Yes! This is absolutely the right move for SBC makers. It has always bothered me that SBCs are sold as entire laptops, which were often subpar because its hard to justify an expensive screen/keyboard/shell for a single cheap underpowered SBC. Then, when you move onto the next SBC, you end up with an entire laptop of e-waste instead of just a single motherboard. Framework has already done an excellent job creating the physical/human interface, so lets just replace the brains inside of it so I can keep using the same excellent matte screen, keyboard, and touchpad and have essentially the same user experience (but with different performance).
Hopefully this board can also operate in stand-alone mode outside of the framework shell just like the official framework motherboards because then we'd get the best of both worlds while also being able to benefit from the coolermaster case. Regardless, I'm going to order one when it becomes available.
Also stoked to see the open source CAD files for the shell.
I don't understand what you mean that SBCs are sold as entire laptops. SBC is Single Board Computer. Like a raspberry pi. A raspberry pi is not a laptop. If you just look up "RISC-V SBC" none of them are laptops. Is the intent that you want the SoC to be socketed and easily replaceable like AM4 Ryzen?
Often SBC makers will release a laptop with the same SoC and specs as the SBC, sometimes literally containing the SBC inside of them.
The Sipeed Lichee Console 4A [0] contains the Lichee Module 4A (one of those SBCs on a module like the raspberry pi compute modules) and it looks like the Lichee book 4A is doing the same thing [1]
The DC-Roma 2 [2] appears to be this sbc [3] adapted to a laptop.
The Spacemit Muse Book [4] seems to be the Spacemit Muse pi adapted to a laptop [5]
Banana Pi posted this youtube video [6] showing a laptop with "Banana Pi BPI-F3" in the title which is the name of one of their SBCs. [7]
> Is the intent that you want the SoC to be socketed and easily replaceable like AM4 Ryzen?
Nope, the intent is I wanted exactly what just happened: swappable board inside a reusable shell. I don't want to get a DC-Roma 1 and then a DC-Roma 2, ending up with 2 entire laptops of e-waste when I should be able to upgrade boards and keep the screen/battery/keyboard/touchpad/webcam/speakers/microphone/fingerprint reader, ending up with 1 complete laptop of e-waste and 1 additional board of e-waste.
Sipeed has announced that they'll release two new modules next months: the LM3A with a SpacemiT K1, and the LM5A with a EIC7700 that has 4×SiFive P550. [0]
All of Sipeeds form factors are supposed to be upgradeable: laptops, tablet and SBC... but personally I would like to have it verified beforehand that all ports and functions are supported by the new module before switching.
I guess I don't understand the problem. If you don't want to buy the laptop, don't buy it?
I'll note that an SBC and a SoM are different things. SoM (system on module)s are, as the name says, just one module out of many (usually with a carrier, I guess in this case that's laptop). An SBC is much more often designed to be standalone. I will note that those SoMs look suspiciously similar to SMARC SoMs. If so, you should be able to swap them out for ARM, x86, I guess RISC-V, etc modules as they get released. Some of those are indeed just laptop-y motherboards.
Industry has had SoM carriers with swappable SoMs for ages. Just look up SMARC carrier. Any one of those should be able to take any SMARC SoM. Same thing with COM-Express. They have different sizes (type 10, type 6, some others I don't remember, where I think backwardsly 10 is the smallest). Interestingly I haven't been able to quickly find SMARC or CoME laptop carriers. So, if that's what these are, they're actually industry leading innovators instead of holding people back. These types of SoMs are even less wasteful. Things like USB hubs, ethernet PHYs, battery circuitry, video converters, all of those can stay on the carrier and the core functionality of the chip (CPU/SoC, flash, RAM) can get changed out as new ones come in. In a way, it may even be better than framework. Again, if this really is a set of SMARC carriers and not just them using the same form-factor or looking similar.
> SBCs on a module
That's an SoC on a module (though nobody says SoCoM) (or even worse, if it's also a SiP, a SoCiPoM). I don't think that's really a proper use of the term SBC. An SBC is a standalone, able to be used by itself product. Usually no need to buy anything else to adapt it to the outside world.
> I guess I don't understand the problem. If you don't want to buy the laptop, don't buy it?
The problem is I want a risc-v laptop, but I don't want many risc-v laptops. Before framework, if I bought one of these risc-v laptops then I would either be stuck at exactly that level of performance, with no ability to get chips with newer instructions (for example, the DeepComputing board that is going in the framework laptop does not have the vector extension), or I would have to upgrade at some point which previously necessitated getting another entire laptop which is needlessly wasteful. This isn't as much of a problem in the mainstream laptop market because advances from year to year are pretty small but risc-v is in a rapid development phase.
> I'll note that an SBC and a SoM are different thing
Are you saying that if I took a raspberry pi compute module and soldered power to the power pins, that it would not boot? Because if it would boot then its an SBC, just with a nifty form factor. I've not run that test myself, but my understanding was that all the necessary hardware was on the module itself and the carrier boards were just routing pins to PHYs for peripherals.
> SMARC
TIL about SMARC. It looks nice, and I agree that it could be less wasteful and therefore could be better. I guess it comes down to which level of interface will have the greatest longevity.
A socketed processor is the least bundled and therefore potentially the least wasteful but then you're stuck with the pinout in your socket, for example, AM4 is stuck on DDR4 ram.
SoMs like SMARC move the interface a bit further out by moving the ram to before the interface but it looks like you're limited to 4 lanes of PCIe, 2 USB 3.0 ports, there are 6 USB 2.0 ports which is silly in 2024, and there are even pins wasted on nearly-dead standards like SATA.
And then there is the Framework option which moves the interface even further out, being more wasteful by having more stuff bundled on the board, but also achieving the most flexibility since you just need to drive 4 type-c ports, and the handful of onboard connectors like the eDP, webcam, battery, and speakers. If, in the future, USB 7.0 is still using the same type-c port, SSDs are using the new m.5 connector, and LPCAMM 4.0 is out using optical magic connectors, then the Framework shell is still good-to-go whereas a laptop SMARC carrier is not (as demonstrated by SMARC 2.0 not being backwards compatible).
Personally, I think Framework is drawing the line at the right level to ensure longevity, but only time will tell. In the case of SMARC time has already told as we can see by them being stuck on USB 3.0.
I understood that sentiment to mean you want them socketed. Maybe that's me projecting. I wasn't around for the x86 boom days where every week speeds would double. So I don't know how it was solved back then for laptops, but I do know some laptops had sockets even maybe 10 years ago.
> Are you saying that if I took a raspberry pi compute module and soldered power to the power pins, that it would not boot?
Not necessarily. Though if you look at the CM3 (and likely previous compute modules), you'll see that:
1. some modules don't have storage. This is occasionally common with SoMs. That would indeed prevent booting. 2. you need a lot of supply rails. Just hacking up a 5v from a USB cable won't do it. You need like 3-4 regulators or something to get full functionality.
The big difference is that an SBC has connectors for users. A SoM has connectors for integrators. SoMs often have a single card edge (like SMARC, the pi CM3, and some others) or a mezzanine (like CoM-e). These are often high density and not meant for soldering to. So, let's say you did have a SoM that only took 1 supply rail (like a good SoM should) If you soldered power, you'd have a powered SoM. Maybe it booted, you have no way of knowing because there's likely no LEDs, no ethernet jack to connect to, no usb to plug into, no display connectors, etc. All of those you'd need to solder in a connector for. That's what the carrier does, if you've seen those.
> I guess it comes down to which level of interface will have the greatest longevity.
> you're stuck with the pinout in your socket
I think this is where Framework falters. IMO, I need more than 4 ports on a small laptop. Just today I was using 3 type-A, charger, and an hdmi. I could have also been using an Ethernet but because reasons I could get away with wifi. That's 6 ports, potentially. I think the framework 16 doesn't have a headphone jack builtin? I took a call during that so that's 7. Framework decided on a card form factor where you can only fit 1 of most things. And they're now stuck with that. There was a good video I can't find on what it would look like to have the cards be 1.5x the width or so. I could also use a hub, but the whole point of this is that any decent laptop these days can have 1 hub give you tons of IO. The framework should be hubless, ideally, in my opinion.
SMARC is kinda industrial. For different users, that's enough. Not everything even uses PCIe. With storage onboard, if you don't have discrete graphics, those wouldn't even be used (maybe for Ethernet but I think SMARC also usually has that before the connector). USB 3 I could see as being an issue, but honestly a hub is likely needed for these lower end chips anyway. This way the hub can be on the carrier, which saves cost on upgrades. On the embedded side, not much is higher than 3.0 these days anyway. Which is kinda disappointing. FTDI and Infineon have 0 released >= 10 gbps interface chips. Lattice has the first FPGA (and it's from this year) I've seen with an integrated USB superspeed PHY and that's only 5 gbps. Versal gen 2 I don't know about just yet, but I could see it going either way. Zynq MP/RFSoC is only 5. The list goes on. I think except for external SSDs, nothing really uses > 5gbps? So there's that for perspective.
To sum up, it's weird straddling the consumer/embedded industrial world. Embedded/industrial products are currently > a decade behind the consumer. The amount of ARM chips that are still coming out with Cortex-A53 or other early ARMv8 (not even the most recent ARMv8) astounds me. It took a decade ish for ARMv8M to get popular and it's still not looking like it's taking over ARMv7. It's incredible how fast phones move relative to devices like pis.
Yes but that's not an SBC, that's a laptop. I don't understand who would call that an "SBC". I guess technically any laptop that's only 1 PCB is an SBC, but that still doesn't answer the original question of SBC -> Laptop.
I agree. If I order an SBC, I expect to be buying a minimum of exactly one bare PCB.
If I'm feeling fancy, I'll add a power supply and a housing to that order -- and maybe a keyboard, a mouse, a display, and/or some storage or other accessories. But it's still rooted as a singular PCB, and that Single Board Computer can be made to do useful work with a minimum of help.
This is profoundly distinct from when I order a laptop, wherein: I'm absolutely not buying a bare PCB. My laptop is, at minimum, absolutely going to include a screen, a keyboard, a pointer, a battery, a power supply, and a housing to tie it all together.
---
If a laptop is an SBC, then a Playstation 5 is an SBC. A pocket supercomputer and an iPad are both SBCs. ONVIF camera? VoIP desk phone? SBC. Asus router? SBC. Wifi-connected light bulb? SBC!
That's madness.
If I'm buying a laptop, then I'm buying a laptop. This laptop may or may not contain an SBC, but it's primarily a laptop. We have a term for these things: It is "laptop."
If I'm buying a light bulb, then I'm buying a light bulb. This light bulb may or may not contain an SBC, but it's primarily a light bulb.
And if I'm buying an SBC, then I'm not implicitly buying anything more than a single-board computer. It is primarily an SBC. (We have a term for these things, too: It is "SBC.")
This kind of excessive and deliberate ambiguation is as dumb as the term "loosing" was ten years ago, and it needs to stop.
They are intentionally misusing the lingo to emphasize the fact that guts of a laptop are basically a single board computer. It is a bit of rhetorical flair.
For a while, it didn’t make sense to replace a CPU. Screens and hard drives were advancing around the same speed as CPUs (if not faster, the biggest advances in laptops until pretty recently were OLED screens and NVME drives).
CPU competition picked up a while ago when AMD got their stuff together, but it takes time to turn the boat around, if it is actually turning.
On the other hand we didn’t see laptop CPU replacement as a really mainstream thing during the 90’s/early 2000’s era when things were really going crazy.
On the other other hand, everybody knew desktop PCs were the way to go for performance back then, so tinkering with laptop performance would have been a somewhat odd thing to do.
Hard to speculate about or even measure, because people who tinker with hardware are in a pretty small niche already.
This is interesting, but it'll have a JH7110 and microsd for storage, so it's basically just a RISC-V SBC but in a Framework mainboard form-factor. SBCs with that specific processor seem to be pretty cheap, like the Milk-V Mars which is selling for around $40 currently.
The idea of a RISC-V laptop sounds cool, but this feels like just grabbing a raspberry pi and sticking it in a laptop chasis. It doesn't seem like this is going to really offer anything new in this space other than maybe some increased visibility for RISC-V (esp. if Linus Tech Tips covers it), and a neat project/option for people with existing Framework parts.
As far as development purposes, I don't see what this offers over an existing SBC or even just a VM. From what I've seen of people running Linux on these things, it is definitely not something you'd want to develop on... plus, it seems like DeepCompute sells their own RISC-V laptops which are (probably) more powerful than this thing: https://store.deepcomputing.io/products/dc-roma-riscv-laptop...
> it's basically just a RISC-V SBC but in a Framework mainboard form-factor.
As someone who's been looking for an excuse to get a Framework, I'm thrilled at the idea that if this doesn't pan out, I can just swap in a different mainboard and convert it to a full-featured x86 laptop, and then donate the RISC-V mainboard to the tinkerers at my local hackerspace (who are more than capable of 3D-printing a nice enclosure for it).
I happen to have 2 of them and they're dog slow, pre-vector extension RISC-V. You cannot do much useful with them, they're slower than a Raspberry Pi 3.
No, the first JH7110 boards (VisionFive 2) started arriving in pre-order customer hands in early February 2023, 16 months ago.
In general use they are much faster than a Pi 3, especially given the 8 GB RAM vs 1 GB which is often far more important than the CPU speed. The GPU is also much better. And the M.2 SSD.
Better to compare to the Pi 4. Like all the Arm A55 boards, it's a little slower than Pi 4's A72s, but also uses less power.
Modern Arm chips will maintain a wide gap in performance for quite awhile.* Arm has poured a lot of money into optimizing their chip IP and it speaks for itself in the benchmarks. Excluding if performance wasn't a concern to begin with - I don't see the argument why an open ISA is really worth the performance / heat / power cost.
I hope you realize that you're talking about a cheap SoC that was released in late 2022, and does not reflect currently announced RISC-V microarchitectures.
e.g. SiFive P870 is competitive with ARM's best, whereas Ventana Veyron V2 and Tenstorrent Ascalon/Alastor compete with the state of the art from AMD/Intel/Apple/Qualcomm.
According to SiFive, the P870 is roughly equivalent to an ARM X2 or AMD Xen 4, so it's about two or three generations back from the Cortex X925.
I've yet to see a catalog part w/ that core, but the guys I know at SiFive say they're pretty sure it should hit 3GHz, which is cortex X2 territory. Not exactly ARM's current best, but they're definitely closing the gap.
One difficulty in evaluating SiFive's cores is their business model is to license their designs to companies who are fabbing their own custom or semi-custom SoCs, so they don't appear in generally available parts until much later.
Or even anything better than an A76, announced May 2018? (Pi 5, Rock 5, etc) Which implements ARMv8.2-A from 2016.
Yes, I know ARMv9 cores are available in high end phones running Android. But not, that I know of, in SBCs or laptops running Linux.
We've got several different RVA22+Vector (latest spec, ratified March 2023) RISC-V cores, SoCs, SBCs and laptops on sale right now, with much faster ones (P670, A78-class) coming around the end of the year.
That requires actual hardware, which naturally tends to be in mass production (i.e. cheap) SBCs around four years after the announcement of the core. That's a constant over Arm and RISC-V e.g. we've just in the last two years seen Rock 5, Pi 5 (nine months ago), Orange Pi 5 with Arm A76 cores announced in 2018.
P870 was announced in October. It'll be a while!
P670 was announced November 2022, and will be in e.g. the SG2380 around the end of this year -- actually remarkably fast. It should leapfrog those A76 boards in performance.
If you don't have the actual hardware, discussion about performance is a waste of time. If you disagree, I urge you to invest in my chip startup. Our chips get 10000x better performance than the any other RISC-V chip on the market. The hardware itself won't be out for 100 years, but that doesn't matter to you :)
They also didn't say anything about the mainboard having RGB LEDs or a quantum computing chip, but I'm assuming someone swapping a x86 mainboard for a RISC-V one on a personal laptop will want some kind of performance for daily tasks, which that chip won't provide even barely.
And also understands JH7110 was already available in VisionFive 2 in early 2022, and is neither new nor representative of current RISC-V compliant microarchitectures.
It's intended for hobbyists and tinkerers. If you want a computer that screams, you get a desktop. Nobody's trying to play the newest GTA on this chip.
StarFive catalog parts use slightly tweaked SiFive cores. So if you were going to license a SiFive core for an embedded design, you may want to have something that uses the catalog part to verify your code will work.
I mean, emulating your core for development is a good approach in general, but at some point you may want to run your code on actual silicon.
So sure, an Intel i9 or ARM Mac is probably going to be faster than a 4 core U74 SoC, but if you're using a RISC-V core for some embedded application, having a RISC-V system to test with is probably a good idea.
And it's cool you can get a RISC-V SBC for a couple hundred bux. It wasn't too long ago that you paid $2k for a 4 core U54 SoC with minimal peripherals. And if you can stuff it in a laptop form factor, it's portable.
You can get a quad core U74 1.5 GHz RISC-V SBC for $40 -- the Milk-V Mars. Or $34 for the Compute Module version. Or $5 for the Milk-V Duo with 1.0 GHz single core and 64 MB RAM, which is more appropriate for many embedded uses and still runs Linux (with a 2nd 700 MHz 64 bit RISC-V microcontroller core for real-time tasks)
> It wasn't too long ago that you paid $2k for a 4 core U54 SoC with minimal peripherals
The HiFive Unleashed was never $2k. The standard price was $999, and the first batch of preorders (which I got one of) was $1199.
It indeed was lacking in peripherals, with only 8 GB RAM, an SD card, and gigE interfaces natively, but an FPGA Mezzanine connector that allowed turning it into a full PC using a standard Xilinx FPGA dev board, or the much cheaper custom MicroSemi expansion board.
But if you didn't mind booting from SD, SSHing/X in, and mounting storage from your PC using NFS (etc) then you could do a lot of software development with the bare $999 board.
I think visibility is probably the primary goal which is not a bad thing.
As for actual use the potential lies in the modularity. You could work primarily from an x86 laptop but swap the RISK-V board in for testing. And whichever board isn't currently in the laptop can be even be loaded into the small desktop shell Framework makes. If the price is right an existing Framework user could possibly get a RISC-V machine for less money than one of the dedicated laptops from DeepCompute without needing to buy another screen, keyboard, battery, etc and end up with a better build quality to boot.
What if, with the space you have for the motherboard, you add not one SoC but, say, seven, all connected through an in-PCB network. One has the external ports while all others are headless. A small cluster in the shape of a laptop.
This would be super cool. I don’t know if it would have any actual use in industry. But it would be really neat for people learning MPI/cluster computing. He said, while waiting for a run to go through SLURM.
I've been thinking of doing something with Octavo parts, because my electronic design skills are totally obsolete now and they are simple to integrate. My idea is a board with 16 parts (for 32 cores) with red LEDs lining up one side mimicking a Thinking Machines CM-1 cube. Not a CM-1, but a lot of nodes if you join 16 boards per cube.
> grabbing a raspberry pi and sticking it in a laptop chasis
OT, but does something like that exist for the Pi 5? I actually loved the Pi 400 Desktop Kit (I hope they make one for the Pi 5!), and I saw quite a few laptop shells for the Pi 4, but I've not seen anything announced for the Pi 5 except for various ridiculous "desktop" cases (like the Pironman 5 that I've actually ordered).
The JH7110 is a multi-year-old SBC that is slower than a Raspberry Pi 3 is. It does not have many extensions for things people today take for granted (no hardware crypto for instance is in practice a massive loss.) So, if you're OK with that, then it will be fine. But most people probably aren't interested in making their expensive laptop perform worse than a 15-year-old device in every way.
It's slow to the point of being outpaced by an ARM SBC from 2016, and it's not even current with today's RISC-V spec. This is a curiosity, nothing more, but it will still be far and away the nicest (but not the only!) RISC-V laptop. Give me a Pi CM5 + 16GB RAM Framework motherboard carrier and I'll get out my credit card.
Completely ridiculous benchmarks for what people will use this board for. The xz compression and SQLite are the only slightly relevant tests -- and on those it's pretty close to a Pi 400.
Comparing an 8 GB JH7110 board to a 1 GB Pi 3 is beyond ridiculous. The Pi might win a few micro-benchmarks that use NEON, but not general purpose C code, and in real use 1 GB is incredibly limiting.
All Arm SBCs at present as far as I know -- certainly including the Pi 5 and RK3588 boards (Rock 5 etc) -- are far behind the current Arm ISA, as they implement ARMv8.2-A from 2016. And none of them even have the optional SVE vector ISA that was defined as part of ARMv8.2-A.
In contrast, the JH7110 implements mid 2019 RISC-V specs, plus some things from late 2021 (e.g. Zba and Zbb).
The SpaceMIT 8 core SoC in the BPI-F3, Muse Book and others being released now implements RVA22+Vector ratified in March 2023. The Canaan K230 (on e.g. the CanMV-K230 board) also implements the same RVA22+Vector spec.
Late this year the 16 core ~2.5 GHz P670 (A78-class) SoC will leapfrog anything available on currently known Arm SBCs. Milk-V say the base model of their Oasis SBC will be $119. Sipeed says a fully-kitted board will be $300.
While Framework's blog post and their marketplace don't have a photo of the mainboard, Deep Computing's press release [1] does. Given the "DC-ROMA" name for the mainboard from DC, and their DC-ROMA laptop (not DC-ROMA II) which seems to have similar specs, I would guess that this is essentially the mainboard of that laptop, in Framework form. For context on price, DC is selling the full laptop for $300.
While it's a niche product, it is great to see other companies actually developing components for the Framework platform, and more, more diverse options for components starting to appear. Yes, as people point out, this product doesn't make much sense for many people, but it doesn't have to: part of the advantage of this platform is that components don't need to appeal to a wide customer base. We are also starting to see this now with the two screen options, the speakers (choice between louder or more accurate), etc.
If, as it seems, this is going to be quite low cost, I might buy one just to play around with it. It would be easy to swap with my normal mainboard, and when not in laptop form, could go in a printed enclosure.
Looking at the photo, it of course has no M.2 for storage, as mentioned in the blog post (nrp explains that choice on Framework's forum [3]), but does look like it has one for the wifi card; the microSD slot is also visible. It also seems to have quite a large fan and cooling arrangement for a JH7110, compared to other boards with it?
Battery on the DC-ROMA appears to have been 48 Wh [2], so not enormously smaller than Framework's 61 Wh and 55 Wh options, and battery life may be comparable plus 15% - 25%.
> Yes, as people point out, this product doesn't make much sense for many people, but it doesn't have to: part of the advantage of this platform is that components don't need to appeal to a wide customer base.
I feel like this has been the promise of the modern manufacturing era for so long, but it’s felt like all the momentum’s gone in the opposite direction - everyone chasing the most beige thing they can to try to get the largest market share. I’m excited to see more products and companies pushing against that trend, and it’s part of what I really appreciate about Framework - their product is absolutely a niche product and will always be, and they’re doing the hell out of it.
For context on the cooling system, it is absolutely overkill for the SoC, but to keep the project simple, DeepComputing is re-using the same heatsink and fan from our 11th Gen Intel Core Framework Laptop 13. This means that it should be able to run fairly quietly.
Does the ROMA actually exist? I don't see any reviews of it. In press releases it sounded like jumble of meaningless buzz words (crypto nfts riscv.. etc.)
This is definite validation of our product philosophy, and we're using this partnership to help develop documentation and processes to make it easier to support more partners in the future.
I got an Intel 13 a few months ago, and I'm very happy with it (I'm using Arch Linux on it). I'm also happy to see news like this, other companies buying into the ecosystem.
The only suggestion I'd make is to get the upgraded screen (sorry, even more expensive). The default screen is OK, but it has a fairly distracting "dot pattern" over it in certain lights that I wish wasn't there, and the lower resolution is tough to go back to after using high-dpi displays. I'll probably get the upgraded screen at some point, but it's tough to justify buying a new screen for a brand-new laptop. Wish it had been available when I bought it, but that's life sometimes :)
I ended up with a preorder for the new screen instead of getting a laptop now and the upgraded screen later. Before making that call, I explored a bit around up-cycling the original monitor into a portable display. There are apparently some boards that allow this, if that helps with your math.
The one downside to Framework is that they're nice laptops and (precisely because they're repairable and upgradable) they hold value really well... which means the second-hand market exists but isn't an easy way to score dirt cheap hardware. I can drop $200 on eBay and get a good thinkpad; I can't get any Framework anywhere close to that.
A potential problem for the used market is that, as people upgrade individual components, they are left with used individual components, which are potentially harder to sell than an entire laptop. And as, so far, the major component upgrades have been of the mainboards, it seems like there could be a glut of those, while it seems possible that a used chassis may never be easily available.
As more component upgrades other than motherboards become available, however, it may be that a more useful used market could develop. And it may be that building a 'used' laptop may end up usually involving buying a few new parts. For example, having, over time, upgraded the top cover, hinges, mainboard, battery, wifi card, RAM, and SSD, if I upgrade the display and camera with the soon-to-be-available new modules, I think someone could build a full laptop with my old parts, a bottom cover kit, an input cover, and some fasteners; it may be that the bottom cover (and fasteners) are the only parts there that would need to be purchased new, as I expect other people have replaced input covers (some of my keycaps are starting to degrade, but they actually degrade rather gracefully)
But still, this would be more complex than simply buying a used laptop, and would need a marketplace for all of those parts. I know there was some discussion from Framework hoping someday to facilitate a used component market; that seems like it would be challenging, but on the other hand, Framework seems to have been steadily, actually pursuing the goals they have laid out.
Eventually it should be possible to get the expensive components secondhand at good prices. It would be cool if Framework started selling a bare-bones kit where you could add your own mainboard, screen, etc. That should be possible to do now by buying parts individually, but I don't think it would be a good experience (or cost-effective).
Framework does sell factory seconds for as low as $500 right now, though they need a few more components to be functional.
> It would be cool if Framework started selling a bare-bones kit where you could add your own mainboard, screen, etc.
I'd second this idea. Could be doable by just making every component in the DIY edition optional during configuration, though I don't know how much that'd impact their current manufacturing flow (given that even the DIY edition laptops are mostly preassembled).
> We’re excited to share a preview of a Framework Laptop 13 Mainboard with a new CPU architecture today, and it’s probably not the one you think it is.
Hahaha. Really curious if they’ll announce a snapdragon x elite board. I love the idea of RISC-V long term, but would love an X Elite board for the near future.
I think that would be cool too, and I would guess is in the oven currently, though personally I'm much more excited about RISC-V. Not for what it is today, but for what it can be tomorrow.
>Looks like nothing much and then suddenly it is everywhere
Already happened. 10+ billion cores already shipped as of RISC-V Summit in December 2022. Assume a much higher number by now.
What's new is the general developer/enthusiast visibility we're starting to see.
For ARM that was Raspberry Pi. For RISC-V, the "Raspberry Pi" moment was the release of VisionFive 2, in January 2023, using this same JH7110 SoC.
There are better SoCs now, and SBCs using them e.g. spacemiT K1, in Banana Pi BPi-F3.
That chip has 8 cores that implement RVA22 with Vector extension.
In contrast, JH7110 has only 4 cores, they are slower, and they only implement RVA20, which is essentially RISC-V as of the original user and supervisor specs ratified in late 2019. It makes little sense today, as RISC-V and its ecosystem have advanced considerably.
Sorry just caught up with this.. 10 billion cores! That is astounding. I remember when Intel passed the 1 billion processor mark in early 2000's and that took them about 25 years to do.
It will only be a matter of time until the performance starts to catch up. I have seen many argue that "it will never catch up" but I have heard that before for many ISA's over the years.
I have a Framework 13, and I'm excited about this. I've been looking forward to moving to RISC-V for years now. I'm one of those weirdos who is willing to pay more for less performance. I guess they know their customers.
Hey, y’all weirdos are the ones who pave the way for the rest of us. I’m not going to buy this - I’ve got a large enough hobby project graveyard to know where I sit on the customer/nerd spectrum, but I’m stoked it exists. I like the idea of RISC-V, and the easier it is for people to develop on it, the closer it comes to being useable by upright apes like me.
I also have a Framework 13! About a year ago or more I put out a request to get an ARM based processor. My reasons being the battery life on x86 are so weak, it almost defeats the mobility aspect of a laptop. I was using linux, and spent far too much time tweaking settings to get the longest battery life possible. When I compared this to a MBPro with the M1, which I use for work, it became nearly impossible to get my mind off of it.
Arm processors are way better with battery. There are these new Snapdragon X Elite Laptops, which verify the better battery life. I think to be competitive with Apple, the battery life must improve.
Also, the speakers are weak, likely due to none glued construction. I can live with that. Love how easy it is to swap stuff out. Screen is awesome, would love OLED if possible. Also, the bigger size 16, would better fit my needs.
Currently that computer is sitting on a shelf. Very cool technology. Love Framework!
Odd, I’ve got a zenbook flip OLED and after some tweaking on Linux, I’ve been enjoying not thinking about battery life at all. I think on any modern device most of the power is consumed by the screen, which is not really a problem as much with OLED+linux (since the terminal background is black).
That said, it is pretty rare for me to go more than 10 or so hours without any access to electricity.
At the moment most of my power appears to be going to Bluetooth and wifi, which seems hard to blame on the instruction set.
Yes! I don't currently have a Framework 13, but this is making it reeeallly hard to hold out any longer. Yes, upgrading parts instead of buying a whole new laptop is more economical, and more environmentally friendly, but I think the real killer feature is that it has the potential to democratize new ideas around computer architecture.
It sort of reminds me when you actually had a choice of whether you would set up a Token Ring or Ethernet network, before the option was taken away by the overwhelming ossification of existing hardware.
I've had three FW 13s (a first-gen for myself, a second-gen for my wife, and a third-gen for myself) and I suspect you will be pretty happy with it. What OS do you plan to run?
We're getting a factory seconds one for my son (he loves to tinker, so the fact that it'll need various upgrades is a feature, not a bug). If I were to get one for work, I'd likely start with Ubuntu, but I've really been looking for an excuse to run Nix for my main dev machine.
That's pretty cool. Kudos DeepComputing, Framework, et al.
Tenstorrent is supposed to be delivering a "high-end" RISC-V CPU soon/this year... Don't know what it will look like, but one can dream: It's Jim Keller, after all. If RISC-V is going to blow up general purpose computing he'll probably be making the explosives.
And then one can imagine a Framework mainboard running it.
The DeepComputing launch page[1] reads:
WORLD'S FIRST RISC‑V LAPTOP GATS A MASSIVE UPGRADE AND EQUIPS WITH UBUNTU
Am I suffering a lingo mishap or is that a rather obvious typo?
Speaking about Jetson, Turing Pi developed an RK1 board that is pin compatible with Jetson, and is based on a Rockchip 3588. Recently its 6TOPS NPU got a first class Linux support. Unfortunately, no Vulkan atm.
It would be a really interesting option as a Framework mainboard. Unfortunately, Framework 13" has rather tight constraints on mainboard geometry.
> it is focused primarily on enabling developers, tinkerers, and hobbyists to start testing and creating on RISC-V. The peripheral set and performance aren’t yet competitive with our Intel and AMD-powered Framework Laptop Mainboards. This board also has soldered memory and uses MicroSD cards and eMMC for storage
Not for me, but I will absolutely switch over to the first board that is consumer-focused, assuming it is competitive on performance and battery life.
We’re being deliberate on positioning this as a Mainboard for enabling developers and tinkerers, not a consumer-ready product. As other commenters noted, the idea is to help accelerate the maturity of the RISC-V ecosystem to prepare it for consumer access in the future.
How much involvement did framework have in this becoming a thing? I'm asking because I want to get a feel for how long a decent (probably third-party) ARM mainboard might take.
I think part of the point is to find that problematic software, and contribute patches to make it work (or whatever you do for closed source software). In my mind this is a board to legitimise early adopters' requests, so that they can say "hey, this doesn't run on this very real laptop from a brand that you may have heard of".
I think this is a crucial step on the way to getting an actually good experience on RISC-V based platforms.
There is already a Debian port available for RISC-V platforms including 99% of its packages: https://wiki.debian.org/RISC-V
IIRC even a port of Fedora Linux.
So you could use it as a daily driver already — unless you need a bit more computing power. A YouTuber made a nice video on how he edited a full episode on a RISC-V SoC using Kdenlive on Debian some time ago: https://youtu.be/7ke7d0iO_-0?si=-rjXdTxfYd_jLQuU
>There is already a Debian port available for RISC-V platforms including 99% of its packages: https://wiki.debian.org/RISC-V
The official RISC-V port of Debian is around 97%[0] to be exact. This is already above "armel", an architecture that's been officially supported for much longer.
Well, it's not the only option so no one's being forced to buy it. The same thing could've been said about ARM laptops just 5 years ago, before the launch of M1. Now, with Windows embracing ARM for real this time, it's properly established across all major platforms.
The neat thing about this specific implementation is that it can be dropped right into an existing Framework laptop — a very viable hardware platform. Thus, it will allow developers (presumably mainly on Linux) to test and iron out issues in their software, the only way to fix those software problems you speak of.
It represents a strong leap forward for RISC-V and open ISAs in general.
> what am I going to be able to do w/ a risc v board?
Be the coolest kid in the block.
Now, seriously, people buy RISC-V boards in order to use RISC-V boards. If it were to do something else, you'd probably be better off with an x86 or ARM design.
It’s not super practical yet, a lot of software won’t run on it and they even write in the article that processing power wise it will probably be worse than their x86 counterparts.
As they write in the article, the point is to make developing for RISC-V more accessible so that it can find wider adoption and get over those hurdles at some point to become more mainstream.
I am assuming (hope I am right!) that the RISC-V board doesn't have the equivalent of the IME on it and thus will be open-source "librebootable" from the start, with no binary blobs.
> I am assuming (hope I am right!) that the RISC-V board doesn't have the equivalent of the IME on it
Actually, it does! Check the block diagram on the first PDF linked to by this article, it has two RISC-V "monitor cores", a 32-bit one and a 64-bit one, besides the four 64-bit main cores.
> and thus will be open-source "librebootable" from the start, with no binary blobs.
There's always going to be a small bootstrap ROM, to configure everything to the point where code can be loaded from the external flash; but other than that, I agree with you: I also hope that the bootstrap code, and the code which runs on these "monitor cores", will be open-source from the start.
There's a default flow with u-boot SPL, opensbi and u-boot.
There's an alternative flow which includes TianoCore (UEFI).
There's another alternative flow, oreboot.
All the tiny ROM inside the SoC does is fetch the next stage from elsewhere, as per the state of some GPIOs. One of the options is XMODEM on the UART.
DRAM setup and such happen in the next stage, which is e.g. u-boot SPL mentioned above.
>doesn't have the equivalent of the IME
There are some extra tiny cores in the SoC. They are available for you to use; you can implement an IME if you wanna. But by default they're doing nothing.
Proprietary software mostly will require emulation or other similar techniques and WILL be problematic.
Anything else can generally be re-compiled. Several distributions support RISC-V already. Alpine has a total of 11378 aports, of which 850 are marked explicitly as "doesn't build for riscv64".
If you're worried that it won't just slot into your electronics ecosystem with zero friction then it's not for you. It's for a small cadre of early adopters that will suffer the papercuts and then cajole, shame and badger an even more minute group of firmware and kernel developers into making everything work. After enough of this has gone on you'll be able to safely get aboard.
It's not pretty, or fair, but that's how it all actually works.
TFA does say they've been working with Red Hat and Canonical to ensure a good base experience. I'm planning to buy one to run Fedora, knowing that a lot of packages may not be available, but hopefully enough so that I can build my own software on it and ship RISC-V binaries/packages.
Currently I'm using qemu for that, but for someone who doesn't live and breath qemu it's an uphill battle.
This seems to be a better alternative: a complete 8-core RISC-V laptop with an M1 core by SpacemiT – similar performance to a Cortex A55 – starting at about $300 (8 GB RAM/64 GB eMMC, versions with 16 GB RAM and PCIe SSD are also available):
Honestly, I'm not sure another JH7110 device is a good thing for RISC-V publicity. Couldn't they have waited another year?
Since it's framework, there will be a decent bunch of people buying this, that will be very disappointed.
> This Mainboard is extremely compelling, but we want to be clear that in this generation, it is focused primarily on enabling developers, tinkerers, and hobbyists to start testing and creating on RISC-V. The peripheral set and performance aren’t yet competitive with our Intel and AMD-powered Framework Laptop Mainboards
Good that they acknowledge it, but that is putting the performance difference very very lightly.
> DeepComputing is demoing an early prototype of this Mainboard in a Framework Laptop 13 at the RISC-V Summit Europe next week, and we’ll be sharing more as this program progresses.
The important thing is the Framework laptop is getting a third party board even if it is under-powered and unusable for most people. Raspberry Pi could take some of their IPO money and do an ARM board or Qualcomm and not have to worry about the keyboard, case, display and the logistics of moving laptops. Seems like a great way to get your platform in front of developers in a very usable and sustainable package.
I know there's quite a big community that buys almost purely old ThinkPads for their Coreboot support and lack of Intel ME/MSFT Pluton.
Considering the processors in those chips are almost 20 years old now, this RISC-V chip may actually offer comparable performance! Would be interesting to see if that groups gravitates towards this product.
The JH7110 is a superscalar quad-core dual-issue design at 1.2GHz and has 2MB of L2 cache. The Core 2 Duo P8600 from late 2008 (around the same era you can buy Coreboot-able Thinkpads like my old X200) has dual 2.4GHz processors with 3MB of L2 on an OoO-superscalar design. It will probably lose handily, if I had to guess (strictly speaking, perf/watt may be better on the JH7110 due to newer manufacturing processes, but you're not using +15yo thinkpads if you care about that.)
A brief and non-reliable search puts the CPU at around 5W.
Unless I'm missing something, it's a 61Wh battery in the chassis, as they all use the same battery.
Therefore... battery specified as reasonable with an intel or amd cpu, but now using a minimal wattage processor, very long battery life?
My use case is basically showing emacs in a terminal with all the compute running on some other server and it seems plausible this would be rather good at that.
(also Lenovo are presently insisting I install windows on a laptop in order to replace a broken keyboard under warranty which makes me especially receptive to framework's messaging)
> Precursor draws less power than most other FPGAs thanks to the "-1L" variant Xilinx Spartan 7-Series at its heart. (The "L" stands for "low leakage.") That efficiency — combined with a super-low-power Lattice iCE40 UP5K FPGA for deep-sleep system management and a Silicon Labs WF200 with integrated network co-processor for Wi-Fi connectivity — allows Precursor to achieve a standby time measured in days and an active screen time of about five to six hours.
Don't count on it. FPGAs are stupidly expensive, and those vendor-provided devboards are often sold below market price. Anything worth your time is stuck behind a proprietary toolchain. Besides, there just isn't all that much you can realistically do with it. FPGAs shine at realtime processing of huge volumes of data, but you lack the necessary IO in a mobile form factor to do anything meaningful you couldn't already do with a CPU or GPU.
The Precursor is a neat device, but in the end it's essentially a toy. The entire concept hinges on emulating a SoC with the FPGA, but a real SoC can offer an order of magnitude more performance at a tenth the price. It only makes sense if your risk model is a chip vendor putting backdoors in the SoC - but at that point why would you trust the FPGA? It's just moving the goalposts, really.
The reason you would trust and FPGA more than a SoC is because putting in backdoors in an FPGA is much harder. If an FPGA is backdoored to the point where any soft core (whose layout kind be randomized to some degree) is compromised, then that would likely be obvious from the kind of inspections Bunnie has written about: https://www.bunniestudios.com/blog/2023/infra-red-in-situ-ir...
Why is it a toy just because it has an order of magnitude less performance? Computers became powerful enough for most people decades ago. I get much more utility out of a computer that grows capability in other aspects like mobility and power consumption.
This sounds really great, I hope that some day we will get an alternative to x86 that is not ARM.
The new ARM processors like Snapdragon X elite and the M processors from apple sound promising, but I don't know what I should think about the company "ARM".
Though what I find a bit weird is that this board only has 4 cores, as this is intended for developers which probably need to compile many things which could benefit from more cores.
Yes that's for sure, and it's amazing that Framework created this, so people which already have a Framework 13 could easily swap their mainboard.
But slower than a pi 3b+ is a bummer, I'm not the target audience for it, but I think this may be a show stopper for developers who want to port software to risc-v and use it for testing. Or at least it would not be a great experience (apart from all the other problems you'll have with a relatively "new" ISA like risc-v)
Did they mention how much this board will cost when it's released? The BeagleBone guys have a board based on the Micro-Semi / Micro-Chip PolarFire FPGA+U54x4 hard core for about $150, so I wouldn't think this one was too much more expensive than that.
SiFive's U74 cores aren't as inefficient as their earlier cores, but ops per megahertz still isn't on par with high end Intel or ARM cpus.
But if you're going for RISC-V, you're likely interested in it for something other than peak performance.
When I did my time in the RISC-V ecosystem, I thought we would never see catalog parts, so its good to see yet another RV64GC board using a catalog part.
I have a VisionFive2 board with the same SoC, and this is going to really disappoint if you expect laptop grade performance from it. Don't get me wrong, it's impressive for what it is (the fist such SoC that can comfortably and inexpensively run Linux), but in a laptop it will feel glacial, and there are quite a few bugs still in its Linux support. I guess they could clock it a bit higher, but there's only so much you can do to hide performance shortcomings here.
The Mainboard itself is going to be a lot more accessible than you (apparently) think! We'll have more to share on pricing as we progress with DeepComputing on the program.
Is it too late to suggest a cluster of JH7110's on a single board? Each SoC has 2 GbE interfaces, so it could even be two networks to make everything speedier.
> We’re excited to share a preview of a Framework Laptop 13 Mainboard with a new CPU architecture today, and it’s probably not the one you think it is.
RISC-V support in a laptop is cool and the future and all, but I still feel the need to say that I'm getting a Snapdragon X based Framework 13 (maybe 16) the minute it's available.
There's a tiny ROM inside, which just downloads the next stage into SRAM (cpu cache as RAM) from eMMC, SD card or SPI flash, or receives it via UART (XMODEM), according to the state of two GPIOs. This tiny ROM remains accessable, can be dumped and examined.
The next stage is typically u-boot SPL, which then loads a payload of opensbi + u-boot, which in turn loads the device tree and kernel. This is all open source.
There are already alternative boot flows implemented, such as oreboot and tianocore.
I find it really surprising that SiFive doesn't come out with their own edition of this but with the latest cores, rather than a 3rd party with an old u74
Just like Arm, SiFive licenses CPU cores to chip designers, they don't make chips or products using them themselves.
Except expensive eval boards in very low volume, aimed at letting people who are licensing their cores to build a chip get a head start on software development while their own chip is under development -- much like Arm's $10,000 dev boards:
That's my perception as well. TFA is clear this is a partner-driven thing, so I suspect DeepComputing drove most of this (obviously with some collaboration from FW though). Win/win for all involved!
It's great to see partner ecosystem developing! Framework is such an exciting company.
An end goal of targeting RISC-V is better for everyone than targeting ARM or x86. No licensing fees, manufacturers could design their own silicon and be completely royalty free, etc...
Right now we are nowhere near that. RISC-V software and hardware is not very mature at all. But much of this can change very quickly once products launch.
Maybe just the availability and maturity of that particular SoC, the PineTab-V tablet uses the same one. An obvious ARM SoC to use in a laptop is the RK3588.
Hopefully this board can also operate in stand-alone mode outside of the framework shell just like the official framework motherboards because then we'd get the best of both worlds while also being able to benefit from the coolermaster case. Regardless, I'm going to order one when it becomes available.
Also stoked to see the open source CAD files for the shell.