entry

Tesla V100 GPU

The V100 remains relevant for legacy HPC and AI training workloads, with 16GB/32GB HBM2 options and Tensor Cores. While outperformed by newer GPUs, it's still used in scientific computing and older inference pipelines.

VRAM 32GB
CUDA Cores 5,120
Tensor Cores 640
TDP 250W
Process 12nm
From
$0.05/hr
across 10 providers
Tesla V100 GPU

Cloud Pricing

Cheapest on Verda 93% below avg
ProviderConfigPrice / hrUpdatedSource
1×2×4×8×
$0.05/hr
4/21/2026
1×2×4×
$0.12/hr
4/21/2026
1×2×4×8×
$0.14/hr
4/21/2026
1×
$0.17/hr
4/16/2026
4×8×
$0.19/hr
4/21/2026
1×
$0.21/hr
4/21/2026
2×
$0.21/hr
4/15/2026
1×2×4×
$0.23/hr
4/21/2026
7×9×15×
$0.29/hr
4/21/2026
1×8×
$0.43/hr
4/21/2026
8×
$0.79/hr
4/21/2026
1×2×
$0.99/hr
4/21/2026
1×
$1.12/hr3mo
3/31/2026
1×
$1.12/hr36mo
4/20/2026
8×
$1.15/hr
4/21/2026
1×
$1.56/hr1mo
3/31/2026
1×
$1.56/hr12mo
4/20/2026
1×2×4×
$2.34/hr
4/21/2026
1×
$2.34/hr
4/21/2026
1×
$2.48/hr
3/31/2026
Direct from providerVia marketplace

Prices updated daily. Last check: 4/21/2026

Performance

FP16
112 TFLOPS
FP32
14 TFLOPS
Bandwidth
900 GB/s

Strengths & Limitations

  • 32GB HBM2 memory provides ample capacity for medium-scale AI models and datasets
  • 900 GB/s memory bandwidth enables efficient handling of memory-intensive workloads
  • 640 Tensor Cores with mixed-precision support accelerate AI training and inference
  • NVLink support allows multi-GPU scaling with 300 GB/s bi-directional bandwidth
  • 250W TDP offers reasonable power efficiency for data center deployment
  • 5,120 CUDA cores provide strong performance for traditional HPC applications
  • Established ecosystem support with mature drivers and software optimization
  • 12nm manufacturing process is less efficient than newer 4nm and 5nm GPU generations
  • Lacks modern features like FP8 precision support found in current Tensor Cores
  • Limited to older PCIe 3.0 interface compared to PCIe 5.0 in current GPUs
  • 112 TFLOPS FP16 performance significantly trails current-generation alternatives
  • Missing hardware support for newer AI model architectures and optimizations

Key Features

NVIDIA Volta architecture with unified memory architecture
First-generation Tensor Cores with FP16 and mixed-precision support
NVLink high-speed interconnect technology
HBM2 high-bandwidth memory with ECC support
CUDA Compute Capability 7.0
Multi-Process Service (MPS) for workload isolation
GPUDirect RDMA for direct memory access
Unified Virtual Addressing across CPU and GPU memory

About Tesla V100

The Tesla V100 is NVIDIA's data center GPU based on the Volta architecture, introduced in 2017. As a previous-generation product in NVIDIA's lineup, it sits below current offerings like the H100 and GB300 series, but established many foundational technologies still used today. The V100 was the first GPU to introduce Tensor Cores for accelerated AI workloads and marked NVIDIA's initial major push into the data center AI market. The V100 features 32GB of HBM2 memory with 900 GB/s of memory bandwidth, 5,120 CUDA cores, and 640 Tensor Cores built on a 12nm manufacturing process. Its 4,096-bit memory interface provides substantial memory throughput, while the GPU delivers 14 TFLOPS of FP32 performance and 112 TFLOPS of FP16 performance. The 250W TDP makes it relatively power-efficient for a data center GPU of its generation. In cloud deployments, the V100 serves as an entry-level option for organizations beginning AI development or running smaller-scale workloads. Its combination of substantial VRAM and Tensor Core acceleration makes it suitable for training smaller models, inference workloads, and traditional HPC applications where cutting-edge performance is not required.

Common Use Cases

The Tesla V100 is well-suited for entry-level AI development, smaller-scale model training, and traditional HPC workloads. Its 32GB memory capacity handles medium-sized datasets and models that don't require the latest architectural optimizations. The V100 works effectively for AI inference deployments, data science workflows, and scientific computing applications like molecular dynamics simulations. Organizations transitioning from CPU-based computing or exploring GPU acceleration for the first time often find the V100's balance of capability and accessibility appropriate for initial deployments.

Full Specifications

Hardware

Manufacturer
NVIDIA
Architecture
Volta
CUDA Cores
5,120
Tensor Cores
640
Process Node
12nm
TDP
250W

Memory & Performance

VRAM
32GB
Memory Interface
4096-bit
Memory Bandwidth
900 GB/s
FP32
14 TFLOPS
FP16
112 TFLOPS
FP64
7 TFLOPS
Release
2017

Frequently Asked Questions

How much does a Tesla V100 cost per hour in the cloud?

Tesla V100 pricing varies by provider, region, and commitment level. Check the pricing table above for current rates across all providers.

What is the Tesla V100 best used for?

The Tesla V100 excels at entry-level AI training, inference workloads, and traditional HPC applications. Its 32GB memory and Tensor Core acceleration make it suitable for medium-scale AI development, data science workflows, and scientific computing tasks that don't require cutting-edge performance.

How does the Tesla V100 compare to newer GPUs like the H100?

The V100 offers 112 TFLOPS of FP16 performance compared to the H100's significantly higher throughput, and lacks modern features like FP8 precision and fourth-generation Tensor Cores. However, the V100's 32GB memory capacity and mature ecosystem make it viable for workloads that don't require the latest architectural improvements.