Librem 5 Fir Batch

[carlosgonz]

@amosbatto

I’m 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.

Yes something strange happened. Because no make sense.

Thanks.

@amosbatto

There is no point upgrading to the Plus,

I still believe on the plus, in fact FIR GPU could be better for Librem 5 because ship display 720p.
FIR SoC contains massive accelerator, which enable high video camera 60fps recording at low power, it can not be on evergreen soc and a lot more good things.

@amosbatto

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.

FIR L2 cache is 512 but (ECC) and a RAM Bus is much faster as 4000 MT/s than 3200 MT/s(8M,QUAD)Evergreen.

Also Fir ship coprocessor Arm Cortex-M7-768KB,OCRAM(ECC)Shared-as-L3 at 800 MHz than 8M,QUAD,Evergreen Arm Cortex-M3-256KB,OCRAM(NoECC)NoShared at 266 MHz

Also 8M,QUAD,Evergreen L2 cache is 1Mb but (NoECC), also is 1MB due to Leakage on 2D Silicon as 28nm is near to 20nm-2D-full-crash.
Also Fir ship 3D Silicon next-gen as RK3588.

@amosbatto

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.

I said gc7000 to reach real gles3.2 and full desktop-class level also gc7000 still for mobile and embedded.

@amosbatto

GPU ARM Mali-G610 MP4 (4x 256KB L2) Vivante GC7000 UltraLite (2 shaders) Vivante GC7000 Lite (4 shaders)

As far i know GC7L is 16 shares on 2 cores and GC7UL is 8 shares on 1 cores.
So looks like you wrong about.

@amosbatto

I think you are confusing “embedded” with “mobile”. The Librem 5 generally wouldn’t be considered an embedded device. See this article:

Librem5 is Hybrid, as ship Mobile and Embedded capabilities, but to me and just to me the Librem5 will be so-much better on Embedded mode than Mobile mode. As Mobile is not ready yet but GDAWN is very promising as it will ship as first time a Main NEWPIPELINE which will unlock all the system bottlenecks.

@amosbatto

What firmware are you talking about?

So as Librem5 it is Embedded, Gnu PureOS Mobile can be a Firmware install, so this mean
close-all-doors-and-windows and any holes that LEAKAGE Librem 5 full performance into Gnu PureOS Mobile Firmware, so this mean gnu pureos is no more generic or compat to Librem14-or-whatever. So PureOS Mobile will Fully ENGINEERED BIN for L5, so waydroid,flatpak,firefox and a lot more sotware just gone(bye,bye for L5)

@amosbatto

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.

I not talking about opensources blackboxes.

@amosbatto

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.

Kinda

---

I would like brakedown more put i full busy a.t.m on PowerPC.

Thanks for reply my post.

[amosbatto]

@carlosgonz

As far i know GC7L is 16 shares on 2 cores and GC7UL is 8 shares on 1 cores.
So looks like you wrong about.

Yes, I should have written “cores” instead of “shaders”. There are 4 Vega shaders in each Vec4 core. I will fix that in my prior post. I was following sites like this one that describe the GC7000Lite as having “4 shaders”.

@carlosgonz

@amosbatto

but I think Purism has to be conscious of the fact that a 5.3 inch screen is considered very outdated in the current phone market.

Do you mean 5.7 inch?

Yes, you are right. It’s 5.7". I will edit that post.

[carlosgonz]

@amosbatto what about: Librem 5 Fir → Librem 5 Solo

Librem_5_Solo by i.MX 8QuadMax SoCs. ??

[amosbatto]

@carlosgonz

Librem_5_Solo by i.MX 8QuadMax SoCs. ??

Unlike the i.MX 8M series which uses an SPL U-Boot, the i.MX 8 series requires a bunch of blobs to be able to boot, which is why Purism rejected the i.MX 8 QuadMax back in Aug. 2017 when it first announced the Librem 5. The 8 QuadMax also has poor single-core performance compared to the RK3399 which was used in the PinePhine Pro. I don’t see much point in using the 8 QuadMax, since it is a 28nm planar chip released in 2016. Its two Cortex-A72 cores are now 9 generations old, and the four Cortex-A53 cores is tech that ARM first released in 2012.

If your goal is software freedom, the best choice for a Linux phone is probably using the RK3399S (which is what PINE64 used in the PinePhone Pro), but you will need to add an extra chip to handle DisplayPort Alt Mode, because the RK3399S requires the proprietary dptx.bin firmware blob at boot for DisplayPort Alt Mode. You get horrible battery life, two outdated Cortex-A72 cores from 2015 and four Cortex-A53 cores from 2012, and a chip that was released in 2016.

The second freest option is probably the 14nm i.MX 8M Plus, which requires the Synopsys DDR timing training blob at boot. You won’t be able to use the NPU, ISP and hardware video encoding without blobs, but you don’t need the NPU, and the image processing and video encoding can be done in software. You will have to add another chip for DisplayPort Alt Mode and the GC7000 UltraLite GPU is crappy and video out is not good quality.

The next freest option is the RK3588S. Like the i.MX 8M Quad and i.MX 8M Plus, the RK3588S needs to execute a proprietary Synopsys DDR timing training blob at boot. The RK3588S can now use the free Panthor driver for the GPU, but it requires the proprietary mali_csffw.bin firmware, which is loaded from the /lib/firmware/ directory at boot. All the other blobs for the RK3588S have either been replaced or they aren’t necessary. There is now a FOSS driver for the NPU. The proprietary MPP library for hardware video encoding can be replaced with software video encoding, and Collabora is working on free drivers for the ISP and hardware video encode, so software video processing and encoding probably won’t be necessary in the future. Collabora has developed a FOSS version for the Trusted Firmware-A (TF-A), that can replace the proprietary bl31.elf firmware. The HDMI blob in the bl31.elf isn’t necessary if using DisplayPort Alt Mode for video out. The proprietary bl32.bin firmware is not necessary if not using the hardware-enforced security features or encryption keys.

The i.MX 95 is less free than the RK3588S, and I don’t think that any group is working on freeing the drivers for the i.MX 95 like Collabora is doing for the RK3588.

The RK3588S doesn’t have a secondary processor like the Cortex-M4 in the i.MX 8M Quad, that can be used to boot the system and handle a firmware-jail, so RYF certification won’t be possible, but the Librem 5 never got RYF certification, so that ship has already sailed. However, with the RK3588S you get a phone with the processing power to actually function like a modern smartphone. Most people simply won’t buy a phone that can’t shoot decent video. If we want Linux phones to ever break out of their tiny niche and actually be considered a viable alternative to Android and iOS, Linux phones need a good processor.

The other thing is that Purism now has real competition. The Jolla Phone (2026) sold out its first batch of 10,000 units in the first 48 hours that Jolla opened preorders for the phone. If the Liberux NEXX or the Dawndrums Divine D. come to market with the RK3588S, I doubt Purism will be able to sell its next gen Librem phone if it has a low-performance SoC like the i.MX 8M Plus or RK3399S.