns 'Slide to Unlock' " and then it sounds like a bad joke to him.
Here is a Chinese firm provising design free-as-in liberty to the world, which ARM/Intel are doing the opposite, I don't know we can capitalise on this opportunity, but we should do so without prejudice.
The RISC-V cores are provided as the Verilog source code files (with the Apache license), so there are no binary blobs.
You can either synthesize the cores for an FPGA, or, if you want to make an ASIC, you must combine them with the many other blocks that are needed for a complete device.
After you have a working FPGA or ASIC device, you could use RISC-V tools from other sources, but the Alibaba cores implement a lot of extensions over the base RISC-V, so for optimum performance you should use the tools also provided by Alibaba.
Besides the cores, the Github repositories contain gcc and binutils for the cores and also standard C libraries for both Linux and Android, a Linux 5.10 kernel, patches for the Android kernel and everything else needed to be able to boot Linux or Android on them.
when i looked at risc V very early on a few years ago it was not going to be able to outperform available cortex-A ARM SoCs. eventually i settled for an 1.8GHz A17. i’m interested in risc-V… but are we at the point yet where it is competitive compared to ARM SoCs? another issue i always deal with is that ARM SoCs come with lots of peripherals i don’t need and the ones i do need are not available in high enough numbers which means i need to attach an FPGA. (for example i2s ports). the GPU situation is also frustrating with hard to use binary drivers and a slow process of open sourcing the drivers with lots of tricky realtime issues that take forever to get fixed.
Open sourcing these sort of designs might be a nice business card, recruiting young Chinese students to stay in China instead of moving oversea. But it is also a nice recruitment advertisement for foreign talent "he we make cool stuff, you want to design and create cool stuff contact us!!!"
Don't forget access too US tech is not guaranteed anymore, so why should Chinese big tech risk it all building their bedrock on stuff the US can just sanction. Seems like a very very big business risk. You can try to do it but just look at Huawei and ZTE, Huawei is too diversified for the US to kill but they have created enough supply fear that other parties are hesitant to deal with them. ZTE was on the chopping block and now has a internal US cell keeping a eye on them to make sure they don't misbehave.
It's either "commoditize your complement" or "integrate your entire supply chain". The latter is especially interesting if you want strong control over your supply chain, see Apple. Apple could not have created such a well-integrated experience if their CPU was designed by a different company.
BYD is also a big vertical integrator, so much so that they their EVs are price competitive with ICE vehicles. They don't seem to have any issue with chips supplies given they are their designer and fabricator.
Reasons for open source vary and are perceived differently by different parties. Sometimes is more than one reason, and there is nothing like common sense when it comes to open sourcing stuff.
The announce of open-sourcing the RISC-V cores on which they have worked until last year was simultaneous with the announcement of a high performance server CPU using the latest ARMv9 cores, to be used in their data centers.
So, yes, I would guess that after playing with RISC-V for a couple of years they have switched to ARMv9 for the real thing.
One of the cores that have been open-sourced now was in 2020 the fastest RISC-V core ever designed, but for the best performance they had to correct a blatant omission in the RISC-V ISA and add a custom extension for indexed addressing.
A company as big as Alibaba isn't putting all of their eggs in one basket. They've probably been playing with both RISC-V and ARM, and will continue to do so.
It makes sense to do that even just to hedge against ARM and be in a better negotiating position for negotiating license fees with ARM. 'Wow, those prices are a bit high, would be a shame if someone released even faster RISC-V cores and took the margin out of your mid range offerings...'
Additionally, these cores are all very competitive.. in their gate count niche. There's plenty of places for Alibaba to use cores like this that aren't top of the line server processors.
So I'd be shocked if they don't continue to play both sides.
I thought the Chinese don't have access to the latest ARM cores anymore - they were cut off because ARM China essentially went rogue and was taken over by Chinese interests?
The new CPU announced by Alibaba uses the ARMv9 ISA and new cores, which were announced publicly only earlier this year, long after the parent ARM company had cut the links with the rogue China subsidiary.
So I assume that Alibaba must have licensed the cores directly from the parent ARM company.
