← All instances

p5.48xlarge

⚡ 8x GPU • 640.0 GB HBM3

Part of the P5 Accelerated Computing family — Hardware accelerators (GPUs, FPGAs) for specialized compute.

Name:pGPU Deep Learning & HPCp = GPU Deep Learning & HPCHigh-end NVIDIA GPUs (Blackwell, Hopper, Ampere, Volta) built for frontier AI/LLM training, HPC simulations, and generative AI.55th Generation5 = 5th GenerationAWS 5th generation hardware platform with updated CPU/system architecture.48xlarge48 Extra Large48xlarge = 48 Extra Large192 vCPUs • 2048.0 GiB RAM • 30720 GB SSD • 3200 Gigabit Network • 8x NVIDIA H100 (640.0 GB VRAM)
Price (Linux)
$55.0400/ hr
~$40,179.20/mo
Compute
192vCPUs
96 cores • x86_64
Memory
2048.0GiB
10.7 GiB / vCPU
Value
$0.2867/ vCPU-hr
$0.0269/GiB-hr
Network
100Gbps
Baseline: 100 Gbps
Storage & EBS
10,000MB/s
Max: 260,000 IOPS

Technical Specifications & Limits

Hardware specs, container thresholds, and system limits

Compute & Processor

ProcessorAMD EPYC 7R13 Processor
Architecturex86_64 (x86_64)
vCPUs / Threads192 vCPUs
Physical Cores96
Clock Speed2.95 GHz
Intel AVX / AVX2No
Intel Turbo BoostNo
Bare MetalNo (Virtualized)

Performance & Benchmarks

CoreMark Score3,900,325
CoreMark / vCPU20,314.2 / core
FFmpeg Transcoding—
CoreMark per Dollar70,863 score / $
Cost per vCPU-hr$0.2867
Cost per GiB-hr$0.0269

Containers & Kubernetes

EKS Max Pods100 pods
Max ECS Tasks182 tasks
Max ENIs64
IPv4/v6 per ENI50
Total Addressable IPs3200 total IPs
CNI Prefix DelegationSupported (High Density)

Networking & Bandwidth

Network Performance3200 Gigabit
Baseline Bandwidth100 Gbps
Burst / Peak Bandwidth100 Gbps
Enhanced NetworkingYes
EFA SupportNo
IPv6 SupportNative Dual-Stack Supported

Storage & EBS Optimization

Storage ConfigurationSSD (30720 GB)
Local NVMe Disk30720 GB
EBS Baseline Throughput10000 MB/s
EBS Max Throughput10000 MB/s
EBS Baseline IOPS260,000
EBS Max / Burst IOPS260,000
Dedicated EBS Throughput80 Gbps

GPU & Hardware Acceleration

GPU ModelNVIDIA H100
Accelerator Count8x GPU(s)
Total GPU Memory640.0 GB
Memory ArchitectureHBM3
CUDA Compute Capabilityv9.0
Acceleration RuntimesCUDA, TensorRT, cuDNN, PyTorch, PyJAX

Platform & Architecture Details

Hypervisor / SystemAWS Nitro System (PCIe Offload)
Nitro EnclavesSupported (Isolated security enclaves)
Normalization Factor384
Generation StatusCurrent Generation (Recommended)
Date Introduced2023-07-26

Regional Pricing Matrix

Live on-demand rates across all global AWS cloud regions
Lowest Price Region
$55.0400/ hr
US East (N. Virginia) (us-east-1)
Highest Price Region
$92.4672/ hr
South America (Sao Paulo) (sa-east-1)
Global Fleet Average
$66.6351/ hr
Across all active regions
Regional Coverage
15regions
Available worldwide
Linux — Available in all 15 regions
Region NameRegion CodeOn-Demand Price
Asia Pacific (Jakarta)
ap-southeast-3$75.9552
Asia Pacific (Mumbai)
ap-south-1$66.0480
Asia Pacific (Seoul)
ap-northeast-2$75.9552
Asia Pacific (Sydney)
ap-southeast-2$71.5520
Asia Pacific (Tokyo)
ap-northeast-1$68.8000
AWS GovCloud (US-West)
us-gov-west-1$66.0480
Canada (Central)
ca-central-1$63.2960
EU (London)
eu-west-2$71.5520
EU (Stockholm)
eu-north-1$58.8928
South America (Sao Paulo)Highest
sa-east-1$92.4672
US East (Atlanta)Lowest
us-east-1-atl-1$55.0400
US East (N. Virginia)Lowest
us-east-1$55.0400
US East (Ohio)Lowest
us-east-2$55.0400
US West (N. California)
us-west-1$68.8000
US West (Oregon)Lowest
us-west-2$55.0400