Home > Rockchip RK33xx > Rockchip RK3399Pro SoC Integrates a 2.4 TOPS Neural Network Processing Unit for Artificial Intelligence Applications

Rockchip RK3399Pro SoC Integrates a 2.4 TOPS Neural Network Processing Unit for Artificial Intelligence Applications

Rockchip RK3399 (aka OP1) SoC was launched in 2016 with an hexa core Arm Cortex A72/A53 processor, Mali-T860MP4 GPU, support for 4K video decoding, and high speed interfaces like USB 3.0 and PCIe, as well as Gigabit Ethernet. The processor is found in Chromebooks, TV boxes, development boards, and other devices.

The company has unveiled an upgraded “Pro” version of the processor at CES 2018. Rockchip RK3399Pro appears to have most of the same features as its predecessor but adds a neural network processing unit (NPU) delivering up to 2.4 TOPS for artificial intelligence and deep learning applications.

The company claims that compared to traditional solution, the computing performance of typical deep neural network Inception V3, ResNet34 and VGG16 models on RK3399Pro is improved by almost one hundred times, and power consumption is less than 1% than A.I. solutions implemented using GPU acceleration.

Based on the information provided in the chart above (source: elDEE on twitter), Rockchip RK3399Pro outperforms other high-end SoCs (for such tasks) including Apple A11, and Huawei Kirin 970, both of which also features an NPU, and even offers better performance than NVIDIA TX2.

RK3399Pro NPU supports 8-bit & 16-bit operations, is compatible with various AI software frameworks and APIs, including OpenVX and TensorFlow Lite/AndroidNN API, and comes with AI software tools capable of handling Caffe/TensorFlow models. An RK3399Pro hardware reference design can also be provided to speed up development, but no details were provided.

Click to Enlarge

 

Via Liliputing

  1. January 8th, 2018 at 14:32 | #1

    I have to say, this is a very wise strategy, adding a NPU in the existing chip instead of a new one. RK3399 PRO is pin to pin with RK3399. So ROCK960 PRO is on the way 😀

    • Alejandro M.
      January 8th, 2018 at 21:07 | #2

      @Vamrs did you get confirmation from RK that it’s pin-to-pin compatible? what about the functionality, did they change anything additionally to adding the NPU? did they say anything about price?

  2. January 8th, 2018 at 14:34 | #3

    @Vamrs
    Do you already have samples or did Rockchip provide you with a devkit? If so, is it working as advertised in the table above?

  3. January 8th, 2018 at 18:51 | #4

    Wondering it’s a in-house design or IP from other vendor(like ceva core in rv1108).

  4. Alejandro M.
    January 8th, 2018 at 19:31 | #5

    sorry for my ignorance but is 2.4TOPS good/average/poor compared to the other ASICs that have come out recently?

  5. January 8th, 2018 at 19:39 | #6

    @Alejandro M.
    Kirin 970 NPU is said to offer 1.92 TOPS, and Apple A11’s Neural Engine can deliver 0.6 TOPS.

    The new Myriad X should be 1.0 TOPS (10x Myriad 2 performance).

    Tegra X2: 1.5 TOPS (single precision). I understand the latest NVIDIA Xavier SoC can do 30 TOPS, but that’s quite a different power envelop.

  6. January 8th, 2018 at 20:29 | #7

    @cnxsoft

    No. We haven’t got the real sample of rk3399 pro. I think it’s just announcement because CES is a good time to announce something new 🙂 I guess it’s still IP from third IP companies. The price for RK3399pro is very competitive, considering the availability and Rockchip’s open-source strategy now. I think RK3399pro will be a success.

  1. No trackbacks yet.