Nvidia CPU Servers and Accelerators –

Enterprise System Pulls (tested working)

Gordon Gekko Holdings offers high-performance NVIDIA GPU servers and accelerators pulled from working AI and data center environments. All units have been fully tested and verified operational, delivering exceptional compute power for demanding workloads at a fraction of new system pricing.

These systems excel in AI training and inference, large language models (LLMs), deep learning, HPC, rendering, and generative AI applications.

| Series / Platform | Popular Models | Form Factor | Typical Configuration Highlights | Est. Selling Price |

|----------------------------|-----------------------------------------|-------------|-----------------------------------------------------------|------------------------|

| NVIDIA DGX Series | DGX H100, DGX H200, DGX B200 | 8U / 10U | 8× H100/H200/B200 GPUs, dual Xeon or Grace CPUs, full NVLink interconnect | $120,000 – $450,000+ |

| NVIDIA GH200 Grace Hopper | GH200 Superchip systems | 1U–2U | Grace CPU + Hopper GPU tightly coupled, 480–960 GB high-bandwidth memory | $85,000 – $220,000 |

| HGX GPU Platforms | HGX H100 / H200 4-GPU & 8-GPU | 2U–4U | 4–8× H100/H200 GPUs with NVSwitch for multi-GPU scaling | $65,000 – $180,000 |

| NVIDIA GPU Accelerators | H100 PCIe/SXM, H200, A100 80GB, L40S, L4 | PCIe / SXM | Individual or multi-GPU cards for server integration | $12,000 – $38,000 (per GPU) |

| NVIDIA Grace CPU Systems | Grace CPU Superchip servers | 1U–2U | Arm-based Grace CPUs with massive memory bandwidth | $45,000 – $95,000 |

Common Configurations Available -

GPUs: NVIDIA H100 (80GB), H200, Blackwell B200, A100 80GB, L40S, L4 Tensor Core

Superchips: GH200 Grace Hopper (CPU + GPU integrated)

CPUs: Dual Intel Xeon Platinum or NVIDIA Grace CPU Superchip

Memory: 480 GB – 1.4 TB+ high-bandwidth HBM3e / DDR5

Interconnect: Full NVLink / NVSwitch (up to 900 GB/s+ GPU-to-GPU bandwidth)

Storage: High-speed NVMe (up to 30+ TB)

Networking: 400 Gb/s InfiniBand or Ethernet options

Power: Redundant high-wattage Titanium PSUs (2000W–3000W+)

Product Specifications - Condition: System pulls – fully tested and working

Origin: Pulled from active AI and data center deployments - Optimized for CUDA, TensorFlow, PyTorch, and major AI frameworks

Note: Prices are estimates for typical configurations and are subject to change based on exact specs, condition, and current inventory. Contact us for current pricing and bulk discounts.

Wholesale & Bulk Terms - Competitive pricing with best rates on 2+ systems or GPU bundles - Significant discounts for larger orders (5+ GPUs or full servers) - All systems undergo full functionality and stress testing prior to shipment

Warranty: 30 days on basic functionality. Extended warranty options available upon request - Custom configurations and bundling with RAM, storage, or networking gear available

High-demand NVIDIA H100, H200, and DGX systems move extremely quickly. Contact us today for current stock levels, detailed specifications, photos, or a personalized quote.

Greed is good — Nvidia enterprise pulls are better.