The Rockchip RK3588S is a flagship processor designed for high-end intelligent computing scenarios. Featuring an 8-core heterogeneous architecture (4×Cortex-A76 + 4×Cortex-A55) combined with a 6TOPS NPU, it supports 8K video encoding/decoding and quad-display output. While maintaining a typical power consumption of just 15W, the RK3588S delivers desktop-class performance, making it especially suitable for edge servers and intelligent NVRs that require high-performance AI computing and multimedia processing.
Based on this processor, BLIIoT has introduced the BL450 Edge Computing Server, supporting up to 32GB LPDDR5 memory and PCIe 4.0 expansion, addressing the high-density computing needs of smart city and Industrial IoT applications.
4×Cortex-A76 @ 2.4GHz + 4×Cortex-A55 @ 1.8GHz
4MB L3 + 2MB L2 cache
Supports ARM DynamIQ technology
6TOPS NPU (INT8/FP16 mixed precision)
Typical AI performance:
• Face recognition: 300fps @ 1080p
• Object detection: 60fps @ 4K
• Natural language processing: 1200 tokens/s
Video Codec:
• Decode: 8K@60fps (AV1/VP9/H.265)
• Encode: 8K@30fps (H.265/H.264)
Display System:
• Quad-display output (2×8K + 2×4K)
• HDR10 / HLG / Dolby Vision support
Dual-channel PCIe 4.0 (8GT/s)
4×USB3.1 Gen2 (10Gbps)
Dual-channel MIPI-CSI (single channel up to 8K input)
Industrial-grade packaging (-40℃ ~ 85℃)
ECC memory error correction
Hardware-level security and encryption engine
Supports 16-channel 1080p video analytics
Typical full-load power consumption: 18W
Concurrently handles data from 200+ IoT devices
Up to 32-channel video input
Supports multiple algorithms in parallel (face, license plate, behavior recognition)
≥7 days of local storage backup during network outages
Real-time rendering of 4K 3D models
Compatible with Unity/Unreal Engine
Latency <20ms
Compute Unit: RK3588S
Memory: 8GB / 16GB / 32GB LPDDR5 options
Storage: Dual M.2 NVMe (up to 4TB)
Standard:
• 2×10GbE SFP+
• 4×2.5GbE RJ45
Optional:
• 5G RedCap module
• Wi-Fi 6E module
Smart campus: Single node managing 500+ smart devices
Smart manufacturing: 20-node digital twin production line
Cloud gaming: 50 concurrent 8K game streams
Best suited for:
✓ Edge computing nodes requiring desktop-class performance
✓ Ultra-high-definition video analytics centers
✓ Digital content creation workstations
Not suitable for:
× Autonomous driving domain controllers requiring >50TOPS
× Ultra-low latency (<1ms) industrial control
× Large-scale AI model training
Note: Actual performance depends on thermal design and system optimization. Please refer to Rockchip official documentation for detailed specifications. BLIIoT provides customizable thermal management solutions.
