NVIDIA Tesla V100 PCI-E 32GB (Passive)

Accelerators

NVIDIA Tesla V100 PCI-E 32GB (Passive) Main Picture


NVIDIA Tesla V100 PCI-E 32GB (Passive) Picture 48537NVIDIA Tesla V100 PCI-E 32GB (Passive) Picture 48538NVIDIA Tesla V100 PCI-E 32GB (Passive) Picture 48539
NVIDIA's Tesla V100 is built for GPU computing. It is based on the Volta graphics chip, and is very similar to a video card - but with an emphasis on pure calculation performance rather than actually rendering graphics on a monitor. In fact, it doesn't have display outputs at all! As such, Tesla cards should be paired with a Quadro video card to handle display output.

On the computation side, the V100 has unhindered double-precision floating point performance (FP64) - a big advantage over most Quadro and GeForce cards, albeit in a niche area. It also supports several lower-precision modes and has dedicated Tensor cores for specialized machine learning applications.

The Tesla is a passive card, meaning it has no cooling fan of its own. Cards like this can only be used in cases which provide the necessary amount of airflow across the PCI-Express slot area. If you want a Tesla but are unsure which of our systems support it, please contact our technology consultants.

Model: NVIDIA TCSV100M-32GB-PB

Specifications

Chipset Manufacturer NVIDIA
Product Line Tesla
Motherboard Connection PCI Express 3.0 x16
Cooling Method Passive Heatsink
Core Specifications
Processors CUDA: 5120, Tensor: 640
Memory Specifications
Onboard Memory 32GB
Memory Type HBM2
Memory Bus Width 4096-bit
Bandwidth 900 GB/s
Performance
Double Precision Floating Point (Peak) 7 Tflops
Single Precision Floating Point (Peak) 14 Tflops
Power Connectors
Plug 1 8-pin PCI-E
Plug 2 None
Dimensions
Length 267 mm (10.5 in)
Height 112 mm (4.4 in)
Width 42 mm (1.7 in)