The EOMA68 Libre Computer Developer Wants To Tackle A Quad-Core RISC-V Libre SoC Design

[scottishduck]

RISC-V has no hope for consumer application without a GPU component.

[Weasel]

Speaking about yourself?

We all know

I'm not misguided like your definition of knowledge is.

[Weasel]

I think outsiders make tech choice a religion, not insiders. I don't see GNOME developers throwing insults at KDE developers or users, or Wayland developers throwing insults at X11 contributors. But J Random Person that runs GNOME and Wayland and is happy to insult everyone that does not.

How do you know J Random Person is not one of the devs? Obviously they don't want to reveal themselves cause they have a shred of wisdom there.

[wizard69]

Actually, I hadn't read up on the so called "standard" so I did. Maybe I got the wrong source but from what I read, "standard" looked like far fetched thing.
To me it looked more like the personal jots of a person with ideas. Hardly a standard by any measure. But maybe I got it all wrong?

Well if you are a designer you can create any standard you want.

My problem is that the card standard they are trying to create is a joke if you consider the targets they have for the card. First off they are using cell phone processors with limited I/O. At that point it makes no sense to pull all your I/O connections of the board the processor is mounted on. The industry has been putting those connections on a bards edge for a reason or two. One of those being reliability! Another being costs. Not to mention is that their connector will not be able to support near term technology improvements.

Now considering all all of that I’d love to see a motherboard standard that works equally well in laptops and desktop. That is completely doable with today’s tech, at least for low end desktops it is. Lots of potential here as many uses do not require ATX sized computers and more. Limit I/O to two edges and you can easily create a small Universal card.

[lkcl]

I would hope more people realized being free has a price :-)

Also royalties due to CPU licenses on lower end ARM based SoC is less than $1.

the target sale price of the SoC is around the $4 mark in volumes of 10 million. a $1 royalty per chip represents a whopping 20% increase. for comparison, the Allwinner A64 is $4 including the AXP803 Power Management IC, which on its own can be bought for around $1.

hope that gives some perspective.

[lkcl]

RISC-V has no hope for consumer application without a GPU component.

or a VPU. and the nice thing is: *as* components, no user will care what's under the hood. so as an opaque "blob" performing two important tasks, RISC-V would be the heart of some important tasks that also happened to end up in mass-production, and thus quietly gain acceptance *and* make some money. this was the argument i presented to sifive's "democratising ideas" proposal which is currently being evaluated.

so what i anticipate is that as a product, the SoC may actually end up as a co-processor to an ARM or even x86-based design. somebody may come along and say, "hey, y'know what, could this design be turned into a low-power PCIe Graphics Card, just like the OpenGraphics Project?" and the answer is "yes, if you're happy to put up the investment to do so".

[lkcl]

Next step is to open up design process and fabs. Let's see how this turns up as it needs billions of dollars of investment.

i can't reveal who it is: i did meet someone who had the funds from crypto-currency investment to actually be able to give this some serious consideration. the justification: people world-wide now have literally billions of dollars in crypto-currencies, in hardware wallets where they have absolutely no proof of the trustworthiness of the "FIPS-approved" so-called "secure" chips, and plenty of evidence that these so-called "secure" FIPS standards have been tampered with by the NSA *and* evidence that the NSA has orchestrated putting spying back-door co-processors into Intel Processors (aka the "Management Engine") for almost a decade.

what this guy told me is that various people in crypto-currencies have analysed the Elliptic Curve constants used in FIPS-approved pseudo-random number generation algorithms, and found them to be highly suspect. i think i recall him saying that bitcoin actually chose *different* constants precisely for this reason.

the next logical step in the chain of compromising security is to interfere with the Founry. substitute compromised masks, compromise the proprietary tools that *create* the masks. in speaking with the Shakti Team in India, they *already* have ruled out using foreign foundries for processors that are used in Defense, Nuclear and National Security-sensitive environments. that means that, at the moment, they can only make something in 180nm, and they're absolutely fine with that.

so yes, there are some big heavy-hitters who are genuinely investigating how to get a secure fab made... or bought. buying an existing fab is easier than creating a new one from scratch.

[lkcl]

Wait, what now?
This project, having trouble shipping a PCB made of available components with thousands of templates to duplicate solutions from wants to build a quad-core RISC-V with GPU and a fully featured SoC ecosystem?
Dude. Assembling a PCB with ready made SoCs from manufacturer templates is not even in the same division as building a full ASIC SoC, testing, qualifying, porting software etc.

This religious war target seems to be way off based on skill-set.

i've written another update to help answer this a bit more comprehensively, it'll be accessible here once approved https://www.crowdsupply.com/libre-risc-v/m-class

there's absolutely nothing "religious" about what i'm doing. you can find any number of efforts now, some of them even funded by Silicon Valley VCs, which tell you that people are only just beginning to wake up to the detrimental social consequences of pursuing pathological profit maximisation in a global connected context. the most poignant of these is the one where people in India were flash-mob murdered based on hysterical viral spreading of rumours and hearsay using Whatsapp.

if you *really* believe that it's okay to leave power concentrated in the hands of a few amoral companies, i am genuinely alarmed.

so that's the first aspect. the second is that there is absolutely no way i would consider spending my time or effort to begin a project that was unmarketable. it's why i picked a mass-market commodity design, *NOT* something straight out of the FSF playbook. this is a *commercial* product... oh that happens to be Libre, and it happens to be Libre for *commercial* reasons.

the third aspect is that you've clearly not been reading the updates from the EOMA68 project. here's a link - https://www.crowdsupply.com/eoma68/m...esktop/updates - where you will find that it's just not as clear-cut and simple and straightforward as you believe it to be. there are specialist parts involved that are extremely hard to get hold of, there are design aspects that had to be done over time, on a shoe-string level budget, and there also had to be alterations to the standard in order to make it successful, each of which cost an average of USD 8,000 a pop. a Corporation would have invested USD $100,000 into this project and had it done within months. however, a Corporation would also have made profit-maximising compromises that would jeapordise and undermine the overall goal of the project.

[lkcl]

With regards to ASIC hardware, it has almost always been a pipe dream.
Hardware in contrast to software actually requires some serious investments.
It is ridiculously tough to build a competing SoC with regards to power envelope, scalability, performance etc.
You need not only brilliant ASIC engineers but shitloads of qualification gear and proprietary software to even begin to think ASIC.

i just wrote another (more comprehensive, technically-orientated) update which you'll be able to see once it's approved by crowdsupply, which helps answer some of this. the summary: you'll see links to conversations on the internet and academic papers where some of the most experienced people in the world have been freely giving insights and extremely valuable advice

How about even the basic of stuff? Like a RTL compiler worth a damn not costing millions of dollars?

funnily enough that does actually exist: magic, yosys, alliance / coriolis2, and many more. the IIT Madras Shakti team is, right now, doing a side-by-side evaluation of magic and proprietary tools: one team will use libre tools to lay out a chip, whilst the other team will use proprietary ones. they plan to tape out both chips, both from the same HDL, and see what happens.

If you're going to have communication you need even more stuff for verification of your transceivers to send data on

an anonymous sponsor bought a ZC706 for me, which should be enough to simulate even a quad-core RISC-V processor. it also happens to have GPIO, SD/MMC, QUADSPI SPI Flash, USB ULPI, RGMII Gigabit Ethernet, and a TFP410 which converts RGB/TTL to HDMI.

as you no doubt know, it's an important milestone, to get to an FPGA-proven level.

RF-verification, RF-design, Power-planning, clock-trees, floor-planning.. Bla, bla bla. I could go on.

one of the things that we will need to do is ensure that enough money is raised in order to cover these things (and, to use creative innovation to bring down the costs). that's why the project is in "pre-launch" phase. it's very very early days. we're at least 12-18 months out from the phase you describe, above, as there is the design and implementation of the core to do.

what i'd like to see happen is for people to get in touch - investors, sponsors and engineers alike - to make this a success. i'm not interested in reasons why *not*, i am however interested to hear constructive advice from people with experience who know what needs to be done to get it done. and because it's an open project, that's already happening. i'm *already* interacting - publicly - with people across the entire world who are more than happy to share their experience and insights, so that we don't fall into the same pitfalls as the billion-dollar company they were working for.

it's a really liberating and humbling experience.

[milkylainen]

if you *really* believe that it's okay to leave power concentrated in the hands of a few amoral companies, i am genuinely alarmed.

Dude. Don't even go there. I evaluated, synthesized free CPU's back in 2003 already. The very first OpenRISC1200 implementation.
I have probably evaluated every reasonable ISA and available "free" CPU implementation.
Don't even begin to think that you are the first on the planet to have these ideas.
I'm not trying to put you down but you're annoying the crap out of me.
Far better people with far higher skills and better funding in Full Custom ASICs have already tried and failed, miserably.

one of the things that we will need to do is ensure that enough money is raised in order to cover these things (and, to use creative innovation to bring down the costs).

All I hear is "gimme money" with 0 HDL lines to show for. People have written bucketloads of HDL code without asking for a single dollar.
Go check the opencores community.

what i'd like to see happen is for people to get in touch - investors, sponsors and engineers alike - to make this a success. i'm not interested in reasons why *not*, i am however interested to hear constructive advice from people with experience who know what needs to be done to get it done.

We already have a community for that. It's called opencores. You're not bringing anything new to the table.
And good luck with not listening to reason. Really.
My experience tells me you havn't got the slightest clue as to what you're doing.
Either that or you're perfectly aware on how to blow sunshine up stupid peoples ass.