ARM: “RISC-V Architecture: Understand the Facts”
riscv-basics.com
riscv-basics.com
So somebody is paying google to make this thing come up on top.
Even if there is some validity here, this is in surprisingly poor taste. I've known people who have worked for ARM. I wonder what they think about this.
"Did you RTFM?"
"Debain CPU doesn't have that issue."
"Just disable the ALUs in firmware, they're not necessary in modern processors anyways."
"That's been patched, just download processor v2.3.4.432 and send it to your FAB, they should spin you a new CPU in 8-12 months."
If everyone is running the same CPU design but sourced from different manufacturers who had them spun up at different fabs, how would that affect support? Today we have one or two device manufacturers we rely on for support and answers about if security fixes can be done in firmware vs requiring a hardware fix. We can pull the CPU ID of our system and then check with AMD or Intel for an authoritative answer. What does that look like when anyone can manufacture a chip?
Will there still end up being one or two predominant CPU manufactures for desktop/mobile/server CPUs to simplify support? Are we going to have to rely on OEMs to support the CPUs they put in their machines?
I'm fearful we might end up in a situation similar or worse to what we have with Android where smartphones are shipped in their final state and never receive anything beyond failure support from the OEM/ODM.
* Serious question. Apart from the freedom and control aspects of the Open Source model, what is solved with an open source CPU architecture or what are the benefits?
I can see being able to spin your own chips would enable an OEM/ODM to keep producing the same old design indefinitely where as reliance on Intel/AMD/Samsung/Mediatek means you have to refresh your design. That could be both a benefit and a detriment depending on how you look at it.
The freedoms from the Open Source model is substantial enough to make it attractive on that basis alone, just as for Linux/BSD. One example is Western Digital transitioning to RISC-V for their hard drive controllers. I would imagine being able to add your own instructions without having to deal with NDAs and non competes with ARM will be attractive to a lot of OEMs.
Just like with Linux, it's difficult to anticipate what intersting things will come out of RISC-V, but one interesting example is a formally verified RISC-V core, picorv32: https://github.com/cliffordwolf/riscv-formal/tree/master/cor... (presentation here: https://www.youtube.com/watch?v=VU97ffHF_IQ). According to the author, this core will power safety critical applications at a synchrotron: https://theamphour.com/374-an-interview-with-clifford-wolf/
Which means that anyone can implement the instruction set and thus many different implementations can be made. Most likely there will not be one common and shared processor.
Notwithstanding your terrible MythTV experience, there's seveal companies offering commercial support for Linux today, and a lot of effort has been put into making it as secure as the Windows equivalent. I suspect the same could happen with RISC-V.
Even when we do get something like a custom ROM for a router or smartphone it will still be reliant on custom kernel modules or other lock-ins hindering open source support.
It just feels like more of the same with a chance for ODMs to save some on the BOM while not passing the savings on to us.
I guess it's just hard to get excited about it without seeing some sort of benefit to the end user. That doesn't mean I feel like we should stick with ARM or x86, I frankly don't care.
>No announced RISC-V silicon is susceptible, and the popular open-source RISC-V Rocket processor is unaffected as it does not perform memory accesses speculatively.
https://www.techpowerup.com/240310/risc-v-foundation-issues-...
Do you have a source?
Does someone actually make a RISC-V processor with out of order execution? Or is this a "They're going to make one eventually, so they'll have to address Spectre attacks before that happens" problem?
Also pretty sure BOOMv2 has been taped out: https://content.riscv.org/wp-content/uploads/2017/12/Wed0936...
Instead, of say, ARM M4+, A7, A53/A55, etc.
I'd be happy to try out a RISC computer, actually it's quite sad that the CPU diversity got so little due to Intel/AMD.
This actually looks like a smear campaign against ARM.
RISC-V may, perhaps, compete in two fields with ARM: education, where ARM doesn't give a fuck, and hacker boards, which is a tiny market compared to smartphones, smarttvs and so on. There have been around 15 million Raspberry Pi (the most popular SBC) sold in 6 years. Samsung sold 321 million smartphones last year alone.
Whoever is thinking today about replacing some production anything with a RISC-V is insane to begin with, but this campaign ups the insanity further.
https://www.wdc.com/about-wd/newsroom/press-room/2017-11-28-...
> Western Digital’s leadership role in the RISC-V initiative is significant in that it aims to accelerate the advancement of the technology and the surrounding ecosystem by transitioning its own consumption of processors – over one billion cores per year – to RISC-V.
Western Digital is moving to RISC-V. I'd say ARM has begun to notice the shift and needs to begin damage control.
Those seem like very similar markets to those that Linux started out with.
AFAIK, Sun never ran FUD/attack ads against Linux, but this feels similar to a hypothetical world where they did so very early in 1993.
I doubt FUD will ultimately be able to prevent the adoption of RISC-V, but it may be able to delay the inevitable for a few years.
And SCO was much more vulnerable to Linux than Sun was. Linux could even run SCO binaries unmodified.
And we know what happened to them later.
And I say this as someone who used to work for a SCO distributor, and later Sun.
You're saying they're all wrong?
I doubt any of those companies (aside from SiFive) will be actually using RiscV within the next few years.
I imagine RISC-V is attractive to them.
A lot of these SoCs become the core of smartphones. Did I mention Google (of Android fame) is also a member?
Just because something is sold often does not magically make it better or useful for specific peruses. How does it help Nvidia that ARM sells millions of cheap when they design an internal chip?
The are lots of places RISC-V can be used productively, and companies like NVidia, Google and WD are already doing so.
Propaganda ("FUD"). SEO.
You'd be surprised...
ARM would be much wiser to embrace RISC-V early, spearheading this effort and bringing RISC-V users into their tools and ecosystem.
Everyone freaking out about how "disruptive" an open source core would be... You still need someone to place it on silicon (unless you want to waste most of your FPGA on a soft core) and you need someone to make peripherals which lets be honest is 80%-95% of what Cortex M-type embedded processors are doing at any given time. These seem like the major barriers that aren't solved by removing a licensing fee for the core.
Maybe I just don't get it, but it seems a LONG way out for RISC-V powered top tier phones.
In the past, people just threw a little bit of money to ARM and called it a day.
But now, I could grab picorv32 or rocket for free off the internet. Both have been taped out in this capacity before, both speak standard bus interfaces, and the former has been fully(?) formally verified.
Sure, it's not a resounding success story, but it helps me get my SoC to market sooner and at a lower cost (no need to even spend time negotiating with anybody).
What do you mean by "fully formally verified"? My understanding is that picorv32 has been verified by riscv-formal, and riscv-formal performs bounded model checking, which is not full formal verification unless you otherwise establish an upper bound on trace lengths that must be considered to find a bug.
Ok, but you're then dumping what at least a million dollars into someone else's firmware that is unsupported if there is an issue? IDK. I can think of lots of cases where it might make sense for the hobbiest, or for softcores or for weird applications - but...
There are engineers from STM with desks at ARM, and engineers from ARM with desks at STM. I just don't see STM saying "fuck it, let's just spool up some core we contributed to as a fully in house project". They already need to support their peripherals and production, maybe they are paying enough to ARM to make it worth it to add core code to that list, maybe not.
ARM makes a ton of money, but they also assume all of the contractual obligations for issues and support. I'm not sure the entire industry would be using ARM if it made sense to spool up your own, even as open source. But again, IDK.
Huh?
As an exercise in FPGA stuff, I decided to use PicoRV32 on a small MAX10 8k FPGA. The core itself consumes easily under a quarter of the logic blocks.
But there are also loads of companies using and producing real RISC-V cores and all the major Linux distros are already shipping on RISC-V. The barrier to entry to RISC-V is far lower since you can quite literally download the cores off github and program them into just about any FPGA (there's even a version which works with an 8k-LUT FPGA that costs $10 in quantity). Or just fire up a qemu RISC-V VM at no cost at all.
As with Linux & Windows servers c.1999, who would you bet on winning in the end?
It was more about UNIX & Windows.
Linux won on the servers thanks to majority of UNIX guys cutting down costs by outsourcing their development costs into Linux.
Had Linux not gotten the money of IBM, HP, Compaq, SGI, Intel, Oracle, Sun, Google, ...., the outcome would have looked much different.
> Had Linux not gotten the money of IBM, HP, Compaq, SGI, Intel, Oracle, Sun, Google, ...., the outcome would have looked much different.
But these companies didn't spend that money because of a fluffy corporate love of the underdog. They wanted to "commoditize their complements". And the exact same desire will apply to the CPU, which is why software companies will want to do the same thing in future once there's a viable open source CPU (whether that is RISC-V or something entirely different).
Linux would win anyway. I've worked in the late 90s-early2k in places which were then already planning the move from proprietary Unix to Linux, some of them also in pretty critical contexts. Most of the companies you mention already had their own Unix: AIX, Digital Unix / Tru64, Irix, Solaris; they shoveled all that money into Linux (apparently against their own business) because they realized that selling products and services would pay in the long run a lot more than selling also the operating system, especially when the OSS competitor was gaining support every day. Soon or later that abundance in developers and testers would turn into less bugs, more features and faster response to vulnerabilities that no corporation in the world could equal.
Apparently people can't read.
For example, when they write that licensing the ISA is only a small fraction of the total cost of designing a processor, that is correct. But they are simply ignoring the broader implications of a non-free ISA and its financial effects, and the fact that this fact in itself may actually depend heavily on their current business model. Think about how the free availability of Linux has changed how people can build software, and how that has brought down the cost of software considerably, and much more so than the licensing costs of a Windows install per server. Also, their argument is sort-of "you don't need competition because the price isn't that high", even though it obviously would be a lot higher if there were no competition.
Similarly, it's not wrong that verification and validation of CPU designs is expensive. But (a) noone suggests that those people who are buying ARM chips now will do their own CPU designs in the future, and (b) ARM chips obviously also require expensive verification and validation, so you end up paying for that regardless where you buy a chip. So, again, simply intentionally missing the point.
Also, a lot of it is obviously just speculation, such as "well, what if the new processor is less secure than our product?" Well, yeah, big if, tons of unjustified assumptions: Are their processors really that secure? Couldn't a new processor designed from scratch be much more secure than their accumulated cruft? Again, not factually incorrect, but highly misleading by implying unjustified assertions. That's what the "(U)ncertainty" in FUD refers to.
The main points seem to be that RISC-V is not there yet and that customization is expensive. I guess for most projects you currently use ARM processors ARM will make most sense for a while. RISC-V will probably slowly eat a growing piece of that market though over the next years to decades.
Also point 4 is pretty useless, and point 5 is basically nonsense. It's implying that you should avoid the higher costs of customizing by using Arm instead of just... not customizing.
The article is not fake
This so far away from the early 00's Microsoft anti-Linux campagin that you can't even really call it FUD
That's a joke, but I think Windows vs. Linux is the right reference frame here, assuming that RISC-V proves out. Those who want to roll their own and control their own destiny may go with RISC-V, while those who want a packaged solution will go ARM. And ARM will have to get better to compete.
Western Digital has gone with RISC-V for a controller, because Western Digital is the kind of company that wants to commoditize its inputs so it can sell a packaged solution to its customers as cheaply as possible. I can also see Amazon or Google finding uses for RISC-V in their data centers.
Phones ... I'm not so sure. Maybe Apple could go that way, as they're vertically integrated and already make their Ax chips in house. But the Android OEMs wouldn't unless Google leads the way.
Still not practical for the masses, but not more expensive than similar extravagant architectures like PowerPCs (IIRC a PowerPC9 desktop system costs about 9000$ today).
Someone joked that you could get a 30$ xeon system on ebay from a couple of years ago that would crush these on pretty much any benchmark, but who knows, maybe some day riscv chips will become cheaper and more powerful. What we are seeing in the market today are the very first such systems, from the very first vendors.
Domain Name: RISCV-BASICS.COM
Registry Domain ID: 2277459939_DOMAIN_COM-VRSN
Registrar WHOIS Server: whois.wildwestdomains.com
Registrar URL: http://www.wildwestdomains.com
Updated Date: 2018-06-20T21:54:44Z
Creation Date: 2018-06-20T21:54:44Z
Registry Expiry Date: 2019-06-20T21:54:44Z
Registrar: Wild West Domains, LLC
...
Registrant Name: Registration Private
Registrant Organization: Domains By Proxy, LLC
Registrant Street: DomainsByProxy.com
Name Server: NS1-07.AZURE-DNS.COM
Name Server: NS2-07.AZURE-DNS.NET
Name Server: NS3-07.AZURE-DNS.ORG
Name Server: NS4-07.AZURE-DNS.INFO
At first sight, I'd assume it's fake.
Link to arm.com though has GA tags for tracking:
?utm_source=riscv-basics.com&utm_medium=website&utm_campaign=18.06_riscv-ds-eval-confidence_riscv_na_na_web_pg2956a1_02
<meta name="server" content="ARMGSCD2" />
<meta name="server" content="ARMBPCM1">
But maybe it's real.
EDIT: It's real:
https://www.arm.com/-/media/global/company/arm-risc-v-infogr...
https://www.theregister.co.uk/2018/07/10/arm_riscv_website/
They sure did the RISC-V community a great service over the last 48 hours ....
Does ARM have this? Does RISC-V have this?
If Management is good, then someone's Remote Management of your computer must be even gooder!
But scary for whom? For ARM maybe? Is it just me, or is this pure FUD?
I don't think there are that many people yet considering risc V seriously for current commercial implementations, rather than keeping an eye on it for future projects, but I'm sure there will be once momentum builds.