ABS Zaurion Pro Workstation - Powered by aiDAPTIV+ & Ubuntu
- Cost-Effective, On-Premise LLM Training
- 80B – 90B Parameter LLM inference
- Fine-Tuning for 70B+ Parameter LLMs
Next-Level AI Infrastructure for Training and Deployment
What is the ABS AI Workstation?
The ABS AI Workstation, powered by PHISON aiDAPTIV+, is a high-performance, scalable AI infrastructure solution designed for cutting-edge generative AI workloads. Supporting both large-scale inference and full-parameter training, it can handle large language models (LLMs) with up to 200B parameters.
Unlike traditional AI PCs or entry-level AI devices, ABS delivers enterprise-grade computing power, featuring advanced memory architecture and multi-GPU scalability. Built for AI labs, research institutions, and forward-thinking enterprises, it provides the efficiency, flexibility, and performance needed to accelerate AI development and deployment.
Key Benefits
Scalable Inference & Fine-Tuning
- Train and deploy Llama-3-70B, Stable Diffusion XL, Whisper, and more.
- Supports both full-model training and domain-specific fine-tuning.
High-Performance Multi-GPU Architecture
- Equipped with 4x 48GB GPUs for demanding AI workloads.
- Scales seamlessly for 13B–200B parameter model training.
Advanced Memory Expansion
- aiDAPTIV+ architecture overcomes HBM limitations.
- Expands GPU memory with NAND for improved efficiency.
Secure On-Premise AI Deployment
- Keep data, models, and IP secure within local infrastructure.
- Customize AI for specific languages, industries, and datasets.
Who Needs AI Workstation?
Advanced Research and Academia
- Provides comprehensive curriculum support for large-scale generative AI education.
- Empowers AI research labs with cutting-edge model development and experimentation.
Enterprise AI Deployment
- Develop and deploy domain-specific AI solutions for healthcare, finance, legal, and retail industries.
- Customize LLMs to adapt to regional languages and regulatory requirements.
Software and Application Development
- Train and deploy chatbots, AI co-pilots, document assistants, and multi-modal applications.
- Accelerate the development of commercial AI-powered platforms.
System Highlights
- CPU: Intel Xeon
- GPU: 4x NVIDIA RTX 6000 Ada (48GB each)
- Memory: 512GB DDR5 ECC
- Storage: High-speed NVMe SSDs + aiDAPTIV+ AI100E Modules
- Middleware: aiDAPTIVLink for seamless AI training, optimization, and fine-tuning
Supported Models
- Llama, Llama-2, Llama-3, CodeLlama
- Vicuna, Falcon, Whisper, Clip Large
- Metaformer, Resnet, Deit base, Mistral, TAIDE
- And many more being continually added
Why Choose the ABS AI Workstation?
- Scalable AI Training – Supports 13B to 200B models with multi-GPU flexibility.
- Privacy-First Design – Secure on-premise model training and inference.
- Breakthrough Memory Innovation – Overcomes HBM limitations with NAND memory extension.
- Education-Ready – Ideal for academic AI research and training.
- Enterprise-Grade Deployment – Optimized for production-ready AI applications.
A Good Way to Harness the Full Potential of AI
The ABS AI Workstation is the foundation of next-generation AI infrastructure, delivering unparalleled compute power, innovative memory solutions, and seamless deployment options. Whether you're training future AI leaders or deploying enterprise-grade LLMs, provides the flexibility, security, and performance needed to drive real-world AI innovation.
