Allwinner MR133 Processor Targets Robot Vacuum Cleaners

It’s been a while since I’ve seen a new processor from Allwinner, so today I got curious and visited the company website hoping to find maybe some news about the first Cortex-A7x processor from the company. But instead I found something else intriguing with the new Allwinner MR133 processor specifically designed for “SLAM/VSLAM intelligent robot sweeper solutions”. In other words, Allwinner has just launched a processor for robot vacuum cleaners…

Allwinner MR133

Allwinner MR133 key features and specifications:

  • CPU – Quad-core Arm Cortex-A7 processor @ 1.8GHz with 32KB L1 I-cache + 32KB L1 D-cache per core, 512KB L2 cache, low-power CoolFlex power management architecture
  • GPU – Arm Mali 400MP2 with support for OpenGL ES 2.0/1.1, Direct3D 11.1, OpenVG 1.1
  • Memory I/F – 32-bit DDR4/DDR3/DDR3L/LPDDR3/LPDDR4
  • Storage I/F –
    • eMMC 5.0 flasj, compatible with eMMC 5.1, support Full Disk Encryption(FDE)
    • 8-bit TLC/MLC/SLC/EF NAND flash, supports FDE
    • LDPC/80-bit BCH/1024bytes
  • Video Engine – H264 HP encoder 1080p @ 60fps, JPEG encoder up to 4096 x 4096
  • Video Input – MIPI-CSI2 V1.00 and MIPI DPHY V1.00.00 up to 13M @ 15fps, or 8M @ 30fps with 4 data lanes
  • ISP – Up to 4224×3168 resolution with adjustable 3A functions, including automatic exposure(AE), automatic white balance(AWB) and automatic focus (AF)
  • Audio
    • 2x audio DAC and 1x audio ADC
    • 3x analog audio inputs and 1x analog audio output
    • Capless stereo headphone driver
    • Up to 2x I2S/PCM controllers for connecting Bluetooth and external audio codec
    • Integrated digital microphone, supports maximum 8 digital microphones
  • USB – USB Host, USB 2.0 OTG
  • Peripherals
    • SDIO 3.0, RSB
    • 4 x TWI, 2 x SPI
    • 6 x UART, 3 x PWM
  • Security Engine – AES,DES,3DES,XTS summetrical algorithms, MD5,SHA,HMAC hash, RSA, ECC Pubic Key algorithms, 160-bit hardware PRNG, 256-bit hardware TRNG, 2.5K-bit EFUSE for chip ID and security applicationConnectivity
  • PMIC – AXP2585 PMU, AXP15060 BMU
  • Package – FBGA 413balls,  12.3 x 12.8mm, 0.5 ball pitch, 0.3 ball size
  • Process – 28nm HPC

WiFi can be added to the chip via XR829 module or other wireless modules. The company provides Tina Linux and Android 8.1 for the processor. SLAM stands for Simultaneous Localization And Mapping, while VSLAM adds Visual to the acronym, and relies on images instead of lasers allowing for 3D mapping.

Allwinner is apparently still operating the company through 4 or 5 business units, where the same processor is used by various units under a different name and supported by a different SDK, because the product page for Allwinner MR133 appears to be filled with the exact same content as the product page for Allwinner R311, except the latter is designed for “intelligent voice interaction with screen, image recognition solutions”.

Share this:
FacebookTwitterHacker NewsSlashdotRedditLinkedInPinterestFlipboardMeWeLineEmailShare

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress

ROCK Pi 4C Plus

19 Replies to “Allwinner MR133 Processor Targets Robot Vacuum Cleaners”

  1. Not sure where they are going with this. This is no AI acceleration. The Mali400 GPU is not capable of GPGPU so you aren’t doing AI on it. You can run AI on the A7 but it is not very fast, and you aren’t going to be processing 4K video with it. Unclear what this chip is targeted at.

        1. Then why does it need 4K sensor support and the ISP to process it? AW has some nice IP pieces but they sure aren’t stringing them together in useful ways.

    1. “Did someone say a new buzzword? Better spin off a new version of an existing chip!” — seemingly every marketing person at Allwinner.

  2. I just went through the specs on every chip AW has listed on their website and I’m not interested in any of them. This splitting into business units is a bad idea, it may even kill the company. I’d guess AW is still being run by chip people, and like all chip people they don’t value software highly enough.

    1. I suspect they don’t want to spend money on anything new. They’ve got a bunch of TV box IP blocks and they are going to just keep slicing them up into new chips with different core configurations. They can’t compete with the high end smart phone stuff so it’s not worth them buying the best cores ARM has and designing a bunch of new or improved perpherials around them.

      If I was Allwinner I’d be using the stuff they have and the opensource community (linux-sunxi) to push out high volume chips that rivals the i.mx6ull on price. They have (crap) wireless IP so they could put out a complete WiSoC for Amazon Echo style smart devices.

      1. It is a form of MLM in many respects, they keep the SoC profit and put the Software, Hardware, design risk on the lower MLM members ( down line distributors )

      2. The camera support is good, but then there is no vision AI hardware that can make use of it. The A7 is simply too slow for running CNNs.

        This chip would be a lot more useful if…
        1) Replace the useless GPU with an AI CNN core.
        2) Current ARM Trustzone, not this 10 year old DES stuff. IOT devices need protected SSL keys. The cypto core is too old and is designed for DES encryption. Current Trustzone supports the needed security features.
        3) 100Mb Ethernet. 100MB Ethernet would turn this into a useful camera chip.

        They can buy IP for a nice CNN core here…

        They already own the current Trustzone and enet since it is in their other processors.

        1. >IOT devices need protected SSL keys.

          While I think that is true (more not putting keys in the same address space as your kernel, network stack etc) not many real world IoT devices actually do that.

          1. That is true because most IOT SOCs don’t have the necessary hardware. Every phone SOC has it.

          2. Every phone soc has the hardware but you can’t get a phone soc on a solder down module or a multichip BGA/LGA package with FCC etc certs for ~$10. The best you can get is a Cortex M3/4 or R4 with an SDK that supports some sort of encrypted boot up. After which all of your secrets are in the same address space as all of the firmware.

            Newer Cortex Ms have the trustzone stuff you want but will anyone use it? Vendor SDKs aren’t implementing basic stuff like marking the stacks and heap NX using the MPU so I don’t have much faith they’ll come up with a Trustzone enabled stack (Like ARM’s PSA) that will actually work.

            That’s why I think Allwinner could do something with the IP they already have but they seem to be obsessed with spinning the same smart TV SoC with HW video decoders over and over.

          3. You can use an external chip like Microchip ATECC508A to secure the SSL keys. But doing that adds $1 to each device.

    1. Both the Roborock AND the NES/SNES classic! Such versatility!
      With this new chip which definitely isn’t exactly the same as five of their other chips just with a different silkscreen, you’ll never be playing Mario Bros on your Roomba again!

Leave a Reply

Your email address will not be published. Required fields are marked *

Khadas VIM4 SBC
Khadas VIM4 SBC