Could RISC-V Processors Compete With Intel, ARM, and AMD? (venturebeat.com) 112
"As promised, SiFive has unveiled a new computer featuring the company's SiFive FU740 processor based on RISC-V architecture," reports Liliputing:
The company, which has been making RISC-V chips for several years, is positioning its new SiFive HiFive Unmatched computer as a professional development board for those interested in working with RISC-V. But unlike the company's other HiFive boards, the new Unmatched model is designed so that it can be easily integrated into a standard PC...
SiFive says the system can support GNU/Linux distributions including Yocto, Debian, and Fedora.
"SiFive is releasing the HiFive Unleashed in an effort to afford developers the ability to build RISC-V based systems, using readily available, off-the-shelf parts," explains Forbes: SiFive says it built the board to address the market need for easily accessible RISC-V hardware to further advance development of new platforms, products, and software using the royalty-free ISA...
A short video demo shows the HiFive Unmatched installed in a common mid-tower PC chassis, running the included Linux distro, with an AMD Radeon graphics card pushing the pixels. In the video, the HiFive Unmatched is compiling an application and is shown browsing the web and opening a PDF. SiFive also notes that video playback is accelerated in hardware with the included version of Linux.
"At the moment, these development PCs are early alternatives, most likely targeted at hobbyists and engineers who may snap them up when they become available in the fourth quarter for $665," notes VentureBeat.
But they add that "While it's still early days, it's not inconceivable that RISC-V processors could someday be alternatives to Intel-based PCs and PC processors." The startup has raised $190 million to date, and former Qualcomm executive Patrick Little recently joined SiFive as CEO. His task will be to establish the company's RISC-V processors as an alternative to ARM. This move comes in the wake of Nvidia's $40 billion acquisition of the world's leading processor architecture.
If Little is also looking to challenge Intel and AMD in PCs, he'll have his work cut out for him. For starters, SiFive is currently focused on Linux-based PCs, not Microsoft Windows PCs. Secondly, SiFive wouldn't build these processors or computers on its own. Its customers — anyone brave enough to take on the PC giants — would have to do that.
"I wouldn't see this as SiFive moving out of the box. It's more like they're expanding their box," said Linley Group senior analyst Aakash Jani. "They're using their core architecture to enable other chip designers to build PCs, or whatever they plan to build."
SiFive says the system can support GNU/Linux distributions including Yocto, Debian, and Fedora.
"SiFive is releasing the HiFive Unleashed in an effort to afford developers the ability to build RISC-V based systems, using readily available, off-the-shelf parts," explains Forbes: SiFive says it built the board to address the market need for easily accessible RISC-V hardware to further advance development of new platforms, products, and software using the royalty-free ISA...
A short video demo shows the HiFive Unmatched installed in a common mid-tower PC chassis, running the included Linux distro, with an AMD Radeon graphics card pushing the pixels. In the video, the HiFive Unmatched is compiling an application and is shown browsing the web and opening a PDF. SiFive also notes that video playback is accelerated in hardware with the included version of Linux.
"At the moment, these development PCs are early alternatives, most likely targeted at hobbyists and engineers who may snap them up when they become available in the fourth quarter for $665," notes VentureBeat.
But they add that "While it's still early days, it's not inconceivable that RISC-V processors could someday be alternatives to Intel-based PCs and PC processors." The startup has raised $190 million to date, and former Qualcomm executive Patrick Little recently joined SiFive as CEO. His task will be to establish the company's RISC-V processors as an alternative to ARM. This move comes in the wake of Nvidia's $40 billion acquisition of the world's leading processor architecture.
If Little is also looking to challenge Intel and AMD in PCs, he'll have his work cut out for him. For starters, SiFive is currently focused on Linux-based PCs, not Microsoft Windows PCs. Secondly, SiFive wouldn't build these processors or computers on its own. Its customers — anyone brave enough to take on the PC giants — would have to do that.
"I wouldn't see this as SiFive moving out of the box. It's more like they're expanding their box," said Linley Group senior analyst Aakash Jani. "They're using their core architecture to enable other chip designers to build PCs, or whatever they plan to build."
frist psot (Score:3, Interesting)
Re: (Score:2)
Well, none of the systems where I've replaced ARM with RISC-V even have an OS, so that might have something to do with the lack of distro support.
Re: (Score:3)
Fedora supporting RISC-V is questionable. I looked, apparently their last image for RISC-V was Rawhide from over a year ago.
Well, isn't RedHat owned by IBM now? I'd question everything they do, especially considering that IBM would probably rather see RISC-V go fuck itself.
Re: (Score:2)
On the other hand, a good chunk of SiFive's money comes from Intel.
Re: (Score:2)
Seriously? You have a link or something?
Re: (Score:2)
Second DDG link [sifive.com]. Or Wikipedia.
(Apologies for not searching for a better link, busy.)
Re:frist psot (Score:4, Insightful)
Clock speed is largely irrelevant when talking about CPU performance. All it tells you is how many clock cycles per second it has to work with, but how many instructions can it execute per clock cycle on various workloads?
Modern CPUs treat machine code as a kind of intermediate language, re-ordering instructions on the fly and doing live performance profiling to optimize flow control. Now we are well into the age of multi-core too performance also depends on the ability of multiple cores to work together and of threads to be dynamically allocated to them. Access to resources like cache and RAM are key.
ARM also makes power consumption a key metric.
On the software side a lot of effort is put into improving compilers to produce better machine code, tuned for different families of CPUs and for different profiles (performance, power consumption, small executable size, debugability).
RISC-V can potentially compete on all these fronts but it's a very big and expensive hill to climb.
Re: (Score:2)
Less relevant that it was in the old days. Not irrelevant, far from it.
Re: (Score:2)
Risc-V stuff is still very pre-production, though the goals for it are the same as ARM.
Like, there is a good potential for , like in maybe 20 years, RISC-V will supplant ARM and Intel parts as there will be nothing left to patent as far as legacy CPU designs go. Everything needs larger GPU/AI compute stuff and neither RISC or ARM have parts that do this. Heck only nVidia does, and by acquiring ARM, they could theoretically design ARM SOC's that scale linearly with the GPU power for AI stuff. However Intel C
Re: (Score:2)
They sure surely trying to get the ball rolling..
https://www.seeedstudio.com/Si... [seeedstudio.com]
Re: (Score:2)
Redhat and all RPM distributions are very much the minority of Linux installation these days, with no sign of that trend turning around or slowing down. So this [debian.org] is what matters.
Is there a Betteridge browser extension? (Score:5, Funny)
Re:Is there a Betteridge browser extension? (Score:5, Informative)
That's not entirely fair in this case. RISC-V was designed as a competitor for the ARM architecture, and it should grow to be a respectable competitor there. A few energy-efficient supercomputers notwithstanding, both of those compete much better in the low-power and performance-per-watt metric rather than total performance for a personal computer, which is where AMD and Intel compete.
Re: (Score:2)
RISC-V looks more like it was a competitor to ARM embedded SoCs. Not with stuff like NeoVerse and the like.
Re: (Score:3)
Some people are looking towards data center and more HPC applications for RISC-V implementations, like Alibaba's XT910, and there is a general vector extension to the ISA that should support scalability in that direction. You are right that it's behind ARM right now, but the royalty-free nature of the ISA is attractive to a lot of companies.
Re:Is there a Betteridge browser extension? (Score:4, Informative)
You are right that it's behind ARM right now, but the royalty-free nature of the ISA is attractive to a lot of companies.
Well, than it's a direct competitor for the 8051!
Seriously tons of embedded chips use 8051 cores because it's completely royalty free. Some of the higher performance ones have moved to Cortex M0 and M4 cores, but the tradeoff is higher cost due to license fees.
Re: (Score:2)
How/why would an INTEL 8051 be "royalty free" - do you make them yourself?
Re:Is there a Betteridge browser extension? (Score:4, Informative)
How/why would an INTEL 8051 be "royalty free" - do you make them yourself?
No one's getting them from Intel any more. All the patents have long since expired so every one and their dog can make them now, and there's loads of implementations out there, including royalty free ones because ain't no one paying royalties with that much competition around. The margins aren't up to that.
Re: (Score:2)
The Chinese ones cost about a nickel.
Re: (Score:2)
Ah, I did not think about that.
Astonishing that for such a shitty thing is still a market.
Re: (Score:2)
Right, and this is the same wall that RISCV will hit:
It's a lot easier for microcontroller makers to build their own custom RISCV, (start with the familiar, then customize), and they should be able to use it effectively.
Now, imagine a Linux Distro that has to add driver support for two-dozen different RISCV implementations? They will just choose the best implementation, and ignore the rest (which will require much infighting, much like every advancement in Open Source).
The idea that RISCV is going to get p
Re: (Score:2)
Reminds me that I actually wanted to set up a VM with Hurd, and one with Plan 9 :P :D!
Thanx
Re: (Score:2)
If the chip makers want support for their devices in the Linux kernel they will have to write it themselves. SiFive has already done that for the HiFive Unleashed board, and it will be available for the HiFive Unmatched as well when it is released.
GCC and LLVM support RISC-V so it should be possible to build a complete Linux distribution. None yet exist, though there has been some work on versions of Debian and Fedora for the Unleashed. The Unmatched is a much more capable board than the Unleashed so it's l
Re: (Score:2)
Do report here on your progress...
Re: (Score:2)
There is an even shittier Chinese MIPS derivative that they sell for about 2 cents at quantity.
Re: (Score:2)
ARM competes with 8051 on TCO though. Sure the 8051 ISA is free but that's only part of the equation. You need to design and test and implementation of it, and then pump resources into developing debuggers, toolchains and support libraries.
8051 already has a lot of that stuff because it's been around so long, but ARM e.g. is constantly developing the CMSIS RTOS and improving the LLVM/Clang ARM target.
That's partly why when you compare M0 and 8051 unit pricing they are pretty similar, if not a slight edge fo
Re: (Score:2)
ARM competes with 8051 on TCO though.
I reckon TCO is a junk metric because it's very easily gamed. A house brick has a lower TCO than, well, any of those but it's ROI is crap because it doesn't work as a CPU.
Sure the 8051 ISA is free but that's only part of the equation. You need to design and test and implementation of it, and then pump resources into developing debuggers, toolchains and support libraries.
Or buy a cell library (many have no royalties, so it's a fixed cost), and use an off the shelf toolch
Re: (Score:2)
I'm currently trying to convince the boss that we don't want an RTOS at all, bare metal is the way to reliability and ultra low power performance.
I have to use Keil at the moment, really not a fan of that.
Re: (Score:2)
Re: (Score:2)
You might want to read up more on it, then. It'd be very surprised to see RISC-V seriously challenging NeoVerse anytime soon, if ever, unless something is done to consolidate the standard around an enterprise or HPC target.
The value of RISC-V as it exists today is that the modular ISA lets you roll out a cheap 32-bit SoC that doesn't have much extra cruft. Just because RISC-V has ISA extensions for things like vector math/SIMD, hypervisors, etc. doesn't mean your RISC-V SoC has to support any of that. Im
Re: (Score:2)
You might want to read up more on it, then.
I've already done that, how is it going to change the design?
Just because RISC-V has ISA extensions for things like vector math/SIMD, hypervisors, etc. doesn't mean your RISC-V SoC has to support any of that.
That's called "moving the goalposts". Just because *some* SoC won't support some features doesn't mean that *no* RISC-V SoCs ever will support the application areas that NeoVerse does.
Re:Is there a Betteridge browser extension? (Score:4, Funny)
Is there a Betteridge browser extension?
No. (Duh!)
Re: (Score:2)
Whoosh
Re: (Score:2)
Oh no, it's a whoosh whoosh
Re: (Score:2)
Well played.
Is there a Betteridge browser extension? No. (Score:3)
See title.
Re: (Score:2)
Has anyone already written a browser extension that adds automatically the word "no" to each headline ending with a question mark?
No.
Depends what they're running (Score:3)
Depends on the target market (Score:2)
In the Windows world, no way. Microsoft won't port Windows to other hardware platforms until they absolutely have to. Without that, RISC-V isn't going anywhere in the Windows desktop and server market. In the low-power market, I think it'll have a hard time ousting ARM.
That does leave the non-Windows server market, though. That market's pretty big and primarily focused on Linux where cross-hardware portability of applications is fairly straightforward. If they can provide either better performance/price or
Re: (Score:2)
Re: (Score:2)
Everyone assumes RISC is "best for low power" because ARM had successes in that sphere,
It's not just that, at all. It's also because all modern processors are internally RISCy. RISC did win where it matters, internally. RISC architectures have proven benefits, but as long as your instruction set doesn't stink on ice (like classic x86 with its total lack of general purpose registers due to all of the instructions which MUST use SPECIFIC registers as operands) then it is much less relevant. amd64 solved that problem for PC processors, for example.
Re: (Score:2)
RISC-V is getting a pretty big share in embedded components. No patent issues as per ARM, so a CPU can be made quite cheaply. No, it may not have the cool stuff like multi-threaded cores, but for some real time applications, that isn't really needed.
Desktops and servers are becoming a fraction of the marketshare. If RISC-V gets enough support, it can easily sell more units than x86. Smartphones come to mind, as well as things like car ECUs, or IoT stuff, which tends to be ARM, but with a patent-free alt
Gotta give em some credit (Score:3)
The ARM world has been plagued by an absence of decent developer hardware. Want a 4c-8c board based on something like A76 or A77? It doesn't exist. You can get one of the SQ1 or 8cx-based laptops. Or you can just get an instance on AWS and develop on one of their Graviton2 systems.That's about as good as it gets.
SiFive is giving us a board with U74 cores and an S7 core, and . . . honestly I don't know enough about the underlying cores to know how much of the RISC-V ISA it supports in hardware. That's one of the sketchy things about RISC-V outside of embedded applications: which parts of the ISA will any given SoC support? How many ISA extensions can I use and reliably expect that someone else's RISC-V hardware will run my software? It's like being an AVX512 developer.
Re:Gotta give em some credit (Score:5, Informative)
So I found the documentation for the U74 core:
https://sifive.cdn.prismic.io/... [prismic.io]
Looks like the U74 core supports only five ISA extensions: M, F, D, A, and C. It's based on 64-bit integer version of the ISA. You can read here to see what is left unsupported:
https://www.cnx-software.com/2... [cnx-software.com]
No packed SIMD, no vector extensions, no bit manipulation. Oh and no hypervisor instructions either. Meh! Why not implement the ENTIRE standard if it's going to be a dev platform?
Re: (Score:2)
The board is a dev platform.
But if it used a purpose-designed dev-platform processor, it would probably cost 600000 € rather than 600 €.
Designing a dev platform around an existing processor is what made this board affordable.
Re: (Score:2)
They need a better existing processor then.
Re: (Score:2)
From what I see on the RISC-V website, the bit manipulation ISA extension isn't even finalized yet.
Re: (Score:2)
Several of the ISA extensions are not yet "frozen".
Re: (Score:3)
The specifications for the V (vector) and the B (bit-manipulation) "extensions" (ISA subsets) are not finished yet.
The SIMD extension is considered obsolete in favour of the vector-extension.
Not having vectors means only that performance is limited, but the current V draft is almost as capable for typical vectorised code as 64-bit ARM with the SVE extension, if not more, IMHO.
Lack of the bit-manipulation extension means that you'll just have to use boolean operations and left and right shifts to get stuff d
Re: (Score:2)
Kind of makes you wonder why they don't just deprecate the SIMD extension? I would like to see how RISC-V's V extensions stacks up against SVE2 (which, admittedly, doesn't exist anywhere in silicon - yet).
Also not having anything like AES or SHA seems like an oversight.
Re: (Score:2)
I was under the impression that most of that stuff wasn't frozen yet. Even though RISC-V has been in development for about 10 years, it's still very much a work in progress.
I have a bad feeling RISC-V will suffer the same problems as planned cities. It doesn't matter how perfectly designed and optimized your platform is if it takes forever to materialize. On top of that, despite how long it's taking to develop, it's not exactly revolutionary or unique, presumably to stave off patent suits and other IP qu
NO (Score:2)
This is obviously a rhetorical question in an attempt to catch people's attention while also dodging the real question. The answer to it is, yes it can, because to compete does of course not imply winning or losing. So yes, it can compete with Intel, ARM and AMD, and in fact it does ever since RISC-V was designed.
The real question is, is it going to win and dominate over AMD, Intel and ARM? This is also an obvious question, because we would have heard and read about it if it came even close in doing so.
Ask
Re: (Score:2)
The real question is, what does winning and dominating mean? Are we talking about overall processing power or processing power per watt?
I think Intel and AMD are in the first category while ARM and RISC-V are in the second.
But you have to keep in mind that the more technology advances in both categories, the less important the first category becomes except for specific cases (such as supercomputers) and the more important the second category becomes because of laptops, tablets, phones, AR (all portable comp
Neither (Score:2)
> The real question is, what does winning and dominating mean? Are we talking about overall processing power or processing power per watt?
> I think Intel and AMD are in the first category while ARM and RISC-V are in the second.
I don't think either matters, unless the new competitor is either MUCH faster per thread, or MUCH faster per watt. I'm not trading in my amd64 laptop for a RISC-V machine to get a 5% or even 10% performance boost on the two programs that the RISC-V machine can actually run. Es
Re: (Score:2)
ARM is heavily patented and ARM ltd aggressively enforces its patents. ARM makes sure there is plenty of patented IP in their current releases.
Re: (Score:2)
Let's be clear here:
The latest chips from Arm Holdings definitely included IP cores.
The Arm architecture itself is open, and open designs do exist.
In a freely licensed design, you can't make an exact duplicate of Arm Holdings' latest chip, you can make a new Arm-compatible chip which with run Linux for Arm.
The main point is if there will be a platform (Score:5, Interesting)
With ARM we are currently in a similar position as with the Z80 in the CP/M age. Yes we have a semi-coherent platform (Android), but it needs to be ported to every single device as they are utterly incompatible.
The main reason for the success of the x86 instruction set is that there was a common hardware platform below the operating system. If you want to run your software/operating system on barebone x86 hardware, you just put it into sector 0 of your disk right after the partition table. You will then find a list of your standard hardware at a pre-defined address. You will also be able to just blindly reach into memory or IO-space to get to your hardware as everything is perfectly well standardized. If you don't want to access the hardware directly, you have a BIOS which gives you a decent abstraction to it.
In short even as an operating system you will run on virtually any PC out there as they are all basically the same design.
If RISC-V manages to define a simple, flexible and widely used hardware platform, so it doesn't matter which SoC you use, then it could take off for serious use. If RISC-V falls into the same trap as ARM did, where every SoC is purposefully different, it'll be doomed for the embedded world.
Notice that CPU speed is not extremely important. A simple and slow computer can be of use, if it has other advantages, like being secure.
Re:The main point is if there will be a platform (Score:5, Insightful)
RISC-V is designed to ensure that almost every SoC will be purposefully different.
Re:The main point is if there will be a platform (Score:4, Informative)
Well not really, while RISC-V does have different instruction set extensions so do many other CPU architectures. That's actually not such a big problem since those things are just compiler options in regards to software development. It's like on x86 where you also have everything from an 8088 to a modern 64Bit CPU.
Re: (Score:2)
Re: (Score:2)
a common hardware platform
This. RISC-V can certainly compete with ARM, which still has no "common hardware" aside from what is part of the standard core (what ARM licenses, not what chip makers add to it). The ISA/VESA/PCI/etc. architecture outside the CPU core is the primary advantage that Intel and AMD have over both ARM and RISC-V. Some of us remember the early days of MS-DOS when you needed a custom version built with drivers for hardware that didn't try to clone the IBM-PC. (for instance, Tandy-2000 and the Sanyo MBC 5xx series
Re: (Score:2)
But, but... Linux on the desktop! Freedom to do your own thing! Choose KDE or Gnome!
You are trying to say the people do not want that?!?!
Sounds good until you reach the price (Score:5, Insightful)
Yeah? Tell me more.
I'm, uh, excited. Go on...
Perfect, that's my use case. And? How much is it?
A raspi with 8GB is $75. The answer to the question of whether RISC-V can threaten other architectures is no as long as it costs eight times as much. These are early days yada yada, but srsly wtf. I appreciate that it has PCIE, and that makes it a very different platform, and there's no economies of scale in RISC-V at this time, but ouch.
Re:Sounds good until you reach the price (Score:4, Insightful)
Yeah, this particular product is dead on arrival. The market of "I want Raspberry Pi level performance but in a giant PC formfactor that requires an expensive PSU at a price point that exceeds even the cost of an 8-core 3.6 GHz AMD processor + motherboard + 32GB of RAM that also has more PCIe and that PCIe being Gen 4".
In terms of whether a RISC-V product could emerge to disrupt servers, desktops, cell phones, and/or embedded? Probably not server (several have tried with ARM and despite having a relatively nice head-start working from ARM, none have really seen success). Almost certainly not desktop/laptop (Microsoft would have to invest to even stand a chance). It would be hard to imagine displacing ARM in mobile devices, since there's a large ecosystem of compiled for ARM out there that RISC-V would screw up. So we have embedded, where there's always a hope as device makers are generally game to try different things to shave a few cents off a chip. It could extend into the 'Raspberry Pi' space at an attractive price point in theory. The question is whether a sufficiently well resourced vendor would see a point in migrating from ARM (nVidia might drive people away depending on their strategy) into an architecture with less likelihood of being a trap down the line (RISC-V).
Re: (Score:2)
Yeah, this particular product is dead on arrival. The market of "I want Raspberry Pi level performance but in a giant PC formfactor that requires an expensive PSU at a price point that exceeds even the cost of an 8-core 3.6 GHz AMD processor + motherboard + 32GB of RAM that also has more PCIe and that PCIe being Gen 4".
Actually, I want a RISC-V on mini-itx like I'm running now, with a small 12V power brick. 16G of ram would be great, but I'd settle for 8G. So, maybe $199 or so if it's pretty low end, more if it's somehow worth more.
That's way more than a Raspberry Pi, but it's a better form factor, especially if there are some standard connectors.
Re: Sounds good until you reach the price (Score:2)
Re: (Score:2)
Of course, for the price delta, you could literally afford over a hundred years of electric bill of a raspberry pi....
Re: Sounds good until you reach the price (Score:2)
Re: (Score:2)
Depends on price. A Raspberry Pi's functionality for a lot of things isn't really tied to ARM. Most stuff, be it the OS, could be moved to RISC-V fairly easily. The issue is economies of scale and price. Even though RISC-V has the advantage of no need to pay patent royalties, having someone make a relatively fast, 4+ core RISC-V chip might be an issue.
As time progresses, I'm sure RISC-V will improve. It may wind up at a speed parity in a few years, just because it has one advantage over ARM, and that i
Re: (Score:2)
Right, that seems to be the consensus of the thread...
There are a myriad of little embedded special purpose or even general purpose without a huge closed-source ecosystem that I expect to jump on RISC-V, just as soon as a few viable high-volume vendors bother to come to the table.
I did bother to go to their site and saw they had a more RPi competitive product in price, but it just was terrible performance wise.
It seems like these boards are less 'product for the masses to buy' and more 'here's some samples
Re: Sounds good until you reach the price (Score:2)
Re: (Score:2)
The HiFive1 Rev B competes with ARM-based Arduinos.
It cannot run Linux and has no external DRAM.
Re: (Score:2)
Re: (Score:2)
Wasn't the original Raspberry Pi marketed as for educational purposes. A Raspberry Risc-V could do the same.
It was, but even the least capable version had all the trimmings of a desktop system, albeit with minuscule RAM. Still, I know firsthand that it was possible to run a minimal desktop Linux on it. Part of the idea was that you wouldn't need an additional PC to do development on it, like the computers of old that booted into BASIC but obviously a lot more capable. The lesser HiFive board you're talking about is better compared to the BBC Micro Bit [wikipedia.org]. Which, ironically, is named after a computer which actually d [wikipedia.org]
Performance? (Score:2)
What's the die area at a given process size, and what's the performance in comparison to ARM and x86? Cost isn't that important, because that is a function of demand and mass production. As in, if it uses less silicon die area for the same or better performance in comparison to an ARM cpu there is no reason it can't cost less if a phone or computer manufacture (raspberry pi?) switches to it and production ramps up.
Re: (Score:2)
Not anytime soon (Score:5, Informative)
Fragmentation (Score:5, Insightful)
Isn't the RISC V architecture highly prone to fragmentation, with each vendor adding its own extensions? That could doom it.
Re: (Score:2)
TL;DR: Yes, in a few lustres (Score:2)
'Nuff said
RISC-V is the future of computing (Score:3)
https://www.youtube.com/watch?... [youtube.com]
This is a link to a Lex Friedman interview with Chris Lattner.
Chris Lattner (born 1978) is an American software engineer best known as the main author of LLVM and related projects such as the Clang compiler and the Swift programming language. He joined SiFive as Senior Vice President of Platform Engineering,[1] after two years at Google Brain.[2] Prior to that, he briefly served as Vice President of Autopilot Software[3] at Tesla, Inc. and worked at Apple Inc. as Senior Director of the Developer Tools department, leading the Xcode, Instruments, and compiler teams.[4][5]
RISC-V sounds awesome. The days of suppliers of chips obsoleting chips would be over with this move. You litterally pick the chip with the features you want/need and stick it into place of the piece of silicon that got obsoleted. Lattner explains it better than I but this interview was definitely worht the listne to those interested in softawre and hardware.
Re: (Score:2)
[1] ... [2] ... [3] ... [4][5]
NullPointerException at lines 3, 4 and 6.
Re: (Score:2)
More like RISC-V is HD-DVD and x64 is Bluray. It's not happening.
Re: (Score:2)
Your argument by analogy failed to persuade me.
This year? (Score:2)
Could RISC-V Processors Compete With Intel, ARM, and AMD this year?
No.
Re: (Score:2)
No, probably not for twenty years, but that's no reason not to investigate its potential. Consumer attention spans are only relevant for consumer products and this is not one yet. Twenty years flies by faster than most think.
Technically, yes it can (Score:2)
But the market response is on a totally different path.
1998? (Score:2)
I know the clocks changed today, but I'm pretty sure that I didn't time travel back to 1998. RISC didn't grab hold then and it won't now. I'm sure it has some cool applications but there is a steep road to mainstream use.
Re: (Score:2)
Re: (Score:2)
Fair enough. That doesn't change the barriers to mainstream adoption.
Re: (Score:2)
Re: (Score:2)
Intel (CISC) architecture is established. It may not be technically optimal, but it works. I presume that RISC does not have sufficient technical merits to be worth investing in. The old stuff works, we know how to keep it working, and there is no point in throwing that out.
This clinging to de-facto standards grates against engineering principles, but we have to be realistic. An example I like to give is ASCII, which is a way of encoding bytes to represent human-readable text. It is a bit of mess, really. M
Re: (Score:2)
My point really is more about the difficulty of displacing the incumbent architectures. It needs to work with things we already use to gain support, unless it does something special, does something at low power, does something really inexpensively, has some stand-out feature, then its going to have a tough time. I'm sure it has places where it shines.
Re: (Score:2)
My point really is more about the difficulty of displacing the incumbent architectures.
In that case, RISC has already won, because every single modern CPU is internally RISC, and breaks multi-cycle instructions down into single-cycle micro-ops.
It needs to work with things we already use to gain support,
RISC-v is already supported by Linux, the world's most popular operating system kernel (used on more than half of all cellphones worldwide, the vast majority of servers, and in countless other contexts.)
I'm sure it has places where it shines.
Right now RISC-V "shines" in low-power embedded cases where the cost of an ARM license is significant. That used to be the only place ARM "shined", too,
Re: (Score:2)
The future could be different that what we expect (Score:2)
Who said that the world needed to be attached to x86, x86-64 (AMD64) or ARM?
Who said that Windows or Linux or MacOS were the end of our computing story?
Computing must evolve, and the best way is to have alternatives. No matter if they can't compete with flagship "current" machines or if they don't run Windows or if they don't have Fedora or whatever specific distribution ready to go.
I think that we still have some hundreds of years in front of us to invent new things. Who knows if Risc-V is the "fl
need more ram and pci-e (Score:2)
X8 video is ok
X4 m.2 good
X1 wifi / other (better to have an full slot)
but no other lanes? and only gig-e on board?
also 8GB ram is low.
A lot of y'all don't get the point (Score:2)
RISC-V is *early* days. gcc just got basic support for the architecture a few years ago.
Don't see this as an attempt to make a disruptive product, it's not. What is *is* doing is getting the architecture out there in a big way.
* 20k enterprise router using this board? Hell yes.
* 20k enterprise storage array using this board? Hell yes.
* Open source developer want to play with something other than ARM + x86? Hell yes.
All of the things above help push the ecosys
Re: (Score:2)
Don't see this as an attempt to make a disruptive product, it's not. What is *is* doing is getting the architecture out there in a big way.
No, it's getting it out there in a small way. It's got only a limited subset of the extensions that people will want on a desktop system, and it's priced high enough that only professionals needing a dev system will buy it. Enthusiasts will buy something ARM-based for half the price or less.
Re: (Score:3)
I just don't think that this is the product that makes sense right now, sorry. They should have gone with a raspi style board instead, ideally with the same footprint as the current model so you can use their cases. I guess the question is whether they needed the power from the ATX connector for their system, or if it's really just there to support the PCIE slot.