R6V51A HPE AMD Instinct MI210 64GB HBM2E PCIe 4.0 Accelerator $14,989IndicativeRequest a live quote for your exact price · ships worldwide from ICD
100-506014 AMD Radeon Pro Ssg 16GB HBM2 + 2TB NVMe PCIe 3.0 X16 Workstation GPU $14,989IndicativeRequest a live quote for your exact price · ships worldwide from ICDP26855-001 HPE AMD Instinct MI100 32GB HBM2 PCIe 4.0 GPU Accelerator (spare)
$14,989IndicativeRequest a live quote for your exact price · ships worldwide from ICD
32GB
Compatible with HPE ProLiant servers. Tested & quality guaranteed.
P26855-001 HPE AMD Instinct MI100 32GB HBM2 PCIe 4.0 GPU Accelerator (spare)
AMD Instinct datacenter accelerator — high-bandwidth-memory GPU for AI training / HPC / scientific compute, NVIDIA Tesla/H100 alternative.
Specifications
- Part Number: P26855-001
- Brand: AMD
- Family: MI100
- Memory: 32 GB HBM2
- Form Factor: PCIe Gen4 FHFL dual-slot (MI100/MI210 PCIe) / OAM module (MI250X/MI300X — Open Accelerator Module)
- Interface: PCIe Gen4 x16 / Infinity Fabric (OAM 800 GB/s)
- TDP: 300W (MI100/MI210) / 500W (MI250) / 700W (MI300X) / 1000W (MI325X)
Compatibility
AMD Instinct accelerators target NVIDIA Tesla/H100/H200 alternative market with strong HPC + AI training perf-per-dollar. Software stack ROCm (Radeon Open Compute) — AMD's CUDA alternative — supports PyTorch, TensorFlow, JAX, MIOpen.
- MI100 (CDNA 1, 2020 — 32GB HBM2): first-gen datacenter accelerator
- MI210 (CDNA 2 PCIe, 64GB HBM2e): mainstream HPC accelerator
- MI250 / MI250X (CDNA 2 OAM, 128GB HBM2e): Frontier supercomputer (#1 TOP500 2022-2024) uses MI250X
- MI300A (CDNA 3 APU, 128GB HBM3): integrated CPU+GPU APU — El Capitan supercomputer
- MI300X (CDNA 3, 192GB HBM3): standalone accelerator — direct H100 competitor for LLM training/inference, larger memory than H100 80GB
- MI325X (CDNA 3 enhanced, 256GB HBM3e): 2024 — even larger memory for trillion-parameter LLMs
- Validated systems: Dell PowerEdge XE9680 (8x MI300X), HPE Cray XD675, Supermicro AS-8125GS-TNMR2 (8x MI300X), Lenovo ThinkSystem SR685a V3
Use Cases
HPC simulation (CFD, FEA, weather modeling — Frontier, El Capitan exascale supercomputers run MI250X/MI300A), AI training (LLM fine-tuning via ROCm + PyTorch), AI inference (LLM serving with vLLM ROCm port), scientific compute (HIP-based CUDA-port codes), genomics. MI300X 192GB HBM3 = best memory-bandwidth GPU on market for memory-bound LLM serving (Llama-3-405B fits in fewer GPUs than H100). Hyperscalers (Microsoft Azure, Meta, OpenAI) actively deploying MI300X.
Ships worldwide from ICD, Cairo, Egypt. All parts are tested, inspected, and backed by ICD's quality guarantee.







Reviews
There are no reviews yet.