p4d.24xlarge

Amazon EC2 · P4d family

The p4d.24xlarge is a gpu & accelerated Amazon EC2 instance type with 96 vCPUs and 1,152 GiB of memory, 8 × NVIDIA A100 40GB, running on Intel Xeon Platinum 8275L with 400 Gigabit networking, starting at $21.96 per hour on demand.

gpu & acceleratedx86-64intelgpu
vCPUs
96

dedicated

memory
1,152 GiB

12 GiB per vCPU

gpu
8× A100 40GB

320 GiB total

on-demand
$21.96

$16,029/mo · us-east-1

spot
$13.07

−40% vs on-demand

1-yr reserved
$13.92

−37% vs on-demand

p4d.24xlarge price by region

15 regions · hourly list prices in USD · On-demand · Spot · 1-yr Reserved · 3-yr Reserved

US East (Ohio)us-east-2cheapestNorth America$21.96$5.734$13.92$9.374
US West (Oregon)us-west-2cheapestNorth America$21.96$6.205$13.92$9.374
US East (N. Virginia)us-east-1referenceNorth America$21.96$13.07$13.92$9.374
EU (Ireland)eu-west-1Europe$23.72$24.48$15.08$10.15
EU (Stockholm)eu-north-1Europe$23.72–––
Canada (Central)ca-central-1North America$25.25–$16.01$10.78
Asia Pacific (Mumbai)ap-south-1Asia$26.34–$17.03$14.18
Asia Pacific (Singapore)ap-southeast-1Asia$26.35–$16.71$11.25
AWS GovCloud (US)us-gov-west-1North America$26.35–$16.71$9.842
EU (Frankfurt)eu-central-1Europe$27.43$14.95$17.39$11.71
Asia Pacific (Sydney)ap-southeast-2Oceania$28.54–$18.10$12.19
EU (London)eu-west-2Europe$28.54–––
Asia Pacific (Tokyo)ap-northeast-1Asia$30.10$7.642$19.58$15.79
Asia Pacific (Seoul)ap-northeast-2Asia$30.41$8.696$19.28$15.94
South America (Sao Paulo)sa-east-1South America$35.56–$22.68$15.05

Specifications

compute

vCPUs
96
Memory
1,152 GiB
Memory per vCPU
12 GiB
Processor
Intel Xeon Platinum 8275L
CPU vendor
Intel
Architecture
x86-64
Clock speed
3 GHz
Burstable CPU
No

storage & network

Local storage
8 TB SSD (8 disks)
Storage detail
8 x 1000 SSD
Network
400 Gigabit
Peak bandwidth
400 Gbps

accelerators

Count
8
Memory (total)
320 GiB
Architecture
Ampere
Vendor
NVIDIA

aws details

API name
p4d.24xlarge
Family
P4d
Category
GPU & accelerated
Generation
Current
Regions
15 (Asia, Europe, North America, Oceania, South America)
Cheapest region
US East (Ohio) · $21.96/hr
Dedicated EBS throughput
19000 Mbps
ECU
345
Enhanced networking
No
Normalization size factor
192
Processor features
Intel AVX; Intel AVX2; Intel AVX512; Intel Turbo
Spot interruption rate (us-east-1)
15-20%
Spot savings vs on-demand (us-east-1)
56%

p4d.24xlarge equivalents on other clouds

closest match by vCPU, memory, CPU architecture, accelerator and class · Linux on-demand in each provider's reference region

compare side by side→
ProviderInstancevCPUMemoryProcessorOn-demandvs p4d.24xlarge
AWSp4d.24xlarge961,152 GiB8 × NVIDIA A100 40GB$21.96–
AzureND96asr A100 v4900 GiB (1.3× less)96900 GiB8 × NVIDIA A100 40GB$27.20+24%
Google Clouda2-highgpu-8g680 GiB (1.7× less)96680 GiB8 × NVIDIA A100 40GB$29.39+34%
a2-megagpu-16g1360 GiB (1.18×) · 16× GPU961,360 GiB16 × NVIDIA A100 40GB$55.74+154%

P4d family sizes

all about P4d→
SizevCPUMemoryAcceleratorLocal storageOn-demandMonthlySpot1-yr
p4d.24xlarge961,152 GiB8 × NVIDIA A100 40GB8 TB SSD (8 disks)$21.96$16,029$13.07$13.92

p4d.24xlarge FAQ

How much does p4d.24xlarge cost?

p4d.24xlarge costs $21.96 per hour on demand with Linux in US East (N. Virginia) (us-east-1), about $16,029 per month. Across 15 regions the Linux on-demand price ranges from $21.96 in US East (Ohio) to $35.56 in South America (Sao Paulo). Spot capacity costs $13.07 per hour (40% less), but can be reclaimed. A 1-yr Reserved commitment brings it to $13.92 per hour.

How many vCPUs and how much memory does p4d.24xlarge have?

p4d.24xlarge has 96 vCPUs and 1,152 GiB of memory (12 GiB per vCPU), running on Intel Xeon Platinum 8275L. It is a gpu & accelerated instance type in the P4d family.

What is the cheapest region for p4d.24xlarge?

US East (Ohio) (us-east-2) at $21.96 per hour for Linux on demand. p4d.24xlarge is sold in 15 AWS regions.

What GPU does p4d.24xlarge have?

p4d.24xlarge has 8 × NVIDIA A100 40GB with 320 GiB of accelerator memory in total.

What are the equivalents of p4d.24xlarge on other clouds?

The closest matches by vCPU, memory, CPU architecture, accelerator and instance class are on Azure, ND96asr A100 v4 (96 vCPU, 900 GiB, $27.20/hour in East US); on Google Cloud, a2-highgpu-8g (96 vCPU, 680 GiB, $29.39/hour in Iowa).