The i.MX 8M Quad only supports a maximum of 4GB RAM, whereas the i.MX 8M Plus will support 8GB RAM, but the i.MX 8M Plus has a weaker GPU and VPU than the i.MX 8M Quad and doesn’t support DisplayPort, so you have to pick your poison.
I personally think that Purism should dump NXP, and use the RK3568 (which supports up to 8GB LPDDR4X RAM and a triple display) or the RK3588 (which supports up to 32GB DDR5 RAM) in the Librem 5 v2. Rockchip hasn’t yet published info about these chips at http://opensource.rock-chips.com, so it is hard to know whether they can boot with 100% free software, but the Mali-G52 EE GPU in the RK3568 should be able to run with the free PanFrost Gallium3D driver in Mesa. Unfortunately, it will probably be years before the Mali “Odin” GPU in the RK3588 is supported by a free driver.
Once PINE64 releases the PinePhone 2 with the RK3566, it is going to be hard to justify buying a Librem 5 v2 with the i.MX 8M Plus. At this point, I think it is likely that the PinePhone 2 will be released before the Librem 5 v2, because PINE64 is already prototyping the Quartz64, an SBC based on the RK3566.
On paper this looks a good open source replacement for i.MX 8M Quad. It says: LOW POWER Consumption. Could not find the REAL POWER consumption. (is it more/les than i.MX 8M Quad?).
I don’t know that, but other related documentation might be out soon. Here is link to the another available one: https://doc-en.rvspace.org/Doc_Center/product_brief_visionfive2.html, telling us that at least 5V/3A (as minimum) power supply will be required for the entire StarFive JH7110 SoC (available as VisionFive 2 on Kickstarter).
Nov. 2020 was the last time that a Purism employee confirmed that Fir was still planned. You might want to email Purism support and ask again. If you don’t want the pain of being an early adapter, you will have a long wait for Fir. You might want to consider switching your order from Fir to Evergreen.
If you want a Linux laptop, Purism is offering $799 discount on the Librem 14 for people who switch their Librem 5 orders to the Librem 14.
Given some of the things that the i.MX 8M Plus lacks (like support for DisplayPort and a weaker GPU), I have long wondered if Purism will use one of the future i.MX 9 chips, Rockchip 3566 or simply keep using the i.MX 8M Quad in Fir, but add 1GB more RAM and a lot more eMMC storage.
I would expect that Fir will be announced (or rather available to buy) some time after shipping parity. It doesn’t make sense earlier.
Also Purism hinted they will not announce new products before they are available to avoid the problems of the past.
And I really hope it will have the i.MX 9
Well it doesn’t hurt to ask, and it makes sense to have as much information as possible in order to inform your decision-making.
Given that the i.MX 9 may still to this day be announceware, I don’t know how much Purism will be able to tell you even if they wanted to.
It looks to me that the only announced chip in the i.MX 9 series only has two cores. I have my doubts that a smartphone would want to go in that direction i.e. backwards. I mean, sure, you will get lower power consumption and hence greater time between charges but … At the very least, it will be a difficult marketing sell against the latest 8-core behemoths in your pocket - and who knows what specs by the time Fir might actually be available.
So if Fir wants to use a chip in the i.MX 9 series, it might be waiting for a higher spec chip that has not been announced yet.
Even if that would be the case, at todays currency rate I doubt Purism will be able to offer L5 Fir at a reasonable price thus will be out of reach of the targetgroup that might consider switching to a Linux phone.
I am skeptical about the RISC-V idea.
It is already a challenge to get ARM Linux versions of some applications. And ARM Linux is now relatively popular thanks to Raspberry Pi (sold in millions of units).
So if Librem 5 moves away from ARM and goes to RISC-V, this would make the device even more limiting in terms of available software. There will be close to nothing available from the closed source stuff. And even from the open source stuff one would often need to compile himself/herself and this would be hindering many potential users.
In my opinion one should not de facto rule out 99,999% of closed source software with the choice of hardware.
It is unwise fanatism to deny all closed source apps. At the end of the day the huge progress in software in the last 30 years is mainly thanks to the fact that it is a profitable business that motivates people to learn software engineering and to develop software solutions.
As far as I know, risc-v is fully supported by Debian. Currently versions of Ubuntu and Fedora are readily available for risc-v. Low power versions of risc-v socs are already developed and this process speeds up stimulated by US sanctions (many universities and commercial research institutes in China and Europe are focused on risc-v).
I think that just moves the trust problem somewhere else.
I mean I totally get it from the perspective of China. It makes sense for them to have a chip where they control the backdoors, if there are any. Unfortunately that makes their particular implementations of RISC-V a questionable chip to use in a privacy-focused phone. (How many Librem 5 USA phones could Purism sell under those circumstances?) That doesn’t mean that RISC-V itself cannot be part of the solution.
At the end of the day though there has to be a compelling reason to change from ARM to RISC-V.
Another complication is that it is about more than just the CPU architecture or even the CPU. The whole CPU ecosystem comes into it e.g. in the scenario of a smartphone whether there is adequate open source Linux support for all the ancillary functions in the SoC e.g. as in the existing Librem 5 where the SoC has HDMI support but we can’t use it.
What does this mean?
You will have the OS up and running.
But an application that has been compiled for amd64 or arm is still not going to run. Or am I mistaken?
As an example. For one project with friends, other people chose pCloud for file sharing. pCloud has Win, iOS, Android, and amd64 Linux Versions. So I can use it on an amd64 Debian, but not on Librem 5. I asked them if there is going to be an arm Linux version, and they said yes, but did not committed on a timeline.
So at some point there will be a version for arm Linux, but if Librem 5 moves to RISC-V the problem will start all over again.
ARM Linux currently has some weight thanks to Raspberry Pi.
For sure I am not against moving to RISC-V in the future if it gets popular on the market and there is more software for it. But I think Purism should not be the pioneers there because pioneering too many topics at the same time becomes too much work and too much risk of failure.
Even then I would be waiting for the compelling reason e.g. more open, more auditable, fewer unfixable bugs, vastly improved performance/Watt, … Something. (For clarity, I am only talking about the CPU and/or the CPU architecture and/or the CPU ecosystem - because I’ve assumed that it is popular enough and widely supported enough that getting a software build for the CPU architecture is not too problematic - as per the premise of your text.)
Best solution? Create a new chip/CPU from scratch (not so strange for a smartphone built from scratch), exactly like is doing Apple with their new chips M1.
Definitely. There are just several hundred million small green problems.
Elon Musk should have invested in Purism instead of wasting time with this micro-blog-whatever thingy that was supposed to cost way more than what would be necessary to build a flagship mobile Linux smartphone.
What backdoors are we talking about? (The only proven government backdoors in communication equipment are from US companies). What makes an opensource RISC-V made outside the USA questionable for a privacy-focused phone?
There are already Linux distributions for RISC-V (Ubuntu, Fedora, a.o.). There is always the possibility that specific applications are not supported. Even on Wayland some X-org applications don’t work (yet). Point is that a certain ISA (in this case RISC-V) gets mainstream. This increases support for not so common applications dramatically.
Yes, prudent behavior. I.m.h.o. due to the openness of the RISC-V ISA this maturity will soon be reached.
I’m pretty sure that the government (any government) isn’t going to make that information public. I did write “if there are any”. We know for sure that governments are creating the legislative framework that makes it legal for the government to demand a backdoor. You have to persuade me that a government would create such a legislative framework and then not use it.
I did go on to write specifically: How many Librem 5 USA phones could Purism sell under those circumstances?
And also there is a difference between “outside the USA” and “inside China”. If it were only “outside the USA” that is the condition to be checked then both the Librem 5 and Librem 5 USA would fail already.
So let me answer the question as if you had asked much more specifically: What makes a chip made in China questionable for the Librem 5 (non-USA edition)?
I expect that at some point in my lifetime China will invade Taiwan. Then we will be at war with China, either hot or cold. Chinese chips will either be an attack vector or unavailable - and there is a good chance that Taiwanese chips will be unavailable. So, yes, it is “questionable” (you will note that I did not say “certain”) and we should be looking to diversify. That’s just commonsense.
There is one piece of text that you introduced and which I have not addressed and that is “open source” (referring here to open source implementation of a chip). Yes, if it were truly open source, designed in China, but could be verified and then fabbed elsewhere, that could be OK. It is unclear though whether “open source” will actually apply for any given Chinese implementation.
Is the ISA itself any more open than, for example, ARM?
What are the open or closed features of the RISC-V ISA? or of the ARM ISA?
From what I’ve seen - and I admit I am not familiar with the x86 ISA or either of the above - some thorny areas could be how unimplemented or undocumented instructions are to be handled - or anything else that is left open to the implementer - and perhaps whether the ISA supports the idea of “enclaves” - but most of the issues that I see are elsewhere in the ecosystem i.e. not in the ISA itself.
If you don’t want the topic to drift and die, maybe stop the discussion at this point.
Jan2 is a Putin and China fan and US hater.
So starting a conversation with his presence including not trusting China, China invading Taiwan, and trust in western made electronics will just blow away the topic.
And on the topic I think everything has been said.
Fact: RISC-V can become successful in the future.
Fact: RISC-V is at the moment way less popular than ARM and way less software applications offer RISC-V versions. This CAN change in the FUTURE.
Fact: Being a pioneer means higher risk, more effort, longer time to market and higher money inverstments.
Fact: Purism does not have unlimited ressources at the moment.
So I guess the RISC-V Vs ARM for the future Librem 5 versions is something to look again at in 3 years or something and review again.
