It will be interesting to see if say a Google will move its custom SOC from ARM to RISC-V at some point. I think that would be a turning point for RISC-V wider adoption.
The problem is that if MediaTek and Samsung stay on Arm, the Google moving to RISC-V on its own would kill Google Android handset business. Any RISC-V chips it produces won't be competitive (for years) and so it will fade into insignificance while Samsung and MediaTek take over the Android market.
@@GaryExplains I think it's likely you'd see a chip alliance emerge where multiple companies design one chip for multiple uses. And there's nothing stopping the ARM companies from also designing their own RISC-V chips as well.
@@GaryExplains Google controls the Android eco system via the play store. It wouldn't matter what any other SOC manufacture does. RISC-V is catching up, I expect China will be pushing RISC-V hard in the next few years, as a way to avoid US/UK sanctions restricting access. The JH7110 is impressive, that we could already be seeing RISC-V SBCs comparable to an RPI4(yes, no where near top of the line) in both price and performance. Phones might not be that many years away.
@@puffin_000 Yes, the specification is open. But designing a high performance, power efficient processor is hard, so it isn't something that "every company" will use. Only a few specialist companies are capable of using it. Companies like SiFive and Imagination are already licensing their RISC-V CPU designs, exactly in the same way that Arm does.
@@GaryExplains I don't think so. If some folks are unite, they can create efficent RISC-V procesor with under "licence free" open source architecture. Because factoring a CPU is hard and can't doable in home but designing an CPU architecture is doable in home if I know correctly. Also because of ARM's or other companies NDA's, people can't collabrate to create an open source ARM design. But with RISC-V's open specification, this open source architecture design is not a dream anymore its doable. Maybe nobody will do that in 5 years but its not impossible too. I think companies is not the only players in RISC-V.
Great points made & also ARM aren't standing still, so however well RISC-V develops - it's playing catch-up with a still evolving (and well resourced) target!
Nobody says ARM is standing still. What truly matters however is results. The time, manufactures will be able to sell RISC-V chips (and the term "chips" here is more generic) that have a better performance per cost and per watt, is the time ARM has lost. Everything is about money in the end of the day!
While i agree with all your points objectively. There is one real thing that is carrying RISC-V and giving it momentum, and its simply based on the fact that its an open source instruction set. As a result, a big majority of the new innovations in computing hardware and architecture is happening on RISC-V. I am working directly with the instruction set to develop hardware, and the amount of resources, tools are growing at a crazy fast pace. For sure the juggernauts will continue to innovate on ARM, x86 as well, I don't expect that to change, what i think changes is all the cool new innovations comming more from RiSCV based projects
Very useful to bring forward a sobered view on things. The invention of the wheel was nice. However, it was the applications of the wheel which made it worthwhile. Same things with RISC-V or other innovations. Thank you for this video.
My enthusiasm about RISC-V is partially about the open standard and mostly about it being another competitor in the CPU market. More competition is better. I like Windows on ARM for the same reason, plus the potential for better power efficiency compared to x86-64
My only concern about RISC-V is that I don't want it to slow down ARM overtaking x86 in the near future. I want to see more competition in the Laptop/Desktop space and x86 only enables Intel/AMD, ARM has far more competitors using their IS. Once x86 see's less than 50% market share in the laptop/desktop space, I am all for a load more instruction set competition going forward.
ARM shot itself in the foot by suing Qualcomm. No one trusts them anymore to be in a position of owning IP anymore. They are now coasting on the momentum from the time before they sued Qualcomm, but everyone wants to get away from them. It unlikely that they will overtake x86 now.
@@FreshSmog Who?, at the moment is the most used recent architecture for servers, amazon, google and nvidia have tons of servers based on arm (nvidia give service to car manufacturers for tons of data process and AI stuff).
Possible Google will make ARM to RISC V translation layer but it likely won’t make everything work well but could be huge step to make it possible for lots of ARM apps to work on RISC V. Outside of China I would think low end devices would be where RISC V would start to get market share which the ARM equivalent wouldn’t be able to run these high demanding apps such as video games (which doesn’t use javaruntime).
@@GaryExplains Let's take 3 companies intel, arm and apple. Best x86 processor is made by 2 companies intel and amd , intel stuck to 14nm for around 7 years because there was no competition from amd until amd came up with ryzen cpus. Qualcomm and mediatek tweaks design of the arm processor to get the best performance for example adreno compared to mali gpus and last one is apple, apple shocked the industry by arm based cpus that have far better performance per watt than x86 and arm. Apple is able to achieve it because it designed its cpus on top of arm and also apple owns exclusive license for arm. With risk v companies will have the ability to modify and tweak the architecture compared to proprietary design, and as we all know chip fabrication is done by few companies like tsmc and samsung. windows on arm is still terrible until apple came and proved them wrong, more competition is better for consumers. I hope this answers the question.
Competition is good, yes, but notice how all the examples you gave happened without us needing RISC-V. Also Intel got stuck at 14nm because it couldn't make the tech work. Not because of lack of competition.
Reminds me of Linux in the mid 90s vs Microsoft. Not really competitive for most uses and users. Now it's different. As innovation rate declines it will be easier to catch up. That said with RISC V you still need a hardware design and layouts for a lithographic process. This is like saying each piece of software has its own compiler and assembler in the software world. Things might change as there are benefits to the industry to have standardised tooling and ways of doing things. Google are pushing this but only on 20 year old process nodes and not TSMC 3nm etc.
Intel’s “Horse Creek” RISC-V SoC - demoed at conferences since September and due out commercially in the summer - is the very first chip made on Intel’s new “Intel 4” process node.
When they said RISC-V was more efficient than other architectures, they talked about the reference implementation of the RISC-V core in that new high-level hardware description language. Those claims were made based on hardware benchmarks on the same process node. So yes, if you believe the benchmarks and data, it is, in fact, more efficient (but this is old information from several years ago, so these things could have changed possibly). That doesn't translate to real-world gains unless you use it as a microcontroller with the appropriate process node and implementation. For servers and desktops, as well as cellphones, there is a lot of work to do in terms of optimisation and tunning of the existing code (video codecs and other performance-sensitive code) for RISC-V.
I am a keen hobbyist and the main attraction of risc-v processors is price. I only need very basic functions and 10p a chip is very attractive for experimentation.
@@GaryExplains being from Yorkshire, I am quite price sensitive... I'm using a 10p risc-v MCU 32 bit 48MHz ch32v003. I cannot find any other MCUs in the same price range with similar specs. At the level that I am coding on, it really doesn't matter the ISA of the chip. I'm cutting and pasting example code and tying it together with very clunky C. Currently building about 50 tiny controllers for addressable LED strips. :)
But the price of the microcontroller is insignificant compared to all the other bits you need to support the microcontroller and the led strips aren't cheap (if they are programmable).
Also I see that chips like the Puya PY32F002A (based on an Arm Cortex-M0+) is $0.08 in bulk. The idea that RISC-V is somehow cheaper is a bit of a myth.
Just a clarification: " What is the license model? - The RISC-V ISA is free and open with a permissive license for use by anyone in all types of implementations. Designers are free to develop proprietary or open source implementations for commercial or other exploitations as they see fit. RISC-V International encourages all implementations that are compliant to the specifications. Note that the use of the RISC-V trademark requires a license which is granted to members of RISC-V International for use with compliant implementations. The RISC-V specification is based around a structure which allows flexibility with modular extensions and additional custom instructions/extensions. If an implementation was based on the RISC-V specification but includes modifications beyond this framework, then it cannot be referenced as RISC-V. Does that mean free for industry to use and play with, but then we pay if we produce a product using this ISA? -The RISC-V ISA is free for product use too. Those who want to use the RISC-V logo should join RISC-V International (see question No. 1). " (from the RISC-V website)
Interesting video. I don't know a great deal about hardware since I work with software but I've heard about RISC-V and it made to sound like it is somehow going to shake up things. Even then, I never bought into that because it's not like any person in a garage can make it, but takes a lot of capital to get a good chip design. It'll be something to see how RISC-V expands, especially with Chinese companies.
They can't manufacture it in garage, but they can design it thete. They can later outsource manufacturing to companies that aggregate many layouts and just make very few wafers (1-10) with the mask. I have read that it is reasonably cheap, e g 0.35 micron process for 650 Euros/mm2
That is just a myth. I have a video where I actually test a RISC-V Microcontroller against Arm microcontrollers for performance and efficiency. It doesn't turn out well for RV.
I think people are checking for alternatives once a chip architecture matures and usually licensing or IP usually pulls down most of them. Not sure if arm would cease and desist sd gen 2 w/ nuvia ip.
technically, chip companies won't have to pay royalty to ARM anymore which saves them money. but initial R&D is higher which brings in newer costs. so user wont see any effect on the front end if a risc v phone did come out. Its boring for users but very exciting for devs
But if you look at the VisionFive 2, the RISC-V SBC the CPUs are licensed from SiFive. So no change. The zero license fee thing only applies to companies that design their own CPUs from zero, and those are few.
I think it would have been great to discuss what RISc-V is and why the hype. You have done a great job throughout the video managing people’s expectations but have not really helped us understand what the big deal about it is. The way I understand it and what has me interested and probably fall in the 45.68% of people saying Yes for sure, really is all about the geopolitics around the semiconductor industry. You mentioned the eagerly awaited Oryon CPU from Qualcomm courtesy of Nuvia. The legal battle between ARM and Qualcomm has made RISC-V gain huge momentum. As a result Google have upgraded it’s status to Tier 1. It is not that we think we will be getting open hardware, but rather the manufacturers we rely on to develop the processors now have an open standard. They are not limited by their license from ARM or Intel/AMD. We have seen the consequences of geopolitics on brands like Huawei when they were blocked from using ARM or TSMC. Having an Open ISA can accelerate innovation and security.
I don't believe RISC-V will bring much to the consumers. However the openness could bring the ability to custom design chip for specific application. The things that pop in my mind are in ear headphones and glasses and watches. They don't need to do everything and power and size is critical
Hi Gary. Big fan. Love your explainers. Lately, I have been struggling understand these terms like 3nm process or 5 nm process. I saw in wikipedia these "nm" values have nothing to do with any actual physical property. These are just marketing terms. Can you please clarify the clutter and actually provide details.
I can see Chinese companies adopting RISC-V in order to avoid future US technology sanctions in at least that area. For those companies already bought into ARM, then it's going to be a balancing act over whether continuing to pay ARM for licensing is cheaper than going down an entirely new development path and having to re-learn what they already know. Also, there's all the other parts of the ARM SoC ecospace that needs to be taken into account. I can't see Apple changing yet again, but then they have surprised me before. Of course none of this is relevant to the area of bespoke processors where it looks like RISC-V has a big commercial advantage as it's much more flexible. That, of course, leads to it's own issues if RISC-V starts "forking" making use of that flexibility.
i think that risc-v has the potential to replace both arm and x86_64 architecures. if all the companies try to go that way, mobile, desktop, microcontroller and even servers can use risc-v in the future. compile once and run everywhere! in this case even technologies like webassembly and jvm can become obsolete and every compiler will directly produce risc-v code.
@@GaryExplains arm would not want that but intel may want because intel wanted to change their architecture once (itanium series) but could not manage to do that.
Gary, maybe I’m mistaken cause it’s been a few decades but I remember the RISC vs CISC debates. I recall that RISC is a classification. Therefore a Qualcomm Snapdragon is a RISC chip by classification. So is RISC-V a brand name of a RISC classification chip or am I misinformed?
I have lots of videos covering this kind of question. RISC-V is a brand name of a RISC instruction set architecture. You may find my series on RISC-V interesting. First part here ua-cam.com/video/4qBKOAv0sBI/v-deo.html
If risc v can compete with arm Maybe after 1 decade atleast it will create competition and companies could buy it for lesser price and then sell the phones or processors at lesser price. All depends on their success. Seems very unlikely for now. But u never know.
Why do you think the price will be lower? The most significant cost is manufacturing. Any license fee that isn't paid to Arm is used to design the processor, that is the point. When a chip maker uses an Arm Cortex-A CPU they pay for a ready made CPU design, no need for their own design team. Why would it be cheaper?
What do you mean by options? I reviewed the VisionFive 2 SBC on this channel. It has a RISC-V processor. The CPU part is LICENSED from SiFive... Just like a company would LICENSE a CPU core from Arm. How is that any different or any cheaper?
I can see risc-V being strong competition to ARM in the micro- controller market at the cortex M and cortex R levels. But I am not holding my breath in the Cortex A7xx and cortex X cores. Not saying never but......
@@godnyx117 from memory , the A78 is early 2020 since then we have had A78C, A710, A715 and A720 in the A line. And X2, X3 And X4 in the high end Cortex X line.
@@patdbean I am aware of that. But still, for RISC-V, that's an improvement. P670 is also late 2022. They may release something new in H2 or H3 of 2024 if we are lucky...
RISC-V is inherently cheaper, but not more performant, so the top tier would be only competitive with the mid or low tier of the others unless it has the same investment.
There is an effort to bring RISC-V to everything from IoT to desktop to server. This is something that is just not available for ARM, there is no ARM desktop, and no Raspberry Pi does not count. SiFive/Intel horse creek is going to be interesting.
ATX is a design for PCs with its hereditary back to the 1980's. Qualcomm didn't get sued because it bought Nuvia to do ARM on the desktop, it got sued because it broke the license agreement. Same would happen if Qualcomm was using RISC-V technology from SiFive or Imagination, it is all licensed stuff.
There are already plenty of open source CPUs (from unviersities etc). However there will never be a competitive open source SoC because why would a company (with employees, rent, energy bills, etc) ever do that? This isn't like Linux where they can offer services to supplement the loss of income from direct sales. Secondly manufacturing the chips still costs millions.
@@GaryExplains A consortium of companies could heavily benefit from such an effort, because in order to build their final product (device, OS, whatever) they can target a common set of features safely. The chips themselves will no longer be the differentiators. lowRISC's stated mission is open silicon, including a 64-bit SoC, so organizations with this goal already exist, but I guess your argument is that they will never be able to compete with the big dogs? Maybe not in the next 2-3 decades, but in the fullness of time, maybe.
It is hard enough to predict what will happen in tech over a 5 year span, predicting about the next 30 years is highly unadvisable. Think back to 1993, Windows 95 wasn't even out. Tell me anyone's predictions from 93 that came true by 2023!
@@GaryExplains I mean, they will still sell you the hardware, right? So I do not see them losing anything but putting out the specs and detailed info about their hardware. If anything, it will make people trust them more and give them a better image. It's similar to open source software that are self-hosted, selling you a set-up server plan. Let me know if I miss something.
@@godnyx117 What they lose is that then anyone else can make a version of their chip and sell that to consumers rather than the company that invested millions in designing it in the first place.
But will apps follow suit? The main thing that fucked Windows RT was the lack of popular software that people use. No-one wants to use Windows RT if a certain browser, media player, downloader software etc etc isn't avilable for it.
You can create a RISC-V softcore processor in a FPGA, so that is a way you can create a processor without spending a small fortune :) So in that way it's more like your $50 pc doing opensource...
@@GaryExplains, but it does give you a path to learn/create/contribute to opensource hardware. Pano Logic's zero client was a Spartan 6 fpga. But they did go belly up...
So you think an amateur is going to help create a world class RISC-V CPU by learning on an FPGA? I don't think you quite understand the complexities of building a modern superscalar processors. But even if a hobbyist contributes to a low-end, simple RISC-V implementation, then what? Who is going to make it? Why would they make it? And what would it be used for?
So your argument is that opensource hardware community members don't possess the needed skills to contribute? And that it is pointless to share works on github or OpenCores as no company uses opensource hardware designs instead of proprietary IP? Everyone has to start somewhere, and I believe it is amazing that for just a few bucks you can learn Verilog or VHDL and implement something that used to be entirely out of reach for the individual. So sure an amateur/hobbyist isn't going to build a world class super-scale processor design, but then neither is a amateur/hobbyist going to fix kernel defects. But we are at a time when hardware tools/ip are within the reach of ordinary people (amateur, hobbies, professionals, past professionals).
Let's say you contribute to an open source CPU. Nice. Who is going to spend the millions to fabricate that CPU? Why would they? What market would it be used in? If a company did fabricate it, then since it costs money to make a processor then unlike software you can NEVER benefit from your contribution unless you spend money to buy the product that is using your design. You give away your time and intellect to design the CPU and then you give your money to the company that used your design for free so you can use it. Nice.
Oh God the cope Gary. He'd love to have it because new stuff is interesting, obviously! I wonder what kind of raving fanboy crap you get in your inbox that made you develop such an incredible bias against RISC-V...
But to try answering Koysdo's original comment: I don't think there will be upstream support for the board, because Google said at the conference that they wanted to wait with an official platform release until a few more extension specs have been ratified and included into a profile. That said, there might be an "unofficial" (as in not Google backed) port for the board and others using that SoC, since there are a few manufacturers making boards based on it and the SoC maker wants to build more of an ecosystem around it.
I just hope that all the software's will be built with hardware accelerator in mind. What RISC-V need now is to take all the programing language libraries and just make any code written in those language can be easily converted to RISC-V with all the Hardware Acceleration built-in. - if you're a programmer with the source code of your program, it would be nice to just select it to compile to RISC-V instead of ARM or x86. - The libraries subs & functions for RISC-V will have to use the same names as the original in ARM & x86.
Frankly I would prefer a phone that was not android or Apple. Yes, a RISC-V SoC would be a very good feature, but it is the intrusive OS that worries me more.
Some people aren't interested in the same performance as current anymore for smartphones. They're plenty fast already, and even if one is as fast as a mid range processor from 3 years ago you can do anything you need. What it does bring is a more open hardware platform, where manufacturers and customers aren't subject to the whims of the overlords, like with the Arm/Qualcomm lawsuit. Maybe for licensing the RISC-V name some can get in trouble, but that's about it. Stop using the name and you're fine again.
Some of that may be true, it is debatable, but let's assume it is. But then, what about the overlords of GPUs, DSP, 5G Modems, Image Processing, audio codecs, video codecs, Bluetooth, Wifi, etc. Having an "open" CPU resolves nothing in any meaningful way.
The problem is this. MediaTek, Samsung, and Apple don't have any problems with Arm. So if Qualcomm goes the RISC-V route it will be doing it alone. I can't see Samsung suddenly switching to RISC-V for the Galaxy S series just because Qualcomm has a gripe with Arm. Any RISC-V chips it produces won't be competitive and so it will fade into insignificance while Samsung and MediaTek take over the Android market.
RISC-V will catch up with ARM eventually. There are too many people working on it, especially China. Also ARM suing one if it's biggest customers is not a good look.
@@GaryExplains Why won't they be competitive? If Nuvia and Qualcomm engineers can design a competitive Arm -ISA chip then they can just as easily design a competitive RISC-V chip, and indeed do it reusing 95% of the work they've already done.
True. I didn't mean to imply that it wasn't possible. I meant to say "first generations". I think I wrote that in another comment to someone else and didn't do it this time. If the first gen chip isn't on par or better than the competition then it is game over for such a switch.
Why have you implied a connection? RISC-V silicon implemented using Apple's design would be the same size (or bigger) than its current implementation. Also, why do you say Apple is approaching a limit? Apple's current designs are quite modular hence the Max and Ultra etc.
I always wonder why the Chinese (mediatek, Huawei etc...) doesn't pick up riscv and build upon it? I think in 5 years time they can have something viable... Well google up there, linux kernel is there i think i saw arm emulator that run on riscv. I mean in the long run they can save million upon millions on licensing for arm... Will see...
@@GaryExplains If you look at your RAM map address is the same principle. There is a video a guy on UA-cam who shows you how to do it. It's like any circuitry pcb you can reverse engineer If you got a picture of it
I believe that we will see the first RISC-V mobile phones in the Chinese entry market as soon as they can actually make them. China needs some sort of chip independence from the West/Taiwan, and RISC-V is an opportunity for them. Making a competitor for the high-end market is a completely different thing.
@@GaryExplains quite, as far as I know china firms can still licence ARM cores. Their problem is that if they are struggling to fabricate at 16nm while TSMC are at 5nm and moving to 4nm It dose not really matter what design/ISA they are using. They are still a decade behind.
I think RISCV on android could've a big boost if Google decided to make the next tensor chip using that instead of ARM, but odds are any support you see for that would be for that RV-tensor chip and not RISCV on android in general.
Yeah, things like that. And don't laugh I have had conversations with people who thought that because they learnt from Linux that everything is free (as in cost).
@@GaryExplains Could you make a video on those conversations? It would be interesting to see what the thought process is. Like, I can download a RISC-V SoC for an FPGA as of today (e.g. based on LiteX), run it, and it costs me nothing - same as with the Linux kernel, to keep the analogy. Is that something that came up in those conversations?
I don't think such people really thought it through fully. Basically the words "open source" mean free (no cost) to many people. By associating RISC-V with open source the result has been confusion.
Actually, in regards to searching for malware on the microprocessor, I don't see why you can't audit the process of going from microprocessor specs to the physical hardware... after all, YOU have the hardware and the specs in your possession... and there have been so many ways these chips get reverse engineered in the last 25 years that it's gotten to the point where the hobbyist can play along... Duo Labs has a cool article where invasive analysis is used after acid etching to recover firmware off of a microcontroller. If that is within the reach of hobbyists, just think what guys in bunny suits can do in a billion dollar lab. I'm sure some independent security researcher funded by the competition would love to nail a huge company for putting malware on their microprocessors.
I think this video's take on why people voted the way they did is completely unfair to the intelligence of the community. People are sick of being dictated to. Sick of having features pulled away in favour of walled gardens. Take the whole bootloader situation for example. Sure RISC-V isn't going to fix that out of the box because it's still down to proprietary implementations per individual chip and whatever each company decides to do. At least with RISC-V, the barrier is lower for it to possibly happen. For a company to enter the space and make a chip that is more open, and have a shitshow of it even running because the platform at least supports it's generic level specification. RISC-V doesn't guarantee a better outcome for consumers at all. But it at least ALLOWS it. There is no way that's ever going to happen with ARM or x86. Let alone the barrier of license fees, the control of an overbearing body who can dictate terms under the gun of yanking licensing means that freedom is quashed from the start. Dismissing excitement for RISC-V as fanboyism really insults the intelligence of the consumer in general given the context (android authority) At least with RISC-V, it's POSSIBLE for an underdog to exist, and niche markets that target specific demographics with different needs to the breadth of consumers.
You seem to contradict yourself. You start by using bootloaders as an example and then say how RISC-V won't fix that. Of course it won't as bootloaders have little to do with the instruction set architecture. Anyway Arm has a bootloader specification called SystemReady which has worked really well in the Arm server market, now it is up to others to implement it as well. How does RISC-V allow for a better outcome for consumers that Arm and x86 don't. What do you think needs improving for consumers that RISC-V will solve? As for the license agreements and guns etc, I assume you are talking about Qualcomm. If you are then you have misunderstood the situation.
“Opensource” seems quite like what my experience of “peer review”. Sounds awesome but when the peers are a bunch of quacks it’s a sham. Have been on GitHub to find specific bits of code from big programs and they are often so badly documented or named that It’s be given up. If you’ve not got a job and hours spare to search it’s fine but how is the average user supposed to find the bit they have time to read… it’s not really open source when it’s just hidden using obscurity, just like it’s not really peer review if every peer is a quack.
@@GaryExplains what specifically would you say the point was? I live near a Listing library where everything published in the U.K. is sent. If no librarians worked there and all the publications were there but you had to spend hours searching for them it’d be near pointless. Isn’t one of the points of open source that people who aren’t developers can check for themselves there aren’t obvious security flaws? Edited to add: besides whoever coined the term chose to use the word “open” and so accepted the responsibility that it would mean what people commonly understood as “open”. If you went to an open day at a school but there was no one there to guide you so only people that worked at schools could navigate or if there was an open day at a golf club but only people with a handicap of 10 or better were given maps they’d both be considered rubbish open days. Similarly if you were trying to get out of a building and someone pointed out a door they said was open but was locked they couldn’t just say ‘never seen a definition of an open door that said anyone but locksmiths could use it’ it’d be a pretty ridiculous use of the word open. If they’d called it “condescending t**t source” there would have been more flexibility to come up with their own definition, but they didn’t. They chose “open”, which means it needs to be open, not closed to everyone but developers.
@@eberger02 "Isn’t one of the points of open source that people who aren’t developers can check for themselves there aren’t obvious security flaws?"" No. There is absolutely no way a person who isn't a developer could check even the simplest code for security threats. The point of open source is that the code is available, plus is free to use and modify. But you need to be a developer to do that.
What do you think that in the USB-C connector the VBus Pin is directly adjacent to the signal Pins? When you are charging you laptop with 20V can it damage the USB Controller if this VBus. Pin accidently comes into contact with other Pins?
We working on a AI primary phone based on RISC 5 Java Hardware Accelerator with new Java AI functions. Maybe a year 18 months out. Multi-billions investment. Compute in memory so Gary from another Gary, this AI CPU is beyond anything you’re discussing here. Already early adoptions of new network resource from another Gary!
I think you should mention two aspects: 1.) China and 2.) Qualcomm. China is looking at alternatives to proprietary western standards like ARM and X86. And RISC-V just fits that bill. Due to political support and pressure, chinese companies will invest heavily in actual RISC-V hardware. Actual RISC-V phone platforms will probably first come from Chinese manufacturers. Google supporting the ecosystem will make this a lot safer bet for them. They can sell the phones with Google services in the west and just use the open-source parts of Android with a Harmony OS-like skin and homegrown services in the domestic market. And even if there is another era of sanctions, China will thus not be cut off from interoperability with the wider ecosystem, as many mainstream apps (Instagram etc.) will certainly be compiled for RISC-V as well. The other story is ARM sueing Qualcomm: This will lead to many chipmakers looking at the possibility of doing the same thing they were doing with ARM with a different ISA (without having to fear legal hassles). This will at first not impact Smartphones and server chips as much, given the dominance of the established players there. But in other speciality applications (Smart TVs, automotive, industrial) where the user experience is much more locked-down anyway and you don't have to worry about installing 3D games, RISC-V might become more interesting if there is a readily available codebase for it like Android.
Absolutely, but I don't think anyone has claimed that Arm is Linux for hardware. Same for Intel, AMD, whoever. But people are claiming that about RISC-V.
Or in other words: Far from everybody can pull a M1 out of their hat, which is why they need to keep their fruit pie in the oven for a bit longer still if they want to wow the market.
I think we have reached the end of an era with the proprietary chip makers and ISA's. A heap have popped up and fallen in the past 50 years. Now we have an open ISA, we are no longer going to have vendor lock-in. Having to use a particular OS that is bound to a specific ISA (windows) is going to be a thing of the past. If there is going to be any proprietary ISA's, its probably going to be limited to the 'Accelerator' space. I can see that we will probably even see a huge shift in the GPU space in the next 5 years. Couple this with the fact that we are approaching the end of a 70-year run of programming languages that are based on ideas that first appeared in LISP, the next big thing will be 'memory-safety'. Things are looking really rosey :)
I think you should watch my series on RISC-V, there is plenty of proprietary stuff that results in vendor lock-in. I know that sounds counter intuitive, but it is true.I explain it fully in the videos.
A few things: Your initial tirade was not needed. No one expected or thought RISC was going to be free. And even if it was, the phone itself would still cost money anyway. That was never a concern for anyone. Neither was security. No one can verify whats in ANY chip. Let alone Risc. However you are WRONG about the fact its not free. It IS from a design perspective. OF COURSE it will cost money to make a chip You gotta pay a fab to built it. But to design and use it, there is no fee like Arm does. With Risc i can design my own Risc chip and make it. And the only cost to me is the cost of physically making it. NO licensing. However Risc for Android does two things. 1. It allows more options. More competition which is not only good for consumers but google as well. They dont want themselves locked into and reliant on a single source of hardware for phones. Which qualcom basically has done. Google wants more variety to ensure THEY have full control of Android designs and usage and its also good for us consumers to have competition. It either will drive costs down or spur more advancement instead of the incredmental bumps qualcom has been doing since they really have no true competition on the high end.. 2. Risc comes with some benefits. Its more modular. And because it has no licensing fees and is open source, you can fully customize it for ANYTHING cheaper and easier than ARM. So forget phones. Were talking SOCs that run machinery, IoT devices, etc... basically you arent a slave to ARMs demands. You can make a super custom super niche very specific chip for your companies needs WAY cheaper than Arm. And because its super customized for what ever niche specific thing it is going to be used for, it will most definitely have a performance gain over arm do to that.
Sorry dude, my comments come from actually having conversations with people who said those things. So yes, people do expect it to be free, and yes people are concerned about security and think RISC-V is the solution. Please don't fall into the trap of thinking that you know what everyone else is thinking. As for "With Risc i can design my own Risc chip and make it. And the only cost to me is the cost of physically making it. NO licensing. " You seem to think that the licensing fee is somehow relevant compared to the huge cost of design and manufacture! LOL, "the only cost"... just happens to be millions of dollars, but I don't have to pay a license 😂
@@GaryExplains hey Gary, let me ask you a question. I read somewhere. I know it's a rumor that Google is looking to acquire some chip company. Do you know anything about that to use to create their own tensor? What do you think about that company? If you know what it is I don't remember the name. Now I'm going to have to look it up. But I just read about it the other day literally less than a week ago
@@GaryExplains no surprises here, they are the biggest supporters of RISC-V since day 1. I do wonder what the future holds for X86 and ARM in the massive Chinese market dominated by homegrown RISC-V chips.
@Shadow Booster The biggest factor is Chinese manufacturing at leading edge process nodes. China doesn't have it yet, but for sure China is trying to be self sufficient in terms of design and manufacture of CPUs etc.
The world have to develop! Are we going to stick to Google for eternity?! There always must be an alternative. it doesn't matter if it takes 4-40 years but it must be done.
Maybe, but I think a decade is a little too far off. You could buy MIPS based tablets back in the day, so it only takes some Chinese OEM to start selling RISC-V phones on Aliexpress.
@@GaryExplains The GPU business is not enough. They will build and sell CPU's for phones, laptops, consoles, desktops even servers. They don't need to spend 40 billion for arm anymore.
Did you consider the possibility that several companies contributing to an open ISA or chips based on it would be tremendously beneficial and competitive? ARM hides things about their designs but RISC-V doesn't. There is good reason that Android phones have done so well, and it isn't that Android itself does all the work. Several companies contributing and using ARM processors make better and better products as long as it makes them money. RISC-V chips will give them another opportunity to make even cheaper devices, to make more money. It's yet another step away from Intel and now ARM. Companies will be able to design their own chips and chipmakers like Qualcomm will be able to design better chips because they'll have full control over the design. There is a lot to get excited about in theory, but like you say, it's not going to happen overnight. It will happen quickly, though, like in the next couple of years, and because it has so much in common with ARM, it will not take a lot to port programs over. No, I'm not suggesting this because I'm actually sure or experienced. It just makes too much sense for it not to be true. What I don't remember you mentioning is why RISC-V is suddenly a thing worth the attention of Google and OEMs. ARM can't be counted on to remain neutral. Just like Valve has invested heavily in gaming on Linux because Microsoft seems to want to become a software vendor, Google and Samsung will invest heavily in RISC-V because they don't want to find themselves at the mercy of ARM for their chips. RISC-V may not actually change everything, but it could. It could make things so much better all around that ARM, Intel, AMD, and Nvidia have to make everything RISC-V just to stay relevant. It could make things cheaper, or it could make phone makers so much more money that they have room to be competitive in more meaningful ways than they could with razor-thin profit margins. This is big. This is huge. If I had $1 million to bet, I might bet it on the success of RISC-V when people like you don't understand what it means yet. This is another big step toward greater control, competition, and profits. I want this to happen because it would be an improvement, a way forward technologically. The greater control for all companies means less control of companies over each other, which means greater competition, which everyone but the greediest benefit from.
I have a whole series of videos on RISC-V, that is why I don't delve too deep into as a thing in this video, here I am looking at RISC-V support in Android. I cover RISC-V as a wider topic in my other videos. However I think you are being a bit premature (and maybe arrogant) by telling me that I "don't understand what it means yet" etc. I can assure you that I do.
As a follow up, I am intrigued by this sentence, "ARM hides things about their designs but RISC-V doesn't." What do you mean by that? By "design" do you mean ISA or microarchitecture. If it is ISA, then no Arm doesn't hide anything about its ISA, that would be counter productive, how can people develop software for a instruction set that isn't public? If you mean microarchitecture, then that simply isn't true, RISC-V chips aren't necessarily open source and there are plenty of RISC-V chips where the design is private and "hidden". Please clarify. Thx.
It will be interesting to see if say a Google will move its custom SOC from ARM to RISC-V at some point. I think that would be a turning point for RISC-V wider adoption.
The problem is that if MediaTek and Samsung stay on Arm, the Google moving to RISC-V on its own would kill Google Android handset business. Any RISC-V chips it produces won't be competitive (for years) and so it will fade into insignificance while Samsung and MediaTek take over the Android market.
@@GaryExplains I think it's likely you'd see a chip alliance emerge where multiple companies design one chip for multiple uses. And there's nothing stopping the ARM companies from also designing their own RISC-V chips as well.
No, that will never happen.
@@GaryExplains "Never" is a strong word. And there are many such alliances that have been made in history.
@@GaryExplains Google controls the Android eco system via the play store. It wouldn't matter what any other SOC manufacture does. RISC-V is catching up, I expect China will be pushing RISC-V hard in the next few years, as a way to avoid US/UK sanctions restricting access. The JH7110 is impressive, that we could already be seeing RISC-V SBCs comparable to an RPI4(yes, no where near top of the line) in both price and performance. Phones might not be that many years away.
ARM suing Qualcomm is like AT&T suing BSD. It is going to give RISC-V an opportunity.
#RISC_V/Open-Source Hardware & Software is the FUTURE!!! by the end of 2023 we might be able to get a RISC-V smartphone to play with!
But you do understand that RISC-V isn't open source hardware, right?
@@GaryExplains Kinda, but design set is Open Source. That means every company can use it and make great stuff.
@@puffin_000 Yes, the specification is open. But designing a high performance, power efficient processor is hard, so it isn't something that "every company" will use. Only a few specialist companies are capable of using it. Companies like SiFive and Imagination are already licensing their RISC-V CPU designs, exactly in the same way that Arm does.
@@GaryExplains I don't think so. If some folks are unite, they can create efficent RISC-V procesor with under "licence free" open source architecture. Because factoring a CPU is hard and can't doable in home but designing an CPU architecture is doable in home if I know correctly. Also because of ARM's or other companies NDA's, people can't collabrate to create an open source ARM design. But with RISC-V's open specification, this open source architecture design is not a dream anymore its doable. Maybe nobody will do that in 5 years but its not impossible too. I think companies is not the only players in RISC-V.
Great points made & also ARM aren't standing still, so however well RISC-V develops - it's playing catch-up with a still evolving (and well resourced) target!
Competition is always better.
Nobody says ARM is standing still. What truly matters however is results. The time, manufactures will be able to sell RISC-V chips (and the term "chips" here is more generic) that have a better performance per cost and per watt, is the time ARM has lost. Everything is about money in the end of the day!
While i agree with all your points objectively. There is one real thing that is carrying RISC-V and giving it momentum, and its simply based on the fact that its an open source instruction set. As a result, a big majority of the new innovations in computing hardware and architecture is happening on RISC-V. I am working directly with the instruction set to develop hardware, and the amount of resources, tools are growing at a crazy fast pace. For sure the juggernauts will continue to innovate on ARM, x86 as well, I don't expect that to change, what i think changes is all the cool new innovations comming more from RiSCV based projects
Very useful to bring forward a sobered view on things. The invention of the wheel was nice. However, it was the applications of the wheel which made it worthwhile. Same things with RISC-V or other innovations. Thank you for this video.
My enthusiasm about RISC-V is partially about the open standard and mostly about it being another competitor in the CPU market. More competition is better. I like Windows on ARM for the same reason, plus the potential for better power efficiency compared to x86-64
Arm is also a risc
@@Arya_amsha Yep, I'm aware, but the ARM company has been acting a bit unpleasantly lately
My only concern about RISC-V is that I don't want it to slow down ARM overtaking x86 in the near future. I want to see more competition in the Laptop/Desktop space and x86 only enables Intel/AMD, ARM has far more competitors using their IS. Once x86 see's less than 50% market share in the laptop/desktop space, I am all for a load more instruction set competition going forward.
After the arm licensing changes, people have only gotten more interested in moving away from arm.
ARM shot itself in the foot by suing Qualcomm. No one trusts them anymore to be in a position of owning IP anymore. They are now coasting on the momentum from the time before they sued Qualcomm, but everyone wants to get away from them. It unlikely that they will overtake x86 now.
@@FreshSmog Who?, at the moment is the most used recent architecture for servers, amazon, google and nvidia have tons of servers based on arm (nvidia give service to car manufacturers for tons of data process and AI stuff).
@@FreshSmog That's an interesting claim. I hope you have a source to back this up, right?
Possible Google will make ARM to RISC V translation layer but it likely won’t make everything work well but could be huge step to make it possible for lots of ARM apps to work on RISC V.
Outside of China I would think low end devices would be where RISC V would start to get market share which the ARM equivalent wouldn’t be able to run these high demanding apps such as video games (which doesn’t use javaruntime).
If there is a beginning there is hope.
What are you hoping for exactly? Genuine question.
@@GaryExplains Risc V mainstream adoption and an alternative to x86 and arm architecture.
Ok, because?
@@GaryExplains Let's take 3 companies intel, arm and apple. Best x86 processor is made by 2 companies intel and amd , intel stuck to 14nm for around 7 years because there was no competition from amd until amd came up with ryzen cpus. Qualcomm and mediatek tweaks design of the arm processor to get the best performance for example adreno compared to mali gpus and last one is apple, apple shocked the industry by arm based cpus that have far better performance per watt than x86 and arm. Apple is able to achieve it because it designed its cpus on top of arm and also apple owns exclusive license for arm. With risk v companies will have the ability to modify and tweak the architecture compared to proprietary design, and as we all know chip fabrication is done by few companies like tsmc and samsung. windows on arm is still terrible until apple came and proved them wrong, more competition is better for consumers. I hope this answers the question.
Competition is good, yes, but notice how all the examples you gave happened without us needing RISC-V. Also Intel got stuck at 14nm because it couldn't make the tech work. Not because of lack of competition.
Reminds me of Linux in the mid 90s vs Microsoft. Not really competitive for most uses and users. Now it's different. As innovation rate declines it will be easier to catch up. That said with RISC V you still need a hardware design and layouts for a lithographic process. This is like saying each piece of software has its own compiler and assembler in the software world. Things might change as there are benefits to the industry to have standardised tooling and ways of doing things. Google are pushing this but only on 20 year old process nodes and not TSMC 3nm etc.
Intel’s “Horse Creek” RISC-V SoC - demoed at conferences since September and due out commercially in the summer - is the very first chip made on Intel’s new “Intel 4” process node.
When they said RISC-V was more efficient than other architectures, they talked about the reference implementation of the RISC-V core in that new high-level hardware description language. Those claims were made based on hardware benchmarks on the same process node. So yes, if you believe the benchmarks and data, it is, in fact, more efficient (but this is old information from several years ago, so these things could have changed possibly). That doesn't translate to real-world gains unless you use it as a microcontroller with the appropriate process node and implementation. For servers and desktops, as well as cellphones, there is a lot of work to do in terms of optimisation and tunning of the existing code (video codecs and other performance-sensitive code) for RISC-V.
I am a keen hobbyist and the main attraction of risc-v processors is price. I only need very basic functions and 10p a chip is very attractive for experimentation.
And there aren't basic Arm microcontrollers in that price range? What are you building that is so price sensitive?
@@GaryExplains being from Yorkshire, I am quite price sensitive... I'm using a 10p risc-v MCU 32 bit 48MHz ch32v003. I cannot find any other MCUs in the same price range with similar specs. At the level that I am coding on, it really doesn't matter the ISA of the chip. I'm cutting and pasting example code and tying it together with very clunky C. Currently building about 50 tiny controllers for addressable LED strips. :)
But the price of the microcontroller is insignificant compared to all the other bits you need to support the microcontroller and the led strips aren't cheap (if they are programmable).
Also I see that chips like the Puya PY32F002A (based on an Arm Cortex-M0+) is $0.08 in bulk. The idea that RISC-V is somehow cheaper is a bit of a myth.
@@GaryExplains point taken! Although the chip I'm using doesn't require any supporting components and takes 5V.
Just a clarification:
"
What is the license model?
- The RISC-V ISA is free and open with a permissive license for use by anyone in all types of implementations. Designers are free to develop proprietary or open source implementations for commercial or other exploitations as they see fit. RISC-V International encourages all implementations that are compliant to the specifications.
Note that the use of the RISC-V trademark requires a license which is granted to members of RISC-V International for use with compliant implementations. The RISC-V specification is based around a structure which allows flexibility with modular extensions and additional custom instructions/extensions. If an implementation was based on the RISC-V specification but includes modifications beyond this framework, then it cannot be referenced as RISC-V.
Does that mean free for industry to use and play with, but then we pay if we produce a product using this ISA?
-The RISC-V ISA is free for product use too. Those who want to use the RISC-V logo should join RISC-V International (see question No. 1).
"
(from the RISC-V website)
Yeah, I cover all that in other videos on RISC-V.
Thanks for explaining!
Interesting video. I don't know a great deal about hardware since I work with software but I've heard about RISC-V and it made to sound like it is somehow going to shake up things. Even then, I never bought into that because it's not like any person in a garage can make it, but takes a lot of capital to get a good chip design. It'll be something to see how RISC-V expands, especially with Chinese companies.
They can't manufacture it in garage, but they can design it thete. They can later outsource manufacturing to companies that aggregate many layouts and just make very few wafers (1-10) with the mask. I have read that it is reasonably cheap, e g 0.35 micron process for 650 Euros/mm2
Hi. I have read somewhere that Apple is interested also by this risc-5 architecture but maybe not for the main chip but for less demanding tasks.
Is it actually any more power efficient when compared to ARM and x86 or is that just a myth
That is just a myth. I have a video where I actually test a RISC-V Microcontroller against Arm microcontrollers for performance and efficiency. It doesn't turn out well for RV.
@@GaryExplains Alright, good to know, thanks
I think people are checking for alternatives once a chip architecture matures and usually licensing or IP usually pulls down most of them. Not sure if arm would cease and desist sd gen 2 w/ nuvia ip.
Is it full support like video accelaration? A hardware that can run everything Arm runs?
Did you watch the video because I thought I covered that.
What was the soc for the Palm OS (Garnet 1996)?
The Palm devices used Arm based processors.
Interesting, let’s see how far it may go.
RANDOM QUESTION! Is the phone in the thumbnail a Pixel 2 XL?
technically, chip companies won't have to pay royalty to ARM anymore which saves them money. but initial R&D is higher which brings in newer costs. so user wont see any effect on the front end if a risc v phone did come out. Its boring for users but very exciting for devs
But if you look at the VisionFive 2, the RISC-V SBC the CPUs are licensed from SiFive. So no change. The zero license fee thing only applies to companies that design their own CPUs from zero, and those are few.
I think it would have been great to discuss what RISc-V is and why the hype. You have done a great job throughout the video managing people’s expectations but have not really helped us understand what the big deal about it is.
The way I understand it and what has me interested and probably fall in the 45.68% of people saying Yes for sure, really is all about the geopolitics around the semiconductor industry.
You mentioned the eagerly awaited Oryon CPU from Qualcomm courtesy of Nuvia. The legal battle between ARM and Qualcomm has made RISC-V gain huge momentum. As a result Google have upgraded it’s status to Tier 1.
It is not that we think we will be getting open hardware, but rather the manufacturers we rely on to develop the processors now have an open standard. They are not limited by their license from ARM or Intel/AMD. We have seen the consequences of geopolitics on brands like Huawei when they were blocked from using ARM or TSMC.
Having an Open ISA can accelerate innovation and security.
As I said, I have lots of videos about RISC-V, what it is and what it isn't.
@@GaryExplains I can’t wait for them. Thanks
You don't need to wait, they are already live!!! Here is my RISC-V playlist: ua-cam.com/play/PLxLxbi4e2mYFTkLsNYqWLrSQZtLB94wnY.html
I don't believe RISC-V will bring much to the consumers. However the openness could bring the ability to custom design chip for specific application. The things that pop in my mind are in ear headphones and glasses and watches. They don't need to do everything and power and size is critical
Hi Gary. Big fan. Love your explainers. Lately, I have been struggling understand these terms like 3nm process or 5 nm process. I saw in wikipedia these "nm" values have nothing to do with any actual physical property. These are just marketing terms. Can you please clarify the clutter and actually provide details.
I am glad you like my videos! I cover the whole "nm" thing in this video: ua-cam.com/video/ZaqcwlZZST4/v-deo.html
@@GaryExplains Great video as expected :)
I can see Chinese companies adopting RISC-V in order to avoid future US technology sanctions in at least that area. For those companies already bought into ARM, then it's going to be a balancing act over whether continuing to pay ARM for licensing is cheaper than going down an entirely new development path and having to re-learn what they already know. Also, there's all the other parts of the ARM SoC ecospace that needs to be taken into account. I can't see Apple changing yet again, but then they have surprised me before.
Of course none of this is relevant to the area of bespoke processors where it looks like RISC-V has a big commercial advantage as it's much more flexible. That, of course, leads to it's own issues if RISC-V starts "forking" making use of that flexibility.
i think that risc-v has the potential to replace both arm and x86_64 architecures. if all the companies try to go that way, mobile, desktop, microcontroller and even servers can use risc-v in the future. compile once and run everywhere! in this case even technologies like webassembly and jvm can become obsolete and every compiler will directly produce risc-v code.
But why would Arm and Intel want that?
Also because of RISC-V extensions, the whole idea of compile once and run everywhere doesn't work.
Plus there is RV32 and RV64. Obviously a RV32 CPU can't run RV64 code, etc.
@@GaryExplains no, risc-v extensions are trapped and can be emulated by software. this is one design idea of risc-v.
@@GaryExplains arm would not want that but intel may want because intel wanted to change their architecture once (itanium series) but could not manage to do that.
Gary, maybe I’m mistaken cause it’s been a few decades but I remember the RISC vs CISC debates. I recall that RISC is a classification. Therefore a Qualcomm Snapdragon is a RISC chip by classification. So is RISC-V a brand name of a RISC classification chip or am I misinformed?
I have lots of videos covering this kind of question. RISC-V is a brand name of a RISC instruction set architecture. You may find my series on RISC-V interesting. First part here ua-cam.com/video/4qBKOAv0sBI/v-deo.html
If risc v can compete with arm Maybe after 1 decade atleast it will create competition and companies could buy it for lesser price and then sell the phones or processors at lesser price.
All depends on their success. Seems very unlikely for now. But u never know.
Why do you think the price will be lower? The most significant cost is manufacturing. Any license fee that isn't paid to Arm is used to design the processor, that is the point. When a chip maker uses an Arm Cortex-A CPU they pay for a ready made CPU design, no need for their own design team. Why would it be cheaper?
@@GaryExplains companies who will be buying those readymade designs will be getting options. So i think prices can get lower.
What do you mean by options? I reviewed the VisionFive 2 SBC on this channel. It has a RISC-V processor. The CPU part is LICENSED from SiFive... Just like a company would LICENSE a CPU core from Arm. How is that any different or any cheaper?
I can see risc-V being strong competition to ARM in the micro- controller market at the cortex M and cortex R levels. But I am not holding my breath in the Cortex A7xx and cortex X cores. Not saying never but......
Sifive's P670 is on par with Cortex A78. There is good progress going on.
@@godnyx117 from memory , the A78 is early 2020 since then we have had A78C, A710, A715 and A720 in the A line.
And X2, X3 And X4 in the high end Cortex X line.
@@patdbean I am aware of that. But still, for RISC-V, that's an improvement. P670 is also late 2022. They may release something new in H2 or H3 of 2024 if we are lucky...
@@godnyx117 absolutely , are risc V geting the latest manufacturing nodes 4nm etc or are they still on the 7+ ?
@@patdbean From what I've seen online, P670 and P470 seem to be 5nm.
RISC-V is inherently cheaper, but not more performant, so the top tier would be only competitive with the mid or low tier of the others unless it has the same investment.
How is it inherently cheaper?
There is an effort to bring RISC-V to everything from IoT to desktop to server. This is something that is just not available for ARM, there is no ARM desktop, and no Raspberry Pi does not count. SiFive/Intel horse creek is going to be interesting.
No Arm desktop... Wow, I guess all those iMac users will be sad to hear that.
@@GaryExplains I mean you can't get an ATX motherboard with ARM on it. Qualcomm bought Nuvia to do ARM on the desktop and got sued.
ATX is a design for PCs with its hereditary back to the 1980's. Qualcomm didn't get sued because it bought Nuvia to do ARM on the desktop, it got sued because it broke the license agreement. Same would happen if Qualcomm was using RISC-V technology from SiFive or Imagination, it is all licensed stuff.
And Qualcomm won't be doing a ATX motherboard with an Arm on it, regardless of it legal issues.
@@esra_erimez why would u use an old atx motherboard technology anyways. move towards the future where its more energy efficient and is more powerful!
Hot take: A completely open source SoC will eventually happen. We just need some organization to do the hard work and nail it. Who will it be?
There are already plenty of open source CPUs (from unviersities etc). However there will never be a competitive open source SoC because why would a company (with employees, rent, energy bills, etc) ever do that? This isn't like Linux where they can offer services to supplement the loss of income from direct sales. Secondly manufacturing the chips still costs millions.
@@GaryExplains A consortium of companies could heavily benefit from such an effort, because in order to build their final product (device, OS, whatever) they can target a common set of features safely. The chips themselves will no longer be the differentiators.
lowRISC's stated mission is open silicon, including a 64-bit SoC, so organizations with this goal already exist, but I guess your argument is that they will never be able to compete with the big dogs? Maybe not in the next 2-3 decades, but in the fullness of time, maybe.
It is hard enough to predict what will happen in tech over a 5 year span, predicting about the next 30 years is highly unadvisable. Think back to 1993, Windows 95 wasn't even out. Tell me anyone's predictions from 93 that came true by 2023!
@@GaryExplains I mean, they will still sell you the hardware, right? So I do not see them losing anything but putting out the specs and detailed info about their hardware. If anything, it will make people trust them more and give them a better image.
It's similar to open source software that are self-hosted, selling you a set-up server plan. Let me know if I miss something.
@@godnyx117 What they lose is that then anyone else can make a version of their chip and sell that to consumers rather than the company that invested millions in designing it in the first place.
But will apps follow suit? The main thing that fucked Windows RT was the lack of popular software that people use.
No-one wants to use Windows RT if a certain browser, media player, downloader software etc etc isn't avilable for it.
You can create a RISC-V softcore processor in a FPGA, so that is a way you can create a processor without spending a small fortune :) So in that way it's more like your $50 pc doing opensource...
Yeah, but that doesn't actually help. That is just a development tool. Are you expecting laptops and servers with FPGA cores?
@@GaryExplains, but it does give you a path to learn/create/contribute to opensource hardware. Pano Logic's zero client was a Spartan 6 fpga. But they did go belly up...
So you think an amateur is going to help create a world class RISC-V CPU by learning on an FPGA? I don't think you quite understand the complexities of building a modern superscalar processors. But even if a hobbyist contributes to a low-end, simple RISC-V implementation, then what? Who is going to make it? Why would they make it? And what would it be used for?
So your argument is that opensource hardware community members don't possess the needed skills to contribute? And that it is pointless to share works on github or OpenCores as no company uses opensource hardware designs instead of proprietary IP? Everyone has to start somewhere, and I believe it is amazing that for just a few bucks you can learn Verilog or VHDL and implement something that used to be entirely out of reach for the individual. So sure an amateur/hobbyist isn't going to build a world class super-scale processor design, but then neither is a amateur/hobbyist going to fix kernel defects. But we are at a time when hardware tools/ip are within the reach of ordinary people (amateur, hobbies, professionals, past professionals).
Let's say you contribute to an open source CPU. Nice. Who is going to spend the millions to fabricate that CPU? Why would they? What market would it be used in? If a company did fabricate it, then since it costs money to make a processor then unlike software you can NEVER benefit from your contribution unless you spend money to buy the product that is using your design. You give away your time and intellect to design the CPU and then you give your money to the company that used your design for free so you can use it. Nice.
I'm expecting an AV1 RISC-V mediaplayer
Wow Good Job
I would love to have an android Visionfive 2 version
Why?
Oh God the cope Gary.
He'd love to have it because new stuff is interesting, obviously!
I wonder what kind of raving fanboy crap you get in your inbox that made you develop such an incredible bias against RISC-V...
But to try answering Koysdo's original comment: I don't think there will be upstream support for the board, because Google said at the conference that they wanted to wait with an official platform release until a few more extension specs have been ratified and included into a profile.
That said, there might be an "unofficial" (as in not Google backed) port for the board and others using that SoC, since there are a few manufacturers making boards based on it and the SoC maker wants to build more of an ecosystem around it.
I just hope that all the software's will be built with hardware accelerator in mind.
What RISC-V need now is to take all the programing language libraries and just make any code written in those language can be easily converted to RISC-V with all the Hardware Acceleration built-in.
- if you're a programmer with the source code of your program, it would be nice to just select it to compile to RISC-V instead of ARM or x86.
- The libraries subs & functions for RISC-V will have to use the same names as the original in ARM & x86.
Frankly I would prefer a phone that was not android or Apple. Yes, a RISC-V SoC would be a very good feature, but it is the intrusive OS that worries me more.
What worries you? Android is Open Source, so you can compile it yourself and know what you are putting on your phone.
Some people aren't interested in the same performance as current anymore for smartphones. They're plenty fast already, and even if one is as fast as a mid range processor from 3 years ago you can do anything you need.
What it does bring is a more open hardware platform, where manufacturers and customers aren't subject to the whims of the overlords, like with the Arm/Qualcomm lawsuit. Maybe for licensing the RISC-V name some can get in trouble, but that's about it. Stop using the name and you're fine again.
Some of that may be true, it is debatable, but let's assume it is. But then, what about the overlords of GPUs, DSP, 5G Modems, Image Processing, audio codecs, video codecs, Bluetooth, Wifi, etc. Having an "open" CPU resolves nothing in any meaningful way.
Aren't image processing and audio codecs easily available? And 4G modems also?
Does the OS come to the architecture, or the architecture come to the OS 🤔...
Either way, it takes two to tango 🤣
"RISC architecture is gonna change everything."
😂
If the problems between ARM and Qualcomm don't get resolved, will Qualcomm take the Risc-V route?
The problem is this. MediaTek, Samsung, and Apple don't have any problems with Arm. So if Qualcomm goes the RISC-V route it will be doing it alone. I can't see Samsung suddenly switching to RISC-V for the Galaxy S series just because Qualcomm has a gripe with Arm. Any RISC-V chips it produces won't be competitive and so it will fade into insignificance while Samsung and MediaTek take over the Android market.
@@GaryExplains same goes with their A series galaxy series also
RISC-V will catch up with ARM eventually. There are too many people working on it, especially China. Also ARM suing one if it's biggest customers is not a good look.
@@GaryExplains Why won't they be competitive? If Nuvia and Qualcomm engineers can design a competitive Arm -ISA chip then they can just as easily design a competitive RISC-V chip, and indeed do it reusing 95% of the work they've already done.
True. I didn't mean to imply that it wasn't possible. I meant to say "first generations". I think I wrote that in another comment to someone else and didn't do it this time. If the first gen chip isn't on par or better than the competition then it is game over for such a switch.
The most basic yet important question is:
Will we be able to run Doom on a Risc V Android phone?
I am an computer engineering student, we just learned to code in risc-V should I put more emphasis on studying it, or focus on other languages
this may be more for embedded android then for a smartphone
Apple silicon is aproaching to a limit. Apple cant make a larger monolitic SOC
Apple is looking for RISC V experts.
Why have you implied a connection? RISC-V silicon implemented using Apple's design would be the same size (or bigger) than its current implementation. Also, why do you say Apple is approaching a limit? Apple's current designs are quite modular hence the Max and Ultra etc.
I always wonder why the Chinese (mediatek, Huawei etc...) doesn't pick up riscv and build upon it? I think in 5 years time they can have something viable... Well google up there, linux kernel is there i think i saw arm emulator that run on riscv. I mean in the long run they can save million upon millions on licensing for arm... Will see...
While Mediatek is Chinese, it's from Taiwan though.
Well you can see what's on a chip You can scan layer by layer put it in a spreadsheet and run it virtually.
😂
@@GaryExplains I don't know what's so funny! They can x-ray chips they can burn every nano layer and expose the printed circuitry
Oh, you are being serious. OK, please show me a "spreadsheet" with CPU that can run it virtually.
@@GaryExplains If you look at your RAM map address is the same principle. There is a video a guy on UA-cam who shows you how to do it. It's like any circuitry pcb you can reverse engineer If you got a picture of it
OK, so there is a guy on UA-cam. Great. As I said, please show me a "spreadsheet" with a CPU that can run it virtually.
I believe that we will see the first RISC-V mobile phones in the Chinese entry market as soon as they can actually make them. China needs some sort of chip independence from the West/Taiwan, and RISC-V is an opportunity for them. Making a competitor for the high-end market is a completely different thing.
The first market that will adopt RISC-V is China, for sure. But it doesn't solve the manufacturing dependence on Taiwan.
@@GaryExplains quite, as far as I know china firms can still licence ARM cores. Their problem is that if they are struggling to fabricate at 16nm while TSMC are at 5nm and moving to 4nm
It dose not really matter what design/ISA they are using. They are still a decade behind.
I wish I had the resources to make hardware. Would love to make stuff based on RISC-V.
I think RISCV on android could've a big boost if Google decided to make the next tensor chip using that instead of ARM, but odds are any support you see for that would be for that RV-tensor chip and not RISCV on android in general.
I get the sense that Gary is upset that RISC-V is popular.
RISC-V isn't popular (yet). I am upset when people hype things based on misinformation.
@@GaryExplains misinformation such as "big companies will be giving RISC-V chips away for free"?
Yeah, things like that. And don't laugh I have had conversations with people who thought that because they learnt from Linux that everything is free (as in cost).
@@GaryExplains Could you make a video on those conversations? It would be interesting to see what the thought process is. Like, I can download a RISC-V SoC for an FPGA as of today (e.g. based on LiteX), run it, and it costs me nothing - same as with the Linux kernel, to keep the analogy. Is that something that came up in those conversations?
I don't think such people really thought it through fully. Basically the words "open source" mean free (no cost) to many people. By associating RISC-V with open source the result has been confusion.
Actually, in regards to searching for malware on the microprocessor, I don't see why you can't audit the process of going from microprocessor specs to the physical hardware... after all, YOU have the hardware and the specs in your possession... and there have been so many ways these chips get reverse engineered in the last 25 years that it's gotten to the point where the hobbyist can play along... Duo Labs has a cool article where invasive analysis is used after acid etching to recover firmware off of a microcontroller. If that is within the reach of hobbyists, just think what guys in bunny suits can do in a billion dollar lab.
I'm sure some independent security researcher funded by the competition would love to nail a huge company for putting malware on their microprocessors.
It all depends on everyone's RISC appetite
I think this video's take on why people voted the way they did is completely unfair to the intelligence of the community.
People are sick of being dictated to. Sick of having features pulled away in favour of walled gardens.
Take the whole bootloader situation for example.
Sure RISC-V isn't going to fix that out of the box because it's still down to proprietary implementations per individual chip and whatever each company decides to do.
At least with RISC-V, the barrier is lower for it to possibly happen. For a company to enter the space and make a chip that is more open, and have a shitshow of it even running because the platform at least supports it's generic level specification.
RISC-V doesn't guarantee a better outcome for consumers at all. But it at least ALLOWS it.
There is no way that's ever going to happen with ARM or x86. Let alone the barrier of license fees, the control of an overbearing body who can dictate terms under the gun of yanking licensing means that freedom is quashed from the start.
Dismissing excitement for RISC-V as fanboyism really insults the intelligence of the consumer in general given the context (android authority)
At least with RISC-V, it's POSSIBLE for an underdog to exist, and niche markets that target specific demographics with different needs to the breadth of consumers.
You seem to contradict yourself. You start by using bootloaders as an example and then say how RISC-V won't fix that. Of course it won't as bootloaders have little to do with the instruction set architecture. Anyway Arm has a bootloader specification called SystemReady which has worked really well in the Arm server market, now it is up to others to implement it as well. How does RISC-V allow for a better outcome for consumers that Arm and x86 don't. What do you think needs improving for consumers that RISC-V will solve? As for the license agreements and guns etc, I assume you are talking about Qualcomm. If you are then you have misunderstood the situation.
“Opensource” seems quite like what my experience of “peer review”. Sounds awesome but when the peers are a bunch of quacks it’s a sham. Have been on GitHub to find specific bits of code from big programs and they are often so badly documented or named that It’s be given up. If you’ve not got a job and hours spare to search it’s fine but how is the average user supposed to find the bit they have time to read… it’s not really open source when it’s just hidden using obscurity, just like it’s not really peer review if every peer is a quack.
That generally is not the goal or purpose of open source.
I have read nowhere in any open source definitions or mandates or mission statements, that the goal is to provide code snippets to budding developers.
@@GaryExplains what specifically would you say the point was? I live near a Listing library where everything published in the U.K. is sent. If no librarians worked there and all the publications were there but you had to spend hours searching for them it’d be near pointless. Isn’t one of the points of open source that people who aren’t developers can check for themselves there aren’t obvious security flaws?
Edited to add: besides whoever coined the term chose to use the word “open” and so accepted the responsibility that it would mean what people commonly understood as “open”. If you went to an open day at a school but there was no one there to guide you so only people that worked at schools could navigate or if there was an open day at a golf club but only people with a handicap of 10 or better were given maps they’d both be considered rubbish open days. Similarly if you were trying to get out of a building and someone pointed out a door they said was open but was locked they couldn’t just say ‘never seen a definition of an open door that said anyone but locksmiths could use it’ it’d be a pretty ridiculous use of the word open. If they’d called it “condescending t**t source” there would have been more flexibility to come up with their own definition, but they didn’t. They chose “open”, which means it needs to be open, not closed to everyone but developers.
@@eberger02 "Isn’t one of the points of open source that people who aren’t developers can check for themselves there aren’t obvious security flaws?"" No. There is absolutely no way a person who isn't a developer could check even the simplest code for security threats. The point of open source is that the code is available, plus is free to use and modify. But you need to be a developer to do that.
Great
What do you think that in the USB-C connector the VBus Pin is directly adjacent to the signal Pins?
When you are charging you laptop with 20V can it damage the USB Controller if this VBus. Pin accidently comes into contact with other Pins?
We working on a AI primary phone based on RISC 5 Java Hardware Accelerator with new Java AI functions. Maybe a year 18 months out. Multi-billions investment. Compute in memory so Gary from another Gary, this AI CPU is beyond anything you’re discussing here. Already early adoptions of new network resource from another Gary!
I think you should mention two aspects: 1.) China and 2.) Qualcomm. China is looking at alternatives to proprietary western standards like ARM and X86. And RISC-V just fits that bill. Due to political support and pressure, chinese companies will invest heavily in actual RISC-V hardware. Actual RISC-V phone platforms will probably first come from Chinese manufacturers. Google supporting the ecosystem will make this a lot safer bet for them. They can sell the phones with Google services in the west and just use the open-source parts of Android with a Harmony OS-like skin and homegrown services in the domestic market. And even if there is another era of sanctions, China will thus not be cut off from interoperability with the wider ecosystem, as many mainstream apps (Instagram etc.) will certainly be compiled for RISC-V as well. The other story is ARM sueing Qualcomm: This will lead to many chipmakers looking at the possibility of doing the same thing they were doing with ARM with a different ISA (without having to fear legal hassles). This will at first not impact Smartphones and server chips as much, given the dominance of the established players there. But in other speciality applications (Smart TVs, automotive, industrial) where the user experience is much more locked-down anyway and you don't have to worry about installing 3D games, RISC-V might become more interesting if there is a readily available codebase for it like Android.
All the points you make about RISC-V not being "Linux for hardware" are true for ARM too, but much worse.
Absolutely, but I don't think anyone has claimed that Arm is Linux for hardware. Same for Intel, AMD, whoever. But people are claiming that about RISC-V.
Or in other words: Far from everybody can pull a M1 out of their hat, which is why they need to keep their fruit pie in the oven for a bit longer still if they want to wow the market.
Id love you to do a video on ChatGPT and your thoughts on the future of... everything! - with AI coming on in leaps and bounds!
Chatgpt is not interesting.
First comment. Great content Gary, keep it up.
sure my next phone will be r-v , being me a penny pinching
Why do you think it will be cheaper?
@@GaryExplains because chinaphones are going to use it en masses
Eh? Chinese phones are cheap now as well. What has RISC-V got to do with it?
Actually a RISC-V phone will probably be way more expensive (at least in the beginning), simply due to economies of scale 🥲
I think we have reached the end of an era with the proprietary chip makers and ISA's. A heap have popped up and fallen in the past 50 years. Now we have an open ISA, we are no longer going to have vendor lock-in. Having to use a particular OS that is bound to a specific ISA (windows) is going to be a thing of the past. If there is going to be any proprietary ISA's, its probably going to be limited to the 'Accelerator' space. I can see that we will probably even see a huge shift in the GPU space in the next 5 years. Couple this with the fact that we are approaching the end of a 70-year run of programming languages that are based on ideas that first appeared in LISP, the next big thing will be 'memory-safety'. Things are looking really rosey :)
I think you should watch my series on RISC-V, there is plenty of proprietary stuff that results in vendor lock-in. I know that sounds counter intuitive, but it is true.I explain it fully in the videos.
Hi
A few things:
Your initial tirade was not needed. No one expected or thought RISC was going to be free. And even if it was, the phone itself would still cost money anyway. That was never a concern for anyone. Neither was security. No one can verify whats in ANY chip. Let alone Risc.
However you are WRONG about the fact its not free. It IS from a design perspective. OF COURSE it will cost money to make a chip You gotta pay a fab to built it. But to design and use it, there is no fee like Arm does. With Risc i can design my own Risc chip and make it. And the only cost to me is the cost of physically making it. NO licensing.
However Risc for Android does two things.
1. It allows more options. More competition which is not only good for consumers but google as well. They dont want themselves locked into and reliant on a single source of hardware for phones. Which qualcom basically has done. Google wants more variety to ensure THEY have full control of Android designs and usage and its also good for us consumers to have competition. It either will drive costs down or spur more advancement instead of the incredmental bumps qualcom has been doing since they really have no true competition on the high end..
2. Risc comes with some benefits. Its more modular. And because it has no licensing fees and is open source, you can fully customize it for ANYTHING cheaper and easier than ARM. So forget phones. Were talking SOCs that run machinery, IoT devices, etc... basically you arent a slave to ARMs demands. You can make a super custom super niche very specific chip for your companies needs WAY cheaper than Arm. And because its super customized for what ever niche specific thing it is going to be used for, it will most definitely have a performance gain over arm do to that.
Sorry dude, my comments come from actually having conversations with people who said those things. So yes, people do expect it to be free, and yes people are concerned about security and think RISC-V is the solution. Please don't fall into the trap of thinking that you know what everyone else is thinking. As for "With Risc i can design my own Risc chip and make it. And the only cost to me is the cost of physically making it. NO licensing. " You seem to think that the licensing fee is somehow relevant compared to the huge cost of design and manufacture! LOL, "the only cost"... just happens to be millions of dollars, but I don't have to pay a license 😂
Android should get rid of Java, and use Julia instead for speed
but risc-v is cool
I think RISCV based chips will take over the entire Chinese market
The first market that will adopt RISC-V is China, for sure.
@@GaryExplains hey Gary, let me ask you a question. I read somewhere. I know it's a rumor that Google is looking to acquire some chip company. Do you know anything about that to use to create their own tensor? What do you think about that company? If you know what it is I don't remember the name. Now I'm going to have to look it up. But I just read about it the other day literally less than a week ago
@@GaryExplains no surprises here, they are the biggest supporters of RISC-V since day 1.
I do wonder what the future holds for X86 and ARM in the massive Chinese market dominated by homegrown RISC-V chips.
The name of the company would be handy.
@Shadow Booster The biggest factor is Chinese manufacturing at leading edge process nodes. China doesn't have it yet, but for sure China is trying to be self sufficient in terms of design and manufacture of CPUs etc.
More like Android coming to RV.
The world have to develop!
Are we going to stick to Google for eternity?!
There always must be an alternative.
it doesn't matter if it takes 4-40 years but it must be done.
An alternative to Android?
@@GaryExplains
An alternative to everything that's on offer, Google, ARM and every other gigantic ecosystem.
But those alternatives just turn into gigantic ecosystems, so how does that help?
@@GaryExplains
Yes it's a cycle, and there will be alternatives to those as well.
Hmmm, normally people try to break out of cycles, not go around them again and again. But each to his own I guess.
It will be another decade before we buy risc v phones
Maybe, but I think a decade is a little too far off. You could buy MIPS based tablets back in the day, so it only takes some Chinese OEM to start selling RISC-V phones on Aliexpress.
Musn’t be too certain.
Lets mark down the date and see if you are right. History will be the proof.
i have did research on riscv and riscv has horrible energy efficiency over arm.
That has way more to do with the microarchitecture and implementation than the ISA though
If Nvidia adopts riscv instead of arm it would become the dominant architecture in the near future.
What would NVIDIA use RISC-V for? Its GPUs? How many CPUs does NVIDIA sell per year?
That wouldn’t make much business sense for Nvidia as they acquired ARM recently
NVIDIA didn't acquire Arm.
@@GaryExplains The GPU business is not enough. They will build and sell CPU's for phones, laptops, consoles, desktops even servers. They don't need to spend 40 billion for arm anymore.
You think that NVIDIA can take on Intel and AMD in laptops, consoles, and desktops with RISC-V? I don't.
Did you consider the possibility that several companies contributing to an open ISA or chips based on it would be tremendously beneficial and competitive? ARM hides things about their designs but RISC-V doesn't. There is good reason that Android phones have done so well, and it isn't that Android itself does all the work. Several companies contributing and using ARM processors make better and better products as long as it makes them money. RISC-V chips will give them another opportunity to make even cheaper devices, to make more money. It's yet another step away from Intel and now ARM. Companies will be able to design their own chips and chipmakers like Qualcomm will be able to design better chips because they'll have full control over the design.
There is a lot to get excited about in theory, but like you say, it's not going to happen overnight. It will happen quickly, though, like in the next couple of years, and because it has so much in common with ARM, it will not take a lot to port programs over. No, I'm not suggesting this because I'm actually sure or experienced. It just makes too much sense for it not to be true.
What I don't remember you mentioning is why RISC-V is suddenly a thing worth the attention of Google and OEMs. ARM can't be counted on to remain neutral. Just like Valve has invested heavily in gaming on Linux because Microsoft seems to want to become a software vendor, Google and Samsung will invest heavily in RISC-V because they don't want to find themselves at the mercy of ARM for their chips.
RISC-V may not actually change everything, but it could. It could make things so much better all around that ARM, Intel, AMD, and Nvidia have to make everything RISC-V just to stay relevant. It could make things cheaper, or it could make phone makers so much more money that they have room to be competitive in more meaningful ways than they could with razor-thin profit margins.
This is big. This is huge. If I had $1 million to bet, I might bet it on the success of RISC-V when people like you don't understand what it means yet. This is another big step toward greater control, competition, and profits. I want this to happen because it would be an improvement, a way forward technologically. The greater control for all companies means less control of companies over each other, which means greater competition, which everyone but the greediest benefit from.
I have a whole series of videos on RISC-V, that is why I don't delve too deep into as a thing in this video, here I am looking at RISC-V support in Android. I cover RISC-V as a wider topic in my other videos. However I think you are being a bit premature (and maybe arrogant) by telling me that I "don't understand what it means yet" etc. I can assure you that I do.
As a follow up, I am intrigued by this sentence, "ARM hides things about their designs but RISC-V doesn't." What do you mean by that? By "design" do you mean ISA or microarchitecture. If it is ISA, then no Arm doesn't hide anything about its ISA, that would be counter productive, how can people develop software for a instruction set that isn't public? If you mean microarchitecture, then that simply isn't true, RISC-V chips aren't necessarily open source and there are plenty of RISC-V chips where the design is private and "hidden". Please clarify. Thx.