The RK3688 looks more interesting https://www.cnx-software.com/2024/10/04/rockchip-rk3688-armv9-3-aiot-processor-to-feature-a-16-tops-npu-ufs-4-0-storage/
Personally I would ditch the NPU and use that die space for as many core GPU as you can fit, as the frameworks vulkan and opencl offer far more interoperbillity than vendor NPU frameworks.
You also get two use cases where both ML and Gamers should be massively impressed than splitting that die space between a low core count GPU & NPU.
I guess the IP for a GPU is much more than vendor created NPU, the link says a 16tops NPU and if like the RK3588 and no where near rated, with a limited number of models that will run and a conversion framework that seems like Voodoo it all becomes very much less impressive.
Radxa 6 under development
Really? the right question Radxa should be asking to themselves is how to make a SoC as powerful (or at least 80% as powerful) as best consumer chips out there, pricing for this Board should be lower than … say a 8cxgen3 phone instead. Not to mention there are others competing product like Dimensity 9300 .
If they failed to do so, there aren’t going be realistic/ competitive enough for consumers to get interested into buying one; let’s be frank right, the only reason people are hype about rk3588 is just because it’s cheap and looks quite powerful and open source; nothing more nothing less. Look at all those crazily super high pricing product with the similar chip, they don’t sell! As ironic as that!
I saw rk3588 product which is selling more than 500, even 1000 dollars; really who you think you are ? Apple?
Even Apple got punished by sales when they got too greedy, look at the failure sales figures of the pathetic Apple Vision Pro!
(just an idea 
).
Amusing, so it’s slower than the CIX-P1 in the Orion-O6 as it currently ships, since it has the same cores in same quantity but at lower frequency. INT4 can be cool, however.
Any progress / news on this matter?
Will it be possible to pre-order?
Cortex-A78 is kind of old even for products launched today. It should at least be the Cortex-7xx. And btw, you don’t just simply mix and match the Cortex-A78 with Cortex-A72 in the same SoC.
Even if a Rock 6 was planned, knowing how Rockchip started acting, I’d rather wait for more information about the Dragons.
Hopefully Qualcomm will be nicer and better…
Dragon may be bit faster than Rock 5B*
but it’s not flagship 
has better AI and less IO. Next flagship Rockchip is expected to be ARM v9.
Most of Rock products features some Rockchips (except one zero?), Dragon is nice name for new series. Cubie for allwinner, orion for cix, nio, x, sirider etc. So we may not see Rock 6 if there will be no something like RK3688 available for such products.
I’m very out of the loop.
What is meant with “how Rockchip started acting”?
Orange Pi products were found in drones used in the Ukraine war. Rockchip freaked out as they could be sanctioned and its rumoured they have tightened control of their chips supply.
Is this for m.2 slots or for network/video cards?
Yeah, yeah. I was unclear. By “better” I still meant the company’s behavior, not chip’s performance.
Technically true, and seeing an ARMv9 SBC in 5B and 5C formats (maybe even a compute module) would be awesome… But what’s the point if software sucks and the community has to work for years with little result in order to make anything work? We’ll get ARMv10 before such hypothetical RK3688 gets full support.
That too, but that’s not what I meant to say. I meant mainly how Rockchip doesn’t care about providing proper Linux support. Which is understandable from some perspectives, but bad for us all here.
There is nothing like “full support” and that includes every electronic device, also those at x86.
When product is released we all see what is in specs, but this does not promise any software yet and usually details includes some bugs and missing support. Developers are able to get into platform and start to work with it, make software compatible, optimal for platform. It’s much better because You can either involve in project or just use it for something simpler.
Linux is not main market, android is. This is the reason for those kernels origin,
Now that’s a useless statement. How about you stop nitpicking?
I am aware. That’s what I was referring to in the very first place in my responses. I was already stating that as a known fact (because it is well-known), only without directly referring to Android.
Why do you feel such a strong, obsessive need to correct the tiniest inaccuracy? You’re visibly mistaking end-goal pragmatism for stupidity / lack of awareness of the process. Treating others like idiots by default is rude and offensive. And I wasn’t even talking to you, but mainly to the other guy.
Believe it or not, the Rockchip is/was the most open source friendly of all SBCs despite the old kernel. TBH Samsung S5P6818 Octa-Core Cortex-A53 was (is??) the only one you could build the firmware entirely from source (100%), but of course, deprecated kernel. I took a glance at Qualcomm BSP, mainline kernel 6.6.52, but the most important part is closed source code, i think if you want the latest mainline kernel with all the board features, you need to wait for Qualcomm to release the closed source code (ko) specific to the mainline kernel version.
Which SOC Radxa 6 possibly have? Any clue? Is MediaTek that open friendly?
I found some more RK3688 news;
A while back Radxa showed a screen shot of it featuring a 16 TOPS NPU, however there will also be a RK3688M with 32 TOPS NPU.
https://xueqiu.com/2704068488/333261184
We are also looking at a 2026 launch for the RK3688
https://xueqiu.com/7796248954/333149684?_ugc_source=ugcbaiducard
Can anyone help to translate the RK3688 info on the picture?
Thank you in advance.
- CPU: 300K DMIPS
- GPU: 2TFLOPS
- NPU: 32T
4nm process
Basically, this is a 4nm version CD8180 with higher cpu freq. For your reference, the G720 MC10 of CD8180 is 2.3TFLOPS.
Sure. There’s 2026 written. That’s a roadmap date while the other 3 dates on the picture are timestamps in hindsight.
Let’s keep in mind that Rockchip presented us with a couple of ‘roadmap dates’ along RK3588’s way. For example in early 2019 we were told: ‘Mass production is expected for Q1 2020’
In reality a few production ready RK3588 SoCs were procuded at the end of 2021, for example on the SoC on ‘my’ Rock 5B was’week 52 of 2021’ silkscreened while vast majority of early SoCs on other RK3588 devices was produced in 2022.
So what to think of the 2026 number in the picture above? At least you won’t be holding anything RK3688 in your hands in 2026 and that may apply to 2027 and even 2028 as well