@leasing I believe you need this
What do we want in the next Rock Pi?
What really is missing is a somewhat affordable option to replace your PC.
Raspberry Pi 4 8GB is close, but CPU and GPU are just too weak.
Rock Pi N10 looks even better, but the form factor is too large.
Suggestions:
- 8 Core ARM CPU, supporting AES encryption (important if you want to encrypt your new ARM machine)
- 8 or 16GB RAM
- 2 HDMI Ports, supporting dual monitor setups
- Decent YouTube performance w/ hardware acceleration for modern video codecs (important!)
- Sound and microphone onboard
- Good graphics (like N10 or better)
- RPi 4 like form factor
- Passively cooled
- M.2 SSD and micro sd card slots
- Ethernet
- Allow users to overclock easily (at own risk)
- Price ~150$ for the 8GB RAM version (twice as much as the RPi 4, so that should be enough for a decent performance improvement)
- A case should be purchasable separately
I personally wouldnāt need any pin connectors - these could be removed to save space. Why? Because for tinkering I could use a slower machine, like e. g. Rock Pi 4.
What would be the advantages for the user?
- Power saving system
- Ideal for desktop usage
- CPU has no strange firmware built in (like most x86), so it is compatible with open source ideals
- Enough performance for coding and browsing
- Can also be used as a modern media center (HTPC)
- No noise
- Smaller form factor than most x86
- Light gaming possible
- Good emulation performance (I guess)
Let me know what you think.
Both are actually quite decent.
The real reasons why its far away from desktop:
- low memory speed
- terrible IO performance
- failed powering design
- optional absence of AES encryption.
That is not in the hardware design domain.
Then you need to sacrifice performance.
Thanks for picking up on my ideas.
I donāt think so. If you look at benchmarks vs e. g. the Intel J5005 it is no match. While that comparison might seem a bit unfair, it should be noted that the Raspberry Pi 4ās CPU is also much slower than the RK3399 in most benchmarks. If you look at YouTube videos of Pi 4 users, a lot of people have overclocked their Pi. Having 8 gigs of RAM with such a slow cpu makes little sense to me.
Regarding your other points, you seem to have more expertise on those and I cannot answer to them.
What is the issue, then?
What makes you think so? There are lots of very performant smartphones. Take my over one year old middle class smartphone for example. It has 8 cores, which all have higher clocks than the Pi 4ās. Why shouldnāt an 8 core SBC with passive cooling be possible?
1 Like
This https://github.com/ThomasKaiser/sbc-bench is more representative benchmarking mainly between SBCs but also compared to some Intel hardware.
I know.
I am not saying Rpi4 is good enough for desktop. What I am saying is that raw computing performance is the last thing you need to worry about. Unless you plan to do some heavy computing, compilation, ā¦
Its possible. But active cooling can significantly improve performance ā¦ on mobile phone / embedded you canāt afford cooling, on desktop you can. Which is why its there. BTW. Play 4k content or play games for some time on your Phone. It will surely become hot ā¦
Solidrun made ARM based workstation, which is quite powerful and it can be passively cooled ā¦ with a large aluminium cooler. That is probably closest to things from this topic, but the price.
1 Like
Igor talking good about RPi??? At least it didnāt last for long 
It is true tho. Only for the latest xxx games youād need something heavier. But it runs 10 year old games with ease.
The RPi4 indeed has too bad IO. Thatās why I love my RK3399ās so much. Iāve got NVMe on them, CPU wise it isnāt bad at all.
Tho the A311D from my Khadas VIM3 is way faster and consumes a lot less. The IO is then again a lot worse vs the RK3399.
Only a small number of tasks these days need more CPU/GPU power.
Iāve been using my RK3399ās for 90% of my tasks. With the M4V2 almost all desktop tasks, watching movies/youtube. The RockPi or M4 for gaming.
And it is great for gaming.
All I want for my next ARM device. The CPU power and power consumption of the A311D or even a bit more. And the IO of the RK3399.
I guess thatās the RK3588? When? I canāt wait for it.
Also the RockPiX for building Armbian. Iām waiting patiently. Greetings
Yeah, thatās why I wrote ācodingā. 
Very true, but that should be manageable with a large heat sink. Components on a phone are also much closer together.
Thanks for posting your experience. Iād probably try out an RK3399 with a small form factor, but there is afaik none with 8 gigs RAM and dual monitor support. 
I donāt know why they all choose 4 GB. You reach that very fast with a few browser tabs and some work related processes (IDE, development server, ā¦).
Now that Apple is also switching to ARM, the eco system will probably benefit from it. Imagine having your whole PC in a Pi 4 form factor, totally quiet and power saving! Not even my NUC could compete with that.
This one really seems promising!
Just for that the IO is so important. I use my NVMe with an 8GB swapfile. Also zram for 2GB extra. Makes 14GB useable memory. The IO speeds of NVMe are not that far from ram performance. So you never notice when using swap. With slower eMMC thatās a big difference. Then youāve got constant slowdowns. I did use it like that with the Odroid C2 for video editing in 720p.
So for that the RK3399 is for me the best desktop ARM SBC.
The NanoPC T4 should be able to do dual monitor. Tho I aināt got one. It has dp over usb type-c.
I use sbcās for video editing on the go. The RK3399 now does video editing with 1080p video with ease. 1440p is doable. But the Khadas VIM3 does 1440p perfectly, while consuming a lot less.
That should be the next step in more cpu power and better power efficiency. So I canāt wait for it. Thatāll maybe be able to do 4k editing/rendering.
At home Iāve saved a lot on my electrical bill from switching from an x86 pc that consumes a lot constantly when used 24/7. While my sbcās consume at least 10x less for the same tasks. I just use multiple devices for different tasks(an OPi+2 for NAS, 3 RK3399ās for gaming, youtube, testing, and the Odroid N2 for video editing at home. While on x86 youād do it all on one. And all devices together are also cheaper than a good x86 pc.
Only software support is lagging a lot behind vs x86. I wonder what Apple will come with. Greetings.
2 Likes
So, if we are not dreaming, but advising realistic improvements, Iād say:
- 2 * M.2 slots (PCIe 3.1 x4)
- 2.5GBase-T Ethernet (better 2, or 5GBase-T)
- More powerful CPU (RK3588 maybe? or overclocked RK3399 if this is possible)
- 8 GB LPDDR4X (and 12 after year)
- Full-featured USB 3.1 Type-C (Data IO, PD, Video Out), or maybe 2
Unfortunately, iām not in hardware market, and my ballpark estimation is over 100$ for top-notch version of this features. or may be even more
4 Likes
Would any one from the Radxa team read this thread?
Currently I am not using one, but am very interested in buying when new models are released. Too bad that I know about Rock Pi 4 too lateā¦ only after I got RPi 4, which I really donāt like it as it does not have crypto engine in the SoC.
Rock Pi 5 (multiple models?) ==> more like āgeneral purposeā SBC
SoC -> RK3588, allowing users to overclock or underclock+under-volt at his/her own risk
keep close to RPiās form factor whenever possible
3+1 / 6+2 / 12+4 GB RAM options for CPU+NPU (still need to split RAM with RK3588?)
have boot SPI, microSD slot, but seems no need of eMMC (save PCB space)
USB-C 3.0 OTG port x1 for PD power (USB 3.0 OTG can substitute USB GbE IMO)
USB-A 3.0 HOST ports x2 (for more ports we can always use USB hubs)
GbE from SoC
GbE or NBase-T or USB 3.0 from PCIe 3.0 x1
m.2 (2242) for SSD / WiFi from PCIe 3.0 x2 + USB 2.0 x1
m.2 (2230 / 2242) for WiFi / LTE from PCIe 3.0 x1 + USB 3.0 x1 + nanoSIM slot
NO NEED of USB WiFi module if we have m.2, right? more PCB space
mDP / microHDMI / eDP or LVDS for more PCB space
4-pin UART header for serial console (debugging)
*** not sure for 3.5mm (HDMI/DP can substitute audio but not mic) and camera MIPI (can webcam substitute for camera+mic?), may need to skip for more PCB space
*** 8GB RAM is generally enough, but 16GB would be nice and allow the platform to use KVM for many other purposes
Rock Pi E* ==> more like network appliance
SoC -> RK3530 (RK3566 seems to support dual monitor, may not really needed here)
1 / 2 / 4GB RAM (4GB is good for running KVM based virtual router)
USB-C 2.0 OTG port x1 for PD power
USB-A 3.0 HOST port x1 or x2
GbE from SoC
GbE or NBase-T from PCIe 3.0 x1
4-pin UART header for serial console (debugging)
optional SATA port(s) from SoC for functioning as NAS (where NBase-T will be appreciated)
*** do we really need USB WiFi? or we just instal dongle if needed?
1 Like
- RK3588
- 8GB+. RAM
- Another m.2 port
- an ethernet interface that can handle large MTUs
- EMMC slot in a more convenient location on the Penta-hat / enclosure
1 Like
Price/performance is important to me. In fact, the $59 price tag of the RockPi4C was what drove me into buying one. Iām quite satisfied with my RockPi4C and will be definitely buying again from Radxa. We need a stronger/beefy CPU with a better CPU nm process like a 14nm S922X CPU from Odroid-N2+ while keeping great I/O features from RockPi4C, specially for those of us looking for a true Linux desktop experience. Iām only interested in ARM64 architecture, already bought a Odroid-H2+ and already gave it away to my wife, it works great but for me, thatās just another x86 box. I personally, would never buy a RockPi-X, even if it were on sale for $20(x86 & 4 year old CPU).
Regarding how much Iād pay for Radxa next ARM64 SBC, $99 tops for new a RK3588 board. Donāt forget after buying official required accessories, shipping, customs, aluminum case, etc, price could easily reach $200. Some features should be copy/implemented from Hardkernel like using a Blob GPU driver for a HW acceleration experience on linux desktop using mainline kernel.
Being able to flash an eMMC module without having to physically remove eMMC module from board via OTG like Odroid-N2/N2+/HC4 do.
A HK āNetboot installerā software solution is also in my wish list because eliminates the need of physical contact to flash OS images on the fly to eMMC modules, NVMe and SSD drives.
Also, a multiboot option would be awesome from either uSD, eMMC or m.2 like RPI Berryboot, Odroid-C4 custom booting option on uSD or Petiboot from SPI like Odroid-N2/N2+ and HC4.
A GPU with Collabora Panfrost/Vulkan drivers is a must. Iād rather have an older GPU supported with HW acceleration on Linux desktop, even if itās w/ a Blob driver only than having a state of the art Mali GPU that takes 2 years to develop into a working state like Mali G77 or G78 GPUās.
Last but not least, ability to boot from USB3.
Well, guys, you were mostly talking about hardware for the Rock Pi and I am going to make serious software part:
Please, @jack , @Stephen note also.
- adapt Ubuntu 20.10 desktop/server (or 20.04 LTS because of LTS), mostly server to Rock Pi 4, because it has mostly new packages, it is well supported, adaption process should mostly easy and the results would be magnificent; do not put much effort in Armbian as it is still WIP and many basic features are not working
- beside SPI, Bluetooth, M.2/NVMe and other supported features I would make a pressure on:
- temperature reading (it is still not working currently for popular DS18B20)
- fixing bugs like writing to SSD connected via USB 3.0 - data from Ethernet
- making general way for most of the hardware, so the less experienced users would have easier way to solve not working basic hardware (e.g. make kernel, drivers work flawlessly but leave OS packages and applications/scripts for user research/development)
1 Like
Armbian shows and gives you what is possible and that is fucking expensive which is why others donāt give you not even that and certainly not more.
Ubuntu doesnāt even boot without troubles and you can only copy functionality that exists in Armbian, the rest will also donāt work. What doesnāt work, you have to develop! Do you really think all this is that cheap? That you can copy āfunctional ubuntuā from intel world and and paste here? That Armbian folks are retarded because they canāt deliver fully functional system.
BTW. Do you perpaphs support open source software? Any? This ācopy pasteā process costs tens of thausands of hours.
Without Armbian, you will be running development kernel 4.4.y and some amateur made pre-alfa Debian/Ubuntu from things that falls into the mainline. Without any testing or support. And a year later.
Once things are mainlined, even Ubuntu can āsupportā this hardware. Usually if vendors pays them.
Ubuntu only works well on 1st class x86 (and other supported) hardware. Which is cheap / no cost to support.
Why I am wasting time telling you this?
With all respect: you are wasting your time telling me that, because I am engineer since over 11 years (formally, not formally twice as that) and I can recognize a good OS and a bad OS. Armbian is a bad OS. A shitty system. This is once. Twice: you are repeatedly advertising a system that some things are not working out of the box. There is much more time needed to configure Armbian than any other OS to the same spot.
As in Armbian some more complicated things are easier to do when no one needs them, the basics are left over for user development with complexity too difficult for less advanced users.
Iāve tried Armbian few times and I can see that through that time (few years) it has not moved forward in basic concepts. You are investing time, money and work into things that are simply not needed.
I donāt need newest kernel when there is NTFS not working, when there is Ethernet not working the same time when Wi-Fi, I donāt need system where there is not working external screen and I do not see any tutorial to make it start and I donāt need fireworks when there is no possibility to fire them up.
If you work on an OS - make documentation for everything. I hardly can find anything about some things in Armbian (listed above) where some things are just not documented and engineer need to go through many thoughts to make basic things work.
Thatās my opinion about Armbian. You can agree with it or not, you can argue otherwise, but you wonāt make Armbian a good OS in its current form. So advertising it is pointless.
1 Like
Between those that works on Rockpi 4 or in theory?
Its not perfect, but more then majority of users and people that understand things, supports us and are happy even we are all frustrated because in this world things donāt just work ā¦ Why worry about a few completly unsatisfied ācustomersā?
True. People need to know, there are better things out there.
BTW. Name one system in general that doesnāt do that. Most of them, especially commercial ones, are full of bullshit - we donāt do that. We tell how things are.
Compared to Windows?
Which basic functions are hard to setup? Again we would hear that - and people would fix that since they reguraly do. Armbian is community project. If something is enoying enough for many people, this is quckly fixed by random user.
If things are not needed for you, this doesnāt mean they are not important.
If you put your money on the table, you will be asked for opinion what to change, but until your money / resources are not in the game, nobody will not asks you what to change for you. Also you can fix things and submit changes and also fork the project at any moment and do whatever you like with it. You can evev make your own distribution that other will be able to throw shit at
OK. Fine. Perhaps your skills are too low? You though about that? Why this would be responsibility of distribution? Get Windows or Raspberry pi that is adopted to less skilled people.
Linux distrubutions are a sum of freely available software made by thausands of people. We are not authors of Network manager, nor network driver. Yes, we fix both if we can, from time to time and we provide you guidance how to overcome troubles (which costs us a lot of money and is a reason why you could easily get told to f* off if you push too much. Learn on your own cost).
What is that so much better alternative where everything works? Why would we invest our private money to fix external screen, which is serious and expensive problem, because someone from forums asked? Think. Think twice. Think once again is that really a problem of nonprofit amateur project. If you donāt provide money for function realisation, please do 180. Even with money, chances are slim. People doesnāt have time for āyourā issues.
Armbian is a build system - that its main goal - that produces Debian like OS and documentation is sufficient. For general things its nonsense to make yet another manual for Debian like OS. There are tons of manual on the net and all works - Armbian is identical to Debian or Ubuntu on user land level. And besides why would we save you time if you give nothing to the open source? You donāt support the project - what gives you the right to complain about?
Thank you. Use alternative or just use Raspberry Pi. Easy.
With your evilness towards me I wonāt argue with you because of pointlessness of your position.
Make better OS, THEN advertise it. Not the other way.
1 Like
Why? There is absolutely no need. You can use alternative and be happy.
Thatās why: because everyone should know that you are advertising a shitty OS.
I am taking responsibility for others, how about you? You donāt.
If I can use other OS and be happy, why are you arguing with me? Where this freedom for me ends? And why it is moving its boundary from post to post?
Because you would name it by now if such option would exists. Just take time and answer to my questions.
Be a man.