The g4dn.xlarge instance from Amazon EC2 is part of the G4dn GPU-optimized family, designed for cost-efficient machine learning inference, graphics-intensive workloads, and GPU-accelerated media processing.
It features 4 vCPUs, 16 GiB of memory, and an NVIDIA T4 Tensor Core GPU, with network throughput of up to 25 Gbps. The instance offers a balanced performance-to-cost profile, ideal for small-to-medium workloads requiring GPU acceleration without overprovisioning.
This reference sheet includes full specifications, pricing, use cases, family sizes, and variant options for informed selection. To compare the g4dn.xlarge with other instances, click here.
Instance Overview
- Instance Type: g4dn.xlarge
- Instance Family: G4dn (GPU-optimized accelerated computing)
- Generation: Current
- Use Case Focus: Machine learning inference, graphics workloads, video transcoding, remote workstations, cloud-based game streaming
- Availability: Multiple AWS regions (US, EU, APAC, Canada)
Feature Overview
- 2nd Generation Intel Xeon Scalable Processors (Cascade Lake P-8259CL)
- 1 NVIDIA T4 Tensor Core GPU (per g4dn.xlarge) — G4 family supports up to 8 GPUs
- Network throughput up to 25 Gbps (up to 100 Gbps for larger G4 instances)
- Local NVMe storage: 125 GiB per instance (up to 1.8 TB in G4 family)
- EBS-optimized with dedicated storage bandwidth
- Supports Elastic MapReduce (EMR) workloads
- GPU-accelerated machine learning and graphics performance
Instance Details — g4dn.xlarge
Compute
|
Spec
|
Value
|
|
vCPUs
|
4
|
|
Memory (GiB)
|
16
|
|
Memory per vCPU (GiB)
|
4
|
|
Physical Processor
|
Intel Xeon Family
|
|
Clock Speed (GHz)
|
2.5
|
|
CPU Architecture
|
x86_64
|
GPU
|
GPU
|
1
|
|
GPU Architecture
|
NVIDIA T4 Tensor Core
|
|
Video Memory (GiB)
|
16
|
|
GPU Compute Capability
|
7.5
|
|
GPU Average Wattage
|
0 W
|
|
FPGA
|
0
|
|
ffmpeg FPS
|
105
|
|
CoreMark iterations/Second
|
55,463
|
Networking
|
Spec
|
Value
|
|
Network Performance (Gbps)
|
Up to 25 Gigabit
|
|
Enhanced Networking
|
true (AWS ENA)
|
|
IPv6 Support
|
true
|
Storage
|
Spec
|
Value
|
|
EBS Optimized
|
true
|
|
Max Bandwidth (Mbps) on EBS
|
3500
|
|
Max Throughput (MB/s) on EBS
|
437.5
|
|
Max I/O operations/second (IOPS)
|
20,000
|
|
Baseline Bandwidth (Mbps) on EBS
|
950
|
|
Baseline Throughput (MB/s) on EBS
|
118.75
|
|
Baseline IOPS
|
3,000
|
|
Devices
|
1
|
|
Swap Partition
|
false
|
|
NVMe Drive
|
true
|
|
Disk Space (GiB)
|
125
|
|
SSD
|
true
|
|
Initialize Storage
|
false
|
Pricing by Region (USD)
|
Region
|
On-Demand /hr
|
Spot /hr (≈)
|
1-Year Reserved /hr
|
3-Year Reserved /hr
|
|
US East (N. Virginia)
|
0.526
|
~0.19
|
0.331
|
0.227
|
|
US East (Ohio)
|
0.526
|
~0.198
|
0.331
|
0.227
|
|
US West (Oregon)
|
0.526
|
~0.235
|
0.331
|
0.227
|
|
EU (Stockholm)
|
0.558
|
~0.178
|
0.256
|
0.153
|
|
Asia Pacific (Mumbai)
|
0.579
|
~0.247
|
0.266
|
0.179
|
|
Canada (Central)
|
0.584
|
~0.194
|
0.269
|
0.160
|
|
Other Regions
|
varies ~0.584–0.940
|
varies ~0.18–0.31
|
varies ~0.24–0.45
|
varies ~0.14–0.29
|
* Prices are approximate; Spot pricing varies based on capacity. Reserved pricing reflects all-upfront equivalent rates.
Converted Period Costs (US East – N. Virginia)
|
Period
|
On-Demand
|
Spot
|
1-Year Reserved
|
3-Year Reserved
|
|
Per Hour
|
$0.5260
|
$0.19
|
$0.331
|
$0.227
|
|
Per Day (24h)
|
$12.62
|
~$4.56
|
~$7.95
|
~$5.45
|
|
30 Days
|
$383.98
|
~$136.80
|
~$238.53
|
~$163.35
|
|
365 Days
|
$4,607.76
|
~$1,742
|
~$2,909
|
~$1,985
|
Use Cases
- Machine learning inference: image metadata tagging, object detection, recommendation systems, speech recognition, language translation
- Graphics-heavy workloads: remote workstations, video transcoding, realistic image rendering, cloud-based game streaming
- Cost-efficient GPU acceleration: smaller-scale AI and graphics projects
- Containerized GPU workloads: batch ML processing, inference pipelines, and GPU-enabled containers
G4dn Family Sizes
Compare g4dn.xlarge to other instances in the same family to choose the right size for your workload.
|
Size
|
vCPUs
|
Memory (GiB)
|
|
g4dn.xlarge
|
4
|
16
|
|
g4dn.2xlarge
|
8
|
32
|
|
g4dn.4xlarge
|
16
|
64
|
|
g4dn.8xlarge
|
32
|
128
|
|
g4dn.12xlarge
|
48
|
192
|
|
g4dn.16xlarge
|
64
|
256
|
|
g4dn.metal
|
96
|
384
|
Instance Variants
Choose the variant depending on whether your priority is ML/AI performance (G4dn) or cost-effective graphics acceleration (G4ad).
- G4dn – NVIDIA T4 Tensor Core GPUs, ideal for ML inference, graphics workloads, and video processing
- G4ad – AMD GPUs, optimized for cost-effective graphics workloads and media encoding
Compare g4dn.xlarge and GPU Instances Across Clouds
Explore g4dn.xlarge in a live environment and deploy easily with no code. Instantly compare with other GPU instances based on performance, cost, and regional availability — not just in AWS, but also across Microsoft Azure, Google Cloud, and other providers.
Compare instances →
Instant access. No setup required. Full pricing insights.
