The options for Purism are:
- Keep the i.MX 8M Quad processor, but offer 4GB RAM and 256GB Flash.
- Switch to the i.MX 8M Plus processor, which is both an upgrade and a downgrade
- Upgrade to the i.MX 95 processor.
- Switch to the Rockchip RK3566
- Use a down-clocked Rockchip RK3588.
- Accept the blobs and use a Qualcomm QCM6490
If Purism wants to sell the Librem 5/Liberty as a convergent PC, it should consider option 1, but it doesn’t address the poor battery life, high heat and poor processing power of the i.MX 8M Quad.
Honestly, the amount of engineering for options 2, 3, 4 and 5 is significant and I doubt that Purism has the financing to attempt them, because I don’t think Purism can ever do another crowd funding campaign due to the bad publicity it received from the crowdfunding of the Librem 5.
Switching to the i.MX 8M Plus would shrink the process node from 28 to 14 nm and reduce the heat, while improving the processing power, but it cuts the GPU processing power in half and eliminates the video out with USB-C DisplayPort Alt Mode (although a separate chip could be added to provide that). I have no idea if the NPU in the Plus can be run without blobs, but the Librem 5 doesn’t need an NPU in my opinion.
I’m really annoyed that none of the chipmakers are bothered to make a competitive Linux chip for phones, except Qualcomm, and you need a bunch of blobs to run the QCM6490. However, I don’t see much difference in terms of freedom between running blobs on separate WiFi and cellular modem chips and running blobs on an integrated SoC like the QCM6490, although there are security/privacy differences because you can cut the power to those chips with hardware kill switches.
The 8nm RK3588 with four Cortex-A76 and four Cortex-A55 cores is a fantastic chip for a Linux laptop, and maybe it can be downclocked to fit within the power envelope of a phone, but that will give you a hot phone with a short battery life, like the PinePhone Pro, which is hardly ideal. The performance of the 22nm RK3566 with four Cortex-A55 cores with a max of 8GB RAM isn’t very impressive, and I don’t see much point in Purism using the chip when PINE64 says that it will use the chip in the PinePhone 2. There is no way that Purism can compete with PINE64’s low prices, and the chip doesn’t have good enough performance to bother, and you need blobs to boot the Rockchip chips.
It kills me that NXP gave the 12nm i.MX 95 six Cortex-A55 cores, but not a single Cortex-A78 core. NXP seems to have purposely crippled the performance of the chip, but it still looks like the best option to me.
Another option is for Purism to go to Fairphone or SHIFT and work out a deal to sell a Linux phone with their hardware containing the QCM6490 chip. Of course, it would have to run blobs on the main processing cores. I can’t see Purism doing that, because it goes against the company’s ethos and brand image, but I would love to see the company do it, because we really do need a Linux phone that can be used as a full Android replacement. The advantage of this option is that Purism could roll it out quickly and it wouldn’t require that much investment.
Given the financial realities of the company, I think that Purism should do both options 1 and 6, because both can be done for limited investment, and I think both would be profitable. Purism would be able to offer one phone for people who want the freest phone possible and a convergent desktop, and it can offer another phone for people who want a decent camera, good processing power, good battery life, etc. to replace their Android phone and don’t mind the blobs.