Introduce ROCK 5A - Gift for the year of rabbit

Only with an USB3 hub needing additional PCB space. Also you should better make yourself familiar with reality before posting such weird stuff.

RK3588S is a castrated mess compared to RK3588 or even RK3568 if it’s about I/O. And 2.5GbE NICs are PCIe attached while almost everything is pinmuxed on RK3588S:

There’s only two Combo PIPE PHYs available sharing PCIe Gen2, USB3 and SATA.

1 Like

The Rock 5A should just about squeeze into a pi-top CEED.
The Rock 3A would just squeeze into the pi-top [3] if it weren’t for a slight difference in position or size of the full size HDMI connector compared to the RPi3B+.
So if the Rock 5C follows the RPi3B+ layout and uses the same HDMI alignment/connector part as on the Waveshare CM4 to RPi3B+ adapter/carrier board then the Rock 5C would indeed just squeeze into the pi-top [3].

I’m not a motherboard or cpu engineer. I’m just sharing what I would like to see in a new SBC, not a faster raspberry pi clone. Raspberry Pi has great software support, horrible port selection/position. Instead of copying the good things about raspberry pi, this board just copies the bad things, IMO.


The rk3588s is supposed to be a cut down cheaper version of the rk3588 with the exact same cpu/gpu/npu.
Its not a new SBC as the core is exactly the same and prob maybe should of been even more minimal as the choice of cost effective A76/A55/MaliG610 and the full blown rk3588 could of had more cost difference.
Rock5a is a $30 cheaper cutdown version of a Rock5b.
Its like Raspberry in a ModelA and ModelB where the ModelA is a cutdown cheaper version of the ModelB approximately.

The bottom components are very low profile and prob would fit in most cases but its still a look alike as the gpio pin mux is likely not the same and they even put on a hardware switch than say just a jumper so a case can be easily modified with a wired switch somewhere, so yeah in reality a look-a-like Pi4 is really pointless apart from the concept that Pi fans expect it that way.

1 Like

Now this is the SBC i have been waiting for. The specs are very attractive. Used radxa SBC-s before and i was very satisfied with them. I ordered one 5A cuz i want to use one :yum:
The second thought is that which Linux OS will be usable first.
Third thought: Really wonder how the hardware will perform.
I am very excited and can hardly wait to het my hands on ot :yum::wink:

This thing is obviously designed as RPi 4B drop-in replacement hardware-wise, isn’t it? Judging by the pictures there’s only two problems:

  • SoC position differs. But not that much so with some copper shims or a thermal pad it should be possible to even make use of those metal RPi 4 enclosures that passively dissipate the heat out of the enclosure via a huge Aluminium block reaching down to the SoC. TBC
  • the position of the SD card slot on the bottom looks off by a little, not perfectly centered

So if that was the design idea how should Radxa not copy the position of each and every connector as closely as possible?

As for ‘the good things’ or ‘great software support’ as you call it… what should Radxa do here? Today and within the next years there’s just a smelly Rockchip BSP kernel forward ported since ages containing lots of unresolved bugs and vulnerabilities making use of all the hardware features.

And then there’s the ongoing effort by volunteers and Collabora (a software contractor in the embedded world) to support as much of the hardware as possible with ‘latest and greatest’ Linux mainline kernel. Let’s see in half a decade what has happened…

The distribution (Arch, Debian, Manjaro, Ubuntu, whatever) doesn’t matter since the only part that matters is the kernel and the accompanying libs (for e.g. graphics/video acceleration). And this is the same on every OS image that will fly around for Rock 5A (or to be more precise: for any RK3588/RK3588S device since they’re all the same software-wise) today and within the next months/years.


The Rockchip BSP is basically the Android kernel which has has most focus from Rockchip.
The RK3588/s likely due to the size of the herd and number of vendors currently offering boards it seems from previous boards it has picked up more momentum, as yeah mainlining can be a painfully slow process.*

A whole rake of new submissions just went in for Linux Next and prob the only thing is Mesa haven’t really seemed to pick up on this and could be a bottleneck (MaliG610) due to some politics and rumours of a rift due to ‘gaslamping’ … dunno the full story.
Things seem to be moving faster than other previous SoCs.

Point taken :slight_smile:

I bought rock 3c to try my luck on Pi-Top[3].

1 Like

I’m looking forward to seeing the schematic. Particularly interested to see all of the I2S1 8 channel lines broken out !

Interesting board as direct rpi replacement, but 5B is just way better and not much more expensive, of course physically bigger, this was first thought.
On the other hand this board can replace rpi in my quad sata hat and probably then I can sell that rpi for about same price as they are rather high now. The only question is stability, when it will be good time to make such switch. It would be interesting upgrade, but of course same thing build on Rock5B would be far much better (and bigger).
Is m.2 slot can be used with radxa sata e card? This would add one another sata port to quad sata kit. Or maybe there is something to connect small nvme? just more storage options when wifi is not needed :slight_smile:

Rpi 4B 8GB was originally priced as $75, Rock5A 8GB is $119 (and now for $89+$5)
Direct competitor - Cool Pi 4B 4GB is about $142 (with shipping), built in wifi6 instead of m.2. Some unknown manufacturer. For now CoolPi is rather just bad choice (and more expensive).


Same M.2 Key E slot on the board as 5B, same Naneng Combo PIPE PHY in RK3588S as in RK3588 -> should work with similar DT overlay as on 5B to switch between PCIe Gen2 and SATA (device address might need to be adjusted between RK3588S and RK3588).

Thanks for this information,
That adds one additional SATA port or two multiplexed, making my quad sata nas even more flexible.


Rockchip’s current SATA implementation is confirmed to support SATA port multipliers even with FIS-based switching so combined with an el cheapo JMB575 you can connect five SATA devices to the single SATA adapter in the M.2 slot.

And honestly I would rather put all my disks behind such a JMB575 PM than using any storage peripherals behind an USB3A receptacle (probably the worst connector ever invented): Quad SATA Hat Disconnects :slight_smile:

1 Like

I’m not complaining about their half backwards mismatched android kernel, this was expected, I’m complaining about their mediocre creativity when it came to board design. All the RPI cases I have will not work with this board unless I choose to not use the emmc. That’s the problem I have with my Rock 3A, along with the m.2 port.
I’m just saying they don’t have the software going for it, so why copy the horrible layout of the raspberry pi when there isnt any benefits?

Did you desolder the bottom M.2 thingy on Rock 3A? If not why are you talking about the eMMC being in the way? This is how 3A looks like (and I fully agree that trying to mimic RPi form factor with 3A is somewhat laughable):

The blocking M.2 thing doesn’t exist on 5A simply because RK3588S is so crippled wrt I/O compared to RK3568 on the 3A.

I was really glad the look-a-like Pi concept for Pi fans wasn’t employed on the Rock5b as really the 3a should of been similar as it does have great I/O for that price point.
The Mali G52Le or whatever it is, is a cut down G52 and not that great as really that SoC is about great budget I/O and so should of been able to mount a full length 2280 card than some convoluted hat affair that is needed to purely give a passing resemblace to a Pi.

I have got the OrangePi5 and its a great little desktop board with true entry level desktop performance, where if I/O is your thing you probably should go for cheaper IO rich RK3568 or more expensive RK3588.
I have ordered a R3 code as seem to be collecting RK3588/s as that Cpu/Gpu/Npu is so sweet and absolutely destroys a Pi4 by a very big margin.
The A type boards are just a minimalist version of their bigger B type siblings and probably should be even more minimal to give a bigger cost difference as in the 80/20 rule that CPU/GPU/NPU likely could fulfil most common uses.

There are a couple of things that continue to bemuse.
The Pi look-a-like as why.
Consumer grade PD power than a simpler fixed rail 12v that would be really handy to also have on gpio.
Also mobo switches that top of the range x86 style mobo just have cheap and cheerful jumpers because it gives the flexability of wired switches.

I think the Pi look-a-like has more effect on the Rock3a as the Rock5a prob has a more common use as a general purpose desktop whilst the great IO of the Rock3a has much of its cost effectiveness negated because of a Hat that is only needed because it is a Pi look-a-like.

Its not that much of a deal on the 5a but why Pi look-a-like is a complete bemusement to me especially when it doesn’t have a single advantage if its not 100% pi format compatible?

Does the RK3588s include CANbus support like its big brother 3588? And is CAN exposed on the GPIO like it is on the 5B?

If so this could be a fabulous board for 3d printer. Removes the need for the USB-CAN interface board, saving space and cost. More so if you start shipping them while raspberry is still supply constrained.

Is it not the other way round and its the rk3588s that gets the bus that can not be named?

the addition of three CAN bus

CANbus is definitely present on RK3588, not just an “addition” on the RK3588s. Radxa calls out assignments on the 5B GPIO pin out as pins 3&5 (CAN M1) and 12&35 (CAN M0). It’s not clear that any Linux disto has driver support for it as the drivers do not appear to be loaded in Radxa’s Linux.

That said - my hope is that Radxa exposes CAN the same way (or similarly) on the 5A as the form factor is more easily installed in existing 3D printer designs without having to create a new mount.