GPU - The trusted cloud

Speeding up treatment

Putting pre-trained neural networks into production plays an essential role in developing responses and recommendations for AI services. Our GPUs can deliver up to 27 times the inference performance of a single-socket server, significantly reducing operating costs.

Innovating with AI and Machine Learning

To boost the productivity of data scientists and implement new AI services more quickly, you need to train increasingly complex models faster and faster. Our NVIDIA A100 and L40S GPUs reduce Deep Learning training procedures to just a few hours.

Improving efficiency

Fully compatible with the Kubernetes platform, the container systems and the virtual machines GPU technology simplifies access to computing resources for all users, whatever the type of workload.

Benefit from an agile solution

Adjust the number of A100 GPU instances to suit your needs: Multi-Instance (MI) GPU technology allows you to partition a GPU into seven separate secure instances, each with 5GB or 10GB of dedicated memory. Your users can access all the benefits of GPU acceleration.

Technical
specifications

FP32
91.6 TFLops

FP32 Tensor Core
366 TFlops

FP16
733 TFlops

FP8
1466 TFlops

RT Performance
212 Flops

Maximum consumption
350 W

Pricing

Use cases

Engineering and simulation

GPU instances are used in the most powerful modelling systems, offering the power required to run simulation tools in real time and accelerate innovation.

Use cases

Artificial intelligence and machine learning

With the high inference performance of NVIDIA L40S GPUs, create and train artificial intelligence neural networks and process massive volumes of data.

Use cases

Scientific research

Fundamental and applied research institutes can harness the power of the GPU to model complex scenarios and reduce the time needed to analyse the data.