Allwinner is currently at CES 2014 showcasing their latest Octa-core A80 SoC, and corresponding development board, the A80 OptimusBoard. Apart from the picture below, there were very little details for the initial announcement, but we now know, the board comes with 2 GB RAM, 8 GB NAND flash, and other features of the board, as well as the company behind the GPU in AllWinner A80.
A80 OptimusBoard specifications:
- SoC – AllWinner Ultra Core A80 4x Cortex 15, 4x Cortex A7 big.LITTLE processor with an unnamed Imagination Technologies PowerVR GPU
- System Memory – 2GB DDR3
- Storage – 8GB NAND Flash
- Video Output – HDMI 1.4a
- Connectivity – Dual band Wi-Fi 802.11 b/g/n, Bluetooth 4.0, and Ethernet (GMAC)
- USB – 1x USB 3.0 OTG HIC, 2x USB 2.0 host ports
- Audio codec – AC100 Codec
- Camera I/F – 12MP MIPI CSI
- Debugging – UART and JTAG
- Expansion – 32-pin GPIO header
- Misc – IR receiver, reset and power LEDs.
- Power management – AXP806, AXP809
- Power Supply – DC IN (5V?), and battery
- Dimensions – 135x70mm
So AllWinner A80 features a PowerVR GPU from Imagination Technologies, most probably one of PowerVR Series 6 GPUs (G6xxx) compliant with OpenGL 3.0 and OpenCL 1.x. This will probably disappoint a few people that hoped for Mali GPU in order to get better support in Linux.
That’s all I know for now, and I could not find details about supported operating systems, although Android is a given (Kitkat?), the company may also have decided to support one or more Linux distributions. Price information is not available either.
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.
51 Replies to “A80 OptimusBoard Features AllWinner Octa-Core big.LITTLE SoC”
Oh no. Not PowerVR again!
PowerVR has drivers for Linux.
Look at beaglebone. They use TI with PowerVR.
Mali don’t have fully working opensource drivers.
We are still force to use binary blobs.
And PowerVR is much faster than Mali
Still no SATA port…
Good choice power vr they maybe find out all the driver problems,better than Mali and keeping the low as I said before,this is different with rock chip about GPU.
Since the time powerVR exists, the bugs are generally not patched, and the linux support is the worst of all the existing GPU on earth. Look at early series of eeePC and other devices using them…
I don’t think that imagination technology are going to be behind in this competition, they are the leader in this platform, Mobil device,way Intel use their technology and apple still using. Mali its young and getting into.
The problems is Android os still young and far to get in this competition about GPU hardware acceleration, GPU compute,Linux
Driver and others if Google want to that,all is just software, they are companies working on this matter.
Android kit kat nothing mentioned about this,some peoples,says its going in the next version
Oh no, not PowerVR 🙁 With Mali at least there was hope with the Lima driver even though it’s not fully working but with PoweVR you’re pretty much doomed.
And no, PowerVR is not faster than Mali, the new Mali GPU as really fast, just look that the new Rockchip announce here on CNX.
Anyway I’ll probably grab one to use as a server so I don’t care about the GPU but SATA would be really nice.
No, PowerVR isn’t the leader, the leaders are Nvidia with Tegra and Snapdragon with Ardreno in term of performances. Closed source official mali driver (and ARM open source X11 driver) are not really optimized, some have already proven this. Open source driver, are not complet, but are alredy faster and should be a gage of stability. No hope of stability with PowerVR drivers, and some opensource hackers that tried to develop Opensource PowerVR have a really negative vision of the thing.
There is only two GPU with fully closed driver and zero open source driver, the IT PowerVR and the Broadcom Videocore IV (used in Raspberry Pi, making the thing unbootable without closed source driver):
All mobile GPU makers are pretty bad when it comes to OpenGL support in their proprietary drivers:
Which is why open source drivers in Mesa/Gallium3D are so important.
At least the Lima project has made a pretty good start at an open driver for Mali. The Freedreno project for Adreno is doing even better – there’s code in the master git branches of Mesa and libdrm. And etnaviv for Vivante seems to be slowly doing something. But there’s nothing for PowerVR, as far as I know.
And as far I know, Intel thrown IT PowerVR and replaced them by their own GPU since at least 1 or 2 years, and their own GPU is serveral time faster, but also consume several time the electric power of ARM plarforms, so they are not adapted at all to the ~5W ARM SoCs devices world.
Good or bad as open source, while they are doing a good and decent hardware with the right software implementation in different platform support is OK to me, look for apple and windows. But I like open source.
As far I search and read;
Windows driver support OK we know,own platform
Windows mobile driver don’t know but I think OK
Android driver (Linux) no good, up top Google
Intel: windows and Linux support driver,with Linux ok (open source).
Amd: good hardware and software implementation drivers no good Linux support (android)
Nvidia: good hardware and software implementation driver for windows (x86) and Android pretty good and own arm license.but some bugs.
Apple:pretty good, even using power vr,good and lightweight OS.no driver problems,maybe some bugs.own system.
Mali: waiting for Lima driver,some CPU maker no APU video engine,their is need or not need about APU.arm recently
Acquired a united engine platform for Mali about GPU compute.
Imagination technology: OK ,no open source, several companies working with them,own drivers.
Allwinner: own libraries and APU video engine. Why they not use Mali,we don’t know the explanation.
I think in mobile graphics in android are getting better.
Correct me if I am wrong
For a status about open source drivers for mobile GPU, you may want to check the upcoming talk at FOSDEM.
Last year report is here: http://www.cnx-software.com/2013/02/14/open-arm-gpu-drivers-fosdem-2013-video-and-call-to-arm-management/
AllWinner uses Mali for 2D/3D graphics, and their own CedarX VPU. That’s because Mali does not support video decode. That a common misconception: http://www.cnx-software.com/2013/12/10/most-embedded-gpus-do-not-support-hardware-video-decoding-acceleration-the-vpu-does/
I guess you already know, but most mobile GPUs only support “OpenGL ES”, that means “OpenGL” app would not work unless they use software emulation. Nvidia Tegra K1 changes that.
Thanks for the information.
So reverse engineering done to the GPU/VPU in allwinner soc,maybe the key for hardware acceleration and GPU compute.
Allwinner use Mali in their own technology gpu/VPU xcedarus,because
Mali doest no support video decoder,that’s why allwinner it’s using power vr as GPU for the video decode engine.
So far the emulation no bad compared to other Chinese soc , some people says about arcade emulation games some, Mali OK,
Power vr too,about android high resolution games, Mali no good and power vr OK.
Arm bought united 3d engine game software platform for Mali,and some video looks good, plus GPU compute.
What I am missing?
What they need?
Linux-android a real true software implementation driver
So reverse engineering or Google android is the key.
Or Google android its no a open source
AFAIK PowerVR does no video decoding either. It’s a 2D/3D GPU. Not a VPU.
Etna_viv is almost stalled. The author was asking for help, but no one is able (or want) to help with GC2000 (wandboard quad and others) development.
After wandboard (freescale / vivante etc…) graphics hell, my respect for opensource drivers have greatly enhanced.
(sorry for my bad english).
I hate PowerVR for their bad Linux support, and their attitude against open source
Apple must be really dumb to go with PowerVR if they are so bad.
Apple tablets must be really slooooow.
You got the money, then PoverVR is talking to you.
But no hobbyist or non-fortune500 company has that sort of cash, so PowerVR is absolute shit for hobbyists or budding companies.
I’d like to start by thanking you all for your feedback. It’s important for us to understand your issues and requirements; let me assure you we are listening and doing our best to address them.
I would also like to take this opportunity to address some your comments and provide some resources for developers looking to build custom ROMs or use our open source binary drivers (yes, we have them just like everyone else) for their projects.
Let’s start with some clarifications on the driver side first.
– PowerVR reference platforms for driver development are Linux-based, and always have been. Our support for Linux is therefore extremely strong, and any concerns about our support for the Linux community are usually based on out of date or incorrect facts.
– We always open source the kernel mode code for our drivers, to enable Linux developers to build kernels without restriction. However, just like every other GPU provider we know of, we do not provide open source code for the user mode section of our driver, as a lot of confidential information resides in that code. Also this code is closely tied to the hardware, which means each driver must be modified by our licensees for each separate chip to link it to all the hardware functions within the chip, and to ensure it works optimally. Imagination could never disclose any proprietary information of our licensees, such as hardware details or address maps.
– Imagination is working very closely with the Linux community, significantly increasing the number of code releases upstreamed over the past year. We are extremely keen to maintain our current level of involvement especially following our acquisition of MIPS and build on our already broad and active links with the Linux community for all of our IP
– We have been aware of various open source initiatives over many years attempting to produce open source GPU drivers for PowerVR and other GPUs. However inevitably these drivers are significantly lower performance than production drivers from Imagination and our licensees, as developers of these drivers have never tried to obtain the architectural information necessary to optimise performance. We know our licensees are always keen to get the best possible performance platform for their end customers, which is why the GPU driver is always provided as part of the BSP provided with their SoCs, and why we always recommend that developers get their GPU drivers directly from our licensees rather than trying to use open source.
Secondly, the links below contain DDK (Driver Development Kit) source packages distributed through our customers:
OmapZoom (up to 1.9)
Android Open Source Project (up to 1.8):
Rhombus Tech (1.10 onwards)
Please note the user mode section of our driver is provided only under the standard commercial license to PowerVR GPU customers. This code is closely tied to the hardware, which means each driver must be modified by our licensees for each separate chip to link it to all the hardware functions within the chip and to ensure it works optimally.
Finally, let’s talk about performance. We believe we currently have the technology that enables our licensees to offer the same or better levels of graphics and compute performance while (and this is very important) keeping the power consumption adequate for a mobile device. Our unique PowerGearing and PVR3C triple compression technologies are essential here. It allows our multi-cluster GPU designs to ramp up raw performance when required and then turn large parts of the design off when unused. This prevents throttling and ensures a reliable, solid user experience. If you add to this the improvements in efficiency we’ve made to our TBDR architecture (all competing solutions are either IMR or brute-force, tile-based designs), our Rogue line-up (Series6/Series6XE/Series6XT) offers the best PowerVR GPU IP available right now.
Technology PR Executive
Imagination Technologies Limited
t: +44 (0)1923 260511
http://www.imgtec.com | http://blog.imgtec.com | @ImaginationPR
But, Where are the UM drivers for the Atom processors? And it was nor Imagination Nor Itel who released a proper KMS driver:
Besides Phoronix is littered with teasers from Imagination that opensource drivers are going to available. Haven’t seen any.
But if there is OSS support now… good for you and hopefully us too. But that doesn’t mean older statements are untrue from now on. Your OSS support has been crappy, period.
Please don’t forget that once your hardware is released without open drivers, it’s your responsibility to release updated, UM, drivers for new Kernels/Architectures/X.Org/Wayland/SurfaceFlinger.
If you do not, your hardware becomes obsolete within weeks. Nobody is going to buy things that breaks within weeks.
If you want to be open source friendly, start by being open. Make a list of available drivers on your site. List their compatiblity. Create a public bug tracker. And if bugs are not going to be fixed be honest. Don’t let the bug reports rot.
Even if it a part soled/used by your customers. If they do not comply with the above. It will reflect badly on you too!
And joining the effort is more cost effective, for both parties, than letting every customer of your muddle trough.
And I can’t imagine that all parts UM drivers need to be closed. You can split is in parts that can and can’t.
There are opensource GPU drivers for AMD, Nvidia, Quallcom (ardreno), Intel. There are a few in good progress. ARM (Mali), Vivante.
They may or may not be on par with the closed drivers. But at least there are. For Imagination there are non, nothing, nada.
Now, why would that be? I think I have a good view on the answers. Do you?
N.B. For Intel there are no closed drivers! So your statement is not accurate. And Intel is already moving to Embedded Linux space with their own GPU’s!
I really hope you get your get to see the light. Because you’re obviously capable of making decent resulting hardware.
But for now your statement is just PR patchwork.
We have imagination technology attentation,the other part and us,good CNX
I like to says everything is valid,but all depending for the way that all we see,means,that me as costumer,we as a community even trying to do something good for us,themselves and the other part the company .
The real true right now, community are getting bigger to HELP and that is what I am looking for as a costumer and for business.
For what I see on the last couples years, open source, its a big thing in this worldwide in this matter.
Arm,Linux, android?, ecosystem are bigger and getting each day,we all know this trough the last CES 2014
Open source I like, because I can see and manage all the differents way can to be apply to a something in this case hardware-software for optimization with Linux, android, community and the companies involved, and results has be something in common ,the benefits.
As we see in this century, windows; struggling(no open source, licensing), Intel; trying to get into this business (expensive, licensing)doing their own thing,good for them.Apple; smart company,smart move,(expensive,quality)but no open source,I don’t know for how long they are going to keep this pace,loyalty customer.
So Linux-android? That’s the key for the future open friendly source or Linux derives, but the best of this its big community.
Get it right open source is the future
OK I was looking forward to see the unveiled of the A80 was nothing like that, same info,no interview with Eva,spokeswoman.
No video either, more for the other chipset.
I want more info about Chinese SOC, can CNX supply that.
There was just the info about AllWinner A80 I mentioned in the post above, as well as some details about Rockchip RK3288 I wrote in another post. For videos, Charbax is your only hope…
@Alexandru Voica Our support for Linux is therefore extremely strong
Where is your API and Register description which to allow open source drivers to be made? Binary blobs are not “open friendly source”
Charbax interviewed AllWiner at CES 2014. Nothing really new about AllWinner A80, but still interesting at they talk about other products too.
I’ve already provided this answer to one of the people commenting on our blog. I think the main problem with your reasoning is you are confusing us with a chipset company. We cannot be compared with Intel, NVIDIA, Qualcomm in this respect because we sell GPU IP, not chips.
If we had sold SoCs, then yes, we would have had more flexibility with releasing our drivers. But we have tens of customers, each with a different microarchitecture, each with different CPU/GPU configurations, to which you add things like VPUs (hardware video encoders/decoders), ISPs, display controllers, memory controllers, sensor hubs and many other elements which need to communicate between each other.
Therefore, it is almost impossible for us to release “generic” drivers. This is how things usually work:
a) customer licenses PowerVR GPU, has access to our DDK (Driver Development Kit)
b) customer starts integrating said PowerVR GPU into an SoC that can have very specific microarchitectural features; e.g. CPU ISA (x86, ARM v7/v8, MIPS32/64, etc.)
c) customer works with partner OEMs/ODMs to target specific OS (Android, iOS, Linux, webOS, Windows RT, Windows 8, etc.)
d) SoC is sold to OEMs/ODMs that manufacture end devices
e) end device is released
You can see how it would be very complicated for us to step in and assume the role of a customer or OEM, mainly because some elements of their work have nothing to do with us. Releasing THEIR drivers without their approval would further complicate things even further and jeopardize our working relationship.
I understand the needs of the open source community but you have to address your requests to the right people. If you want a certain OS to be supported, talk to the silicon vendor or the OEM/ODM and see what can be done. We can help by sending them your feedback (and obviously supply our customer the DDK for that OS). But the DDK is something that is a part of our GPU IP package, therefore we cannot release it publicly.
Let me try and give a clearer view of how drivers for mobile SoCs work with one simple example.
We’ve recently partnered with Hardkernel and provided full access to the PowerVR DDK for their ODROID-XU boards. Thanks to this collaboration, the ODROID-XU board supports the full OpenGL ES 2.0 and OpenCL 1.1 EP APIs on Android for the Samsung Exynos 5410 SoC.
In theory Hardkernel could have supported Linux too since we granted full access to our OpenGL DDK, but Exynos 5410 was designed as an Android part and certain drivers, like the display driver, were not written with Linux support in mind.
It’s common in mobile to just target one OS. The Atom SoCs you mention are in a similar situation; Intel has said that it designed those platforms for Windows; it might well be that they’d be good on other OS but they weren’t designed or implemented for those OS and there’s nothing we, as the IP supplier, can do about it.
So you are saying that mostly of the SOC, like, rock chip,allwinner, and others, that they don’t have their
Own GPU,are doing or using the wrong OS,mean that it’s no optimization between CPU and GPU,because its a
Mobil platform, no desktop and that’s why they are using android drivers design and,no Linux,so they are not doing a good job.
So nvidia and Qualcomm they allways are going to better,because for the bond between CPU/GPU .
Saying this its a bad business for imagination technology, even making a powerful GPU ,doesn’t make any
Sense to me,sorry.therefore
Which should be solution a open source
No, you have misunderstood me completely. What I was explaining was the fact that those open source drivers you ask for may not help you in the end because other parts of a mobile SoC might be completely closed. So having ONLY open source drivers for the GPU will not help if you can’t get the display driver, for example, to work.
Our customers (Allwinner, Rockchip and others) have and continue to do a perfectly good job at integrating the graphics core in their designs.
Please don’t try to twist my words to justify your personal opinions. I am simply trying to explain how things work in this business while also attempting to come up with a reasonable solution.
Again, just throwing code out there and hoping it would somehow magically work by itself is not a solution.
> “Also this code is closely tied to the hardware, which means each driver must be modified by our licensees for each separate chip to link it to all the hardware functions within the chip, and to ensure it works optimally. Imagination could never disclose any proprietary information of our licensees, such as hardware details or address maps. (…) This code is closely tied to the hardware, which means each driver must be modified by our licensees for each separate chip to link it to all the hardware functions within the chip and to ensure it works optimally”
@Alexandru: I am no expert so I don’t precisely understand why there can’t be a “generic” driver. I perfectly understand ARM does not have ACPI and cannot self-describe, but as long as we have proper device-tree support (GPU I/O address, etc.), why would a GPU driver have to be different across different SoCs ?
Second question: can anything be technically made (e.g. collaboration with ARM) in order to solve this, such as standardize something (e.g. mandate read-only device-tree at a standard location in a small ROM on all ARM SoCs/reference-designs) so that future SoC can have generic drivers (and kernels) ?
First my appreciation on continuing to reply. Too bad I didn’t get any from you. I wonder why… Something with nails and head perhaps
You’re comlaining that you can sit on the OEM chair. While true. That doesn’t mean Imagination can be the deliverer/maintainer of the driver, albeit with the OEM or not.
The other complaint is about diversity. Let cosider.
CPU Architectures: x86 32/64 BIT,MIPS,ARM,SPARC,etc.
GPU Interfaces: Direct3D,OpenGL,Gallium,etc.
VPU Interfaces: DrirectVideo,XV,V4L2,va-api,vdpau,etc,
While that seems a lot. The hardware in question does not change. You IP core gets reused. Thus the driver doesn’t need to change. Mostly the interface to the driver, with is roughly about 10% of the code.
Intel, NVIDIA both stated that there drivers for Windows and Linux are 90% the same.
The Linux kernel is ported to almost every type of CPU there is. The code difference/glue code is less than 10% of the whole kernel.
There are midways too. eg. Imgantion replease only OpenGl drivers and the vendor releases specialized drivers like Direct3D.
The real issue is revenue here.
1. There is no clear visible return on investment.
2. Maintaining ‘old’ hardware, might hold of new custumer purchace. Why buy new if the old one is still working.
3. If the OEM does the porting than you don’t need to invest.
But it’s like I said before. If the OEM f*** up there “due diligence” in support it reflect bad on you too.
It like building a good and cheap car engine and only provide documentation to a single OEM. And then shield of the engine compartment. And forbid any one other that the OEM to even look at the engine.
1. The OEM is the only one who can do repair and maintenance
2. The OEM can price as it pleases. It has monopoly
3. The OEM can refuse at it pleases.
4. Anyone can fill in the rest.
The above mentioned practice is illegal in the EU. The owner must be able to bring his car to an independend garage en the sellers->manufacturers are obliged by law to provide documentation.
In the the end the product that can be fixed maintained by anyone is favoured the most. Doesn’t lock you in and insures more reasonable prices.
As soon as one GPU vendor realizes that providing full documentation and open source drivers is actually good for profit. the rest is has a very simple choice. Either accept a smaller cut or Open up. Not choosing simply end in bankruptcy.
Being open and unambiguous about it tells you are sure of your product and shows that your customers are not lock-in to you. And in turn their customers in the OEM.
Review the agreements with your vendors and tell them that openness will create more revenue. Your trying to sell hardware and if it breaks by crappy drivers and support nobody likes it. Make it about the hardware not about the software.
Work with your OEM’s. Just putting a SDK in their face and telling it’s on them is ridiculous. Nobody understands that.
Either I’m on the right track or only down talk leads to reply.
Or like Achmed said: Silence!…. I kill you!
PowerVR is bad m’kaay.
Good falcon 1
Nvidia make statement for initial open source, for people and community,a very good first step.
If anybody interested in wayland and powerVR GPU
it is wayland 1.3 compatible
Vaporware? At MWC the marketing folks are anouncing it again. But it one ask directly (when and how much?) the comapny says it is an internal dev board with no plan to sell it.
CubieBoard4 will use this SOC but I’ve not idea when it will be launched.
Strange. I understood AllWinner promised some board to linux-sunxi developers. If the Opimus board is not readily available, I’m not sure they’ll really want to work much on these.
At linux-rockchips we are making some progresses regarding graphics support. Not to be said about the GPU part which, thanks to ARM who released the different parts of the stack and different developers who integrated them, has come to work under linux.
But perhaps soon about the VPU: it comes to be a On2 Hantro H1 Decoder + G1 Encoder(s).
Having On2 been purchased by Google, who has integrated these products into the WebM project, has given us the right direction to go:
ask Google/WebM Project for the IP drivers, ask them to open source them, regardless of the fact that we will have to adapt them.
For sure we will have to adapt them and, FOR SURE, we will do it!
(latest progresses at http://linux-rockchip.info/mw/index.php?title=VideoAcceleration)
I want 8GB of RAM to use the board as a server running FreeNAS. FreeNAS.com
Haha, good one but I got a Tegra K1 board ( Jetson TK1 ) for less on ebay. And Tegra K1 has mainline kernel support, SATA, open source GPU driver … shall I continue ?
@Marius Sure – just noticed it for sale. Does seem a little pricey for a board with an Allwinner SOC…
Jetson TK1 on Amazon: http://www.amazon.com/NVIDIA-DEVELOPMENT-PROPRIETARY-Motherboards-CD575M/dp/B00L7AWOEC 193USD