Still not a thing for the RK3588 but RK356x is getting a decent flow of submissions at kernel.org.
Not heard much of an update apart from early Android tests but still looking Linux wise it will be a trail blazing experience than having anything complete.
Introduce ROCK 5B - ARM Desktop level SBC
1 Like
I see. Thanks for the reply. In time i hope RK3588 would get the same love as RK356X.
RK3568 and RK3588 both have USB 2.0 OTG support only. as per the datasheet and this recent post on the LKML.
Itâs not quite that simple, though - the controllers are set up so that one type-C OTG port runs at USB 3.1 speeds in host mode, but is limited to USB 2.0 speeds in device/gadget mode.
It also looks like the 3568 and 3588 have a lot of IP blocks in common - peripheral drivers etc. for the 3568 are likely to not require much in the way of changes for the 3588.
Yeah I think there is a lot shared IP and quite like what Rockchip did where each one has quite specific applications.
Radxa have got the the Rock5 bang on the button for me as probably one of the best RK3588 packages out there at also the best price.
The 3566/3568 board implementations have confused me but hey as maybe that is where Rockchip has drawn a line where its to get those complete and then the RK3588 is just incremental additions to that base and why we have seen nothing 3588 wise as work is being done on that IP but using names of others from the same series.
The 3566/3568 have USB3.0 OTG and its no bother to me that the 3588 doesnât but it does show the IP between them does have difference. The 3588 just isnât a gadget type device whilst 3566/3568 are sort of perfect for that role.
I havenât seen much about the G710/610 either but I presume the current apple efforts are focusing resources and its on the next todo list (hopefully).
I guess this current lull and not hearing much is making the wait much worse 
I was rather hoping the Radxa Zero2 would be out now to help alleviate the Rock5 wait as generally a pessimist 2 quarter often means 3rd or later as releases often do.
Anyone got any updates on the Radxa Zero2 as that is a really interesting board also?
It also looks like the 3568 and 3588 have a lot of IP blocks in common - peripheral drivers etc. for the 3568 are likely to not require much in the way of changes for the 3588.
yes. we can see it from the features, not even need to check the IP blocks.
So when are we getting our preorders? shipping June 30th?
Second Quarter:
Friday April 1 and ends on Thursday June 30
Hi all, please share your clarifications about the Rock5 SBCâs PCIe lanes design here!: Radxa Rock5 RK3588 SBC: how many PCIe lanes to the M.2 slots? - ROCK 5 - Radxa Forum
-
It appears to me the Rock5 SBC actually has five PCIe lanes, four of them are v3 and one is v2. I would guess the lower M.2 slot gets 4 lanes and the upper slot gets the last 1 lane. Please confirm.
-
Also, some people here have said the Rock5 SBC has the RK3588S. That was actually incorrect, only the Rock5 Compute Module (which is totally different from the Rock5 SBC) uses the RK3588S, isnât it?
-
Also what about the SoC difference between Rock5 SBC model A (does not exist?) vs. Rock5 SBC model B and Rock5 Compute Module
Please share your insights in the linked to thread.
I have been reading and posting on the thread with eagerness from the start as you will see by the number of posts and sort of confused to your post.
Yes Rock5 has a x4 that are PCIe V3 and is on a M.2 connector it supports bifurcation so could have a splitter and be setup as 2x x2 or 4x x1 lanes also but likely the default will be 1x x4 on a single M.2.
The top M.2 is pcie 2.1 x1 for the sort of standard wireless cards you can get and the pcie layout is as it was 1st announced.
Rock5 model A does not exist apart from possible plans of a âcredit cardâ size so maybe but likely a compute module as RK3588 is just âtoo bigâ for the compute format.
https://wiki.radxa.com/Rock5/CM5
Hoping maybe they may also do a very cut down model A as for many would be such a cost effective entry level desktop as never been that interested in the compute modules that have far more commercial applications, but as a compute module it beats the PiCM4 quite convincedly.
Same brilliant PicoItx format no rear connectors but cut the cost of carrier boards with a simple selection for a basic âdesktop/set-topâ
1 Like
you may just check it up with the RK3588 datasheet available online
Basically RK3588 has 3 âPIPEâ PHYs that can be configured as SATA (0~3) , PCIe 2.1 x1 (0~3) and USB 3.0 (0~1), and ROCK 5 should be using PCIe 2.1 x1 (2) and USB 3.0 (1), where one of PCIe 2.1 goes to RTL8125 for the 2.5GbE, and the other goes to the top m.2; there is PCIe 3.0 x4 and all go to the bottom m.2
for RK3588S it is a scaled down version that cut the PCIe 3.0 and just retain one or two PCIe 2.1 lanes, and be used in CM3 which is compatible with Piâs CM4. Rock5 has RK3588 for sure, as it has HDMI IN.
Donât think there will be a ROCK 5A, unless it uses RK3588S instead. But given that we can get a CM5 based on RK3588S, it means we can use whatever carrier board made for Piâs CM4, and I donât see why we need a separate ROCK 5A in this case.
1 Like
Cost is why I see a market for a Rock5A as when you do get a CM & Carrier it adds cost.
The Rock5B may have a M.2 x4 pcie 3 that many might fit a nvme its still a Cortex-A76 & Mali 610 and even though its blisteringly fast as a SBC its still only in the entry level arena that losing the M.2 to eMMC is not going to be that much of a loss.
For many a low cost fit for purpose âdesktopâ could be with eMMC and think its likely for Radxa to produce it would be cheaper than a CM + Carrier.
I am going for the Rock5B but I do think many be it exchange or situation have limited budgets and it would be great to see what Radxa can do as a minimum whilst retaining the core of a âdesktopâ of cpu + gpu in the guise of a Rock5A.
Hi all, please see my request here Product request: Stackable 2x/4x M.2 E key extension board for Rock5B SBC for a stackable 2x M.2 bifurcation board. Please share your thoughts there. (Also for a stackable 4x M.2 PCIe switch board.)
Also, please have a look at this discussion of USB-C power delivery voltage & watt consumption: Rock5B SBC: Voltage on the USB-C Power Delivery input. Rock5B+peripherals watts consumption
Will be possible have an extension for GPU?
eGPU for ROCK5B with G710MP10
for a better graphic experience
1 Like
@Oneofmany888 Regarding the Rock5B SBC working with a Radeon/AMDGPU dGPU:
What I read is that there are certain limitations in Rockchipâs PCI interface design. Specifically there is a DMA cache-incoherency issue that can be worked around in the DRI/DRM drivers if you know where to put the flushes/invalidations. Maybe something about coherency both with DMA and somehow GIC.
The issue is fixable but Linux may choose not fix it. Please do share if it will work for you.
Discussion here https://wiki.pine64.org/wiki/Quartz64_Development#Partially_Working&j akllsc hLibera.
1 Like
Enoch, I hope you are incorrect in your prediction, and that we will indeed see a Rock5 Model A.
Its not just about Linux fixing things as the cards that work are plain dumb non bios cards but on a uefi boot and all the things like adjustable BAR or at least a BAR big enough then its possible as its the u-boot mechanisms we use for ARM sbcs that is problematic not Linux.
I still see it as relatively pointless though as we are talking about a very entry level CPU in terms of âdesktopâ and it will be the bottle neck for many fairly modest discrete GPU cards.
The Mali is coming on though and I think the G710 MP10 & above is nearing GTX1050 sort of levels but Rockchip went for the G610 as that is a good balance to the CPU.
I am sure we will get enthusiasts trying because they can as the RK3588 is a great SoC but Apple M1 its not.
@Larry yeah as low a cost stripped down single all-in-one rock5A as would be such a super board for many who are excluded from the majority of western $ levels.
yep I could be wrong, itâs just a guess based on how I see it; if there would be a Rock 5A, it would likely need to use RK3588S because of the PCB real estate limit, and I donât think it would be that much cheaper than Rock 5B.
personally I am really looking forward to order my 16GB Rock 5B and use it to replace my laptop (i3-7100U 32GB) that I mostly dock it;
if there would be another âSBCâ based on RK3588, I really hope it would be something like a âdesktop replacementâ in the form of mini-ITX (3.0 x4) or mini-DTX (3.0 x4 + 2.1 x1) with m.2 (M=22110 x4, A+E=2242 x1, B=2242 USB3) and 32GB option (4 chips), and PCIe bifurcation for slot and m.2 (become two x2). if mini-ITX only, I still hope it would support cabling to make available that PCIe 2.1 x1 slot. I think this will make everyone happy (router, NAS etc. all can be done with this), but of course cost is a concern then.
If you look at relatively close other product then the Mediatek Kompanio 820/824 is a good ballpark which is sort of considered mid level chromebook or basic entry level desktop.
You could do like firefly and have your ITX motherboard RK3588 but for $450 you would have to be very casual about spending to pay that for what is still a RK3588.
I think some other manufactures have also similar ITX offerings and prices seem to be $250 to fireflyâs confusing price as with $450 there is a hell of a lot you can buy for that.
I think Radxa have the format of the Rock5 as near perfect you can get it and have amazingly competitive price tag, but its still a mid level chromebook style processor and anything from the Rock5B isnât gaining anything but cost.
Rockchip have announced a lower cost RK3588s will be available and the loss of the PCIe x4 for the likes of a eMMC budget desktop is not that big a deal especially if its cheaper.
The PCIe x4 on the RK3588 for NAS is sort of pointless as its hugely bottlenecked by the network available whilst Rockchip just released some perfect NAS style Socs with the Rk3566/68.
I will never understand why the RK3568 was released without a single USB3.0 OTG, 2.5Gbe & 1x sata so that with a x5 port gives great options of 6x sata on a budget board that isnât that stupid pi format needing a riser.
Or even a RK3566 with x3 sata and the normal 1gbe as a super low cost ready to go NAS.
The RK3566/68 where incredibly unique SoCs primed for router / NaS roles and still scratching my head to the formats released and would find a SoC that is primed for entry desktops/set-tops being used as a router/NAS as equally as strange.
The RK3566/68 have that cut down mp1 Mali-G52 EE as I donât think rockchip had any intentions for a desktop role apart from video modes.
1 Like
Stuff like you described already exist, they are sold as Network Version AD Player, by Mekotronics.
RK3588 Run Android 12 and Debian OS 11.
Also Rupa H01 RK3588 8GB