When I learned about devTerm, it only supports rk3566 and raspberry pi cm3. Now I’m happy to find out that it works with RISC-V and cm4.
However, the performance of these chips is still relatively weak, especially now that the rk3588 has been released and the rk3566 has begun to be used in China’s open source handhelds.
I wonder if there will be an opportunity for devTerm to upgrade to these SoCs in a similar fashion to cm4 and RISC-V in the future.
If not, what is the problem.
Well I think maybe the clockworkpi team is just not big enough to handle so many new SoCs. And I’m wondering how much additional performance we can get from these new chips. And the battery lifetime is also need to be considered.
I’d be interested to know if CM4 alternative devices work as well. Such as the Banana Pi BPI-CM4, which has not released yet, and might not for a while, but would be a HUGE step up in power over the A-06 module (4x A-73 + 2x A-53 in the BPI-CM4, vs the 2x A-72 + 4x A-53). It’s not as powerful as the RK3588, but, it is a possibility seeing as it will be the same form factor as the CM4.
Yeah, rk3588 mainline support is in a relatively early stage, but it’s very active – lots of pull requests incoming everyday.
I’m wondering when that board is going to be available. Currently more like vaporware
Power is another concern… after seeing many rk3588 sbcs equipped with 45W power supply. We’ll need to throttle it down to rk3399 level I guess. It will be embarassing if there’s not much performance gain after that
I think “S” means “slim” in Rockchip naming schema. A throttled down version with different max frequency etc. Even so, down the page it lists the power supply is 5V 4A, which is I believe beyond what the DT pmu (axp228) can handle. Not sure about mainboard “version 5” improvements though!
Edit: not relevant. Extra current will be drawn from the batteries instead.
In theory, 3588 will deliver more on the same power budget due to better fab (8nm vs 28nm), and improved core designs. We can avoid busting out pmu by programming the operating points and remove the max performance states – which is already done for A06 for both cpu and gpu.
Anyway, I will get the board once it come out.
Official radxa page: Rock5/CM - Radxa Wiki
That’s 3588S too – good to know actually.
No onboard wifi/bt though – not sure if it’s pin compatible with our SDIO/uart wireless modules.
Edit:
1 x SDIO 3.0
Uh oh. That goes to microsd. kuku.
Edit: and … no onboard audio. Gonna need a much more elaborated EXT.
“IPS” is AXP228+Battery output, which is fed into a step-up regulator of up to 15A. This chip is going to be fine. If we set V_ips=3.7v, we need 5.4A current from the batteries (2.7A each), so the batteries would be fine too – not sure about the overcurrent protection on the battery module – it could cut off power if we push the batteries too hard!
Battery would draw very fast though! Expect some 1.5h battery life if RK3588 kicks in full power.
The main issue I think of with these boards is driver support for the LCD, keyboard, etc. Am I being dumb? Or are these different ARM chips close enough that developing drivers would be easy?
This is maybe counterintuitive, but peripheral drivers are the easy part.
More problematic is the support for the SoC itself, e.g. video codec, gpu, suspend etc. Sure these are available in the vendor supplied kernel (part of what’s called BSP) but these are usually outdated and increasingly hard to use with the current OS userspace.
