Funding a FreeBSD Risc-V board, would you partecipate?

[freezr]

Its not much different on Linux with the exception of Armbian.

The point is that most of those hardware manufacturers/assemblers at least provide official Linux, and often also Android, support even though most of the time Armbian is just better.

 

[freezr]

I think the Arm maker board market is very fractured.
Some want purely embedded boards like BeagleBoneGreen some want a desktop with HDMI.

They try and jam all these features into a 100x75mm footprint for the least amount of money.

This is why I said that it would be cool a board (but risc-v) designed by FreeBSD to FreeBSD...

 

[SirDice]

Some want purely embedded boards like BeagleBoneGreen some want a desktop with HDMI.

Lots of different SBCs and the other end of spectrum, really expensive Ampere boards. But there's very little in between.

 

[Phishfry]

Yes very expensive ThunderX boards too.

How do you make a 100x75mm board that pleases everybody.

Some want 4 LANS, Some want 4 SATA and some want 100 GPIO pins.

Ask GonzoPancho at Netgate how well Arm has worked out for them....

 

[jbo@]

Ask GonzoPancho at Netgate how well Arm has worked out for them....

I'd be interested in knowing more on this - would you happen to have a link or similar you can share?

 

[Phishfry]

I use the word 'Arm board' liberally here. I literally mean any arch but amd64.

From my take the HiFive boards look just like Arm boards. u-boot is in ports for FU540 and FU740.

 

[Phishfry]

I think you could call their first two ARM attempts total flops.

Netgate SG-1000 (FreeBSD based pfSense on ARM) First Look

Sipping power and at a price under$150, the Netgate SG-1000 running FreeBSD based pfSense on ARM looks like a winner in our initial impression

www.servethehome.com

Not enough ass for a firewall. Excellent hardware for embedded.

The EspressoBin was a POS with high failure rates..

Netgate SG-1100 Launched Higher-Speed Arm pfSense Firewall

The Netgate SG-1100 is being launched as a low-cost, Arm firewall with 3-5x more packet filtering performance for pfSense than previous generations

www.servethehome.com

 

[Deleted member 43773]

most of those hardware manufacturers/assemblers at least provide official Linux

Of course.
Guess why.

I'm with you as you already pointed out, the ones from the FreeBSD community want such a board would support it.
But as jbo pointed out, being realistic doesn't mean bringing things down,
same as pure enthusiastic hyped hopes alone will not result in a success either.

One needs to be reasonable.
And the crucial core point here is simply numbers.

How large do you think the Linux community is tinkering with such boards?
5M? 10M? Even more? (I don't know - I'm just blindely guessing.)
And how large you estimate the numbers of FreeBSD users would tinker with such?
5k?... I don't know. Could be completely wrong.
But this is something you might get knowing pretty sure before you spent any real effort in it.
And unless you're really sure you will sell at least a 1/4M (need 1/2M buying undertakes at least) you don't need to start a real project on this.

What you may could do is to develop a board yourself - just for your own purposes, at first.
It'll cost you maybe 300...500 $ - and many hours of work.
But at least then you've already done development, tested and proved it.
And from that you'll have a complete other start.
Don't need no crowfunding anymore - keep the profit for yourself.
Even may sale small lot charges.

I checked Mouser, they don't have a single RISC-V CPU or MCU in their assortment

Maybe not on stock.
We still suffer from "chips-shortage".
Last time I checked (couple of weeks ago) also Raspberries still are not really full available again yet.
Those have no high priority.
x64 CPUs first, then GPU,...also automobile market had a large lack of electronics...

 

[SirDice]

Maybe not on stock.

Plenty of things that aren't in stock. There's just nothing, not even out-of-stock components. Mouser website shows everything, even the stuff they don't sell anymore. And you can keep track of how many they have in backorder and when their new stock is expected too.

We still suffer from "chips-shortage".

Yes, I'm pretty sure these new fangled RISC-V CPUs are somewhere at the bottom of the stack. It's probably also the reason why you can't even buy any of the few RISC-V development boards that do already exist.

 

[Deleted member 43773]

Mouser website shows everything, even the stuff they don't sell anymore.

Yes, that's normal.... Interesting - I wasn't on any large distributer's page (DigiKey would be another one) for a while.
I have no explanation - maybe they just removed them from their catalogue, because it's embarassing having empty stocks while rising enquiries?
...I don't know, cause at normal terms they have any IC at least within their catalogue....

 

[SirDice]

riscv - FreeBSD Wiki

Both the HiFive Unleashed and HiFive Unmatched are sold out everywhere. And SiFive isn't planning on making a new batch of Unmatched boards. They're busy developing new boards and CPUs but no word when one could expect to see any.

HiFive Unmatched - SiFive Boards

Based on the success of the HiFive Unleashed and Linux software ecosystem enablement, the HiFive Unmatched ushers in a new era of RISC-V Linux development with a high performance development platform in a standard PC form factor.

www.sifive.com

 

[covacat]

lol it sold for $700 and has 1/2 cpu performance of a $80 pi4/400
"Powered by the SiFive Freedom U740, a high-performance multi-core, 64-bit dual-issue, superscalar RISC-V processor."
that means it will beat your washing machine in selected benchmarks

 

[msplsh]

Yeesh, didn't know the perf per dollar was that bad. Ok, done with this pie in the sky thread.

 

[ralphbsz]

This is why I said that it would be cool a board (but risc-v) designed by FreeBSD to FreeBSD...

Can you explain how a board (or CPU chip) designed to run FreeBSD would be different from a board/chip designed to run Linux, at the same workload? When I say "same workload", I mean compare laptop to laptop, embedded to embedded, desktop to desktop, server to server.

There is considerable effort in various parts of the computer industry to get RISC-V servers to work, trying to compete with Arm, Intel/AMD, and Power. I know of efforts to build large servers using these chips. I think in the area available to amateurs, there is very little other than one line of chips made by a subsidiary of Alibaba.

And that word "Alibaba" is key to this discussion. Right now, RISC-V is mostly a question of geopolitics, not one of computer architecture. With the various embargos, China is looking at a future where they have no supply of processors. One possible end-run would be to design and fab their own chips. But CPU design is notoriously hard, and requires decades of experience and thousands of person-years of design time for high end chips. So why not start with an open source (license-free) instruction set? It was tried with Arm, but that maneuver seems to not have worked (look at who owns Arm now, it is Softbank). Building CPUs using RISC-V might be another attempt, but the issue here is: while the instruction set is open, there are no open implementations of interesting performance.

 

[Phishfry]

I wanted to say not all Arm firewalls are bad.
The newer Netgate offering seems to be a success.

Netgate 2100 BASE pfSense+ Security Gateway

The Netgate 2100 security gateway appliance with pfSense Plus software delivers unbeatable performance and flexibility in its class. It is ideal for home, remote worker, and small business deployments

shop.netgate.com

Turris Omnia is pretty nice hardware too. I want one.

FreeBSD Foundation strives in managing frameworks for the software cause.
Manufacturing hardware is nowhere in that sphere.

EspressoBin is a perfect example of a terrible Arm experience.
Crowdfunded too.

Track Marvell ESPRESSOBin Board's Kickstarter campaign on BackerTracker

by Globalscale Technologies Inc ESPRESSOBin is a high performance 64 bit dual core low power consuming networking computing platform based on the ARMv8 architecture

www.backerkit.com

I don't see RISC-V being some miracle platform. ralphbsz made some excellent points there.

 

[freezr]

Can you explain how a board (or CPU chip) designed to run FreeBSD would be different from a board/chip designed to run Linux, at the same workload? [...]

Basically because Linux and FreeBSD do the same workload differently you may consider a design that fits better the latter rather than the former. For instance ZFS on Linux doesn't shine brighter as on FreeBSD and you may design a board with may allocate four NVMEs...

By the way for next time I'll refrain to write such topic... ?

 

[SirDice]

For instance ZFS on Linux doesn't shine brighter as on FreeBSD

This has nothing to do with the design of the hardware.

 

[Deleted member 43773]

By the way for next time I'll refrain to write such topic..

Nah, I wouldn't say that.
Just because it took another turn as you expected doesn't mean the idea was bad.

 

[Beastie7]

Let's assume the hardware already exists.. who's going to write drivers for these boards?

A mere pipedream.

 

[shkhln]

lol it sold for $700 and has 1/2 cpu performance of a $80 pi4/400

Afaik, part of Raspberry Pi's magic lies in their relationship with Broadcom. That's not something that other people can easily reproduce.

Eventually Risc-V will take over because open-source is more prolific than closed source development...

The instruction set is almost irrelevant in the modern world. The things that actually matter: performance, price, availability. All these three points favor large corporations due to astronomical development and manufacturing costs. No one else can compete.

 

[tingo]

One difference between ARM IP (ISA and building blocks) and RISC-V is the license cost. For that reason alone, RISC-V will win out. Eventually.

 

[kpedersen]

The way I see it is the general pattern for crowdfunding is, *if* it is successful, the pledgers get hold of the product as part of a first limited batch (yay!) But...

then for a second batch, there is rarely enough interest to fund it; half of us nerds are happy with one toy to play with and those looking for an evaluation board in preparation for large orders are not in the crowd funding market (too unreliable). So effectively the second batch never comes.

So now since few people have the device, there is no real community built. No real development continues. And it ends up as yet another bit of hardware in the bottom of our drawers.

Raspberry Pi's success is rare (and as shkhln mentioned, ties with Broadcom helped). However even that is starting to dwindle; partly because of parts shortage but also they are becoming too high spec and now having to compete with second hand amd64 laptops in terms of price and performance.

 

[Beastie7]

Forget about crowdsourcing any hardware endeavors, Linux is still the go-to for open source driver development for most hardware. Until this changes, ideas like this will most likely never come to fruition. The fact that we have to provide a system call table and KPI for user space applications and hardware drivers says a lot here. The Foundation could do more work with seeking collaboration with various IHVs, and maintaining it.

How could they fix this? Simply ship a desktop version of -RELEASE. I know, it's idealistically stupid, but Canonical/Red Hat was successful with it.

 

[kpedersen]

Simply ship a desktop version of -RELEASE. I know, it's idealistically stupid, but Canonical/Red Hat was successful with it.

-RELEASE is already a desktop version. I am just guessing that our ideas of a desktop differ.

 

[Beastie7]

All nerd bias aside. I think we all know what a conventional desktop experience is. Developers like it, mere mortals like it. Driver development is easier with one.