Suggestion to support Risc-V CPUs and Open Source GPUs

Purism is known to use linux-libre kernel rather than using the Linux kernel in its vanilla form, as Purism believes in complete free and open source technology with no proprietary firmware.

Unfortunately however, the laptops that are produced uses hardware that contains non-free firmware such as the Intel CPUs and the integrated Intel Graphics card. While I get it that Purism has neutralised the ME component found on the CPUs which increases privacy and security, it still doesn’t change the fact that the hardware shipped has non-free firmware on certain hardware, which does goes against Librem’s own philosophy.

I believe that future Librem products should use the Reduced Instruction Set Computer (RISC-V) architecture for its CPUs. I believe that this will save battery power while using free and open source firmware. As discussed with other people, it can be just as powerful with Intel/AMD CPUs as well.

I also believe that Librem should invest in NEOX|V GPU or the RV64X, or some similar graphics cards, which are open source graphics card which afaik does not require any binary blobs.

2 Likes

You may wish to review the many existing topics where RISC-V is discussed as a more open option than Intel or AMD (or ARM).

1 Like

Oh i didn;t realise it was also discussed else where.

Is purism planning to implement the CPU in the near future?

In no way do I speak for Purism but I would guess the answer is “no”.

The purpose of reviewing all the existing discussion is just to see what other community participants have already said about it.

I think everyone can see the privacy / security (/ even functional) weaknesses in existing blackbox core hardware components. If and when there is a realistic better option, I am sure it will get looked at.

1 Like

Thank you very much for your encouragement and enthusiasm - very much appreciated!

I can assure to you that we (Purism) are watching the RISC-V ecosystem very carefully as we agree, I think with everyone here, that RISC-V seems at the moment the most promising path to get as free as possible chips at some point in time.

The challenge for us right is now is the actual hardware implementation, i.e. the silicon. Approaches like the ones mentioned here are announcements by companies developing so called IP blocks (really a bad name since IP stands for ‘intellectual property’ but that’s what these are called in the industry for decades now). These design blocks can and need to be licensed from these vendors and need to get integrated with more of such IP blocks, like the CPU cores themselves, more peripherals like DDR controller, USB/PCI controllers etc. So creating a fully functional SOC (System On Chip) design is a pretty huge enterprise. And if this wouldn’t be enough already the next step isn’t any easier, taking this design into a silicon design for tape out and finally doing a silicon run.

All in all this is a many million $$$ enterprise, just to create one SOC design. I am not super familiar with the silicon industry but from the few bits and pieces I know this is in the range of at least $50 million and upwards. And that’s just the investment you would need to put into it. After that you need to recoup that, i.e. make chips and sell them. It is pretty clear that we would not be able to seriously recoup such an investment by Laptop or phone sales, i.e. we would need to start selling silicon, the chips themselves. And this is a totally different story and market than what we address so far.

So I totally share your sentiment and if we would be in a position to fund such a project, rest assured, we would totally do it, right away. But right now? I’m afraid we can’t unless someone can get us in touch with an investor good for $50m++ just for creating the silicon and some $$m more for making and marketing it.

In some other context I also wrote about my view on the RISC-V ecosystem as I see it right now:

The problem with SiFive (and similar companies) is that they want to become the ‘ARM Ltd.’ of the RISC-V world, i.e. they only want to create, sell and license the IP, not make chips. So someone would have to pickup their design (=license =royalties), add a bunch of peripheral IP blocks (=license =royalties) and tape out silicon. The freedom of the added IP blocks is then a really big question. Just because the ISA is open does not mean anything for a chip, sadly.

The situation, as it occurs to me currently, is pretty mixed. There are some very nice really open and free designs, mainly in collaboration with universities, that use them to teach computer architecture and design. They run on big (=expensive) FPGAs. And then there are the commercial players who license stuff from folks like SiFive (and there are a couple others already) and build their stuff on top. These are mostly companies that already have a track record of making specialized silicon for specific purposes, like Western Digital is making controllers for their harddisks and SSDs. They are just replacing the ARM cores with cheaper RISC-V cores and the resulting designs are not free nor are these chips available on the general market.

There are only very few by now that make general purpose RISC-V CPUs and AFAIK all of them are more in the microcontroller range, i.e. no DDR RAM interface, no fast bus systems (be it PCIe or USB) etc. The ‘biggest’ I know so far is still the Kendryte K210 (IIRC) which is a pretty nice AI enhanced controller, pretty open but not usable as a Linux host, more for special purpose applications like an IP camera or smart speaker or such kind of stuff.

But I think this situation will change, I just can not fully predict when. One big driving push is USA trade war with China and cutting off large companies like Huawei (and potentially ZTE and more) from USA IP. So far China spent little development effort into own CPU IP, I guess simply because it would be hard to sell. They also did not have the silicon production capabilities, TSMC is in Taiwan and in mainland China they only have silicon production for larger structure sizes, good for microcontrollers and flash chips but not so much for large complex designs 14nm and down.

Now with really huge companies like Huawei cut off from USA IP they feel a huge need for China made CPUs, IP and silicon. They can not license ARM because this could put them into the same situation in some years again and they need to have their own production to be independent. And that’s exactly what they are massively ramping up now. And if China is ramping up something then it means it is operational within a year or two, not a decade later.

So I expect to see first announcements or at least rumors from China within about a year and a bit and first silicon about a year and a bit later. I am very sure it will happen, it’s just a matter of time.

The next interesting question is, what it will be what they make? They are recognizing more and more that they have a huge market directly at their door step so it may very well be that the first products will go massively into the Chinese and Asian markets, not so much to the west. This could mean that even if there will be great chips there, that we may not get them. And eventually we might not even want to because all code, documentation and support could be Chinese only, including questionable SDKs with blobs all over the place and ignoring free software licenses. In their markets they don’t have to care, but we have to in ours.

It will be interesting to watch how western companies like ARM, Qualcomm etc. will react to this. One reaction we could already observe, ARM is, first time ever, giving away licenses for some parts of their IP for free under certain circumstance. There are quite some possible outcomes from this ‘war’ taking place right now and most are not very favorable. Let’s hope openness, freedom and good terms prevail.

Cheers
nicole

13 Likes

are they just rumors or is nVidia looking to buy ARM ?

What about the power 9 CPUs though? I heard that IBM was trying to release it for the desktop market? Would that be possible to implement it inside the Librem laptops? It runs on free and open source firmware?

Edit: Here is a link to a video that talks about this

HiFive seems to have no interest in producing SoC’s that can be used by free software. I haven’t seen anything out of China that looks usable in the next couple years (but that situation may change quickly as @nicole.faerber points out).

The best hope that we have for a usable RISC-V processor in the next couple years is the Chassis SoC on the i.MX platform being created by NXP. You can read my assessment here:

Currently the POWER9 chips are too expensive for the desktop PC market, and they suck way too much power to be used in laptops. For years, I have heard that IBM would offer cheaper POWER chips for the desktop PC market, but it has never happened and I doubt it ever will. IBM has walked away from the PC market several times because it doesn’t have good margins.

The best hope for a truly free PC in the near future in my opinion is to build a 2-in-1 tablet/laptop based on the RK3399 (or maybe the i.MX 8M Quad or Plus), which is underpowered, but available today. The RK3399 can now be booted using only free software. The future RK3588 also looks very promising, but it remains to be seen whether it can be booted with free software and whether the Lima drivers can be extended to support its new “Natt” GPU.

See this thread:
https://forums.puri.sm/t/freedom-computer-manufactures/10151

3 Likes

Speaking of China, there was this post on Phoronix last month which was kind of interesting:
https://www.phoronix.com/scan.php?page=news_item&px=Alibaba-XT910-RISC-V-Speed

12nm design, 16 RV64GCV cores and clock speeds up to 2.5Ghz sounds pretty good. It is aimed at running Linux. Some more information in this paper. This is only about the processor, though, it’s not clear if it’s part of a SoC and if so, whether other parts of the SoC will be free.

Regarding availability:

By the end of 2022, a total volume of 15 million units is expected

5 Likes

apologies, it has been a while, our team has been very busy. quick back of envelope figures: DDR4 PHY, $1m per 32 bit interface. PCIe around USD 250k, USB3 likely around the same. 28nm production mask charges around USD $1m and it doubles for each half-node drop ($2m for 22nm)

the solution to the cost of doing an ASIC is: partnerships. and multiple markets. typical SoCs in no way are done for just one purpose (unless there are customers willing to order QTY 5 million for e.g. TV STBs) so they are designed for 10 to 15 different markets where multiple 50k and 100k orders from completely different OEMs will cover the NREs (Non-Recurring Expenses).

so it’s doable: there just has to be collaboration rather than competition.

indeed. and with NVIDIA hell bent on buying ARM, against the wishes and advice of its Founder, after Softbank already jacked up royalties by USD 1.50 (the Allwinner A64 only sells for USD 4 including the PMIC) you can see pretty clearly where that’s going.

hey, if they want to fight amongst themselves, let them see how far it gets them. i have no problem approaching VCs with a story that, based on how things went with Imagination Technologies buying then fire-selling MIPS, explains why there’s a real opportunity to do things differently.

as someone else mentioned, the IBM POWER9 is high end. it’s 200W per processor. desktop systems are available in the form of the TALOS-II and hoo-boy do they do the job. yes, really, people do buy them as a desktop unit.

LibreSOC http://libre-soc.org which is the project i am heading is going after the area that IBM can’t. the 2.5 to 3.5 watts “Pi” style market, a la Allwinner A64 and Rockchip RK3399. except, due to a request from a customer we want to include ECC LPDDR4/DDR4. you will have seen Linus’s latest rant at Intel for holding things back, there, propagating Rowhammer.

LibreSOC is a hybrid 3D GPU-VPU-CPU. Similar to how the Sony PS3 Cell processor worked, except rather than dedicated custom sub-processors that require special software, we are developing, under the watchful eye of the OpenPOWER Foundation, a new 3D and Video augmentation of the OpenPOWER ISA, pretty similar in concept to the RV64X initiative except full SMP capable. In other words rather than shipping the Shader GPU binary off to a separate set of cores it gets executed on the main core(s), right there, right then, just like any other userspace multithreaded application.

that drastically simplifies driver development, allows direct single-step debugging from the application straight into the Vulkan driver, cutting out the insanity and reducing 3D execution latency in the process (queues of 100,000 tasks to foreign arch GPUs is not uncommon)

the ISA augmentation on its own is a huge task so if anyone would like to help here is the starting point
https://libre-soc.org/openpower/sv/

the other fascinating thing about what we are doing is that if both PCIe host and client as well as USB3 and HDMI are included in the same SoC, then it can not only be the basis of a Pi style SBC it can be the basis of a 3D Graphics Card and a DisplayLink-style USB Video adapter. this is what i meant by intelligently targetting different areas, amortising the NREs across multiple customers.

2 Likes