The NVIDIA A100: Available virtualized or as bare metal

Vultr has made accelerated computing affordable by virtualizing the NVIDIA A100 PCIe. Our unique approach partitions physical GPUs into discrete virtual GPUs, each with their own memory and compute. Perfect for AI inference, NLP, voice recognition, and computer vision.

NVIDIA A100 PCIe
Starting at
$1.29 / Per hour
Accelerate AI and data workloads with NVIDIA A100 PCIe GPUs
Pricing

NVIDIA A100 PCIe Starting at $1.290 / hour

Key features

Powered by 3rd generation NVIDIA Tensor Cores and multi-instance GPU technology, the NVIDIA A100 provides unmatched acceleration for diverse applications, including deep learning, data analytics, and scientific simulations.

The NVIDIA A100 GPU is specifically designed to accelerate deep learning applications, such as natural language processing, computer vision, and recommendation systems.

Spin up fast with ML tools and stacks

With Vultr, it’s easy to provision NVIDIA A100 GPUs with the end-to-end, integrated NVIDIA hardware and software stack. The NVIDIA NGC Catalog image provides full access to NVIDIA AI Enterprise. An end-to-end, secure, cloud native suite of AI software, NVIDIA AI Enterprise accelerates the data science pipeline and streamlines the development and deployment of predictive artificial intelligence (AI) models. Vultr makes NVIDIA’s latest AI innovations accessible and affordable for everyone.

Unleashing the full potential
of high-performance
computing and AI

This state-of-the-art GPU delivers breakthrough performance for accelerating high-performance computing (HPC) and AI workloads.

No information is required for download

Low latency through global availability

Vultr offers a global cloud GPU platform, allowing you to place your GPU servers close both to your applications’ end users, and to the regions where training data is first originated.

All Servers
GPU Accelerated
Chicago, Illinois United States
Miami, Florida United States
Amsterdam Netherlands
New Jersey United States
Dallas, Texas United States
Paris France
Mexico City Mexico
São Paulo Brazil
Madrid Spain
Warsaw Poland
Tokyo Japan
Seattle, Washington United States
Los Angeles, California United States
Silicon Valley , California United States
Singapore
Atlanta, Georgia United States
London United Kingdom
Frankfurt Germany
Sydney Australia
Melbourne Australia
Toronto Canada
Seoul South Korea
Stockholm Sweden
Honolulu, Hawaii United States
Mumbai India
Bangalore India
Delhi NCR India
Santiago Chile
Tel Aviv-Yafo Israel
Johannesburg South Africa
Osaka Japan
Manchester United Kingdom
9 regions

Specifications Our easy-to-use control panel and API let you spend more time coding and less time managing your infrastructure.
 A100 80GB PCIe  A100 80GB SXM
FP64 9.7 TFLOPS 9.7 TFLOPS
FP64 Tensor Core 19.5 TFLOPS 19.5 TFLOPS
FP32 19.5 TFLOPS 19.5 TFLOPS
Tensor Float 32 (TF32) 156 TFLOPS | 312 TFLOPS* 156 TFLOPS | 312 TFLOPS*
BFLOAT16 Tensor Core 312 TFLOPS | 624 TFLOPS* 312 TFLOPS | 624 TFLOPS*
FP16 Tensor Core 312 TFLOPS | 624 TFLOPS* 312 TFLOPS | 624 TFLOPS*
INT8 Core 312 TFLOPS | 624 TFLOPS* 312 TFLOPS | 624 TFLOPS*
GPU Memory 80GB HBM2e 80GB HBM2e
GPU Memory Bandwith 1,935 GB/s 2,039 GB/s
Max Thermal Design Power (TDP) 300W 400W
Multi-Instance GPU Up to 7 MIGs @ 10GB Up to 7 MIGs @ 10GB
Form Factor PCIe
Dual-slot air-cooled or
single-slot liquid-cooled
SXM
Interconnect NVIDIA®NVLink® Bridge for 2 GPUs: 600 GB/s
PCIe Gen4: 64 GB/s
NVLink: 600 GB/s
PCIe Gen4: 64 GB/s
Server Options Partner and NVIDIA-Certified
Systems™ with 1-8 GPUs
NVIDIA HGX™ A100-Partner and NVIDIA-Certified Systems with 4, 8, or 16 GPUs
NVIDIA DGX™ A100 with 8 GPUs
*With sparsity

Additional resources

Docs, demos, and information to help you succeed with your machine learning projects.

FAQ

What are the key features of the NVIDIA A100 GPU??

  • 80GB HBM2e memory with ultra-fast bandwidth
  • Multi-Instance GPU (MIG) support to optimize multi-workload performance
  • Third-generation Tensor Cores for improved AI and ML acceleration
  • NVLink support for faster GPU-to-GPU communication
  • PCIe and SXM versions for cloud and on-premise deployments

What workloads benefit the most from the NVIDIA A100?

The NVIDIA A100 is ideal for:

  • AI training and inference (machine learning and deep learning)
  • High-performance computing (HPC)
  • Data science and big data analytics
  • Scientific simulations and research
  • Video processing and rendering

How does the Multi-Instance GPU (MIG) feature work?

The MIG technology allows a single A100 GPU to be partitioned into multiple smaller GPUs, each running separate workloads simultaneously. This increases efficiency and enables multi-tenant use without performance interference.

What operating systems are supported on NVIDIA A100 GPU servers?

The A100 GPU supports Linux-based operating systems, including Ubuntu, CentOS, Debian, and custom AI/ML environments like TensorFlow and PyTorch.

Does the NVIDIA A100 support CUDA and TensorFlow?

Yes, the A100 GPU supports CUDA, TensorFlow, PyTorch, RAPIDS, and other AI frameworks for machine learning and data science applications.

How does NVIDIA A100 improve AI model training?

With third-generation Tensor Cores and high memory bandwidth, the A100 significantly reduces AI training times, enabling faster model iterations and more accurate predictions.

What is the NVIDIA A100 GPU?

The NVIDIA A100 Tensor Core GPU is a high-performance GPU designed for AI, machine learning, deep learning, data analytics, and high-performance computing (HPC). It features multi-instance GPU (MIG) technology, high memory bandwidth, and powerful tensor cores for AI training and inference.

How does Multi-Instance GPU (MIG) on the NVIDIA A100 GPU impact workload isolation and efficiency in cloud environments?

MIG enables the A100 to be partitioned into up to seven isolated GPU instances, each with dedicated memory and compute resources. This makes running multiple workloads simultaneously without resource contention possible – ideal for multi-tenant environments or serving several models at once. Vultr supports MIG, letting users balance performance and cost while maximizing GPU utilization.

How can the A100 GPU accelerate large-scale model training compared to traditional GPU architectures?

With up to 20x higher performance than previous-generation GPUs, the A100 is optimized for training large-scale AI models. It supports FP64, Tensor Float 32, and mixed precision, dramatically speeding up matrix operations central to deep learning. Combined with Vultr’s fast networking and storage, the A100 enables more rapid time to results for even the most complex models.

How do A100 GPU memory bandwidth and HBM2e architecture affect data-intensive analytics workflows?

The A100 features high-bandwidth memory (HBM2e) with up to 1.6 TB/s of bandwidth, enabling ultra-fast access to large datasets. This is crucial for analytics and scientific computing workloads that involve high-throughput data processing. On Vultr, A100-powered instances allow data scientists and engineers to unlock the full potential of high-memory bandwidth for demanding applications.

What are the key features of the NVIDIA A100 GPU??

  • 80GB HBM2e memory with ultra-fast bandwidth
  • Multi-Instance GPU (MIG) support to optimize multi-workload performance
  • Third-generation Tensor Cores for improved AI and ML acceleration
  • NVLink support for faster GPU-to-GPU communication
  • PCIe and SXM versions for cloud and on-premise deployments

Get started,
or get some advice

Start your GPU-accelerated project now by signing up for a free Vultr account.
Or, if you’d like to speak with us regarding your needs, please reach out.