8D6B8AA HP NVIDIA Rtx 2000 Ada 16GB GDDR6 4MDP B04G2AV Graphics Card $1,949IndicativeRequest a live quote for your exact price · ships worldwide from ICD
APP-MPRO-MPX-VEGA-II Apple Radeon Pro Vega II Mpx Module (32GB HBM2) $359IndicativeRequest a live quote for your exact price · ships worldwide from ICD4X97A84510 Thinksystem SR675 V3 GPU Supplemental Power Cable 4
$589IndicativeRequest a live quote for your exact price · ships worldwide from ICD
Compatible with Dell, HPE & Lenovo servers. Tested & quality guaranteed.
4X97A84510 Thinksystem SR675 V3 GPU Supplemental Power Cable 4
Lenovo-validated server-class GPU / accelerator — NVIDIA / AMD datacenter accelerator for Lenovo HPC / AI / virtualization deployments.
Specifications
- Part Number: 4X97A84510
- Brand: Lenovo
- Family: NVIDIA Tesla / V100 / A100 / H100 / H200 / L40 / L4 / RTX 6000 Ada / GH200 / AMD Instinct MI100/MI210/MI250/MI300
- Memory: See PN (24GB / 40GB / 80GB / 96GB / 141GB HBM2/HBM3/HBM3e)
- Form Factor: PCIe FH (full-height dual-slot) / SXM / OAM module
- Interface: PCIe Gen4 x16 / Gen5 x16 / SXM2-5 / NVLink
- TDP: Datacenter-class — 250W to 700W; chassis must support sustained thermal load
Compatibility
Lenovo datacenter GPUs require GPU-validated chassis with proper power + cooling + airflow. NOT every server slot accepts every GPU — verify chassis GPU-support matrix.
- Server GPU chassis types: R740xd / R750xa / R760xa (Dell) / DL380a Gen11 / Apollo 4500/6500 (HPE) / SR670 / SR675 (Lenovo)
- Power required: 8-pin / 16-pin EPS connectors — high-TDP GPUs (H100/H200) need 2× 8-pin or 16-pin 12VHPWR
- NVIDIA-Certified Systems: Dell / HPE / Lenovo offer NVIDIA-Certified configs with validated airflow + driver stack + NGC compatibility
- Multi-GPU configs (4×/8× H100): require NVLink fabric + SXM-baseboard or specialized GPU servers (DGX, HGX)
Use Cases
AI/ML training (large-model fine-tuning, distributed PyTorch / Megatron / DeepSpeed), AI inference (LLM serving, vision, recommender), HPC simulation (CFD, FEA, molecular), VDI acceleration (vGPU), real-time rendering, scientific compute (CUDA / OpenCL / OpenACC). H100/H200 = transformer-engine optimized for LLM. Mi300X = highest memory bandwidth for memory-bound workloads.
Ships worldwide from ICD, Cairo, Egypt. All parts are tested, inspected, and backed by ICD's quality guarantee.
This Part Fits These Servers
Click a model to browse all compatible parts for that server








Reviews
There are no reviews yet.