You get a big boost in what your device can do with the rk3588s2. This soc uses advanced 8nm technology. The rk3588s and rk3588 work fast and save power. They are better than older chips. You find important parts like an octa-core CPU, Mali-G610 GPU, and built-in NPU. These help play 8K video smoothly and run ai tasks right away. The rk3588s2 and rk3588s work great in devices that use batteries, small computers, and smart screens. The chart below shows how rk3588s stacks up against other socs in how they are made and how much energy they use.
Key Takeaways
-
The RK3588S2 chip makes devices faster and saves power. It uses an advanced 8nm process. It has a strong CPU, GPU, and NPU.
-
It can do quick AI jobs like image recognition and robotics. It works well with popular AI frameworks and tools.
-
The chip can play and record smooth 8K video. It supports more than one screen. It handles many video and audio formats for great multimedia quality.
-
Radxa boards with RK3588S2 give strong performance and use little power. They are easy to expand for smart devices and embedded systems.
-
The RK3588S2 helps make fast, flexible, and energy-saving devices. These devices are great for AI, multimedia, and industrial projects.
AI Performance of RK3588S2
NPU and CPU Architecture
The rk3588s2 makes devices much faster. It has a strong cpu, a good gpu, and a special npu. The cpu uses two Cortex-A76 and four Cortex-A55 cores. This helps with hard tasks and ai computing. The 8nm process saves power and works well.
The Mali-G610 MP4 gpu handles graphics and ai jobs. It supports OpenCL HDR tone-mapping for video and ai. The npu can reach 6 TOPS speed. This means it is great for real-time ai jobs like image and speech recognition. The LPDDR5 memory, up to 16GB, lets all parts share data fast. This setup lowers wait times and helps everything run better. It is important for tough jobs.
| Component | Architecture & Features | Contribution to AI Task Acceleration |
|---|---|---|
| CPU | DynamIQ: 2x Cortex-A76 + 4x Cortex-A55, 8nm process | Strong multi-core performance for ai inference and multimedia workloads, balancing high performance and efficiency |
| GPU | Mali-G610 MP4, OpenCL HDR tone-mapping | Efficient graphics and ai processing, accelerates multimedia and ai tasks with low power use |
| Memory | Shared LPDDR5 (up to 16GB) | Fast data sharing among cpu, gpu, and npu, critical for real-time ai workloads |
| Integration | CPU, GPU, and triple-core NPU (6 TOPS) | Complements ai processing with computational and graphics acceleration, enabling real-time ai tasks on-device |
The npu in the rk3588s2 is over three times faster than older chips like the ODROID-M1S. This means you can do hard tasks and ai computing much quicker.
AI Frameworks Support
The rk3588 and rk3588s work with many ai frameworks. The npu supports TensorFlow, MXNet, PyTorch, and Caffe. You can also use Caffe, TensorFlow Lite, ONNX, Keras, and Darknet on boards like Banana Pi. This lets you pick the best tools for your ai needs.
-
RKNN-Toolkit2 helps you change and run models on the rk3588s2 and rk3588s.
-
You can use Python or C/C++ APIs to set up models.
-
The toolkit works with ONNX models (OPSET 12-19) and custom operators.
-
It gives special boosts for the rk3588 series, like int4 * int4 to int16.
-
You can use it on Ubuntu 18.04, 20.04, or 22.04 with Python 3.6 to 3.11.
Note: This means you can use your favorite ai models on the rk3588s2 easily.
Real-Time AI Applications
The rk3588s2 and rk3588s are good for real-time ai in many areas. The npu, with 6 TOPS, gives fast ai results for edge devices. You find these chips in boards like Radxa ROCK 5C and ROCK 5C Lite. These boards help with machine learning, computer vision, robotics, automation, and ai gaming.
-
Machine learning inference
-
Computer vision
-
Robotics
-
Automation
-
AI-powered gaming
The rk3588 and rk3588s are great for edge ai jobs. In places like factories and smart cameras, the npu gives quick ai answers and real-time choices. The small size and low power use make these boards good for built-in systems. You can get up to 32GB LPDDR4x memory, HDMI 2.1 for 8K video, and many ways to connect. These features help hard tasks and ai computing run well, even when things get tough.
The rk3588s2 gives top performance for real-world ai jobs. With a TDP of 5-10W, you get strong ai power without using too much energy. This makes the rk3588s2 and rk3588s great for smart sensors, robots, and cameras that need to be fast and save power.
Multimedia Capabilities of the SoC
8K Video and Multi-Format Support
The RK3588S2 lets you enjoy amazing video. It can play 8K video at 60 frames per second. It uses H.265 and VP9 codecs for this. You can also record 8K video at 30 frames per second. The chip can play and record video at the same time. This is good for live streaming or editing video right away. It can handle up to 32 video channels at 1080p for playing. It can record up to 16 channels at 1080p. This makes it great for high-quality video projects.
| Feature | Specification |
|---|---|
| 8K Video Decoding | H.265/HEVC, VP9 at 8K@60fps |
| H.264 at 8K@30fps | |
| AV1 at 4K@60fps | |
| Multi-channel Decoding | Up to 32 channels at 1080p@30fps |
| 8K Video Encoding | H.265/H.264 at 8K@30fps |
| Multi-channel Encoding | Up to 16 channels at 1080p@30fps |
| Display Output | HDMI 2.1 / eDP 1.3 up to 8K@60fps |
| DisplayPort 1.4a up to 8K@30fps | |
| Additional Codec Support | MPEG-1/-2, VC-1, VP8 at 1080p@60fps |
You can use more than one screen at once. The RK3588S2 works with HDMI, eDP, DisplayPort, and MIPI DSI. You can run one 8K screen at 60Hz. You can also use two 4K screens at 60Hz and one 1080p screen at 60Hz at the same time. This helps you make video walls or pro AV gear.
| Video Output Interface | Maximum Resolution & Refresh Rate | Notes |
|---|---|---|
| HDMI TX | Up to 8K @ 60Hz | Supports 4K and 8K outputs |
| eDP TX | Up to 4K @ 60Hz | Supports 4K output |
| DP TX | Up to 8K @ 30Hz | Supports 4K and 8K outputs |
| MIPI DSI lanes | 2-lane and 4-lane configurations | Additional display outputs |
Tip: The RK3588S2 is good for streaming, security, or any project that needs smooth 8K video.
Dual 16MP ISP and Imaging
The RK3588S2 has two 16MP ISP for better pictures. It can take images up to 48MP, which is much higher than old chips. You can connect up to six cameras at once. It uses MIPI DPHY or CPHY to do this. Each lane can go up to 4.5 Gbps, so you get fast, clear pictures.
| Feature Category | Specification Details |
|---|---|
| Image Signal Processor | Dual ISP supporting up to 48MP input resolution, advanced features like AE, AF, AWB, HDR |
| Camera Interfaces | Supports up to 6 cameras via MIPI DPHY/CPHY at 4.5 Gbps per lane |
You get auto-exposure, auto-focus, and auto white balance. You also get high dynamic range. These help you get sharp photos in any light. If you use smart cameras or robots, this chip gives you a big advantage.
Note: The RK3588S2 is great for AI cameras, robots, and smart devices that need good video.
Audio and Display Features
The RK3588S2 gives you great sound and clear pictures. It supports I2S audio with up to eight channels. You also get SPDIF, PDM for microphones, and a digital audio codec. The audio can be 16 to 32 bits and up to 192KHz. This means you can use it for pro audio or voice devices.
| Feature Category | Supported Audio Codecs and Interfaces | Supported Display Interfaces and Capabilities |
|---|---|---|
| Audio Interfaces | - I2S up to 8 channels (TX and RX) | - HDMI 2.1 up to 8Kp60 |
| - 2x 8-channel I2S, 2x 2-channel I2S | - eDP 1.3 up to 8Kp60 | |
| - 2x SPDIF | - DisplayPort 1.4a up to 8Kp30 | |
| - 2x 8-channel PDM (microphone arrays) | - MIPI DSI output up to 4Kp60 | |
| - 2-channel digital audio codec | - Dual LVDS up to 1080p60 via RK628 chip | |
| Audio Quality | - 16 to 32 bits, up to 192KHz | - Up to four independent displays |
You can set up four screens at once. The chip supports many ways to show video, like one 8K at 60Hz or two 4K at 60Hz. The Mali-G610 MP4 gpu gives you smooth graphics and fast video. This is good for games, signs, or kiosks.
-
You get support for many video codecs, like H.265, VP9, AVS2, H.264, and AV1.
-
The chip works well for both video and audio, so it is good for big multimedia projects.
-
You can use the RK3588S2 in pro AV gear, smart screens, or any device that needs top multimedia.
Pro Tip: The RK3588S2 helps you make devices that are fast, flexible, and high quality.
Integration and Performance in Radxa and RK3588 Devices
Developer and User Experience
Radxa devices with the rk3588 chip are strong for work and play. The radxa rock 5c and rock 5c lite have wifi 6 and bluetooth. You can join fast networks and use wireless gadgets easily. Many people say these boards play lots of 1080p videos at once. This shows they are good for multimedia projects. You also get more ram and better IO than old single board computers.
But, using AI models on radxa rock boards can be tricky. For example, to use YOLOv8 with the rk3588 NPU, you must first export to ONNX. Then, you use rknn-toolkit2. This can be hard to do. The toolkit does not work with some popular frameworks right away. So, you need to read the instructions closely. Forums like radxa forum and Ultralytics GitHub help you fix problems and share ideas. You can also use the PC simulator in rknn-toolkit2 to test your models before using real boards.
Power Efficiency and Thermal Design
Radxa rock boards with the rk3588 chip save power and work well. They stay cool even when you use them a lot. The radxa rock 5c uses 8nm tech to lower power use and stop overheating. You can play 8K video or run AI tasks without slowing down. The design has wifi 6 and bluetooth, so you get fast, steady connections and save energy. Many people see that these boards are quiet and cool. This makes them good for home labs or factories.
Scalability for Embedded Systems
You can grow your projects with radxa rock boards and the rk3588. These boards have many ways to add hardware and expand. You can choose different ram and storage sizes for your needs. The radxa rock 5c and other boards let you add wifi, bluetooth, and more for custom builds. You can use many cameras, screens, and sensors. This is great for robots, smart kiosks, or factory controllers.
Here is a quick look at how radxa boards with rk3588 help you add more parts:
| Feature Category | Details |
|---|---|
| SoC | Rockchip RK3588S octa-core (4x Cortex-A76 @ 2.4GHz + 4x Cortex-A55 @ 1.8GHz) |
| GPU | Arm Mali-G610 MP4 |
| Video Decoding | Supports 8Kp60 H.265, VP9, AVS2, 8Kp30 H.264 AVC/MVC, 4Kp60 AV1, and others |
| Video Encoding | 8Kp30 H.265/H.264 |
| NPU | 6 TOPS |
| Memory Options | 4GB, 8GB, or 16GB LPDDR4x-4224 SDRAM |
| Storage Options | eMMC flash from 8GB up to 128GB (up to 250MB/s) |
| Connectors | 3x 100-pin 0.4mm pitch board-to-board connectors |
| Storage Interfaces | 2x SATA, SD card via SDIO |
| Video Interfaces | 1x HDMI 2.1 (up to 8K60), 2x 4-lane MIPI DSI, 1x eDP (up to 8K60) |
| Camera Interfaces | Multiple MIPI CSI camera interfaces |
| I/O Interfaces | GPIOs, UARTs, I2C, SPI, PWM, ADC channels |
| Compatibility | Compatible with Raspberry Pi Compute Module 4 ecosystem |
| Industrial Suitability | Long-term supply commitment until at least 2032 |
You can use radxa rock boards for many kinds of embedded systems. They are strong, easy to expand, and have wifi 6 and bluetooth. This makes them a smart pick for hobby or work.
You can find the rk3588s2 in real devices like Radxa’s ROCK 5C and AIBOX-3588S. These devices run AI jobs quickly and play 8K video without lag. The rk3588s family helps them work well, use less power, and be easy to design.
-
AI tasks finish faster and cost less money.
-
You can make small, smart robots and machines.
-
One strong chip means you need fewer parts.
The rk3588s is important for new AI and multimedia. It helps make smart devices better for the future.
FAQ
What makes the rock series boards different from other single-board computers?
You get more power and flexibility with the rock series. These boards support 8K video, advanced AI, and many connections. You can use them for smart displays, robots, or edge AI. The rock boards work well for both home and industrial projects.
Can you use the rock boards for AI projects?
Yes, you can run AI models on the rock boards. The built-in NPU gives you fast results for tasks like image recognition or robotics. You can use popular AI frameworks and the rknn-toolkit2 to set up your models on any rock device.
How do you connect multiple displays to a rock board?
You can connect up to four screens to a rock board. Use HDMI, eDP, DisplayPort, or MIPI DSI ports. The rock boards let you show 8K video or use several 4K screens at once. This helps you build video walls or smart kiosks.
Is the radxa rock board good for embedded systems?
You can use a radxa rock board for many embedded projects. The rock boards offer strong performance, low power use, and many ways to expand. You can add cameras, sensors, or storage. This makes the rock boards a smart choice for custom devices.
Where can you find help for your rock board project?
You can visit the radxa forum or check the rock community pages. Many users share guides, tips, and code. You can also find answers for AI, multimedia, or hardware problems. The rock community helps you solve issues and improve your project.
Written by Jack Elliott from AIChipLink.
AIChipLink, one of the fastest-growing global independent electronic components distributors in the world, offers millions of products from thousands of manufacturers, and many of our in-stock parts is available to ship same day.
We mainly source and distribute integrated circuit (IC) products of brands such as Broadcom, Microchip, Texas Instruments, Infineon, NXP, Analog Devices, Qualcomm, Intel, etc., which are widely used in communication & network, telecom, industrial control, new energy and automotive electronics.
Empowered by AI, Linked to the Future. Get started on AIChipLink.com and submit your RFQ online today!
