Huawei Atlas 300I Pro
Overview
The Huawei Atlas 300I Pro is a half-height half-length inference card built on a single Ascend 310P processor, rated at 140 TOPS INT8 and 70 TFLOPS FP16 within a 72 watt envelope. Because the card carries exactly one chip and no co-processor, its published figures are also the clearest public statement of what the Ascend 310P itself does, and they are exactly half the dual-chip Atlas 300I Duo. It pairs 24 GB of LPDDR4X at 204.8 GB/s with an eight-core 1.9 GHz host CPU and a PCIe 4.0 x16 interface, and includes H.264, H.265 and JPEG encode and decode blocks. The board is 169.5 by 68.9 millimetres and weighs 280 grams. Note that Huawei hosts this product at the atlas-300-ai address rather than an atlas-300i-pro one, and that it should not be confused with the Atlas 300I A2, which is a different card built on an Ascend 910B4.
Performance Metrics
Peak theoretical throughput by precision type
| Precision | Bits | Peak TFLOPS | Efficiency |
|---|---|---|---|
| INT8 | 8 | 140.0 | 1.944 TFLOPS/W |
| FP16 | 16 | 70.0 | 0.972 TFLOPS/W |
Power Specifications
TDP
72 W
Max Power
72 W
Power Connector
PCIe Slot
Cooling
Air
Memory Specifications
Capacity
24 GB
Type
LPDDR4X
Bandwidth
204.8 GB/s
Interface
--
Hardware & Design
Form Factor
PCIe
Architecture
Da Vinci
Process Node
--
Launch Year
2021
Variant
Standard
Market Segment
Professional
Chiplets
1
Full Specifications
| Chip Design | |
|---|---|
| Chiplets | 1 |
| Memory | |
| VRAM | 24 GB |
| Memory Type | LPDDR4X |
| Bandwidth | 205 GB/s |
| Interconnect & I/O | |
| PCIe | 4.0 x16 |
| Power & Thermal | |
| TDP | 72 W |
| Max Board Power | 72 W |
| Enterprise Features | |
| Compute APIs | CANN, MindSpore, MindX, PyTorch (torch_npu), ACL |
| Physical & Media | |
| Card Length | 169.5 mm |
| General | |
| Form Factor | PCIe |
| Architecture | Da Vinci |
| Launch Year | 2021 |
Documentation & Resources
Common Use Cases
The Huawei Atlas 300I Pro is optimized for high-performance computing tasks with Da Vinci architecture delivering high TFLOPS of compute power.
Where to Rent
Compare cloud providers offering on-demand GPU instances for AI training, inference, and HPC workloads.
Browse GPU Cloud ProvidersSimilar GPUs
Stay Updated on GPU Releases
Get notified when new GPUs are added or specifications are updated.
No spam, unsubscribe anytime.