AW-V64
Standard AI Inference
The AW-V64 doubles the memory and storage of the entry-level V32, providing headroom for larger models and concurrent inference workloads. With 256GB RAM and 4TB NVMe, it can cache multiple models simultaneously and handle batch inference requests without reloading. The liquid-ready cooling system allows future upgrade to liquid cooling for sustained high-load operation. Ideal for production inference pipelines serving multiple endpoints.
Technical Specifications
Detailed hardware configuration of AW-V64
Performance Benchmarks
Real-world performance data from our test lab
Key Features
What makes AW-V64 the right choice
Single NVIDIA V100/A100 GPU with 256GB system RAM
4TB NVMe SSD — cache multiple models simultaneously
Liquid-ready chassis for future cooling upgrade
Handles 50+ concurrent inference sessions
Pre-loaded Ubuntu + ALWork Model Manager
2U rack-optimized form factor
Ideal For
Production AI inference serving, teams that need to cache multiple models for fast switching, and workloads requiring higher memory bandwidth for data preprocessing.
Frequently Asked Questions
Common questions about AW-V64
How many models can AW-V64 cache simultaneously?
With 256GB RAM, the AW-V64 can hold 3-5 7B models or 2 13B models in memory simultaneously, allowing instant model switching without disk loading delays.
What does liquid-ready mean?
The chassis includes mounting points, plumbing pass-throughs, and thermal sensors for liquid cooling. A liquid cooling kit can be installed onsite without chassis modification.