A new RISC-V Mainboard from DeepComputing
frame.work
frame.work
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.
And I totally agree; this is the perfect way to try out something like this.
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.
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.
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]
[0] https://liliputing.com/lichee-console-4a-is-a-mini-laptop-with-a-risc-v-processor-and-7-inch-display/
[1] https://www.cnx-software.com/2024/06/19/sipeed-lichee-book-4a-14-inch-modular-linux-laptop-launched-with-th1520-quad-core-risc-v-processor/
[2] https://store.deepcomputing.io/products/dc-roma-riscv-laptop-ii-with-octa-core-cpu
[3] https://store.deepcomputing.io/products/risc-v-development-board-spacemit-muse-pi
[4] https://www.spacemit.com/spacemit-muse/
[5] https://www.spacemit.com/spacemit-muse-pi/
[6] https://www.youtube.com/watch?v=vzod1yUzPhc
[7] https://wiki.banana-pi.org/Banana_Pi_BPI-F3
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.
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.
Thanks!
> 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.
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.
[0]: Twitter: <http://x.com/SipeedIO/status/1801460880481169735>
Otherwise we could call an arduino an sbc
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...
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.
It would make far more sense to go the traditional Raspberry Pi cluster way, either as discrete nodes[1], daughterboards[2], or as blades[3].
[0]: https://deepcomputing.io/a-risc-v-world-first-independently-...
[1]: https://www.raspberrypi.com/tutorials/cluster-raspberry-pi-t...
[2]: https://www.youtube.com/watch?v=ecdm3oA-QdQ, https://www.youtube.com/watch?v=h6zt8KeXFdA
[0] https://deskpi.com/products/deskpi-super6c-raspberry-pi-cm4-...
Or was, apparently? Seems to be out of stock and there's no restock or backorder date shown.
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).
https://www.starfivetech.com/en/site/boards
https://pine64.com/product-category/star64/
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.
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.
I thought I saw someone else publish specmarks to corroborate this assertion, but i can't find it right now.
*: https://www.phoronix.com/review/visionfive2-riscv-benchmarks
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.
RISC-V enables the best processors.
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.
That's the point I was making.
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 :)
Nevermind them not being startups, neither SiFive nor ARM make chips.
As you do not actually make chips, you are also not their client.
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.
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.
> 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.
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).
Can you please elaborate? I'm a programmer with a linux framework laptop (NixOS specifically).
Benchmarks of the CPU in question: https://www.phoronix.com/review/visionfive2-riscv-benchmarks...
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.
The chip there (spacemiT K1) is available in Banana Pi BPi-F3, which is already shipping. It is RVA22 with the ratified V extension.
A laptop with JH7110 makes little sense today.
0. https://linuxgizmos.com/musebook-riscv-v-laptop-with-spacemi...
Yeah. spacemiT K1 is a new chip, 8core, RVA22 with Vector extension.
JH7110 is old in contrast. Was first seen in VisionFive 2 in January 2023. Only 4 cores, slower, and RVA20 without Vector.
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%.
[1]: https://deepcomputing.io/a-risc-v-world-first-independently-...
[2]: https://linustechtips.com/topic/1496701-world%E2%80%99s-firs...
[3]: https://community.frame.work/t/introducing-a-new-risc-v-main...
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.
https://www.cnx-software.com/2022/07/02/roma-linux-laptop-to...
Weird they didn't use the chip from the DC ROMA II - though that laptop is a rebadge of a different laptop from the chip manufacturer itself
https://www.cnx-software.com/2024/04/30/muse-book-laptop-spa...
I'm impressed they managed to get press releases from framework and canonical in the space of a week bc they seem like a shell company
https://canonical.com/blog/worlds-first-risc-v-laptop-gets-a...
This 2015 Macbook Pro is getting rather dated.
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 :)
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.
Framework does sell factory seconds for as low as $500 right now, though they need a few more components to be functional.
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).
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.
Please make more weird partnerships.
I'd love to have a pen tablet color E-Ink screen with a mechanical keyboard on a 13 inch laptop.
It is a great compromise between the readability of e-ink and the performance of LCD.
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.
But you could say the same thing about ARM in the 90's/early 2000's. 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.
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.
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!
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.
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.
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?
We're using this program with DeepComputing to help develop our Mainboard documentation further as well.
https://github.com/FrameworkComputer/Framework-Laptop-13/tre...
It would be a really interesting option as a Framework mainboard. Unfortunately, Framework 13" has rather tight constraints on mainboard geometry.
Relevant links:
https://turingpi.com/product/turing-rk1/?attribute_ram=16+GB
https://lore.kernel.org/lkml/20240612-6-10-rocket-v1-0-060e4...
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.
I think this is a crucial step on the way to getting an actually good experience on RISC-V based platforms.
Currently I'm using qemu for that, but for someone who doesn't live and breath qemu it's an uphill battle.
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.
(I know you can use a virtual machine for that)
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".
Ironically, with JH7110, there isn't much driver development left to do.
Most of it is already upstreamed[0].
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.
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.
Already is.
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.
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
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.
Current build process status is available[1].
>IIRC even a port of Fedora Linux.
Arch, too. I run Arch on my VisionFive 2.
0. https://buildd.debian.org/stats/
1. https://buildd.debian.org/status/architecture.php?a=riscv64&...
It's not pretty, or fair, but that's how it all actually works.
JH7110 is RVA20 w/o vector.
> 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.
I'll definitely check it out.
Hard agree. Not with spacemiT K1 (8 core, faster, RVA22 + Vector) already in devices on-sale.
I have had a JH7110 board (VisionFive 2) since January 2023. It was nice at the time. Now it's aging, with only RVA20 and no Vector extension.
RISC-V is moving really fast.
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)
edit: there's a vesa mount case for these things at https://frame.work/gb/en/products/cooler-master-mainboard-ca...
Similar to the AMD/Xilinx Zynq family of boards ($150 to $3K), https://www.xilinx.com/products/boards-and-kits/device-famil...
Or Bunnie Huang's Precursor dev kit for RISC-V with handheld kb/display ($600), https://www.crowdsupply.com/sutajio-kosagi/precursor
> 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.
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.
AMD/Xilinx will use US-mainland TSMC fab for some FPGAs.
> you lack the necessary IO in a mobile form factor
Framework has pluggable I/O expansion modules.
Framework's IO modules are just USB-C. That's not really flexible or fast enough to really make a high-speed FPGA shine.
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.
Edit: looks like Jeff Geerling tested this same SOC in the vision 5, TLDR; slightly slower than a pi 3B+
https://www.jeffgeerling.com/blog/2023/risc-v-business-testi...
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)
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.
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.
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:
https://www.digikey.com/en/products/detail/arm/V2M-JUNO-0317...
It's a way for other companies to piggyback off of Framework rather than the other way around, while strengthening the Framework ecosystem.
It's great to see partner ecosystem developing! Framework is such an exciting company.
Yup, I've got the RK3588 module preordered for my Reform and I'm looking forward to it. Would be neat to see an RK3588 board for the Framework too.
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.