Loomio
Wed 2 May 2018 11:35PM

Partnership: Purism Librem laptops?

J JC Public Seen by 299

What is the feasibility of cooperating with Purism to incorporate RISC-V chips on their Librem line of laptops? RISC-V is already on Purism's radar:

PURISM:
"Intel, AMD, and ARM seem to suffer from the same issues that proprietary software suffers from: a lack of transparency that results in an unethical design which shifts us further away from an ethical society. RISC-V is something we are closely following in the hopes that it can create a future where processor hardware can be as ethical as Free Software—meaning that the user is in control of their own hardware and software, not the developer.

Purism, as a Social Purposes Corporation, will continue to advance along the best paths possible to offer high-end hardware that is as secure as possible, in alignment with our strict philosophy of ethical computing."

Source: https://puri.sm/posts/meltdown-spectre-and-the-future-of-secure-hardware/

A

axd Fri 22 Jun 2018 2:22PM

I think this is an incredible proposition. Don't other people care about having a workable laptop based on a libre ISA?

J

JC Sat 23 Jun 2018 5:41AM

During a teleconference with Purism, it is clear that RISC-V definitely remains a consideration, however they are easing into it by first approaching product applications that do not necessitate video/audio processing. Many people care about making the "most libre" computing platform possible, but we must walk before we can run. At any rate, the future looks bright for consumer adoption of the RISC-V ISA.

L

lkcl Sat 23 Jun 2018 8:13AM

yes, RISC-V does not have SIMD or Bit-manipulation, those are extensions under development and without them it would be necessary to use hard macros for video and audio processing, or to have external ICs (external GPU) that took care of it. if you speak with purism again do mention that i am looking for sponsors for a Video Extension and also a 3D GPU Extension, both based on RISC-V.

NF

Nicole Faerber Sun 24 Jun 2018 2:40PM

What we (Purism) would need first before a GPU or special support for multi media (like with SIMD) would be general purpose fast interfaces like PCIe and/or USB (preferably USB3) to attach things like storage, video or more network interfaces. So if there is is a chance to influence the development roadmap of future silicon implementations I would suggest to add PCIe with as many lanes as possible. This would enable adding peripherals via PCIe as one sees fit - SATA controllers, USB controllers, graphics controllers etc.
Yes, this will limit the use cases of such a CPU to either small servers, desktops or laptops. But to get off the ground with RISC-V consumer products I thinks this would be a good first step before targeting low power devices for which more integrated peripherals would be desirable.

L

lkcl Sun 24 Jun 2018 2:56PM

hi nicole, i've been trying to reach you! your email came through absolutely fine but i did not receive a reply about FOSDEM, and helping you with component sourcing for the phone.

a PCIe PHY is available here: https://github.com/enjoy-digital/litepcie/blob/master/litepcie/phy/s7pciephy.py

that knocks at least a minimum of USD $100,000 off the development cost of a chip. i've also tracked down SATA (libre), a USB3 PHY (can be made libre for around USD $300,000 - including doing the PHY's layout).

these numbers however are nothing compared to the NREs associated with doing the layout ($USD 2m+) and paying for the masks (USD $1.5m for 28nm and escalating exponentially).

basically these are damned expensive things to develop, so you better have a damn good justification for throwing money at them, such as a market that really, REALLY wants the resultant chip, and i don't mean a couple of hundred thousand, i mean a million minimum and preferably 10 million and above.

and i really don't see anyone looking to buy RISC-V processors that have been specifically designed for desktop markets right now, when the OSes are nowhere near complete.

on the other hand i do happen to have one anonymous sponsor that wants a specialist chip made where they will do the OS entirely themselves. they don't want to sell chips: they just want to sell "product with chip in it".

so this is why i am leveraging existing opportunities - the IIT Madras Shakti Group - to work with them to get something done. if purism would like to help by being one of the commercial sponsors, such that a chip ends up existing such that they can sell products based around it, i'm happy to hear from you. if that is of interest, here are the existing scenarios:

http://libre-riscv.org/shakti/m_class/pinouts/

one of them is already a laptop / netbook / chromebook: if you believe there's anything missing from that i'm happy to talk and add it to the pinouts (as long as it's digital and not high-speed. PCIe, SATA, USB3, these are all off-limits for this particular chip, in order to keep cost and risk right down). it's designed for target pricing in volume around the $4 mark (NOT $25 and NOT $100 to $200 like you get with intel chips), and it's designed for a maximum of a 3 Watt power budget (NOT 15 to 150 like you get with intel chips).

USB2 however is not a problem, neither is SD/MMC (and we may also have eMMC in time, as well - all Libre). i have three USB2-ULPI interfaces on the pinmux: these may move to straight USB2 PHYs (2-wire instead of 8 ). just have to see how it goes with tracking down a Libre USB2 PHY.