32-core Threadripper PRO 7975WX and dual RTX 5090 GPUs for high-performance AI training, simulation, and inference. ECC DDR5 memory, high-speed NVMe storage, and tuned thermals for 24/7 workloads.
Overview
The HydraCore Proxima Lite is a streamlined, high-performance AI workstation for researchers and developers who need top-tier GPU acceleration and robust CPU performance at an optimal value point. Featuring AMD’s Threadripper PRO 7975WX (32 cores / 64 threads, up to 5.3 GHz) and dual NVIDIA® RTX™ 5090 GPUs, this system is engineered for AI labs, production teams, and advanced compute users who want fast model training, responsive inference, and reliable 24/7 stability without overspending on unnecessary core counts.
Key Features
Threadripper PRO power: 32 cores / 64 threads for rapid data preprocessing, simulation, and orchestration workloads
Dual RTX 5090 GPUs: Next-gen CUDA, Tensor, and RT cores for unparalleled deep learning and AI acceleration
ECC DDR5 stability: 128GB DDR5 ECC RDIMM ensures consistent, error-free performance during long training runs
High-speed storage: 4TB PCIe Gen4 NVMe SSD (≈7,000 MB/s) for large datasets, checkpoints, and real-time caching
Thermals for heavy load: 360 mm liquid CPU cooling + tuned high-airflow chassis for sustained multi-GPU performance
Developer-ready stack: Optional preinstall of CUDA Toolkit, cuDNN, TensorRT, PyTorch, TensorFlow, and NVIDIA Container Toolkit
Specifications
CPU: AMD Ryzen™ Threadripper™ PRO 7975WX — 32 cores / 64 threads, up to 5.3 GHz
GPU: 2× NVIDIA® RTX™ 5090 (GeForce RTX 50 Series)
Memory: 128GB DDR5 ECC RDIMM, 8-channel (upgradeable)
Primary Storage: 4TB NVMe PCIe 4.0 SSD (≈7,000 MB/s read)
Motherboard: WRX90 workstation-grade board (PCIe Gen5, dual-GPU optimized)
Cooling: 360 mm liquid CPU cooler (optimized for multi-GPU thermals)
PSU: 1600 W 80+ Platinum, fully modular, low ripple
Chassis: Full-tower workstation case with high-airflow design
Networking: 2.5 GbE onboard (10 GbE options)
OS Options: Ubuntu LTS (CUDA stack) or Windows 11 Pro
Warranty: 3-Year HydraCluster Limited Warranty (Parts & Labor)
Note: RTX 5090 does not use NVLink; multi-GPU scaling is achieved via PyTorch DDP/NCCL, TensorFlow MirroredStrategy, or similar frameworks.
What’s in the Box
HydraCore Proxima Lite workstation (assembled, burn-in tested)
AC power cable
Motherboard & GPU accessory kits
Quick-start & recovery guide
Warranty & support card
Options/Upgrades
Memory: 192 GB, 256 GB, or 512 GB DDR5 ECC RDIMM
Storage: Additional NVMe drives (2–8 TB); RAID 1/10; dedicated scratch/cache SSDs
Networking: 10 GbE / 25 GbE NICs for high-speed data environments
GPU scaling: Upgrade to triple/quad RTX 5090 or switch to RTX 6000 Ada for ECC VRAM & pro driver support
Thermals: Acoustic damping kit, high-static-pressure fans for dense GPU configs
Software Image: CUDA Toolkit, cuDNN, TensorRT, PyTorch, TensorFlow, Docker + NVIDIA Container Toolkit, W&B, Hugging Face CLI
Warranty
3-Year HydraCluster Limited Warranty (Parts & Labor) with lifetime U.S.-based technical support. Advanced-replacement options available.
Lead Time
Built to order. Standard lead time 7–10 business days (includes dual-GPU burn-in and thermal profiling). Rush build available.