13 ms·
RISC-V on the Verge of Broad Adoption
- snazz 8y agoI wonder what kind of BIOS/EFI replacement there will be on RISC-V motherboards. A brand new ISA seems like a great time to throw out any legacy features and focus on a future proofed low level firmware.
- tyingq 8y agoWe want something like this. But I suspect "broad adoption" is more along the lines of Western Digital not paying ARM royalties for embedded SATA controllers (or similar). Edit: Intel could use it for their management processor just for the irony :)
- dejaime 8y agoI suppose that's true, it'll probably get adopted on those "embedded yet invisible" uses. Still, I believe that does bode well for more general adoption, once it starts replacing ARM on the OEM side of things, by having a positive effect on its hardware price.
- wyldfire 8y agoSadly everything that's come since x86 has done bootstrapping much more device-specifically and it always comes with signed bootloaders. If you're lucky you can opt-out. RISC-V will likely be designed with the goal of eliminating royalties, not making extensible/portable bootloaders.
- chriswarbo 8y agoI was a fan of OpenFirmware, which AFAIK was royalty-free and cross-platform (with a layer of Forth abstracting between software and hardware)
- zzo38computer 8y agoAnother thing that it is useful with Forth is that it means that even if there is no operating system it is able to boot, still the computer will be usable.
- est31 8y agoI think Microsoft was actually a force for standardization here: they didn't want to make multiple versions of Windows for multiple IBM PC implementations. So everything is designed in a way that you can have one OS image that boots on a wide range of hardware. Now compare that to ARM :). Here basically every device/SoC family needs its own linux build.
- pjmlp 8y agoMicrosoft used to have PC specification guides. https://en.wikipedia.org/wiki/PC_System_Design_Guide https://en.wikipedia.org/wiki/PC_System_Design_Guide
- nickik 8y agoIts absolutely a goal of RISC-V to make it friendly to good bootloaders and there are lots and lots of discussion going on how these interfaces should look. This stuff is still in flux but go into working groups for privilege architecture and security and you will see all these discussions. RISC-V was never and will never be designed with the goal of eliminating royalties.
- ronsor 8y agoUEFI/EFI is that "future proofed low level firmware" -- I think there's already a port of it to RISC-V processors
- nickik 8y agoUEFI is a pretty terrible firmware and a huge mess. HP ported it for RISC-V but they seem to have lost interest. The open source firmware people like are however doing a lot, coreboot, u-boot, linuxboot and so on. There are discussions ongoing about how to design the low level interfaces.
- StillBored 8y agoI don't see how you can say a specification is a mess. Particularly since it (and openfirmware) tend to implement far more of what is required to boot a "generic" OS than any of the firmware projects you mention. Projects that are all mostly linux bootloaders. Further, there is an open source mostly complete UEFI (tianocore) implementation. Some people dislike tianocore for mostly religious reasons (it has 2 character tabs, and an odd build system), but those reasons actually have nothing to do with base UEFI. Finally, UEFI as the OS interface is actually being embraced by u-boot and coreboot as the standard OS loader. Thats because they have realized that the services provided by UEFI actually do solve many of the problem users of uboot/etc systems experience. For one, it standardizes the update process for the actual firmware, as well as providing services/controls for managing the OS boot process following update/etc. It also has interfaces for plug in cards (PCIe option roms) and many other features that turn out to be critical to building a generic computing device.
- nickik 8y ago> I don't see how you can say a specification is a mess. A specification can absolutly be a mess. Over-complicated for what is needed 90% of the time and in the other cases its also not optimal. > Further, there is an open source mostly complete UEFI (tianocore) implementation. Tinocore is not really complete for what you actually need and most venders are so far down stream that the advantages of real open source don't apply. > Thats because they have realized that the services provided by UEFI actually do solve many of the problem users of uboot/etc systems experience. I understand that. But there is a reason why Facebook, Google and other providers move away from UEFI. > For one, it standardizes the update process for the actual firmware, as well as providing services/controls for managing the OS boot process following update/etc. It also has interfaces for plug in cards (PCIe option roms) and many other features that turn out to be critical to building a generic computing device. The way the update process is implemented is incredibly sub-optimal and I have heard people from Intel agree that it is so. The problem is that it creates and unnecessary parallel universe that is far more insecure, far harder to understand and with way worse tooling. Check out this talk by one of the people who wrote UEFI and he admits many of the issues: https://www.youtube.com/watch?v=1XDYORK2z_M https://www.youtube.com/watch?v=1XDYORK2z_M Then check out this by one of the Linuxboot people going into many of the existing problems from his perspective: https://www.youtube.com/watch?v=ZyZfS00LZ70 https://www.youtube.com/watch?v=ZyZfS00LZ70
- zik 8y agoMost embedded systems which run linux use u-boot as their boot loader. There's no BIOS really, just a boot loader. https://www.denx.de/wiki/U-Boot https://www.denx.de/wiki/U-Boot
- yellowapple 8y agoI hope OpenFirmware makes a comeback here.
- 0x1DEACAFE 8y agoWhere is the "LIKE" button when I need it? +1
- ChuckMcM 8y agoHah! EEtimes 2002 "Infiniband on the Verge of Broad Adoption" whoops :-) But more seriously, I think having a license free CPU core with software support is a huge win, it will enable people like TSMC and GF to make 'jelly bean' CPUs that can be low cost and high volume. But they won't be that much different than the low end ARM cpus which are basically the cost of packaging these days anyway. The market forces in a 14nm (last year's process node) world are pretty interesting. 90% of the cost is the dicing, testing, and packaging the parts. So the other market force is the Western Digitals of the world who will make RISC-V embedded SOCs with exactly the right set of peripherals they need to make a billion disk drive motor controllers. But most people won't see those chips in the marketplace. And there probably won't be a 'design to order' chip house that will make small volumes of these things. Imagine you are a Microchip or a ST Micro, what does RISC-V do for you? It lets you avoid the few cents you pay for the ARM license per chip, does it let you differentiate any more? Look how well that worked out for the AtMega16 for Atmel. So that is the embedded market, kind of a wash. But what about "bigger" systems? Laptops or desktops or servers? Do the pareto analysis on those systems, compare an ARM SoC to a RISC-V one. The costs are in memory, support chips, PCBs, tooling etc. Not the CPU license. Will it let people build chips for phones like Apple does with their own GPUs and microarchitecture? Sure, and at less cost, but what then does that give the average consumer? A cheaper phone with market features? Ok that's a win but they won't care it is RISC-V versus ARM v8 or what not. Bottom line is that I think it is great to have the ability to create computers that are not beholden to a certain vendor but I don't see the vector for getting them into general distribution where consumers actually benefit from the change.
- smallstepforman 8y agoThe info everyone is missing is code density comparison with ARM. Risc-V is more efficient and has about 10% denser code, which translates to more instructions fitting in i-cache, less memory pressure, and ultimately better performance and battery life. Long term roadmap, thats a win for Risc-V.
- duskwuff 8y agoIs "10% denser" comparing RV32 or RV64 against A32, T32, or A64? And is that with or without the Compressed Instructions extension?
- jeffdavis 8y agoDumb question: is an ISA really proprietary? Like I can't make an ARM emulator without violating some IP law? Or just an ARM physical chip?
- writepub 8y agoyes - ARM ISA is copyrighted. To implement it, you need an ISA license. ARM typically makes this license cost almost the same as the license for an already implemented core, which is why most vendors drag-and-drop ARM cores into their designs
- FPGAhacker 8y agoBut does this cover software emulation?
- wmf 8y agoQEMU exists so I guess they're OK with software emulation but Arm seems to go on the attack at the slightest whiff of an unlicensed hardware implementation.
- mcbits 8y agoAre they really OK with it or are they just legally powerless to stop it? That is the question.
- bluejekyll 8y agoIt may not be in their interest to stop it, as those are generally used for testing and building software to target ARM, which helps with their potential market share. Similarly, MS offers free VMs to developers for Windows so that you can build and test against Windows without cost.
- est31 8y ago
- lsllc 8y agoAll that's missing now is an affordable Linux RISC-V development board.
- epx 8y agoExpecting some sort of Riscberry pi to spend my money on
- dejaime 8y agoIn hopes for a RISC-V Pi
- drudru11 8y agoI would like to buy SiFive’s unleashed dev board. It has everything I need: RV64GC, gigE, 8 mB ECC, and usb UART. I just won’t pay $1000. I wonder what the economics would have to be to get it to $300.
- ahmedalsudani 8y agoOof that's steep. Assuming ARM doesn't try to kill RISC-V, I imagine the economics will improve significantly over the next few years. Probably larger than 75% drop for similar specs over two years as the low hanging fruit has not been picked yet... But it's important that there's still demand for RISC-V, which is where ARM might be able to choke it off.
- imtringued 8y agoBelow 100 thousand units it doesn't make any sense to have custom silicon. The ARM SBCs are all based on existing TV boxes, including the original Raspberry Pi so you are unlikely to see a development board before a proper laptop or NUC-like desktop is released.
- 0x1DEACAFE 8y agomeh. i think the unit count is down to about a thousand. also strongly dependent on what node size you use. 180nm and 350nm are pretty cheap these days. but you're probably only off by an order of magnitude if you're talking about 28nm (which is what the SiFive HiFive Unleashed chip uses.) but... your point was probably more along the lines of "getting custom silicon is still going to need a market of greater than several thousand," and that's probably always going to be true.
- JohnJamesRambo 8y agoCan someone eli5 why people keep wanting to make RISC-V happen? I'm not in tech and I keep seeing posts like this.
- AnIdiotOnTheNet 8y agoWell for one, techs love hype. Otherwise it is mostly the association with the word "open".
- pcwalton 8y agoThe dominant machine languages—the language compilers compile programs to so that the code can run on the silicon—are all proprietary and protected by patents. RISC-V offers a free, open-source machine language, as well as open designs for the hardware. You can more or less download their designs for free and send them off to a factory to make high-quality processors.
- tenebrisalietum 8y agoIntel and AMD have put backdoors in their CPUs - Intel with the Management Engine and AMD with the Platform Security Processor. These are separate sub-CPUs required by both platforms to boot and run. Both of these CPUs run proprietary, encrypted, otherwise-inaccessible and inauditable code that is always running and has full access to everything the CPU has access to. The Intel ME in particular is part of Intel's offerings that allow remote access to a system even if the operating system fails--it's accessible remotely by design, and flaws have already been found in it. Only Intel can update the ME. Only Intel knows what's in the ME and what it's doing, same for AMD and the PSP. I think people want RISC-V to succeed because it gives privacy-conscious individuals a chance to have a truly secure platform not 100% under control of a single company with 100% auditable code from each stage of the boot process.
- mda 8y agoI think Risc-V does not have this goal and ISA can not enforce it. There could be a Risc-V processor with worse characteristics than Intel's.
- 8y ago
- steamport 8y agoOpen source CPU? where can i buy one :O
- Newtonip 8y agoSearch for "HiFive1". It will cost you around ~$100. If you want something cheaper, for around $20 there are a bunch of mini FPGA based boards available on Aliexpress purposefully made to load into them a soft-RISC V core.
- lambda 8y agohttps://www.seeedstudio.com/Sipeed-MAIX-I-module-WiFi-version-1st-RISC-V-64-AI-Module-K210-insid-p-3206.html https://www.seeedstudio.com/Sipeed-MAIX-I-module-WiFi-versio...
- IshKebab 8y agoI'm still unconvinced by the modular nature of the ISA. Are we going to end up shipping fat binaries with a dozen binaries for all the different variants that are allowed? On x86-64 there's a reasonable minimum supported set of instructions, and you only really need runtime CPU detection for the newer vector extensions.
- audunw 8y agoI wouldn't expect server/desktop/phone variants of RISC-V to drop any of the current extensions of the ISA. Why would you, if gate count isn't what you're optimising for? There could be further, more advanced extension that may not be in every processor, but that's just the same as x86 The modularity in the ISA as it is now, is something that's more geared towards embedded devices or FPGAs, where you'll be compiling code specifically for that target anyway.
- kabacha 8y agoEasy solution: Just GPL it - every extension of RISC-V has to be open as well.
- IshKebab 8y agoI don't see how that is a solution. The problem is the variability in what the hardware implements, not the availability of extension specifications.
- BrandonMarc 8y agoEven ESR himself praised them a short while ago. http://esr.ibiblio.org/?p=8242 http://esr.ibiblio.org/?p=8242
- walrus01 8y agoWhere can I buy a motherboard and CPU, in super low volumes, right now?
- reportingsjr 8y agohttps://www.sifive.com/boards/hifive-unleashed https://www.sifive.com/boards/hifive-unleashed You have to get the extension board if you want PCI-E, SATA, and M.2 connections, but it doesn't look like there are any available right now.
- walrus01 8y agoA thousand bucks? Seriously? This is why the Qualcomm centriq arm server cpu never caught on for use with Debian, centos, or any mainstream distribution. Same problem. Unavailability of normally priced cpu+motherboard. I'll check again in six months...
- baobrien 8y agoThe HiFive board really isn't aimed at the hobbyist or datacenter. It's a dev board with a low volume part made on a high end process. The goal is more to get silicon in the hands of people who want to work with SiFive cores and those who are working on the RISC-V software ecosystem.
- openloop 8y agoYou asked for it.