NVIDIA and AMD AI servers

La plus rapide des cartes 16 Go de la série RTX 50, pour l'IA locale, la création et la location GPU. Caractéristiques Architecture NVIDIA Blackwell, cœurs Tensor de 5e génération Cœurs CUDA 10 752 Mémoire 16 Go GDDR7, bus 256 bits, 960 Go/s Interface...

The H200 GPU in PCIe form factor: 141 GB of HBM3e memory for large-model inference in standard air-cooled servers. Key features NVIDIA H200 GPU, Hopper architecture, PCIe NVL version 141 GB of HBM3e memory, 4.8 TB/s of bandwidth PCIe 5.0 x16 interface, dual-slot form...

Our most powerful AI workstation: AMD Threadripper processor, 1 to 3 NVIDIA RTX PRO 6000 Blackwell 96 GB cards (up to 288 GB of video memory) and 128 or 256 GB of ECC memory. Built to order, installed and tested by us. What you...

AMD's alternative for large-scale AI: 8 Instinct MI355X GPUs and 2.3 TB of HBM3e memory, with the open ROCm software stack. Key features 8 AMD Instinct MI355X GPUs (CDNA 4 architecture) with 288 GB of HBM3e each, for a total of about 2.3 TB...

A complete rack that works like a single giant GPU: 72 Blackwell Ultra GPUs and 36 Grace CPUs linked by NVLink, built for AI factories. Key features 72 NVIDIA Blackwell Ultra GPUs and 36 NVIDIA Grace CPUs in a single rack A single NVLink...

The NVIDIA HGX B300 platform: 8 Blackwell Ultra GPUs and 2.3 TB of HBM3e memory for AI reasoning and the largest models. Key features 8 NVIDIA B300 Blackwell Ultra GPUs, 288 GB of HBM3e each, i.e. around 2.3 TB of GPU memory 5th-generation NVLink...

A server built on the NVIDIA HGX B200 platform: 8 Blackwell GPUs for training and inference of the most demanding AI models. Key features 8 NVIDIA B200 GPUs with 180 GB of HBM3e each, for a total of 1.44 TB of GPU memory 5th-generation...

A server built on the NVIDIA HGX H200 platform: 8 H200 GPUs and over a terabyte of HBM3e memory for training and inference of large models. Key features 8 NVIDIA H200 SXM GPUs, 141 GB of HBM3e each, i.e. 1,128 GB of GPU memory...

The power of an AI server in a desktop tower: the NVIDIA GB300 Grace Blackwell Ultra superchip and 748 GB of memory to train and serve large models without a server room. Key features NVIDIA GB300 Grace Blackwell Ultra Desktop superchip 748 GB of...

A book-sized AI supercomputer: the NVIDIA GB10 Grace Blackwell chip and 64 or 128 GB of unified memory to run, test and fine-tune AI models right on your desk. Key features NVIDIA GB10 Grace Blackwell Superchip: 20-core Arm CPU (10 Cortex-X925 and 10 Cortex-A725)...

Professional GPU server optimised for generative AI, inference, 3D visualisation and compute-intensive workloads. Configuration proposed on the basis of a 4U rack server with 8 GPUs, recommended by our server partner. Indicative configuration 8× NVIDIA L40S, 48 GB GDDR6 ECC per GPU 384 GB...

High-density GPU server custom-built for artificial intelligence, LLM inference, scientific computing and professional rendering. Configuration proposed on the basis of a 4U rack server with 8 GPUs, recommended by our server partner. Indicative configuration 8× NVIDIA RTX PRO 6000 Blackwell Server Edition, 96 GB...