Introduce ROCK 5B - ARM Desktop level SBC

#44

Yes, but that’s 2 x Gen x1 and not 1 x Gen2 x2. Both lanes are behind separate Combo PHYs and I don’t think you can let them operate as x2. But if you’re only keen on bandwidth and don’t care about latency and two SSDs are an option you can solve the problem one layer above at the OS with mkfs.btrfs -d raid0 -m raid1.

AFAIK are both PCIe lanes multiplexed with SATA and as such you could get even higher throughput with SATA (6 Gbps vs. 5 GT/s per lane both with inefficient 8b/10b coding).

I really hope Tom/Radxa can confirm this and they will already provide a simple SATA adapter for the M.2 key E slot on Rock 5B…

1 Like
#45

That’s very sad then.
One of the best things (imo) happened to new SOC is PCIe Gen3, and if CM5 will be based on RK3588S - it will loos big part of the it’s power =\

1 Like
#46

I think you are referring to 802.3af PoE (up to 15.5W) and 802.3at (POE+, 25.5W). I have the same question.

@hipboi How much power is drawn by the board itself, how much is reserved for onboard m.2 SSD and Wifi? Is there support for dynamic POE (LLDP, 802.3at)?

#47

Are the PCIe bugs present in RK3399 fixed? I.e.:

  1. BAR limited to 16x1MByte (seems so, but I would like a definitive answer)
  2. ABORTs when probing the bus (see https://lore.kernel.org/linux-pci/CAMdYzYpOFAVq30N+O2gOxXiRtpoHpakFg3LKq3TEZq4S6Y0y0g@mail.gmail.com/)
#48

Are there any R3 code options/distributors for EU customers?

3 Likes
#49

Is there any plan to cooperate with a case manufacturer for making cases?

From what I have seen the pico-ITX form factor is not standardized and to be user friendly the manufacturer needs to at least punch holes in the correct locations for the board ports.

2 Likes
#50

Were they even bugs?

There was no conclusion to that thread.

#51

@hipboi does bottom m.2 PCIe 3.0 x4 slot support PCIe bifurcation?

#52

Sure.

1 Like
#53

Will it come with casing, heat sink and power adapter?

1 Like
#54

https://browser.geekbench.com/v4/cpu/compare/15648450?baseline=16426661

Summary

FGyPfFeVQAE4N8_
FGyPfFeVQAE4N8_.jpg1080×1440 424 KB

#55

@hipboi any idea what power inputs this will support?
For instance would 5V 2/3 A be enough?

#56

I think we should reserve 15W for CPU, 9W for USB 3.0, 5W for USB 2.0,10W peak for SSD. So 65W should be enough for full peripheral at full load.

For normal usage, I think 24W should be enough, 12V/2A.

1 Like
#57

There is no barrel jack on this board so we need to use USB-C with USB PD. Does the USB PD circuitry support QuickCharge 4.0 AKA USB PD 3.0 with flexible power rules or are we stuck with power profiles? If the latter it looks like:

  • Profile 4 for 60W (20V/3A)
  • Profile 5 for 100W (20V/5A) but needs 5A rated cables

For 65W… is there a way to do this with 3A PSUs/cables?

1 Like
#58

Hey guys, I already bought the discount coupon, really hoping this SBC is going to be awesome :slight_smile:

I would like to ask someone from the team:

What’s the situation with GPU/VPU drivers for linux going to be? I have a rockpi 4c and gotta say, there’s no good solution to have a linux desktop with GPU-accelerated video (there’s a patch for armbian with legacy kernel, even that is kind of hacky and the legacy armbian has weird issues with HDMI), i’ve tried like 4 different distros – it’s either old kernel and patched (hence updates held back) packages or current kernel and no VPU support (because the patches, afaik, were a one-time community effort)

Is RK3588 going to be an improvement, VPU-support-wise? I mean, like, set up a distro like armbian or manjaro and have Firefox play 4k youtube accelerated, have mpv/vlc/kodi use the GPU properly, and all that.

Cheers, thanks in advance <3

@hipboi

1 Like
#59

Yes it is confirmed :smiley:

PCIe 3.0 can be bifurcated to 1x4lanes, 2x2lanes, 4x1lane and 1x2lanes + 2x1lane
it can operate with Root Complex and End Point.

Please wait for official confirmation from radxa until rk releases the full specification and what all is utilised by the vendor.

In ARM GPU can only fo 2D and 3D rendering, Video is processes by VPU and sadly there is still not stable and robust support for this in any ARM device yet.

Hopefully with RK35xx series if RK can provided all this in uefi and then push the driver to linux driver then only it can be fully utilized. Till then we have to wait for team to reverse-engineer it which takes years to reach a stable stage.

1 Like
#60

Yeah, that would probably instantly put the RK35xx devices on top of everything in the media box world. So sad to see these powerhouses not living up to the full potential of the hardware bc of the whole closed binary driver package thing. So far it’s either “get an apple tv” or “get a highly vendor-customized android box and hope the firmware is not total nightmare”

#61

@hipboi will the board support dumb 12V power supply like RockPi4? What about dumb 20V?

#62

Sad, as RK3588S may not cut it: yes it has two PCIe 2.0 x1, but I don’t think it would be “sufficient”, say if we want to make a more flexible platform. Really recommend a bigger CM5 based on RK3588 instead, and consider carrier boards in the size of (better be exact for “compatibility”, as Rock 5 is a bit wider at 72 vs. 69.85)

  1. 3.5" HDD (147x101.85), and 2) 5.25" ODD (203x146)

With these carry board sizes, it should have enough space to put “everything” back that was left out because of 5’s limited size (e.g. 2xGbE). Say, one can even create a 8-port “router” if the carry board has PCIe slots (PCIe 3.0 x4 for 2x10GbE, PCIe 2.1 x1 for 4xGbE, onboard 2xGbE => 8-port), or 16-ports NAS (x4 => 8-port HBA, 2x1 => 2*4-port HBA), or anything in between?

And of course, I know if Radxa keeps it “Pi CM4 compatible”, sales could be better perhaps, as I saw applications like TPU enabled surveillance etc. where RK3588S is more than sufficient.

3 Likes
#63

Dumb 12V and 20V PS are supported.

1 Like