As expected, Pine64 has now unveiled its own Rockchip RK3588 SBC with the QuartzPro64, joining many others from Radxa to Banana Pi, and lesser-known companies such as Mixtile and Mekotronics.
The QuartzPro64 single board computer will be equipped with 16GB RAM, 64GB eMMC flash, and with a large 180×180 mm form factor, offers plenty of ports including two HDMI outputs, one HDMI input, two SATA ports, two Gigabit Ethernet RJ45 connectors, and more.
QuartzPro64 specifications:
- SoC – Rockchip RK3588 octa-core processor with four Cortex-A76 cores @ 2.4 GHz, four Cortex-A55 cores @ 1.8 GHz, an Arm Mali-G610 MP4 GPU, a 6TOPS NPU, 8K 10-bit decoder, 8K encoder
- System Memory – 16GB LPDDR4x
- Storage
- 64GB eMMC flash
- 2x SATA ports
- MicroSD card slot
- Video Output – 2x HDMI 2.1 ports up to 8Kp60, 2x MIPI DSI “DPHY” interfaces
- Video Input – 1x HDMI port up to 4Kp60, 2x MIPI CSI “DCPHY” camera interfaces
- Networking
- 2x Gigabit Ethernet RJ45 ports
- 2x SMA antennas for wireless module
- USB – 1x USB 3.0 port, 2x USB 2.0 ports1x USB Type-C port with DisplayPort alt mode, 1x USB Type-C port for debug
- Expansion
- 1x PCIe 3.0 socket
- Misc – RTC + battery holder, fan header, various buttons for power, maskrom, system key, etc…
- Power Supply – 12V via DC jack
- Dimensions – 180 x 180 mm
As we previously discussed in previous announcements, Rockchip RK3588 software support will take time, and Pine64 will only sell QuartzPro64 developers initially since the device is not “not end-user ready”. The plan is to see both BSP-based support in the initial months, with mainlining efforts running in parallel.
As we’ve already found out, Rockchip RK3588 is not exactly cheap, and systems based on the processor usually costs about the same as Intel Jasper Lake-based hardware, with lower performance, but support for features such as 8K video output and decoding, 4K video input, MIPI CSI cameras, and possibly lower power consumption. Pine64 QuartzPro64 is expected to cost over $300, and it may even be subsidized at that price. The board should become available to developers in the next few weeks. Rockchip RK3399 powered Pine64 RockPro64 will not be phased out considering the price difference between RK3399 and RK3588 hardware and should remain available for years to come.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.
Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress
OMG that is huge board. Can we get Pi size board PLZ
As if you would be willing to pay +300 bucks for a smaller board 🙂
But for 300USD I can have intel or AMD mini ITX board that “JUST WORK”
Smaller boards will have less IO and be more bothersome to cool.
18×18 cm are very nice dimensions imo. Still fits easily in the hand but big enough to have everything needed.
17×17 would allow to match the mini-itx form factor which would have been even smarter IMHO.
I was about to comment to that effect. 18×18, how silly is that?
This is just yet another unusable product from Pine64 that isn’t compatible with anything on the market. I really don’t get their business idea.
Yea. The (relatively strong) pricing of Jaspar Lake CPUs essentially turns the RK3588 obsolete for anyone but diehard arm fans.
Still, I think the 3588 will mature quicker than previous arm boards. The fact it actually has a strong enough CPU+GPU will mean it’s actually usable for day2day tasks. Unlike a Pi4 or RK3399 which will both choke on youtube and heavier webpages. (Also a problem shared with the Khadas Vim3 and Odroid N2) No optimization can solve their lack of horsepower.
Since optimization will actually pay off, I expect it to happen faster than before.
People also like their arm alternatives to x86, even if at this pricepoint they’re not really competitive next to a Jaspar Lake based cpus.
We’ll probably be also seeing the Pentium 8505 and Celeron 7305 soon which will make this board look rather weak. But they have to release first.
I overall agree with your points. But what Arm fans seek in Arm products is not exactly performance but efficiency (i.e. the best consumption for a task being completed in an acceptable time). Having powerful enough cores that are slightly slower than x86 ones while eating much less power usually is the tradeoff users are willing to make. But some mid-end x86 cores are equally good if not better on certain tasks, And they work out of the box, and without the need for USB-to-TTL adapters when things go wrong, nor rebuilding one’s kernel or boot loader, so that’s already a serious competition that often justifies to pay more for x86. But when facing a highly priced Arm board, the customer should really think twice: does this board offer something I value that I don’t have on a cheaper x86 one ?
The near future will be interesting to observe because x86 tends to increase the very low-end performance (atom z8350 was dog slow but new 6413 and friends look nice), while Arm’s increasing performance risks to reach the point of making a fan mandatory if they go too far, which would void one of their persistent advantage. There’s definitely something wrong in the supply chain where it takes 4 years for an Arm core to land into a CPU that can be found on an SBC. It’s something like 6 months in the x86 world, and not everyone is willing to pay more for something 4 years old.
> But what Arm fans seek in Arm products is not exactly performance but efficiency (i.e. the best consumption for a task being completed in an acceptable time)
The biggest difference in efficiency comes from x86 boards generally having a ton more IO and way higher power limits (because they can actually reach 5ghz). With reasonable power limits and limited IO efficiency increases massively.
Phones have also enjoyed a node advantage for a while now. Which also fuels the efficiency myth.
But an A76 core on Samsung’s 8nm node isn’t going to be more efficient than a Golden Cove core with a reasonable power limit. And a task that requires a good amount of power.
> slightly slower than x86 ones
> mid-end x86 cores
Low-end x86 cores*, An A76’s single core is only about as good as a N5095. Probably lower in non-synthetic benchmarks.
So what would be an equivalent jasper lake based SBC with similar I/O? HDMI input, MIPI-CSI interfaces, 2 gigabit ethernet ports, 2x sata, PCIE3.0x4 slot, eMMC, 16GiB DRAM at a similar price point?
> HDMI input, MIPI-CSI interfaces
I might be wrong but IIRC Freekiedecakie is a Linux user so why should he care about interfaces that will work now (and most probably also the next years) only with Android? Or in general: if the use case doesn’t include video capturing why should the device in question have these capabilities?
Rockchip usually has good Linux mainlining support for it’s driver, I don’t think there will be Android limitation. 4K HDMI input is a killer feature, especially for streamers. They know, and this is a goal of Pine64 that there is a large public that want a light and efficient computer with GNU/Linux and don’t want NSA backdoors (TPM, AM bios, Microsoft, Google, etc…) or other BigData thieves, on their computer anymore, but they want opened CoreBoot, mainline kernel and drivers and stable system. Intel is now totally unusable, they cumulate bugs and unreliability, from CPU (with so much security flaws, bugs and useless complexity) to SSD. Most Atom based server boards, die in one or two year, SSD are totally unreliable, even with a RAID you could loose your 2 server class SSD at the same time, Samsung is far more reliable on this last point.
> Intel is now totally unusable, they cumulate bugs and unreliability
Actually given the kilograms of Arm-based e-waste we all have on our shelves it’s hard to point the finger at intel first for this. And the reliability problem is mostly a software issue (lack of interest from vendors to sell something that works), resulting in devices that are not fit for long-term use as “production” machines even at home, for many of them.
> Most Atom based server boards, die in one or two year
For once that doesn’t match my experience at all. I’m currently typing this on a N2600 laptop which works pretty fine:
Vendor: American Megatrends Inc.
Version: 1025C.1105
Release Date: 07/12/2012
and my file server running 24×7 uses a Supermicro board featuring an D510:
Vendor: American Megatrends Inc.
Version: 1.0c
Release Date: 05/26/2010
This one was installed on 2010-07-11 and has already reached 4yr of uptime before being put down by its UPS which died, then by the external power block which died as well. But the board has remained flawless over the years, which is why I’m extremely picky when it comes to choosing a replacement solution.
I’ve yet to find that reliable a machine that doesn’t cause me any trouble in 12 years (I predict that the clearfog and macchiatobin will easily last that long). My insistence for fanless designs counts a lot for this observed reliability.
And even regarding more recent models, on my desk at work there’s a z8350-based UP-board that’s been running flawlessly 24×7 for the last 5 years with the fan unplugged and which I’m very satisfied with.
So no, I don’t have examples of Atom boards dying in two years. There certainly are at low-cost vendors, likely those who also cut costs by using bad capacitors, but it turns out that this doesn’t match those I have among my machines and that this cannot be imputed to the chip vendor.
There was a problem with Atom servers stopping working after a year or so. But I understand this was limited to specific parts.
https://www.cnx-software.com/2017/02/08/older-intel-atom-c2000-series-server-chips-may-stop-working-after-a-while-and-theres-no-fix/
> Most Atom based server boards, die in one or two year
BS. A certain stepping of Avoton server SoCs was hit by a specific problem. One of my customers got his servers replaced in time and the replacement SoCs/mainboards work until today. I recently replaced my own D270 board in my ‘archive server’ with a RK3399 one. Worked flawlessly for +11 years and still works but software support for 32-bit x86 sucks too much now…
> SSD are totally unreliable, even with a RAID you could loose your 2 server class SSD at the same time
FYI: everybody who builds a RAID-1 out of two identical SSDs especially with identical firmware and identical power-on hours is a moron since ignoring technology (concepts for spinning rust are useless with flash storage). Last confirmation of this fact: HP Enterprise SAS SSDs fail after exactly 32768 hours.
So many nice RK3588 boards, but i don’t think they will sell many at those prices.
Indeed.
QuartzPro64 180 x 180mm (about 7.1″ x 7.1″)
Mini-ITX 170 × 170 mm (6.7″ × 6.7″)
Pine64 this is the dumbest thing ever. Fix it before is too late.
As if you would be willing to pay +300 bucks for a smaller board 🙂
That talking point doesn’t work when it comes to shrinking it just 11% to fit Mini-ITX (compared to Pi which is like 85% smaller).
I would be a bit concerned about choice for PCIe lan (anything by Intel or Broadcom would be better than Realtek), Wifi usually ages poorly and/or uses some weird controller with no open source driver and/or documentation and sata controller can be a bit of a hit or miss.
PCIe LAN? Nope, RK3588 contains two internal GbE GMACs that are usually combined with external RGMII attached RealTek PHYs (these days more likely MotorComm).
Wi-Fi ‘with no open source driver’? Sure, what else.
SATA ‘hit or miss’? Hopefully RK licensed same IP block in RK3588 as in RK356x since then at least we know ‘SATA works’. What could be a miss here? SATA PM support? Anything else?
Look at the diagram. It shows that one of the GbE ports is going through one of the PCIe 2 lanes. AIUI realtek GbE works fine nowadays, though.
I don’t understand the WiFi complaint either… The broadcom WiFi in previous Pine boards is not great, but it is supported by mainline. Also it looks like they are literally just going to hand you a PCIe 2 lane (and hopefully USB/UART) on an M.2 E key so it should be possible to use whatever Intel or Mediatek chip you want.
> Look at the diagram. It shows that one of the GbE ports is going through one of the PCIe 2 lanes
it looks like a failed copy-paste from the PCIe box on the right. Pretty sure it was supposed to be GMAC0. The chip under the port is marked “RGMII Phy”, so this port is an RGMII port hence is connected to a GMAC.
“Block diagram of a typical system based on RK3588 processor”
It’s not the block diagram for QuartzPro64.
> “Block diagram of a typical system based on RK3588 processor”
Not even this or since when do you connect HDMI in to a display? 😉
Haha, I missed the part where a monitor is connected to the HDMI Rx port 🙂
> sata controller
Now with the RK3588 TRM in the wild we know that each SATA lane supports port multipliers (up to 5 disks on each controller) and that FIS-based switching is also supported.
Possible to choose from CBS (command based switching) or FIS (frame information structure) switching for each port’s data transfer?
Impressive board, but clearly out of reach for most. I personally wouldn’t spent more than $180 for a RK3588 board or any ARM SBC. I don’t know what Pine64 was thinking. Ironically, everybody is jumping into the Rk3588 wagon and most will fail simply because of their high price. Price matters and price talks. Seems like the only “affordable” RK3588 board will be Radxa Rock5B board(under $160).
> I don’t know what Pine64 was thinking
They were thinking that software situation around RK3588 really sucks (there’s only a Rockchip provided Android BSP based on a forward ported 2.6 kernel) so it will take a long time until there’s something ready for end-users.
As such they’re now doing an RK3588 board for potential mainline developers exposing all interfaces that costs them +300 bucks per piece to sell it at a lower price once it’s ready.
The high price and strange form factor are unappealing to most consumers, but as I understood Pine64’s blog post, this is not aimed at consumers but very much a board for dedicated developers to explore the RK3588 with. They repeatedly & clearly state that the software isn’t there yet, and is not going to be for a long time, and that this won’t be sold to people who don’t know what they’re getting into.
I assume that they prioritised exposing every possible interface of the RK3588 (for developers to experiment with), and to that end made the board as big as it needs to be to cram everything on it comfortably, and as expensive as the BoM demands. In that light, this is a reasonable product; and once the Soc matures, they’re likely going to follow it up with a cheaper, smaller, more focused board for the broader market. (But I too wonder if it would have demanded too many compromises to try and hit the mITX 17×17 dimensions).
> as I understood Pine64’s blog post, this is not aimed at consumers
Exactly. Since it’s easy to ask I did exactly that and TL Lim confirmed: the board dimensions are meant to show that this board is supposed to sit on a developer’s desk and nowhere else. Once software support situation for RK3588 has improved there will be consumer devices as well (most probably following their usual form factors – assumption by me).
The design/dimensions emphasize on their March update: ‘To say that the device is ‘not end-user ready’ would be an understatement’
Straight copy from RK3588 EVB only with JTAG port removed.
QuartzPro64 pre-orders will open soon, and you can register your interest.
https://preorder.pine64.org/#/quartzpro64
The board will be subsidized and sold for $150 to developers despite the ~$300 BOM cost.