Open Source CPU Architecture RISC-V Is Gaining Momentum (insidehpc.com) 41
The CEO of the RISC-V Foundation (a former IBM executive) touted the open-source CPU architecture at this year's HiPEAC conference, arguing there's "a growing demand for custom processors purpose-built to meet the power and performance requirements of specific applications..."
As I've been travelling across the globe to promote the benefits of RISC-V at events and meet with our member companies, it's really stuck me how the level of commitment to drive the mainstream adoption of RISC-V is like nothing I've seen before. It's exhilarating to witness our community collaborate across industries and geographies with the shared goal of accelerating the RISC-V ecosystem...With more than 420 organizations, individuals and universities that are members of the RISC-V Foundation, there is a really vibrant community collaborating together to drive the progression of ratified specs, compliance suites and other technical deliverables for the RISC-V ecosystem.
While RISC-V has a BSD open source license, designers are welcome to develop proprietary implementations for commercial use as they see fit. RISC-V offers a variety of commercial benefits, enabling companies to accelerate development time while also reducing strategic risk and overall costs. Thanks to these design and cost benefits, I'm confident that members will continue to actively contribute to the RISC-V ecosystem to not only drive innovation forward, but also benefit their bottom line... I don't have a favorite project, but rather I love the amazing spectrum that RISC-V is engaged in — from a wearable health monitor to scaled out cloud data centres, from universities in Pakistan to the University of Bologna in Italy or Barcelona Supercomputing Center in Spain, from design tools to foundries, from the most renowned global tech companies to entrepreneurs raising their first round of capital. Our community is broad, deep, growing and energized...
The RISC-V ecosystem is poised to significantly grow over the next five years. Semico Research predicts that the market will consume a total of 62.4 billion RISC-V central processing unit (CPU) cores by 2025! By that time I look forward to seeing many new types of RISC-V implementations including innovative consumer devices, industrial applications, high performance computing applications and much more... Unlike legacy instruction set architectures (ISAs) which are decades old and are not designed to handle the latest workloads, RISC-V has a variety of advantages including its openness, simplicity, clean-slate design, modularity, extensibility and stability. Thanks to these benefits, RISC-V is ushering in a new era of silicon design and processor innovation.
They also highlighted a major advantage. RISC-V "provides the flexibility to create thousands of possible custom processors. Since implementation is not defined at the ISA level, but rather by the composition of the system-on-chip and other design attributes, engineers can choose to go big, small, powerful or lightweight with their designs."
While RISC-V has a BSD open source license, designers are welcome to develop proprietary implementations for commercial use as they see fit. RISC-V offers a variety of commercial benefits, enabling companies to accelerate development time while also reducing strategic risk and overall costs. Thanks to these design and cost benefits, I'm confident that members will continue to actively contribute to the RISC-V ecosystem to not only drive innovation forward, but also benefit their bottom line... I don't have a favorite project, but rather I love the amazing spectrum that RISC-V is engaged in — from a wearable health monitor to scaled out cloud data centres, from universities in Pakistan to the University of Bologna in Italy or Barcelona Supercomputing Center in Spain, from design tools to foundries, from the most renowned global tech companies to entrepreneurs raising their first round of capital. Our community is broad, deep, growing and energized...
The RISC-V ecosystem is poised to significantly grow over the next five years. Semico Research predicts that the market will consume a total of 62.4 billion RISC-V central processing unit (CPU) cores by 2025! By that time I look forward to seeing many new types of RISC-V implementations including innovative consumer devices, industrial applications, high performance computing applications and much more... Unlike legacy instruction set architectures (ISAs) which are decades old and are not designed to handle the latest workloads, RISC-V has a variety of advantages including its openness, simplicity, clean-slate design, modularity, extensibility and stability. Thanks to these benefits, RISC-V is ushering in a new era of silicon design and processor innovation.
They also highlighted a major advantage. RISC-V "provides the flexibility to create thousands of possible custom processors. Since implementation is not defined at the ISA level, but rather by the composition of the system-on-chip and other design attributes, engineers can choose to go big, small, powerful or lightweight with their designs."
RISC V with a risk (Score:2, Troll)
Really cool, cheap gadgets for geeks and hobbyists ...
https://www.aliexpress.com/ite... [aliexpress.com]
Re: (Score:1)
There are almost as many STM32 clones as authentic devices produced. Most hobbyists are buying the knock off devices produced without any license.
And of course totally legitimate vendors can pack pretty much anything they want around a licensed ARM core. Which includes bus monitoring or even modules actively running side channel attacks on the CPU. It's almost impossible to have 100% confidence in the security of the silicon you use, even if everyone is legal.
Re: (Score:2, Insightful)
You make it sound like allowing companies to make a processor without needing license from someone is a bad thing.
Separate from that, though, given the number of chip factories in countries outside the current American dominance zone, I think it is fairly safe to say that if a "rogue" company or country wants access to a chip design, I'm fairly confident that they can get it without paying a licensing fee to ARM anyway.
And as a separate point: Huawei is a corporate behemoth with a lot of resources to throw
Re: (Score:3)
ARM is British owned by a Japanese company. When the US trade war started they began moving some R&D away from their US office to ensure they could keep supplying it to China.
IIRC they do GPU development in the US.
Re: (Score:2)
If a company has no qualms of messing with the design to inject such backdoors / nefarious wares why would they suddenly have morality to pay a license? They are likely just straight up copying the ARM designs not giving a damn anyway.
Dethroning ARM? (Score:5, Interesting)
I think the question that has to be answered here is why is ARM popular?
There are multiple open source SPARC designs have been around for many years. Yet ARM is going strong despite the costs, companies must purchase designs or jump through contract hoops in order to design their own. And either way pay per chip they make.
Will the low cost nature of open source win? Not having to pay a license to ARM saves you a tiny bit of money. It doesn't save you enough that a small run of RISC-V would be cheaper than a large run of ARM, so end-users, developers, and hobbyists aren't likely to benefit there. I think it's safe to say that with small runs and with other free architectures out there never catching on, that money isn't a primary factor.
In a world where compilers are cheap and a huge chunk of real world software runs in as a scripting language under JIT, the processor architecture you use is often irrelevant. Unless you're specifically aiming for Android or iOS support, there is not a strong advantage for any particular architecture. ARM still dominates the 32-bit and 64-bit embedded world, with MIPS trailing a distance second.
At my work it, RISC-V displaces in-house control and monitoring processors on our ASIC. This is because we can build it more barebones than ARM, but not as bare bones as our in-house processor, so RISC-V still costs us chip area. It's not a slam dunk decision, and I suspect other fabless silicon companies face the same decisions.
Re: (Score:2)
ARM has better per-clock performance and lower power usage, so SPARC loses out. RISC-V has a lot of nice features, although some have said the design (ISA) may be inherently more power-hungry than ARM.
I'd like to see AMD make a competitor to Intel's Cyclone V SE SOC with a RISC-V core. If as you say you can implement RISC-V in less area, that's significant: more room for FPGA fabric.
Re: (Score:2)
Does SPARC scale? RISC-V and ARM seem to scale down and up fairly well, which is why you're seeing them in hard drive controllers and desktop machines.
SiFive's Core Customizer: https://www.sifive.com/core-de... [sifive.com] lets you a base chip and start customizing the inner bits.
Re: (Score:1)
As all the parts of a computer system will exist for RISC-V.
End-users, developers, and hobbyists will enjoy an actual desktop ready computer system they can use and buy into.
All the chips and parts needed to get everything to work. Not just the CPU.
Re: (Score:2)
It's the support that keeps ARM popular. There are many high quality compilers and code libraries available, and support for things like Linux and Windows CE. It's easy to hire ARM developers and the skills are transferable to many different manufacturers.
Also ARM put a lot of effort into optimising the architecture for the needs of the market. You have various versions of ARM for everything from tiny low cost microcontrollers to desktop grade CPUs and everything in between, all with full support (compilers
Re: (Score:2)
Where I work, the real value for RISC-V (and we're just starting to use it) is the ability to modify the instruction set, which can give 10X improvements in performance (or, equivalently, power efficiency) when running specific algorithms.
With RISC-V and Chisel, we can spin a new hardware accelerator for a specific image processing task incredibly quickly. That is very valuable.
Unless you're Apple, ARM doesn't want to talk to you, but with RISC-V you can do it yourself.
Re: (Score:3)
In a world where compilers are cheap and a huge chunk of real world software runs in as a scripting language under JIT, the processor architecture you use is often irrelevant.
People would like to believe that, but it's just not true, porting optimizing, debugging and maintaining compilers (including JIT compilers) is a lot of work. The further you stray from the mainstream, the more problems you are likely to run into with toolchains that are buggy, outdated, poorly optimised or just plain not available.
If anything the rise of a bunch of new languages has made things worse. In the old days almost everything you would want to run on your Linux system was written in either one of
Re: (Score:1)
* There was a project to make openjdk use LLVM as a backend, but AIUI it's not really maintained since an open source native arm backend was written.
Perhaps you are talking about some old project that maybe tried. Nowadays we have GraalVM and Polyglot (https://www.graalvm.org/), an official project that is very much maintained and that allows to both include LLVM binaries in a Java application (something that used to be very painful using JNDI), or compile a Java application to LLVM bitcode (making it native but making it run much faster).
Not trolling. Just thought this piece of information could be pertinent to some.
spokesmodel pumps RISC-V (Score:3)
"a growing demand for custom processors purpose-built to meet the power and performance requirements of specific applications..." says the head of the RISC-V. How is this news? Yes, it's wonderful architecture and all that, but what else would be expected to say.
Re: (Score:2)
Fake Open Source (Score:5, Interesting)
The CEO of the RISC-V Foundation (a former IBM executive) touted the open-source CPU architecture at this year's HiPEAC conference, arguing there's "a growing demand for custom processors purpose-built to meet the power and performance requirements of specific applications..."
Q: will it be possible for *end-users* to program any of these custom proprietary processors?
A: almost never. look at NVIDIA's new GPU which uses RISC-V, or WD's new proprietary embedded cores. WD even specifically tells people, at the beginning of its talks, that end-users will *not* be permitted to program their own firmware on WD SSDs, USB Sticks, or HDDs using RISC-V.
in other words, the "open-ness" is entirely misleading. the *design* houses have "open" access to the ISA, with zero royalty payments.
so whilst it's very nice to see this article on slashdot, we, here on slashdot, as technical users, are not the target audience to whom the CEO of the RISC-V Foundation is speaking. he's speaking to ARM's customers (ARM Licensees), "come and use RISC-V and save yourself ARM license fees".
Re: (Score:2)
Its not fake open source. The ISA is open and allows you or anyone else to design a modern processor without licensing. How is the fact that WD or anyone else can implement them in closed systems different from any established open source project? Firefox has plugins for viewing closed DRM infected content. Many Linux distros ship with closed source blobs every day.
Re: (Score:3)
Its not fake open source. The ISA is open and allows you or anyone else to design a modern processor without licensing.
this is factually incorrect. Trademark Law applies: you must still apply for a License. You may not use the RISC-V ISA without obtaining that permission in a commercial product, and that permission involves paying for a full Compliance Suite test to be run on the final product.
given that Trademark Law applies even *that alone* tells you that it's "Fake Open Source". it's a misleading use of the word "Open".
but it's worse than that. if you *use* the ISA - as dictated by the RISC-V Foundation - you are ab
Re: (Score:2)
So don't put a "RISC-V" logo on your RISC-V-compatible product. You're all set.
Re: (Score:2)
So don't put a "RISC-V" logo on your RISC-V-compatible product. You're all set.
yes, i forgot to mention that. you could go that route: it's perfectly acceptable. the problem comes if you are then expecting the upstream gcc, binutils, llvm, linux kernel, u-boot, libc6, debian, fedora, suse, gentoo, freebsd, qemu and literally hundreds of other software developers to blithely accept patches supporting the hardware that you spent tens of millions of dollars developing.
if you're happy to be COMPLETELY isolated from the rest of the software community, then "not putting a RISC-V logo on y
Re: (Score:3)
the problem comes if you are then expecting the upstream gcc, binutils, llvm, linux kernel, u-boot, libc6, debian, fedora, suse, gentoo, freebsd, qemu and literally hundreds of other software developers to blithely accept patches supporting the hardware that you spent tens of millions of dollars developing.
Well, if all of those things need patches, then the hardware you're developing isn't compatible with existing RISC-V implementations. Therefore, it's not RISC-V.
Having the attitude that you should be able to call this new incompatible thing RISC-V without coordinating with the RISC-V project is just silly.
You have something different, so now you need to apply a few of those tens of millions of dollars to developing your own brand identity.
Re: (Score:2)
the problem comes if you are then expecting the upstream gcc, binutils, llvm, linux kernel, u-boot, libc6, debian, fedora, suse, gentoo, freebsd, qemu and literally hundreds of other software developers to blithely accept patches supporting the hardware that you spent tens of millions of dollars developing.
Well, if all of those things need patches, then the hardware you're developing isn't compatible with existing RISC-V implementations.
that's (from the previous post - the parent-of-parent - which wasn't included in the quoting) exactly what i said. if you want to *INNOVATE*, then of course it won't be "fully" compatible (just "partly" compatible).
Therefore, it's not RISC-V.
because you've missed the context, you've misunderstood. you can have an *extension* on *top* of the core - which *IS* compatible and therefore *IS* RISC-V compliant: the *extension* however is not compliant and thus is not subject to the Compliance Suite Testing.
Having the attitude that you should be able to call this new incompatible thing RISC-V without coordinating with the RISC-V project is just silly.
You have something different, so now you need to apply a few of those tens of millions of dollars to developing your own brand identity.
you're misunderstanding, unfort
Re: (Score:2)
So your extension has nothing whatsoever to do with RISC-V. Just like if you had added a new innovative RS-232 port to the system. So maybe you slapped some new instructions on top of the CPU core. That's an implementation detail that doesn't make any difference. Your new instructions are not part of RISC-V either.
Like I said, create your own brand. All those other projects will add patches for it if you convince them that your brand of NOT-RISC-V is something great.
Re: (Score:3)
You're not making any sense. The trademark (and requirements for using it) exist precisely because of the things you are complaining about. If the trademark did not exist, we would have exactly fragmented mess you envision - everybody could call their thing 'RISC-V' even though it is different from everyone else's RISC-V. Is that really difficult to understand?
If you want people to use your product BECAUSE it is RISC-V, then get permission to use the trademark. If you want to USE RISC-V, but don't use t
Re: (Score:2)
You're not making any sense.
it takes approximately twenty minutes to half an hour of face-to-face conversations to go over this extremely complex and in-depth issue, clearing up misunderstandings and going through the "but this is how it's normally done, i don't understand" things, *only* which, when you have those misunderstandings, is it possible to then state the *exceptions* which allow full understanding to take place.
The trademark (and requirements for using it) exist precisely because of the things you are complaining about. If the trademark did not exist, we would have exactly fragmented mess you envision - everybody could call their thing 'RISC-V' even though it is different from everyone else's RISC-V.
yes. as a Trademark Holder myself i fully understand this (where most people do not).
Is that really difficult to understand?
you would be amazed at the
Re: (Score:2)
Re: (Score:2)
So Linux is also 'fake open source'?
Re: (Score:2)
Trademark of the name has nothing to do with how open the ISA is. RISCV is under a permissive license, the only way you can seriously consider it closed is if you are one of the diehard copy left types, and I suspect you may be. In which case, most of the world disagrees with you. kindly get over it and stop bringing up a dead old argument. Linux is also trademarked. Do you consider it to be closed? The only restriction related to the trademark is if you want to use the RISCV name, and the restriction is to
1984 (Score:2)
Not the book, but the year I begun to work on a mainframe that use these processor.
Re: 1984 (Score:1)
Also the year Apple released its closed design personal computer.
Re: (Score:1)
And the year IBM released the AT. And the year Dell was founded. And the first 3D printer was demonstrated. ... lots of stuff happened in 1984 [computerhope.com]
What does it help us? (Score:2, Interesting)
When there isn't any open-source fab.
I mean it's nice and all,
but *one* of the machines you need for your clean room costs $15 million easily!
Can't we make a kickstarter to buy a lights-out fab (with everything)? And then offer every backer to make chips at the price of the materials, utilities and maintenance, with zero profit or management bonuses etc.
That would really help humanity!
Intel's 14nm might be up for sale soon. ;)
Re: (Score:1)
GF first.
Re: (Score:2)
What would an open source fab be like? Even if you had the space, all the stuff that goes into making a chip is highly proprietary (chip simulation/testing, equipment, chemicals, plastics, ventilation etc) - some company spent a fortune building it and they want their cut.
The next best thing - companies like TSMC will take your design and make a chip - it's not even that expensive (as far as making chips goes) - they publish all the manufacturing guidelines and everything so there's no pre-flight issues (ca
It's just China pushing for it (Score:2)
China is driving this as an alternative to US/EU technology - don't be fooled.
Pure advertising (Score:1)
How did a content-less buzzword ad get to the Slashdot front page?