NVIDIA

P100

Name: NVIDIA P100 PCIe
Brand: NVIDIA
Availability: InStock

PCIe

FP32 TFLOPS

9.3 TFLOPS

CUDA Cores

3,584

TDP

250 W

Memory

HBM2

Provider Marketplace

Cheapest

$0.11/hour

Starting from

Prime Intellect Visit

Best Value

—

Awaiting listings

No additional provider yet

Enterprise Choice

$1.87/hour

Starting from

Alibaba Cloud Visit

All Cloud Providers

2 Options available

Prime IntellectCheapest

On-Demand

$0.11/ hour

Estimated Cost

Provision

Alibaba Cloud

On-Demand•China (Shanghai)

$1.87/ hour

Estimated Cost

Provision

Estimates only — rates are collected automatically from public provider pages and may be out of date. Prices vary by region, commitment term, and availability, and typically exclude storage, egress, and tax. Confirm current pricing with the provider before purchasing. Last collected Sep 1, 2026.

Compute Performance

FP644.7 TFLOPS

FP329.3 TFLOPS

FP1618.7 TFLOPS

Architecture

MicroarchitecturePascal

CUDA Cores3584

Base Clock1189 MHz

Boost Clock1328 MHz

Dynamic PrecisionSupported (FP64/FP32/FP16)

Memory & VRAM

Memory TypeHBM2

ECC SupportYes (native support, enabled by default)

Unified MemoryYes (unified virtual memory and page migration engine)

Connectivity & Scaling

InterconnectPCIe

GenerationPCI Express 3.0

PCIe InterfacePCI Express 3.0 x×16

P2P Memoryunified virtual memory

Power & Efficiency

TDP250 W

Peak Power250

ConnectorsOne CPU 8-pin auxiliary power connector

Thermal LimitsPassive heatsink (requires system airflow); Operating temperature 0 °C to 45 °C; GPU maximum operating temperature 80 °C; GPU slowdown temperature (50% clock slowdown) 82 °C; GPU shutdown temperature 85 °C

Physical Design

Form FactorPCIe

FHFLYes

Slot WidthDual-slot

Weight1177 Grams

CoolingPassive

Thermals & Cooling

Airflow{"30 °C":12,"35 °C":15,"40 °C":19,"45 °C":26,"50 °C":37}

Temp Range0 °C to 45 °C

ThrottlingGPU slowdown: 50% clock slowdown at 82 °C; GPU shutdown at 85 °C; GPU maximum operating temperature 80 °C

Software Ecosystem

Compiler StackCompute APIs CUDA, DirectCompute, OpenCL ™, OpenACC

Driver StabilityWindows Hardware Quality Lab (WHQL) certified (Windows 7, Windows 8.1, Windows 10, Windows Server 2008 R2, Windows Server 2012 R2)

Server & Deployment

Rack-ScaleTesla P100 for PCIe enables a single node to replace half a rack of commodity CPU nodes by delivering lightning-fast performance in a broad range of HPC applications.

System Compatibility

Required PCIePCI Express 3.0 ×16

MotherboardConforms to NVIDIA Form Factor 3.0 specification

Rack PowerTotal board power 250 W

BIOS LimitsUEFI Supported

OS CompatWindows 7; Windows 8.1; Windows 10; Windows Server 2008 R2; Windows Server 2012 R2

Multi-GPU Scalability

Scaling Characteristics

Network BottlenecksThe old approach of deploying lots of commodity compute nodes requires huge interconnect overhead that substantially increases costs without proportionally increasing performance.

ParallelismCompute APIs CUDA, DirectCompute, OpenCL ™, OpenACC

Workload Readiness

Vision Training

18.7 TeraFLOPS (half-precision, FP16)

HPC / Simulation

4.7 TeraFLOPS (double-precision, FP64)

Scientific Computing

4.7 TeraFLOPS (double-precision, FP64)

Market Authority

Community Benchmarks

Performance comparisons for NAMD, VASP, MILC, HOOMD, AMBER, and Caffe/AlexNet vs K80

Key Strengths

Limitations

Also in the Lineup

Expert Insight

The P100 represents a strategic leap in AI compute. When comparing cloud providers, consider not just the hourly rate, but also the interconnect bandwidth (InfiniBand/NVLink) and regional availability which can significantly impact total cost of ownership for large-scale training.

Glossary Terms

FP32 TFLOPS

VRAM

TDP

Cores

Information updated daily. Cloud pricing subject to vendor availability.