Introduce ROCK 5B - ARM Desktop level SBC

#148

Business model with development boards is to provide hardware and community will sponsor most of the support. In a couple of years support could probably get to decent levels, but future is always hard to predict … HW vendors design, sell hardware and ship some generic OS with (slightly modified) private HW interface from Rockchip SDK.

Which is good enough to show what HW can do and you might get some support within from a vendor & Rockchip but it is far away from what you might expect - you get what you pay. Do you support at least productive groups that does software support? If not, start with R&D and code maintenance on your own … No, it’s not cheap nor simple but this is how it works.

tl;dr;
It depends on how much you and community at large will invest into software support.

1 Like
#149

The short answer is no. If you expect to work out of the box, then come in a year or a year and a half, and even that is not a fact.

If you want an android that works normally out of the box, then also wait a year, or even better, wait for an android console from Ugoos and buy it in six months. The latter option is also good because there are many times more owners of such consoles. These are the realities of arm boards.
p.s.
If you need android for games and videos, you need powerful hardware (more powerful than AL992 and NV Shield). Buy a used Asus Rog or Samsung smartphone on a Qualcomm chip and an hdmi adapter.

#150

Its sort of strange how often things progress as often there is some sort of BSP image that is manufacturer hack collection often with specific blobs that work for evaluation and can be part of bespoke product.

You end up with a community hacking away at removing blobs for a working mainline kernel version, where often initially the more functional image is the maintained BSP whilst efforts try to go mainline.
Hacks, drivers and functionality often comes from many sources especially commercial adoption as WolfQwerty says as methods maybe adopted from the likes of Ugoos if they employ a RK3588 and the more popular a SoC the bigger the herd is and often faster moving.

I often think aiming 1st at a static release style distro such as debian 1st often has problems where backports can be problematic but give a base of a release to aim at whilst maybe a rolling release of the latest and greatest such as Arch has advantages but is rolling and at other times working with a static release is easier to co-ordinate.

Much of that 1st stage is just purely transitory a collection of hacks and fudges to get as near a working image as possible whilst nobody really wanting such an image as an end goal.
The RK3588 has strong parallels to the RK3399 and likely will follow a similar route but have an advantage much is similar to what has gone before.
One of the biggest waits for the RK3399 was opensource non blob GPU drivers and Collabora are already on that one and its purely the incremental aditions that Valhall add rather than starting from scratch with Panfrost as what Collabora do with reverse engineering Arm Mali they have got extremely good at it.
That might be a common theme with the RK3588 as it is a complex SoC but much is just an incremental addition to what has gone before.

Its still a how long is a piece of string question as often things only become apparent as things unwind and are implemented.
It could be in 6 months a relatively good working image is available but working changes can be an element of luck and timing as they are upstreamed to linux-next to be incorporated in each release.
So being incremental on the RK3399 you could of expected the RK3588 to be quicker, but this time it seems Radxa are trailblazing RK3588 adoption as not much else yet is employing it so with a smaller herd it could be slower.
So another factor is how well Rockchip manage to sell thier new flagship SoC as if its just a trickle to a few SBC providers that also will not bode well.

1 Like
#151

Thanks for your answers.

I’m just trying to understand where to set the expectations and if a purchase make sense.

I’m just a hobbyist and my current SBC is getting old.

What i expect/hope:

Being able to install Debian or Ubuntu using a conventional method.

It will boot and work fine as desktop computer for browsing and basic office work. Install some apps like libreoffice or openoffice.

First year i may even only use it as a headless VPN server/gateway by installing NordVPN app for debian.

This board is total miss for this purpose?

#152

If honest currently its a coin toss, could be / could be not. Everything from the community is speculation until boards are released.

1 Like
#153

True.

Was about to purchase an odroid n2+ but little too pricey at the local dealer.
I think I will make a purchase of the middle brother.

At the current price positioning it’s worth a try :blush:

#154

put it on a tray, bold down with magnetic bolts

#155

would there be a debug port for m0/a76/53 cores？ it would be necessary for m0 cores

#156

Will it be possible to install and use OpenMediaVault (OMV) on the Debian Buster?

#157

Its highly likely but to be honest the Rock5 with the RK3588 would mean much of its graphical client power would go to waste as it really is a tablet / desktop monster.

The Rock3 is probably a better choice and likely more cost effective as even though a relatively high Mali by number the GPU – Mali-G52 EE is actually a cut down G52 but its network, sata & nvme options are excellent.
Its also got a head start on the Rock5 which could be 6 months before early adopters see boards.

Rock 5 is naturally a client board for tablets, desktops maybe even phones and the Rock 3 has great I/O for NAS with a 5 port m.2 sata card.
Always wished the Rock 3 would provide 2 directly focused versions specially for Routers & NaS and we almost got there but would of liked to see Sata Controller 0/1/2 on board and a single high speed 5/10 Gbe for NAS and the router to have dual balanced high speed Gbe but the Rock3 and compute versions would be my pick.
I think OMV & TrueNAS default offering is a custom X86 image but often its ported to various devices if Radxa had implemented 3 native sata on the Mux with a single fast Gbe it really would of been a nobrainer as we salivate of the possibilty of x8 native sata with high speed Gbe.

#158

On the software support and function reliability, I suggest stems from the business model Arm and Rockhip, other SoC venders use. I feel a more enlightened model is needed. It needs a forum of some sort to bring Arm, hardware, software vendors and end users business, embedded and public consumer together. To discuss the software support problems and to getting generic drivers Linux, Android mainlined.

I favour a small percentage charge to the designs to fund the needed GPU, VPU work.

Keeping it civil and amicable discussion what are others peoples on topic thought to the hardware software reliability issue ?

2 Likes
#159

If you look at a recent release RK356x that has boards out in the wild and initial prototyping has been done then a small percentage is funding GPU, VPU work.
All you have to do is read the long and extensive discussions around Kernel.org and the sponsors of actors in that involvement.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/?h=v5.17-rc1&qt=grep&q=rk3568

This work is submiited to each linux-next release schedule and that is the nature of things as commercial android devices widen there sphere to opensource use.
To post here on the RK3588 that a fund is needed to work on GPU, VPU before boards are released whilst vendor EVBs are still very fresh ignores there is funds and work is being done as the above link will show and how that work has to be submitted into the Linux release schedule.

The forum already exists and its very active, very efficient and does remarkable work and its all documented in the Kernel Archives and there is no advantage at all to flooding that discourse with consumer level ignorance.
That may sound condescending but when you ask at consumer level what could they add to the software engineering process of low level drivers that often needs reverse engineering, what could a consumer offer apart from $ and compliments, whilst still essentially pre-release.

When Warren East was replaced by Simon Segars at Arm it signalled a direction and calls of dictate for a billion $ company that creates revenue from licencing is an act of the tail wagging the dog and its not going to happen and why the sale to Nvidia should definately not.
Android again is similar as you can discuss and make offers but the Dog is Google and the dictate is thiers.

So in the context of what is possible and available what you are calling for is already happening and it is driven by a very small fund and due to the amazing work and small core of Linux it is distributed very efficiently to a very large herd.
The transparency of this probably doesn’t work well to the herd (aka consumer) as what is provided as funds is likely what is currently available and maybe it could be made more visible with a better donation model where consumer $ and compliements can be made.
Many donations lack any metadata to why or what may of prompted a donation and lack the user led nature of opensource as the diverse array of offering of work a provider may give often only has a single branded unitary donation model.

There are many forums, many providers and often many actors involved in specific work that should make a donation model more transparent by a simple link for donation that includes the metadata of specific consumer interest.
You can not dictate to a vendor that they are not charging enough to provide enough funding for hardware and software reliability issues as they work to whatever business model they think is appropriate as again that is the tail wagging the dog.

I do think vendors can employ a better donation model so that in any discourse there is a simple link so a consumer can make an expression of ‘I would like some of that’ or ‘Thanks for what your doing’ that actually provides metadata feedback to what the consumer appreciates and the dog might wag its tail.
If consumers want to donate to show specific interest or provide specific appreciation then maybe it isn’t the tail wagging the dog to ask vendors to provide a better, more granular donation model to the feedback they apreciate in terms of specifics and who by and champion the individual providers a tad more with a bit more transparency.

1 Like
#160

Just to double check: PoE is built in?
no need of a hat or drivers using outdated libraries? :wink:

#161

This is very cool, and I’ll probably get a 16GB model if the Linux support is solid. However, now that these “little” Arm single board computers have reached the point where it’s perfectly reasonable to use them as desktop or x86 server replacements, I think it’s time to offer a version in a mini-ITX form factor. Someone posted a link above for a Pico-ITX 1U chassis, but those are damn hard to find, whereas mini-ITX support is ubiquitous.
I hope I’m not the only one who would love to put a Rock 5 on a proper mini-ITX 1U server case, thus avoiding all the messiness with cables that comes from having to put it in a tray.

#162

I am sorry but the idea of growing to a board 4x the area so it can fit cases for me is some strange chicken & egg logic.

The Pico-ITX is a near perfect format if all connectors are either on the backplane or vertical on the topside of the board. Maybe you could argue for Nano-ITX and maybe be the first in adopting popin/popout backplates for that format as they have served well as going to a standard format is a really good idea but it would be a mini-itx with maybe more than 50% bare purely to fit a case?

I have a feeling embedded GPU’s and M.2/MiniPCIe is likely here to stay and compact formats are a preference to toaster style cases.
I was a bit of a Pi format fan but hell no as the Radxa5 apart from is it the PD power (the micro usb on the side) and power/reset buttons is very near to perfect as plugable connectors on one plane means simple backplates can be use.

I presume Radxa do have a case in the offering and maybe will be doing a 1U x4 R5’s front panel that you can use with any rack.
I am a fan of compact as the Pi totally fails when it comes to full length m.2 and uses multiple planes for I/O which is just bad even if it has become popular its still a total pain.

Radxa if you would do a revision where the buttons are via a connector and just supply the 2 pins on board than switches maybe? Also if that is power then guess it gives choice of a relatively simple extention back mounted choice.

I think wait for a case than grow the board by 4x the current area.

3 Likes
#163

I think that the release of commercial hardware, without normally working software and drivers, is not what buyers want to get.

1 Like
#164

It is logical to expect that it is the hardware developer (ARM and RockChip) who should do the drivers, and not the board collector.

3 Likes
#165

Radxa are a design house who also make there own boards and demonstrate actual product released to a makers community but that is not their only market.
That is what you do not want to get and your expecting something that Radxa don’t do.

Again this is very chicken and egg as how do you have drivers 1st before hardware exists for a community to work on?

Send an email to Arm and Rockchip or maybe find another product maybe a Steamdeck or something as that reflects the cost of commercial finnished product than developer/community releases.
Or pick an older board 1/2 years old from release that now is pretty much complete with drivers and images (Choose wisely as there are some stinkers :slight_smile: )

If you want a real experience in not very opensource boards suffering driver and hardware problems maybe try some of the offerings from what I call the (other fruits) but early adoption opensource hardware is what it is and would seem not what you think it is.

Nobody wants to release hardware in this manner but it does provide a very efficient and lower cost method as in the above its very hard otherwise and there is nothing really to test that it works until hardware exists.

1 Like
#166

Linux and Android support is announced. In my opinion, the operation of the operating system is something more than a simple download to the desktop. Or not?

And what does the community have to do with it? Drivers, for example, for video cards are made by Ati/AMD and NV, not by Asus or GigaByte.

I want to get out of the “box” right away. :slight_smile: Especially for money, and not for nothing.
I don’t mind donating money for software development, I placed an order for a 16GB version, but I wanted to get a working version, not a toy.

According to your link, it is not clear why the end user needs it at all? I need hardware acceleration of video and 3D, in Linux, not the source code made in the garage.
Well, again. I was surprised, asked why there are no drivers from the manufacturer of the ARM/RockChip hardware.

Humorist.

I looked at what it was. Some kind of strange and useless thing.

#167

Things just are not as you envisage, I do think maybe things could be a little more transparent to the nature of opensource and that you have to have an element of knowledge to what is and what is not out there in terms of opensource.
Radxa do want to sell product and I guess if people purchase without a true understanding of what they are purchasing then I guess the problem lies with them.
Arm actually don’t make product they just license design and adopters make product and why ARM provides SDK’s and not end user drivers.
Rockchip and others provide services to make end product and will design and assist to complete them and its how they make revenue.
The above is Collabora reverse engineering an Apple M1 to provide an opensource Linux OS for it and its being done because Apple has no interest and if they could legally they would stop it but at least as users we do have some rights.
Opensource drivers do not exist because huge revenues are created from providing the latest and greatest with closed source that those vendors provide.

That is reality and I am sorry if you think otherwise.

Its because of the M1 work its likely we will have a driver for the G610, but so far no it does not have a driver apart from the android focussed Rockchip blobs.

1 Like