AW-A120
Large-Scale AI Cluster
The AW-A120 is our flagship AI compute server, packing 8 NVIDIA V100/A100 GPUs with 1TB of system RAM and 16TB NVMe storage in a single 4U chassis. This is a data-center-grade AI cluster engineered for the most demanding workloads: large-scale model training, multi-model serving at enterprise scale, and high-throughput inference pipelines. With 256GB total GPU VRAM (8× 32GB), it can train 13B+ parameter models or serve dozens of concurrent LLM endpoints. The 4000W redundant PSU and full liquid cooling ensure rock-solid stability under maximum load.
Technical Specifications
Detailed hardware configuration of AW-A120
Performance Benchmarks
Real-world performance data from our test lab
Key Features
What makes AW-A120 the right choice
8× NVIDIA V100/A100 GPUs — 256GB total VRAM
1TB DDR4 ECC system RAM for massive data pipelines
16TB NVMe SSD enterprise storage array
4000W redundant PSU with full liquid cooling
Designed for large-scale distributed training
ALWork cluster management with multi-node orchestration
Ideal For
Enterprise AI teams, large-scale model training, multi-tenant AI service platforms, and any workload requiring maximum GPU density in a single chassis.
Frequently Asked Questions
Common questions about AW-A120
Can AW-A120 train 70B+ parameter models?
Yes. With 256GB total GPU VRAM and model parallelism (FSDP/DeepSpeed Stage 3), the AW-A120 can train 70B parameter models. For extremely large models (100B+), multiple AW-A120 units can be clustered via high-speed interconnect.
What is the power and cooling requirement?
The AW-A120 requires 4000W (dual 2000W PSUs) and liquid cooling with facility-grade chilled water or closed-loop cooling system. It is designed for data center deployment with proper cooling infrastructure.