AW-V96
High-Throughput Inference
The AW-V96 is a dual-GPU inference powerhouse for production environments demanding high throughput. Two NVIDIA V100/A100 GPUs work in parallel, delivering 2× the inference capacity of single-GPU models. With 512GB RAM, 8TB NVMe, and liquid cooling, it can sustain 7×24 high-load inference for LLMs, image generation, and video analytics. The 4U chassis provides ample airflow and expansion room for additional networking or storage controllers.
Technical Specifications
Detailed hardware configuration of AW-V96
Performance Benchmarks
Real-world performance data from our test lab
Key Features
What makes AW-V96 the right choice
Dual NVIDIA V100/A100 GPUs for parallel inference throughput
512GB DDR4 ECC RAM for large model footprints
8TB NVMe SSD enterprise storage
Integrated liquid cooling for 7×24 sustained operation
Load-balances inference across both GPUs automatically
Hot-swap redundant PSU for maximum uptime
Ideal For
High-traffic AI APIs, video analytics pipelines, production LLM serving with multiple concurrent users, and workloads requiring sustained high-throughput inference.
Frequently Asked Questions
Common questions about AW-V96
How does dual-GPU inference work?
ALWork Model Manager automatically distributes inference requests across both GPUs. For model parallelism, supported models can split across GPUs for larger context windows. For request parallelism, each GPU handles independent requests simultaneously.
What is the power consumption at full load?
The AW-V96 has a maximum power draw of 2200W under full GPU load. The liquid cooling system keeps temperatures below 65°C even at sustained maximum output.