Does someone have latest news to Fir? I mean, are there official statements concerning progress or decisions for Fir or did Purism even stop following this path?
Happy for any hint, because I still wait for Librem delivery (yes, after 9 years).
Does someone have latest news to Fir? I mean, are there official statements concerning progress or decisions for Fir or did Purism even stop following this path?
Happy for any hint, because I still wait for Librem delivery (yes, after 9 years).
Librem 5 Fir orders will be fulfilled using version 2 of the mainboard, which is actively being designed. Progressive milestones are dependent upon funding projections (backwards-compatible mainboard upgrade vs partial mechanical design revision vs full redesign, etc). At this time, Purism is not pursuing NXP as the CPU manufacturer the next revision given available options. There are a lot more exciting details to share as milestones are crossed.
Is @ekuzmenko and @dos involved on L5v2 development?
![]()
Or next Librem 5 v2 it is a chinese clevo phone with a fancy notch???
Thanks
Wow! Letâs go back to Todd Weaverâs announcement of the âFir batchâ. Librem 5 Shipping Announcement â Purism
So:
2, 3, 4 are just âmarketing fluffâ in my opinion.
5, 6 really should be combined. And is unchanged even 6 years after the original shipping window.
1 is now on hold. Clearly they were thinking along the lines of a newer and more energy efficient NXP in the i.MX line at that time. Now we know it will be a completely different chipset. Of course this should be expected given how much time has passed.
Librem 5 Fir cpu it is more powerfull and heat-and-energy efficient then Librem5 Evergreen so everyone know this.
This is a superbad news.
Purism can not demonize FIR and EVERGREEN SOCs as not good, because it is good-and-enough for mobile. The current software for Librem5 still SUPERBAD which make no work as to be. Do you want to prove me that i superwrong @dos ??
It Purism loooking a google mobile socs for next L5v2?
I doubt it.
Speculation: Rockchip might be a decent alternative. The RK3566 or RK3588 look fairly promising in terms of open drivers (Collabora is a key player in panthor + panfrost).
Wild Idea: Since Pine64 is not currently (due to high RAM prices) doing much (Pine64 halts all Linux hardware manufacturing through at least mid-2027 due to shortages â memory crunch forces open-source maker to freeze SBCs, tablets, and phones | Tom's Hardware) , Purism might partner with them. Their HW expertise is very strong. Consider the speed at which they developed the Pinephone and Pinephone Pro hardware vs. what Purism did.
[Edit: I recall that Amos talked a bit about RK3566 and RK3588 in 2024 in this thread. Librem 5 Fir Batch - #160 by amosbatto ]
[Aside: Also, one should note that GrapheneOS is partnering with Motorola to develop a non-Google phone for GrapheneOS. While itâs unlikely, itâs possible that just for their own ease-of-development, they might have more open drivers. Itâs worth keeping an eye on that.]
FirstâŚbe carefull with vendors socs marketingâŚ
To me and just to me RK3566 or RK3588, RK3399S it just not good for Librem5.
The best promise of these SoCs is that they can be Libreboot.
So i guess too that Purism it want to use RK3588S as Divine phone, however as on divine docs it clear they know all obstacles that need to cross to achieving a battery life acceptable for a phone.
Personally i prefer Librem 5 Fir SoC than RKXXXX.
Also i would like a full-statement from @dos @JCS explaining why RKXXXX is full-better on battery-life-performance than i.MX8M Plus. We need to tackle battery life first not cpu marketing horsepower.
Personally i would like a Librem 5 v2 Evergreen soc but refined for battery life, like 5.4â premium display - low-power, premiun desing, premiun electronics, better h..k.s, vapor-chamber, add nfc, more.
We already know a really bad experience on powersave on pinephonepro-rk3399s.
If Purism take divine phone socs why to have 2 phone with same socs? even worse divine phone still not on market to test battery life firstly.
What is it so obsession with CPU power to run plain software? that will turn my phone into a museum piece, because âBad by Designâ.
Also current librem5 evergreen still do not work as to be because missing all STACK software ACCEL. SO WAIT GNU DAWN FOR.
Also release a new more resistence premium battery L5, to support power drain.
Found a way to boot gnu gnulinux more faster..
Also kick-out all the software that is not compliance with gdawn full pipeline acceleration. So firefox is gone, moreâŚ
I like it.
Please add a LED on PCB to turn ON RED if the module cover is open and the phone is ON, Turn Yellow if phone is off but battery inserted.
I suspect NXP told Purism that a 14nm version would be coming, but they neglected to mention that the i.MX 8M Plus would have a much weaker GPU and wouldnât support DisplayPort Alt Mode, so Purism had to stay with the i.MX 8M Quad.
Yes, I suspect the next Librem phone will use the RK3588S, which as has a typical TDP of 8-12W vs 12-15W for the RK3588. I doubt Purism is planning on a phone with 32GB of LPDDR5 RAM, so the RK3588S that supports up to 16GB of LPDDR4X RAM should be enough. The smaller package size and lower power consumption are more important for a phone.
From what I call tell from my web searches, there is no longer much difference between NXP i.MX 8M Quad and Rockchip RK3588 in terms of software freedom. Both the NXP and Rockchip chips require proprietary Synopsys DesignWare blobs to train the DDR PHY at boot, but can run on a FOSS stack after that. Collabora has done a ton of work to get the RK3588 properly supported in mainline Linux. They just added video capture with MIPI CSI for the RK3588 in Linux 7.2 and are now working on hardware H.264 encoding, which will save Purism a lot of work. See Collaboraâs table of what is currently supported on the RK3588.
Given that NXP, Amlogic and Nvidia simply arenât interested in making the kind of chip that Purism needs in its next gen phone, and Allwinner no longer collaborates with the community to support its chips, and Qualcomm requires lots of blobs, it looks to me like the RK3588S is the only viable choice.
Iâd add, so they donât get overtlooked:
I expect that the AI boom will crash and memory prices will return to sane levels. Even if that doesnât happen, the memory companies are massively expanding their factories, so eventually they will catch up with demand. If Purism is going to market the Librem phone as a convergent desktop device, it needs to have a decent amount of RAM and storage. More RAM also helps to be able to run a Waydroid container, so that people can use Android apps.
It would be nice to be able to upgrade the Librem 5 by putting in a new motherboard, but I think Purism has to be conscious of the fact that a 5.3[1] 5.7 inch screen is considered very outdated in the current phone market. According to Gsmarenaâs Phone Finder, only 12 out of the 388 phones with touch screens introduced in 2026 have screens that are 5.7" or smaller. Today, most phone screens are between 6.1 to 6.5 inches. To appeal to the broader market, I think that Purism is going to need a different form factor with a larger screen, which probably means that it will need to change the shape of the motherboard.
Edited to change 5.3" to 5.7" âŠď¸
Seems like the best estimates is the ram price reductions are at least a year away if not longer. Iâm trying to get a 32gb DDR5 SODIMM and will have to fork over $500 for a single stick. Itâs kind of ridiculous.
It would appeal me more if it was just a tad smaller - again, something that would set it a part (but not so small as to be inconvenient or become a different type of device). May be harder to get, though..?
NXP i.MX 8M Plus is 14 nm for long time, do you mean 14nm for NXP i.MX 8M?
The weaker is that i,mx-8m plus gpu is not support 4k, but 2k and the Librem 5 display is 1.3k so this why removed DisplayPort Alt Mode.
Not really as NXP i.MX 8M Plus is 14 nm is so much bettter for Librem 5, so even with NXP i.MX 8M Plus gpu weaker purism stilll can use it a converter to enable DisplayPort Alt Mode to suppport 1080p only for Dock Mode.
Librem 5 line can not use it RKXX SoCs, if so it will evil as opensource-&-vendors, however Purism may enable it RKXX on L4 or L6. But Divine phone will ship it RK3588Soon
NXP i.MX 8M Plus is already supported and mature.
NXP i.MX 8M Plus is already supported and mature. Even more with 2 block-ip-isp
.H.264 is legacy, however NXP i.MX 8M Plus is already supported h.264enc libre-blobs and mature.
Yes and Not.
We do not need any next-gen socs chipsets for L5 because even NXP i.MX 8M it is sufficient; its just that we havent seen L5 working wonder due to a wrong and de-embedded firmware.
However i wanted a Librem 5 Special Edition i.MX 8M Plus+ SOC featuring full gc7000 gpu
Liked
So Librem 5 it is an embedded device, not a desktop device.
Embedded devices requires Embedded software.
So i know too that on last years evils opensources oses stacker is moving the desktop software compatible to Embedded software but i far i know this is not complete, nevertheless the Librem 5 requires special handling due to its architecture to achieve peak performance, and this may even necessitate deviating from the mainline stack.
So Flatpak, Waydroid, Whatever it most for desktop devices, however as librem 5 is so wonder you may use this evil software on dock-mode as extra, but dont expect it to function like a desktop device
So mobian and pmos treat Librem5 as desktop device because is cheaper but enable a superbad experience.
At the end Librem 5 mission was for a gnu gnulinux mainline and a pure gnu system mobile device, not dock mode or etc per se.
What if Purism decides that they need to provide an option of running models privately locally to compete with popular privacy-invasive products? They could probably reuse Alpaca, for example. So, there could be a 32GB RAM version as an option. But having an option likely means that the base version has the same SoC.
Or Librem 5-Special-Edition i.MX 8M+ SoC shipping 14nm and full gc7000 gpu.
Note that i fan for TEXAS Semiconductors, like Texas Instruments and Freescale so do not expect i speak good for Rockchips. But RK3588s it just not good for L5 serie.
Do you mean 5.7 inch?
PainphonePro RK3399s with a TDP 5 watt, do not worked on anything to save battery overall.
Also Opensource killed PP and PPP despite that PP,PPP was Opensource device. So if we continue supporting opensource it will kill the Librem 5 too eventually, in fact opensource already started to kill the librem5, however the L5 it is special device like ghidra level, so it will no so easy for opensource to kill it. Dont think i am delirious because i not.
Excuse me my big-endian english.
Sent from ppc64be.
Iâm talking about September 2019, when Purism announced the Fir batch. I suspect that NXP told Purism that a 14nm die shrink was coming soon for the i.MX 8M Quad, which is why Purism thought it could ship the Fir batch in Q3 2020. Then, NXP changed its mind and decided to design a new 14nm chip with a weak GPU and no DisplayPort Alt Mode support, and that was the i.MX 8M Plus, which was announced in March 2021.
However i wanted a Librem 5 Special Edition i.MX 8M Plus+ SOC featuring full gc7000 gpu
The i.MX 8M Plus has a lot of limitations, as I already explained on the Community FAQ.
There is no point upgrading to the Plus, when you get half the L2 cache and half the GPU cores, but only gets a 20% faster CPU. With the reduced cache, you are probably talking about 10% better CPU performance. Nobody sells a separate GC7000 chip, and most separate GPU chips are not designed for mobile processors. Adding two more chips (a separate GPU chip and a HDMI->DisplayPort converter chip) means redesigning an already crowded circuit board.
We do not need any next-gen socs chipsets for L5 because even NXP i.MX 8M it is sufficient; its just that we havent seen L5 working wonder due to a wrong and de-embedded firmware.
âŚ
So Librem 5 it is an embedded device, not a desktop device.
I think you are confusing âembeddedâ with âmobileâ. The Librem 5 generally wouldnât be considered an embedded device. See this article:
Is an Android smartphone an embedded system? It is small, self-contained, and has limited input/output capabilities. Nevertheless, you can personalize it and download âappsâ to your heartâs content. In that sense, itâs really more of a general purpose computing appliance. So I would say no, an Android phone is not an embedded system.
The Librem 5 is a convergent smartphone that supports both a mobile interface and an external desktop interface (over USB-C).
What firmware are you talking about? If you are talking about firmware-imx-8m, then I doubt there is much that can be improved there. Of course, we will never know since it is proprietary code.
Note that i fan for TEXAS Semiconductors, like Texas Instruments and Freescale so do not expect i speak good for Rockchips. But RK3588s it just not good for L5 serie.
The RK3588S is much better than the i.MX 8M Plus in terms of performance and is arguably better in terms of software freedom due to the work of Collabora and Ideas on Board on the RK3588.
| SoC | Rockchip RK3588 | NXP i.MX 8M Plus 1,2 | NXP i.MX 8M Quad |
|---|---|---|---|
| Process | Samsung 8nm Low Power (8LPP) FinFET | Samsung 14nm LPC (Low Power Consumption) FinFET | TSMC 28nm planar |
| CPU | 4x 2.4GHz Cortex-A76 (2MB L2), 4x 1.8GHz Cortex-A55 (0.5MB L2), 3MB L3; | 4x 1.8Ghz Cortex-A53 (0.5MB L2); 800MHz Cortex-M7 | 4x 1.5Ghz Cortex-A53 (1MB L2); 266MHz Cortex-M4F |
| PassMark multi-core CPU | 4707 | 918 | 831 |
| PassMark single thread CPU | 1533 | 455 | 390 |
| GPU | ARM Mali-G610 MP4 (4x 256KB L2) | Vivante GC7000 UltraLite (2 cores) | Vivante GC7000 Lite (4 cores) |
| Graphics API | Vulkan 1.2, OpenGL ES 3.2, OpenCL 2.2 | Vulkan 1.0, OpenGL ES 3.1, OpenCL 1.2 | Vulkan 1.0, OpenGL ES 3.1, OpenCL 1.2 |
| Shaders | 256 | 8 | 16 |
| Texture fillrate (GT/s) | 32 | 0.8 | 1.6 |
| FP32 performance (GFLOPS) | 512 | 32 | 64 |
| NPU | 6 TOPS (3 cores) | Vivante VIP8000 2.3 TOPS | none |
| ISP | 48MP or 2x 24MP, 8K@30, 4K@60, HDR, 3DNR, 2DNR, LSC, fisheye correction, dehaze | 2x 12MP, 4K@45, HDR, 3A (auto-exposure, auto-focus, auto-white balance), fisheye correction | none |
| Camera interface | MIPI-CSI2 / DCPHY and DVP lanes (6 cameras) | dual 4-lane MIPI-CSI2 (2 cameras) | dual 4-lane MIPI-CSI2 (2 cameras) |
| Display | 8K@60, 4K@120, 2x HDMI 2.1, eDP 1.3, 2x DP 1.4 (USB-C), 2x MIPI-DSI | 4K@30, 1080p@60, HDMI 2.0b, 4-lane MIPI DSI, LVDS | HDMI 2.0a (4K@60), 4-lane MIPI DSI (1080p@60), eDP/DP 1.3 (4K@60) |
| VPU decode | 8K@60 (H.265, VP9), 8K@30 (H.264), 4K@60 (AV1) | 1080p@60 (H.265, H.264, VP9, VP8) | 4K@60 (H.265, VP9), 4K@30 (H.264, VP8) |
| VPU encode | 8K@30 (H.265, H.264) | 1080p@60 (H.265, H.264) | none |
| RAM | RK3588: 32GB LPDDR4x/LPDDR5, RK3588S: 16GB LPDDR4x | 8GB LPDDR4 | 4GB LPDDR4 |
| USB | 3x USB 3.1 Gen1 (5 Gbps), 2x USB 2.0 Host | 2x USB 3.0/2.0 | 2x USB 3.0/2.0 |
| Typical power | 5 - 8 W (active) | 2 - 4 W (active) |
According to PassMark, the RK3588 has 5.1 times the multicore CPU performance and 3.4 times the single core CPU performance of the i.MX 8M Plus. iTayga says that the RK3588S should have the same CPU performance as the RK3588 over short bursts (under 30 seconds), but is 10% - 15% slower over long periods.
With an i.MX 8M Plus, you get the CPU performance of a low-end smartphone from 2015, but it will shoot video even worse than the current Librem 5, because the video will have to be processed by a GPU, which is frankly garbage. With the RK3588, Purism can offer a smartphone with CPU performance that is just a little slower than todayâs low-end smartphone with a Snapdragon 4 Gen 4 processor, but it will be able to take photos and shoot video like a normal smartphone, and it will be able to offer a good external desktop, so you will want to use it as a portable PC.
Historically the i.MX processors were the best chips for embedded or mobile applications in terms of software freedom, because NXP made commits to mainline Linux to get its i.MX processors supported. NXP also provided good documentation without an NDA and a public forum for its i.MX processors.
However, NXP did a poor job of getting the i.MX 8M Quad chip supported in mainline Linux, which is why Purism had to do so many commits to mainline Linux for the chip. NXP also poorly documented how to use its MIPI-CSI2 interface, which is why it took Purism many months of painful trial and error to get the cameras in the Librem 5 to work.
Most of the new features in the i.MX 8M Plus are not supported in mainline Linux, because NXP has not bothered to work on getting them supported. The NPU only works because Ideas on Board paid Tomeu Vizoso to work on it. There are no open source drivers available for the Hantro H1/VC8000E VPU, so all video encoding will have to be done in software using the GPU and/or CPU. Because the GC7000 UltraLite GPU in the Plus is so weak, shooting video will likely be even worse than todayâs Librem 5 with the i.MX 8M Quad.
In contrast, Rockchip has been working with Collabora to get the RK3588 supported in mainline Linux. At this point, the RK3588 is arguably better supported by FOSS drivers than the i.MX 8M Plus, and Collabora reports that it is working to get the ISP and H.264 hardware encoding supported in the RK3588 and lists H.265 encoding as a âTODOâ.
The RK3588S will also be better for thermals and power consumption because it uses less energy and generates less heat per unit of processing power than the i.MX 8M Plus.
Thanks for this 2 big and detailed post. I will check tomorow and reply.
re-thanks.
Despite everything that I mentioned in my prior post, there are some downsides to using the Rockchip RK3588S. Rockchip never published the documentation and source code for the RK3588 at its opensource.rock-chips.com site, and Rockchip recently took down that site at the end of 2025. On December 21, 2024, Jeff Geerling â the most influential reviewer in the Raspberry Pi alternative space â reported on X:
I was told Rockchip no longer considers RK3588 to be an âopen sourceâ chip. In practice, that means Rockchip no longer contributes RK3588 code to open-source projects and they also decided not to sell RK3588 directly to SBC vendors.
See: Rockchip RK3588 No Longer âOpen Sourceâ â What It Means for SBC Community
On the other hand, Rockchip is still updating the code for the RK3588 at Github.
The other downside is that the RK3588S doesnât have a real-time, low-power coprocessor, so it will be necessary for Purism to add an extra processor to the circuit board to boot the RK3588S and pass it the proprietary blob to train the timing on the Synopsis DDR PHY, similar to what Purism does with the Cortex-M4 coprocessor on the i.MX 8M Quad when booting the Librem 5.
Another possibility which I didnât consider is the Allwinner A733 or the upcoming A838, because Allwinner has a bad reputation of not working with the Linux community. However, CNX Software reported in December 2024: " I was told that Allwinner was now back to the âopen-source marketâ with the A733 and upcoming A838 octa-core Cortex-A78/A55 SoCs."
However, Collabora has done so much work on the RK3588/RK3588S, that I really canât see Purism going with any other chip. Collabora appears to have the funding (probably from one of the SBC manufacturers) to keep working on the RK3558, so it will take very little work on Purismâs part to use the chip. It will be much easier than all the work that Purism had to do in order to use the i.MX 8M Quad in the Librem 5.