Slashdot Mirror
Open Source RISC V Processor Gets Support From Google, Samsung, Qualcomm, and Tesla (seekingalpha.com)
An anonymous reader writes:
Google, Qualcomm, and Samsung "are among 80 tech companies joining forces to develop a new open-source chip design for new technologies like self-driving vehicles," writes Seeking Alpha, citing a (pay-walled) report on The Information. "Western Digital and Nvidia also plan to use the new chip design for some of their products," while Tesla "has joined the RISC-V Foundation and is considering using the tech in its new chip efforts."
MIT Technology Review adds that while Arm had hoped to bring their low-power/high performance processors to AI and self-driving cars, "The company that masterminded the processor inside your smartphone may find that a set of free-to-use alternative designs erode some of its future success."
MIT Technology Review adds that while Arm had hoped to bring their low-power/high performance processors to AI and self-driving cars, "The company that masterminded the processor inside your smartphone may find that a set of free-to-use alternative designs erode some of its future success."
So it'll spy on you?
that there are so many to choose from.
We went through an era of tons and tons of CPUs. An open source CPU is very nice, and would be useful for it to be adopted, but is there something wrong about ARM based CPUs that they couldn't be used for this task? ARM is no slouch when it comes to performance, and it it is pretty thrifty when it comes to wattage.
Is there something ARM can't do that a whole new CPU design is needed?
So these chips are 'open source' does that mean that if I have the technology that I can create said chips at home? Does this also mean that the means of fabrication logistics are included somewhere like github eg: "wtf do I need to create this thing at home?"
All the stuff about how chips now a days have back doors for the NSA, and that is just the one country we know about. It makes me not trust any companies. I used to sort of trust them, they were idealists bringing us the future but reality is they are trying to sell us down the river like we're pariah-babies on the nile.
I really don't want corporate anything, I want machines capable of building machines and machine parts and I need the software and hardware to do it. It would be a boon to humanity to create a much more distributed economy, as we have seen with software distributed anything pretty much hands down wins over centralized anything. Not that a mixture of the two would be a bad thing but at this point we are all just staring behind glass at technology handed down to us by the almighty powerful and it reeks of dystopia.
I saw a guy try to make his own toaster once from raw materials he mined, I think it caught on fire when he tried to make his first piece of toast with it. We should be able to make frigging toasters though, and fridges, and induction cookers, and goddamn near anything else we genuinely need. Could you imagine how hard it would be to take down a nation where everyone could rebuild just about anything and everyone stored basic environmental energy (wind, solar, rf radiation etc) such that destroying a chunk of a city didn't mean the lights in the other chunk went out?
I get the feeling though that this chip is about as 'open source' as my elbow is my asshole. There likely isn't any information about how to machine the damned things probably just specs that only someone with a multi million dollar techno-jazz factory could create. After all I read the article, it has no links to websites with data about this 'open source' chip. It has no data about how to actually create it even if you magically get the specs...so how exactly is this open source if only the elite can get the design and only the elite can create it?
No baloney about global warming. This is real tech news.
Stories like these RISC stories cause by penis to twitch and tingle. These stories make me want to solder something.
and licensing costs while still being able to hide their secret proprietary sauce.
Watch and see, the ABI for these things will be fucked to all hell within 5 years from all the proprietary forks of the architecture, and pretty soon we will be back to 20 effectively different architectures, even though they were all based off the same core design.
Why don't we make it really hip and run PLCs too!
Fucking doofus electrical engisneers.
Who has the say and responsibility for the design and in making trade-offs in design? There are going to be plenty to make, and different applications will need different things. Just a starter list includes things like: power requirements, instruction set and special instruction, performance, area, I/O, memory, self-test, coprocessors, design technology and tools, fab technology, supporting libraries and tool kits, and so on. Who will support design integration at the application level? Who wins when the performance hungry AI side runs up against the power and heat sensitive mobile device side?
That is before you get to manufacturing. Who will fab them? In what technology? Who will qualify parts? Who will stock the inventory? Who closes the loop between hardware defects and design?
Who will build and support the software tool chain? (Including the special tweaks needed for niche applications?)
Who will hold the patents? Copyrights?
Forming a grand coalition is the easy part. Getting something useful out of it is a much bigger challenge, especially when it comes down to actual hardware instead of just high level "standards".
much of left-wing thought is a kind of playing with fire by people who don't even know that fire is hot - George Orwell
lol - it will be bad compromises by committee
I've been working for a couple of years on Fedora and Linux on RISC-V and the "Seeking alpha" article is the strangest thing. The RISC-V Foundation offers BSD-licensed specs and multiple CPU designs (and a lot more besides). WD, Google, and many more are members. But they are not in any real sense "joining forces to develop a new open-source chip design". The design and chips are already out there, you can make your own FPGA or (if you're very rich) ASIC and have been able to for years. WD are going to switch all their hard drives to RISC-V soon. Google are likely interested because it could be used for their TPUs of their own design. "Joining forces" just means the companies subscribed to the Foundation for a very nominal fee, back-of-the-sofa loose change for these companies.
libguestfs - tools for accessing and modifying virtual machine disk images
Take a look at how big companies develop Linux kernel modifications. It's not that hard. You do the things you want for yourself, and then you share those with the others.
That is before you get to manufacturing. Who will fab them? In what technology? Who will qualify parts? Who will stock the inventory? Who closes the loop between hardware defects and design?
These cores will not be used as standalone CPU devices. They will be integrated in an application-specific SoC design that each company will have fabbed for themselves according to their own needs.
Most of them don't publish, thank you ghod. Been there, done that, I've seen a *lot* of bad kernel "optimizatons", correctly discarded by Linus Torvalds and the core kernel team show up on customized environments and completely screw up their systems. I still remember the former NetApp bozos claiming they could "backport any kernel changes needed" to keep their customized kernel. Turned out the only relevant change they'd actually made in 3 years wasn't even written by them, it was an optimization for the ext file system that had already been done better, upstream, in the next major kernel release, and they'd been churning whitespace and comment changes in the source tree to hide their lack of actual work. I wound up taking apart the source tree history to get it back into real source control, and I was *shocked* at the complete crap. Cost all three of them their jobs and saved the company almost a million dollars a year, counting benefits, to dump the bozos.
The point of this is not for one central party to manufacture some canonical RISC-V CPU. The point is having a base design than be tweaked if need and then added to a SoC, etc. by the designer to be fabbed by a foundry. Just like how ARM does not actually manufacture it’s core but licenses ISAs and core designs to third parties.
The requirements and consequences are pretty different between:
>>make world
and
>>fab chip --with $$$$$
One of those is trivially and cheaply repeatable, variable, and testable. The other is not.
much of left-wing thought is a kind of playing with fire by people who don't even know that fire is hot - George Orwell
Please don't make abusive replies. They waste everyone's time.
This is just the chip design. We would still need ASUS, MSI, Gigabyte, Intel (I would LOL if they released a board for a competing chip), etc, to provide a working platform to integrate these into. A standard socket, PGA or LGA, or soldered on to each board. Maybe we'll see them in a littany of SoC or SBC designs first, that seems much more likely honestly.
/* * pope1 */
Your point is well taken. Even so that only makes part of the problem go away. I don't see how one design is going to do it all even if you can mate it with different I/O, ram, etc. I doubt they can parameterize the design sufficiently to cover the range of uses this seems to be intended to cover. I expect there will have to be multiple variants to choose from, and the associated design support, etc.
But hey, if there is one thing we're short of its processor core IP. :|
much of left-wing thought is a kind of playing with fire by people who don't even know that fire is hot - George Orwell
Alright, so you can have open source linux contributions from large companies.
Meanwhile, Qualcomm is basically a patent troll with a sideline in silicon design. They never operate in good faith, and they have a long and inglorious track record of getting a standard that, surprise! they have a patent on, or simply using their patents to screw their competitors and collaborators with lawsuits. When they're not using their patents, they're just not serving users, they don't bother to innovate unless it's to squash their competition.
The reason why GSM wasn't a global standard for many years but CDMA stuck around? Qualcomm, because they got more patent revenue that way. ... many more
The reason why Android Wear sucks? Qualcomm.
Qualcomm, much more than any other large tech corporation with lawyers and patents, is toxic. An "open standard" with their involvement will have a trap.
"For a successful technology, reality must take precedence over public relations, for Nature cannot be fooled"
One design won’t do it all. That’s why people will take the base design and tweak it their specific application. Just like how Apple and Qualcomm take the base ARM64 ISA and make their own CPU designs.
They're a fad just like pet rocks, chia pets, cabbage patch dolls, and spinners. Any time or money invested in them is money lit on fire. Buying one is literally throwing your money away. Write your congressperson and demand reforms in driver education, training, and testing, then we won't NEED shitty self-driving cars.
Yes but nobody is forcing you to buy it. Just like you get to choose a Linux distro and which versions of tls your website will support.
The open part is the design of the CPU, and there will likely be many variations on it.
ARM generally makes nearly nothing on a CPU/SoC. Some were less than a penny (US) and the max were closer to a dime. Many companies, Intel, Apple, and such have paid their fee and they're done.
So exactly WTF was ARM expecting long run? ARM is very good at low power. Very, very good in fact. But high functioning isn't their forte. ARM core designs are slow moving beasties. If a company needs a faster solution they need to be as heavy into the solution as building their own 64bit ARM a full year before anyone else, e.g. apple.
RISC V will allow faster design cycles. RISC V will let those champing on the bit to move forward without their high demand sitting on ARM's low power bus. One doubts they will all agree to use the same version of the silicon, just as one wouldn't expect them to share an ARM solution.
Even if it's only a dollar, you have to figure in the hassle of licensing, and the risk of getting hit with
increases later once the design is done and the product is in production.
A 64-bit CPU more simpler, faster, polished and easier to fabricate is https://en.wikipedia.org/wiki/... dated from 1992.
The first "1 TiB terapage" (the page #0) maybe reserved for NULL page's exception (e.g. segmentation fault of some programs).
Starting the next terapage (the page #1), how long are their offsets required from any instructions?
Do they affect the performance of caches/TLBs?
Is there any form to not lose the availability of this precious memory (the page #0)?