Hacker News new | past | comments | ask | show | jobs | submit login
HiFive – RISC-V-based Linux development board (sifive.com)
240 points by hucker on Feb 3, 2018 | hide | past | web | favorite | 133 comments

For those complaining about price, you might want to look into costs that go into ASIC's. Adapteva has a nice write-up on that with their strategies for getting cost down. Note that it was critical to have a few million with a strategy that their own write-up said couldn't work if priced for consumer market. They'd have to charge more or have ridiculous volume.


The HiFive product uses a CPU they developed on 28nm. That's one of most expensive nodes you can use. Depending on if and how they split cost, fabbing those chips could've cost over a million dollars in masks. Low-volume, RISC workstations used to cost five digits each on older nodes. Getting a low-volume CPU on cutting-edge node for a grand is a great improvement. So long as they ship products that work as claimed.

There's been a lot of vaporware in FOSS-type hardware...

I was under the impression that (albeit as of a year ago or so) 28nm was the cheapest node you could work on. The end of Moore's law is starting to hit the industry and smaller nodes were more expensive per transistor still.

My prior data said most designs were being done well above 28nm (even 90nm) with mask costs increasing as the feature size got smaller. eASIC even still offers 45nm S-ASIC's (Nextreme-2's) for those who can't afford ASIC's at that node. So, I'd be surprised if 28nm cheaper. If 28nm has gotten that cheap, then I'd love to have references on that to give to CompSci folks I've seen doing semi-open or FOSS designs from 90nm-32nm.

Note: I'm also talking for complex designs like a CPU or SoC instead of Bitcoin-mining primitives or something.

Edit: see my response to wmf as I partly answer thoughts on your other question about volume or eating up costs.

Ohhhh, ok. I think you're totally right for lowish volume runs like this. My information was in the context of where you have enough volume to amortize the mask cost out.

Maybe SiFive is eating the cost here either with the hope that they can sit at 28nm for a while and treat this chip as gravy train money in the long term? Or maybe they just want 28nm experience for higher volume runs in the future?

The problem is the lack of IO. If they had at least one SATA port and a PCIe port it would make sense to sell this long term. The only way they could possibly fix this problem is by connecting their chiplink to an external SATA and PCIe controller which means they need to fab another chip. This is going to be expensive so I don't see it happening.

The board specs mention an FMC interface. Isn't that used (with appropriate module) for PCIe connectivity?

Or use an FPGA which'll be cheaper than any chip they could fab.

Do those costs somehow not apply to ARM SoCs?

ARM SoCs aren't made in low numbers for just a bunch of dev boards, or when they are made that way it's for the first initial testing runs and not publicly available at actual prices.

All these ARM single-board computers use chips that are produced for other purposes in way larger numbers (some of the Raspberry Pi models might be an exception and actually have a slightly customized design just for them, but I'm not sure about that).

The rule of thumb is to not do an ASIC unless you're selling at least 100,000 units if you want it cheap per unit. The ARM SoC's that get used in a lot of products probably are selling in really high volume. To illustrate the effect, a product that cost $100,000 to design would break-even at $1 a unit in that volume but $1,000 a unit if only 100 were sold. That's idealized: you generally have to pay for the masks, a wafer per set of chips that fit on one (number varies), absorb/charge the cost of broken chips, IC packaging for the chips, shipping, storage, etc. That's not counting boards, assembly or whatever.

So, selling something in high volume like big-name ARM's gets the prices way down per unit. Getting to high volume is a marketing and product development problem more than a technical one. Good luck on the startup. :) The ecosystem benefits will give ARM-based solutions an advantage there for a while into the future. The RISC-V chips will cost more due to lower volume unless all the heavy costs are absorbed at a loss by whoever builds them. I've recommended Universities or foundations attempt that to get good, FOSS chips started on good nodes with them sold at material, assembly, and distribution cost from there.

OpenPITON is another one with potential since they have prototypes for 32-core, OpenSPARC CPU's at 32nm. Leon3 GPL was an older one in SPARC. J2 (SuperH-style) is a compact one that's about 3 cents a core in 180nm with who knows what performance, energy usage and pricing could be achieved at 28nm-32nm. CHERI, a capability-secure extension of MIPS, could be ported to one to give it a security advantage to secure sales from defense sector. Draper is aiming for that with SAFE architecture (crash-safe.org) added to RISC-V. If wanting max reliability (eg safety-critical), VAMP was a 32-bit DLX CPU formally-verified for correctness that could be done on older node with mods for lock-step, triplicated redundancy, or board-level fault-tolerance.

So, there's possibilities to justify higher prices long enough to recover upfront costs or get them down on simpler designs. The HiFive is a serious core on a relatively-recent node, though. It's not going to be cheap unless selling boatloads. They probably didn't expect this product to sell boatloads, either.

don't forget the cost of testing - that adds up, often people do simple testing on the wafer to avoid the cost of packaging bad die, then full testing once packaged - remember a bad chip for a semi manufacturer is a whole bad board for their customers

Most SBCs are based on SoCs that are pretty much end of life. They have already amortised their costs years ago by selling them in smartphones, tv boxes, tablets.

This is really cool, and I'd love to buy one, but it's an order of magnitude too pricey for me to justify the purchase when I already have a (admittedly probably slightly weaker) raspberry pi gathering dust.

This board isn't really intended to be a "raspberry pi for everyone" style release, its pretty much soley aimed at developers and companies that want to port their software to RISC-V so when the cheaper boards do arrive, they will have software to run on them.

I'm not sure there will be cheaper boards, to be honest. Just more fully functional ones with more peripheral types and added features I'm sure. This is a workstation-class CPU they are developing, not a RasPi competitor.

I’m curious to know what leads you to consider the RISC-V to be a “workstation-class CPU”. As I understand it’s an ISA that aims to be pretty broad, and is expected to compete (at least initially) with low-end embedded MIPS chips that are pretty ubiquitous everywhere in all sorts of devices.

I’m talking about U54 from SiFive specifically.

The U54 is squarely in the Pi competitor category. Claimed IPC is slightly lower at 1.7DMIPS/Mhz vs 2.3 for the Cortex-A53 in the Pi and this SoC clocks slightly higher than the Pi at 1.5Ghz vs 1.2Ghz, so performance should be quite similar. I certainly hope it manages to catch on and we get higher performance RISC-V chips at a reasonable cost in the future but these are clearly not it. If by "workstation class" you mean Skylake/Zen levels of performance there's a long way to go.

The U54 is a single-issue in-order core. Again, what makes you label that a "workstation-class core"?

The company's stated plans for the product line.

Well... this certainly isn’t workstation-class performance. This is more like a smartphone SoC.

Hmm, my google-fu seems to be failing me, do you have a link where the company states that?

No, I’ve learned about SiFive from the talks they’ve given at the RISC-V workshops. I think some of those are recorded, maybe look there? I’m on mobile and it’s hard to provide a link.

There are tons of other people who are working on RISC-V as well, LowRISC for example. The software should run everywhere.

I’m talking about the Freedom Unleashes platform from SiFive.

> I'm not sure there will be cheaper boards, to be honest.

a) No need to assert honesty - we assume everyone is trying to be honest with each other here.

b) History is littered with people believing or claiming that some thing can not possibly become smaller / cheaper / faster ... it's a dangerous predictive path to wander, with basically no recorded precedents to cite.

While my Pi isn't gathering dust, I too don't have the budget for a 1k dev board... I get that they probably have a lot of cost to cover, but at this point only investors can really buy this hardware (i.e. a company needing open CPU hardware for research or future products might invest in a 1k board)

I'm gonna buy one, because I'm quite used to spending $1k on computers .. been doing it for decades .. and this machine is the kind of machine I'd rather be spending my next $1k on, than say, the current commercial crop.

So I hope this won't be the first rev, and that I look back in 2 years or so and go 'well, time to upgrade it' .. this would be an economic as well as practically good investment. You know, like any computer you might purchase ..

> 'well, time to upgrade it'

Won't happen. It's a dev board, not a computer like you'd purchase. If you come in expecting a computer instead of a sort-of-not-broken work in progress, you're setting yourself up for regret.

The RISC-V community is trying to make real useful practical computer. The next revision of this board would be something that is of practical use. There is lots of practical things you can do with it alrady, it has a networking. Maybe not the cheapes option but still useful.

It's a dev board. It's not intended for end users, it's intended for people prototyping new systems based around the SiFive.

Totally agree. One of the main advancements in the last several years IMO has been bringing the cost of this type of thing (ie; dev boards for microcontrollers, RPi, Arduino, etc) way, way down. I remember back in the day to get the equipment needed to program your own MCN or FPGA required a substantial investment, nowhere near approachable for a typical hobbyist or generally interested person. Now days you can order 3.3 and 5v arduino pro mini's by the caseload for $1.50 apiece.

Same here. Kinda bummed, I was looking forward to it :(

Looking at their crowdsupply page, they did make a smaller arduino-like board [0] at a very reasonnable $59 price point.

I'm not sure if those are still available, but you might want to check it out!

[0]: https://www.crowdsupply.com/sifive/hifive1

Pretty different board. RISC-V ISA has various well-defined subsets. The cheap board is the microcontroller end of the spectrum. This CPU has an MMU, etc.

Never mind that said RPi have multiple USB and video ports to go along with ethernet (and on later models, wifi).

Also availability, availability, availability. The RPi can be bought from a multitude of places across the globe.

All that said, if we could get a version with a bit more onboard IO options, and preferably something like SATA, i would be all over it.

Having an actual RISC-V ASIC is cool, but note that you can run a RISC-V softcore on a significantly cheaper FPGA board - or even in software simulation.

This is more interesting for showing progress in the RISC-V ecosystem than for hacking around with.

Honestly, if you want a full MMU with a Linux-capable RISC-V implementation and everything, and a few cores like this -- you'll need a fairly decent FPGA, as well as the clocking resources/peripherals so things aren't unusably slow. You can get cheap FPGAs for the microcontroller-class devices/cores, but the FPGAs SiFive recommend for U-series cores of theirs are quite expensive, like $4000. lowRISC can run on something like a $1500 Kintex-7 devkit, I think. There are better boards coming out soon at better price points (such as this one[1]) that might fit the bill, but then you have to do board setup yourself... It's not worth it IMO unless you plan on reusing the FPGA.

If you want a microcontroller class chip so you can just play with the 32-bit instruction set -- you can get that with QEMU and a $60 HiFive1 board. You can run picorv32 on a $50 FPGA, even! But if you want a Linux-class chip now (with full features you expect e.g. from an ARM-class device) -- this is about as good as it gets, I think. You won't get a 1.5ghz quad core performance like that cheaply. For now. Next year it'll be different, hopefully.

I imagine the real reason this board is pricey is due to limited volume runs for the ASIC on 28nm, which is the bulk (how many people will really buy one? these are early adoption systems), and, I speculate, some of the weird material choices. 8GB DDR4 ECC, ok sure that's cool and unique, but no SATA port? Really, so I'm stuck on a stupidly large microSD card? And they specify FMC, but FMC cards are typically very expensive for high-throughput devices (think HDMI, ADC/DACs, SFP/SFP+ breakouts). Maybe they'll just go all out and have the FMC mount the system directly onto a broken-out PCIe carrier board, or something? I dunno.

As for hacking... all that said, the rather large amount of RAM and relatively fast cores do excite me -- it means you can actually use the parallelism offered for things like actual compilation. And it's real privileged silicon, so for system porters/distros/etc I think it's probably more reasonable of a purchase. I'm eager to get NixOS running on a real silicon device like this, so I've supported it (I had a lot of fun with my HiFive1). If they had just included SATA, this would be almost a no-brainer for integrators/distro porters. I just hope they'll follow up on a decent expansion option... In the mean time until it ships, QEMU should be ironed out enough by now to start a real port...

[1] https://shop.trenz-electronic.de/en/TEF1001-01-Kintex-7-PCIe...

Your speculation is quite accurate...there will be FMC expansion cards specifically built for the HiFive Unleashed that will break out most of those peripherals you're talking about (PCIe, SATA, to name a few in particular). Unfortunately we weren't quite ready to announce those details today but will be making those updates ASAP.

The reason for the large RAMs and microSD card is specifically to help software developers. That's who the board is for until we can drive the cost down even further for even more folks.

-Jack Kang, SiFive

Thanks for the confirmation! The microSD card's speed is my main complaint, relative to everything else. But I'm just complaining I guess -- because otherwise I think the board looks very good, without any expansions. :)

It seems every board always has a catch somewhere, but if an FMC expansion can work around this, that'd be excellent. I'm looking very forward to putting this machine through its paces once I get my hands on it!

Without the expansion board you'd likely use iSCSI or NFS root for the OS. The SD card would only be used to boot.

Gigabit ethernet though, so network storage is presumably an option. The Pi suffers from relatively poor network throughput (although much improved since the first release way back when)but I'm assuming this won't suffer the same problems.

There's already a plethora of cheap Pi-like boards out there from various companies with gigabit ethernet that's not USB-based and other nice features, many of which have the advantage of having more established software support. If you're willing to give up features like graphics which this board also lack, you can even have a 100% open source software stack on them.

I guess that implementing SATA for a small batch of board would have been cost-prohibitive. SATA isn't simple.

The board is already cost-prohibitive for most people! They probably lose money on these alone -- the tape-out for a 28nm ASIC is millions alone, they're never getting that back off niche development kits.

Plus, given it has a custom ASIC and rather newer components like DDR4 (DDR4 controllers aren't exactly trivial on their own) and FMC (not simple, high cost), I'm having a hard time feeling sympathetic for this particular complaint, especially considering bottom-of-the-barrel boards like the OrangePi & networking competitors like MacchiatoBin can stack in multiple SATA ports on devices in the $50 to $350 range.

I mean, the board is already $1k, and I doubt they're going as far as home grown USB/UART/JTAG chips -- probably FTDI chips, so some of the open hardware claim is a bit fluff in practice, I'm guessing (I don't think this is a huge deal, but many people do). You'd probably use an off the shelf SATA controller & chip, it's not like you really get a lot from rolling your own.

Then again, I've never taped out a board with SATA, so what do I know? But I find it hard to believe the difficulty/cost of acquiring the chip/controller, or integrating it, is a limiting factor in a run like this. Unless they actually planned on rolling it themselves, and I don't know why they would. I'm honestly guessing they're just leaving it to expansions for whatever reason, but we'll see.

See slide 38 in the slide link at https://fosdem.org/2018/schedule/event/riscv/

It mentions low speed peripherals like UART being open source, but high speed interfaces like DDR and ethernet are 3rd party licensed IP blocks.

The FPGA boards capable of running a U54 image (what this dev board is) are much more than $1k. And the resulting clock rate would be orders of magnitude less.

That said, FPGA boards are useful for doing interesting things with RISC-V, such as writing custom accelerators.

It all depends on what you're doing. If you make do with pure software, you should stay there. But if you want to execute a lot of instructions, for example to use CRASHME or descendant to look for bugs in the cpu implementation or Linux or whatever, being able to execute several orders of magnitude more instructions by paying more for an ASIC can be very worthwhile.

And before you think this is a silly example, I found a bug in a new microprocessor once using CRASHME. I was attempting to evaluate the quality of the supplier's testing.

> I was attempting to evaluate the quality of the supplier's testing.

That looks like mission accomplished to me.

From the PDF: "Each U54 core has a high-performance single-issue in-order 64-bit execution pipeline, with a peak sustained execution rate of one instruction per clock cycle.". I.e. lower performance per clock vs an ARM Cortex A53 (2-way in-order execution 64 bit CPU).

As prototype, the price is OK, in my opinion. In volume there is no reason for not selling a tiny board with those specs for under 40 USD (e.g. 2GB RAM Pine 64 quad Cortex A53 boards with gigabit ethernet are below 30 USD, including 3D and video acceleration).

While I love the Pine64 concept, the crappy docs and closed graphics blobs means that Linux tends to have all sorts of weirdness, let alone BSD. I can't wait for a first-class open platform.

Sure. I would love to buy that RISC-V board. But 1000 USD is way too much. Anyway, it is not a bad start, e.g. exotic ARM dev boards are very expensive, too.

There is a little more information on the crowdsupply page: https://www.crowdsupply.com/sifive/hifive-unleashed

What are the 8 SMA ports for? I've looked at the sifive page and the crowdsupply site as well; haven't been able to figure it out. Not obvious why a RISC-V development board would have SMA ports.

Labels are the following:









These are just SoC debug headers. Eg: GEMGXL comes from Gigabit Ethernet MAC, and GXL is a marketing term from Cadence.

Six appear to be test points for monitoring memory bus clocks.

Excited about RISC-V hardware hitting the market, less excited that it's $999. I mean, I get that it's not going to be as cheap as a Raspi, and I'd happily pay $200-300 for something like this. However, at $999, I think I'll wait.

Any peripherals on this thing? They mention Ethernet and some SPI flash. How about USB or even a UART? I didn't expect a frame buffer, but what does it actually have? I thought they were going to have PCI via an FPGA, but no sign of that either.

They presented this at FOSDEM today, and ran the entire presentation (and a demo afterwards) on the actual hardware. They used PCI via an FPGA to interface with a standard ATI gfx card iirc. I don't know if the video is out yet, but have a look!

No video yet that i can find, but it seem the slides are available.


That's what I hoped could be done. Did they use an open source ATI driver compiled for risc-v? That would make for a very interesting development rig.

The page mentions an FMC connector for future expansion, a standard FPGA expansion interface supplying 34 to 160 10Gb/s signals.† I don't know about varieties and compatibility, but Hitec makes a great number of interface cards for it, including an 8 lane PCI Express.‡



I wonder what signals they actually route out the FMC card. The website doesn't give a lot of information, but it seems like probably just ChipLink which they claim is their serialized TileLink bus protocol. So, you probably wouldn't be able to just plug a low-cost FMC module into the board. But with something like https://www.xilinx.com/products/boards-and-kits/ek-s6-sp605-... for another $500, you could theoretically use the DVI connector to provide a framebuffer, the PCI connector to interface to an NVME drive, and the FPGA to implement the logic of a TileLink endpoint as well as a minimal framebuffer and PCIe root complex.

ChipLink is indeed routed out to the FMC. The demo during FOSDEM was done with a FPGA card attached to the FMC which then drove additional peripherals.

Jack, SiFive

The instructions for some of the previous dev kits point to "known good" FMC compatible adapters.

The one that caught my interest right away is the PCIe module:


There's also a PCIe expansion that works with it, so increase the number of slots to five:


With those, it seems like enough to test out development for storage applications. eg NAS, etc.

There must be a UART because there's a FT223H and USB port at the top left.

For a FT232H on early development hardware, my first thought would be JTAG, not UART. Could be both if it's a '2232, though.

That micro port may only be for power.

Or perhaps a client port to allow a PC to talk to the board without a network connection.

Yeah, that's totally a console port. The micro-B device connector gives that away.

What does "Linux-capable" mean? Are there other RISC-V systems that are not "Linux-capable"?

RISC-V is defined by two standards. The older user level one is pretty stable but the newer supervisor level one is still changing so many existing RISC-V systems don't implement it. That doesn't necessarily keep you from running Linux on them but it does mean they need a special port. This new product will be able to boot "standard" Linux binaries as will any future system implementing the supervisor spec.

Yes, the same company sells arduino compatible RISC-V microcontroller boards that are not capable of running Linux.

What does "arduino compatible" mean in that context?

AFAIU it means that the I/O pins are compatible with Arduino shields.

Disclaimer: I have neither a HiFive1 or an Arduino myself.

I think they mean it has a real MMU capable of memory protection and virtual memory. Earlier RISC-V cores came out before this was agreed to as a standard.

The other ones are just microcontrolers. This is the first one that is capable of running a full operating system.

So the price is high, but I'd actually consider getting one if only they'd exposed PCI ports and put it in, say, a MicroATX form factor or the like.

Maybe someone will make a daughterboard to make that happen for a reasonable price. Then I'd maybe have fun making a little machine to run it. I really like the idea of RISC-V.

Anybody have any idea how many MIPS we're talking here?

The FMC adapter item on the specs is an standardised interface adapter. Looking at the instructions for some of the previous dev kits points to an FMC PCIe module that seems like the right kind of thing:


Obviously you'd want to check a bit more carefully. I'mjust pointing out that PCIe seems like it's a go-er with the hardware right now even. :)

I'll probably be grabbing one as soon as it makes fiscal sense again. Kinda bummed for the price point, but they aren't a charity.

It'd also be nice if they had an exposed PCIe slot and/or exposed DIMM slots at this price point, but GbE is good enough for rapid testing.

They might as well put $10,000 instead $1,000 and it wouldn't change things much - most people interested in the board won't buy it as the price is prohibitive. They would have to go below $100 at the very least.

Data point of 1: I just purchased an early-access dev board ($1,250), but I wouldn't have at $10k. $1k is too expensive for most individual hobbyists, unfortunately, but cheap enough that it can be considered a petty cash expense in my employer's hardware R&D effort. I think it was perfectly priced.

Academic buyers won't have a big problem with that price. It's a drop in the bucket in the scope of all but the smallest research budgets.

Eh, if they have a defined project maybe...

$100 = Would buy this on a whim on the off-chance it's usable. Also, would consider buying 12x of them to outfit an undergrad lab.

$1000 = Have to have a clear project/vision in mind of how the board would be used and how other methods (FPGA/emulation) couldn't suffice.

That's a good way to sell half a dozen boards. :-(

It may work to bootstrap a larger market, but this board will not be a hot seller.

Anybody knows why it is so expensive? Is it just the low quantities?

I was talking to the SiFive team yesterday and it's down to the relatively small run of wafers they made. The price will come down, likely very dramatically, but not for a while.

This board is very much to get hardware into the hands of developers who are porting software (such as me), it's not really something for end users at this time.

What are you porting?

We are working on Fedora RISC-V.

That, and customers for this kind of board are typically expected to be working on a project with a significant budget where management wouldn't even bat an eye at dropping $1000 on a development tool.

Maybe the pricing is to keep out people who would only buy low volumes of chips later?

Sony used to price Playstation 2 dev kits at $thousands. I always figured that pricing was to ensure only rich companies with e.g. marketing budgets were going to develop games.

Frankly they are probably losing money at a $1k price point, with the millions they have probably spent in 28nm production runs. They're making a long play here.

It's a bit bizarre that people are expecting RaspPi prices this.

Rpi has really mutated what people expect out of EVKs.

What started as an "educational" board has pretty much turned into a turnkey cheap Linux board for a lot of people. So they expect that out of every board now.

I also wish Broadcom hadn't dumped these things under market price. That's another thing that has really messed up expectations.

Keep in mind that console dev kits of past were devices produced almost manually in very limited numbers. They also actually included a lot of hardware required for debugging only and was built to endure almost 24/7 work without overheating.

So there are reasons why this hardware was expensive. And there absolutely no reason for console vendors to sell them at loss simply because otherwise companies will buy more devices than absolutely needed to get job done.

As for the limiting access to console Sony and others managed it fine without making devkits pricey.

> Keep in mind that console dev kits of past were devices produced almost manually in very limited numbers

You might be thinking of the alpha dev kits. Those were basically gigantic metal wirewraps, and if you were at a large game dev you might get one or two locked away in a conference room. I can't remember off hand if those cost anything actually? You had to be a substantial game studio to get one in the first place, and Sony took them back.

Either way, that's not what I was thinking of. The production PS2 dev kits (called "TOOL") cost $thousands and were not manually constructed. A game team would have a lot of these. I can't imagine these cost as much to build as they charged, especially years after the PS2 launched.[1]

[1] - I wish I could remember/find the price for the TOOL, I want to say it was like $10 or $20K, even 2-3 years after PS2 launched?

It seems a little expensive but not unusually so. A beagle board is 270. This has double the processor count, and is 64bit vs beagle board's 32 bit. It has 4x the ram of a beagle board as well.

It looks like it uses an ASIC, which in low quantities are fairly expensive.

It's a great step towards making these chips more generally available. I am sure they will release a low cost dev board like the rpi pretty soon.

Could anyone suggest other boards like this with intel cpus equivalent in terms of performance?

With intel... any number of PC motherboards. With ARM... any number of quad-core ~1.2GHz boards.

Either of the above can easily be had for a few percent of the cost of the RISC-V board, will likely out-perform it and they're losing money on each one of them.

You might have to go back a few years to find an Atom this slow. But for <$200 you can get a much faster Atom board like the UP Squared: http://linuxgizmos.com/sub-100-dollar-maker-board-packs-apol...


Though, as the other commenter mentioned, there are dozens.

Can anyone ellaborate on how this compares to say ARM, intel etc?

RISC-V is a open specification (actually a couple of them) backed by a the RISC-V foudnation, which "anyone" can use to implement a design on their own.

So they are closer to what ARM is, which licenses IP for others to implement.

It’s an open source design.

It is not AFAIK, this chip contains E51 and U54 cores, E51 from what I seen (when prices were public) costs $595000.

--- Upd: some news articles still mention launch prices [0]

- [0] https://www.eetasia.com/news/article/chinese-start-up-sells-...

To put that into perspective, reports I read put MIPS at $700-900k with ARM starting at $1+ million. Buyers of those get their ecosystem of existing tools, OS's, libraries, etc. RISC-V doesn't have much of an ecosystem yet. However, if FOSS software covers application requirements, then a SOC using the RISC-V chip saves a lot of money upfront. Even more down the line since SiFive doesn't charge royalties.

(Edited to change part about royalties after reading article more closely.)

From what I understand, the RTL of these cores and a good chunk of the stuff around them is open [1]. The cores themselves are basically all specialized Rocket cores. What you're paying for is probably the ready-to-fab routed cores in whatever CAD format and support.

[1] https://github.com/sifive/freedom

They charge for the peripheral IP and system design. The actual computation core architectures are freely available on GitHub under open source license.

is it an fpga?

No, as far as I can tell, it's an actual RISC-V ASIC; it has 5 cores on a 28nm die.

I'm afraid the amount of FPGA required for 4 beefier cores + 1 simpler core + 2 MB coherent L1 cache could cost more than $999, but I may be wrong.

I get the difference between FPGA and ASIC. But...

What's the difference between a CPU core on an ASIC and a normal CPU core?

You seem to be confused about the terminology. A typical CPU core is an ASIC.

An ASIC (application specific integrated circuit) is exactly what it says on the tin - an integrated circuit designed to perform one function - in this case, executing machine code. Most of the chips you see inside electronics are ASICs (everything from operational amplifiers to ethernet PHYs and CPUs).

I don't usually see the two terms used together, because of the "application specific" vs "general purpose" thing. So I thought there might be something I was missing.

FPGA is more "general purpose" because you can reconfigure it for almost any purpose. A CPU, compared to it, is more "application-specific", though not as specific as e.g. an ASIC for handling crypro.

I probably added too much contrast into my original comment.

woop woop! RISC-V is excellent and I hope this is the first of many!

Why so pricey?



Virtualization instructions aren't finished yet: https://riscv.org/specifications/privileged-isa/

What's a good way to get involved in RISC-V


HiFive is the cheapest non-vaporware upgrade path for Amiga enthusiasts (if they bring their own AGA GFX)


Sorry could someone explain to me why this is $999? ( Apart from because they can.)

28nm custom CPUs. They've spent millions on setting up the production run. They are probably losing money on this board alone, to be honest, at that price point and expected sales.

My ideal SERVER motherboard for home use: Raspberry Pi style.

1. Compute: A Quad core processor. A73 maybe? 2. Storage: 1 Sata Port to host 1 Hard disc. for Boot / Storage. 3. Network: 1 Gigabit Ethernet port. 4. Power: 1 Micro USB port. 5. RAM: 1 GB or optional 2 GB. 3. 1 USB for everything else.

That's it. No Audio port, No HDMI port, No GPIO Port. No Camera port, No extra USB ports. No MicroSD connector. Now considering Raspberry price of $35, I guess board could be done for <$20.

Every few days, I try to find one.

The Banana Pi M2 Berry fits the bill, unless you mind all the extra stuff it has.


(Costs around $40: https://www.aliexpress.com/store/product/New-products-BPI-M2...)

The PCEngines APU2C0 is their most stripped down APU2 model, it's $99.


Quad-core, 2GB RAM, 2x Gigabit Ethernet, m-SATA, 12V DC power

No audio port, no HDMI port, this model leaves off the GPIO headers, no camera, 2 total USB ports, 1 MicroSD connector.

It's about $99 and the case is another $10. I have several.

What does that have to do with this?

What do you think of the new stuff from HardKernel?

Awesome. Thanks. ODROID-HC2 is on top of my list now. Waiting for my local retailer to catchup!

Perhaps a cheaper route to have an OS up and running on a reference design for hardware that's too new is to implement it on top of MAME.

Even if it takes an 8-way Xeon Platinum to run it at a reasonable speed, I think more people will be able to experiment with it.

And the emulation provides some neat debugging opportunities.

Various RISC-V emulators are available, e.g. you can run Linux on a simulated risc-v processor in qemu.

True. I wonder how much of this board's hardware is emulated in there.

Quite a lot, see item 3 below:

    $ qemu-system-riscv64 -machine \?
    Supported machines are:
    none                 empty machine
    sifive_e300          RISC-V Board compatible with SiFive E300 SDK
    sifive_u500          RISC-V Board compatible with SiFive U500 SDK
    spike_v1.10          RISC-V Spike Board (Privileged ISA v1.10)
    spike_v1.9           RISC-V Spike Board (Privileged ISA v1.9.1) (default)
    virt                 RISC-V VirtIO Board (Privileged spec v1.10)

Applications are open for YC Winter 2020

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact