Beilai Technology's ARMxy Series ARM Embedded Industrial PCs utilize both Cortex-A55 and Cortex-A72 processors. To help users select the most suitable ARM processor, we provide the following performance comparison between the Cortex-A55 and Cortex-A72.
ARMxy BL410: Industrial-grade RK3568J (Quad-core Cortex-A55) Click
ARMxy BL440: Industrial-grade RK3576J (Quad-core A72 + Quad-core A53) Click
| Aspect | Cortex-A55 (2017) | Cortex-A72 (2015) |
|---|---|---|
| Design Goal | Power efficiency (small core) | High performance (big core) |
| Market Role | Mid/low-end devices, co-processor | Flagship device main processor |
| Typical Config | 4-8 core clusters | 1-4 core high-performance clusters |
graph LR A55[A55 Microarchitecture
• 3-issue OoO
• 10-stage pipeline
• Dynamic branch prediction] -->|vs| A72[A72 Microarchitecture
• 4-issue OoO
• 15-stage pipeline
• Advanced branch prediction]
Key Differences:
✔ Decode width: A72 has 1 more decode port (4 vs. 3)
✔ Pipeline depth: A72 has a deeper pipeline (15 vs. 10 stages), supporting higher frequencies
✔ Execution units: A72 has more ALU/FPU resources
| Benchmark | A55@2.0GHz | A72@2.5GHz | Difference |
|---|---|---|---|
| SPECint2006 | 6.5 | 11.8 | +81% |
| FP32 Performance | 28 GFLOPS | 60 GFLOPS | +114% |
| Memory Latency | 24 cycles | 18 cycles | -25% |
Note: At the same frequency (2.0GHz), A72 still leads by ~40% in single-thread performance.
| Parameter | A55 (16nm) | A72 (16nm) |
|---|---|---|
| Typical Power | 0.8W | 2.5W |
| Efficiency (DMIPS/mW) | 3.4 | 1.9 |
| Idle Power | 30mW | 80mW |
✔ A55 is ~80% more power-efficient at the same process node.
A55-Specific Features:
✅ ARMv8.2 ISA (FP16/int8 support)
✅ Memory Tagging Extension (MTE)
✅ Accelerator Coherency Port
A72 Advantages:
✅ Larger L2 cache (1MB vs. 256KB)
✅ AMBA4 ACE bus (higher bandwidth)
| Application | A55 Use Cases | A72 Use Cases |
|---|---|---|
| Smartphones | Co-processor (efficiency core) | Main processor (performance core) |
| Industrial Control | Sensor aggregation nodes | Machine vision processing |
| Automotive | Infotainment systems | Autonomous pre-processing |
| Edge Computing | Lightweight AI inference | Complex algorithm processing |
A55 Models:
Rockchip RK3568 (4×A55 + NPU)
Qualcomm Snapdragon 6 Gen1 (4×A55)
A72 Models:
Rockchip RK3576 (4×A72 + 4×A53)
Huawei Kirin 950 (4×A72)
Choose Cortex-A55 When:
✓ Best power efficiency is needed (battery-powered devices)
✓ Running lightweight AI workloads (int8 quantized models)
✓ Cost-sensitive projects (30%+ lower BOM cost)
Choose Cortex-A72 When:
✓ Desktop-class CPU performance is required
✓ Processing complex algorithms (e.g., SLAM)
✓ Large cache support (1MB+ L2) is needed
A55: Still the mainstream efficiency core (paired with A78/A710)
A72: Gradually replaced by A76/A78 but remains in industrial applications
Technology Evolution: Modern DynamIQ designs typically combine A55 efficiency cores with A7x performance cores for optimal power-performance balance.
