Skip to content
CloudPlay
  • Pages
    • Project Overview
      • Product roadmap
    • Documentation
      • Transformation
      • Server Requirements
      • System Requirements
      • Own Servers vs Cloud Rental
      • Design Related
      • Cloud GPUs
      • icon picker
        AWS Cloud Gaming
      • Harvester Integration Node
      • Operational Costs
    • Team
    • Tasks
    • Funding
    • Funding Cadence
      • Month 1
      • Month 2
      • Month 3
      • Month 4
      • Month 5
      • Month 6
      • Month 7
      • Month 8
      • Month 9
    • Updates
    • Discussions
Share
Explore

AWS Cloud Gaming

Cloud gaming needs a balance between GPU performance, cost, and licensing. On AWS, the two most affordable GPU-powered instances for gaming workloads are g4ad.xlarge and g4dn.xlarge.

error

All prices are calculated for AWS EU-Central-1 (Frankfurt) region using . Prices vary between regions.

Instance Overview

Feature
g4ad.xlarge
g4dn.xlarge
vCPU
4 vCPUs (2nd Gen AMD EPYC)
4 vCPUs (Intel Cascade Lake)
Memory
16 GiB
16 GiB
GPU
1x AMD Radeon™ PRO V520
1x NVIDIA T4 Tensor Core
GPU Memory
8 GB GDDR6
16 GB GDDR6
Network
Up to 10 Gbps
Up to 25 Gbps
Storage (Instance)
125 GB NVMe + EBS
125 GB NVMe + EBS
Hourly Price
~$0.473
~$0.658
Windows License Fee
+$0.184/h
+$0.184/h
Performance 3DMark
5100
4300
Price-Performance
10782
6534
There are no rows in this table
These two instance types are different: g4ad.xlarge often delivers better raw gaming performance with its AMD GPU, but for video encoding and cloud streaming, NVIDIA ( g4dn.xlarge ) is generally preferred thanks to NVENC which is faster, more reliable and lower latency.

Boot Time & Instance Startup

One advantage of using AWS AMIs is fast boot and setup time, compared to providers like Vultr.
Current demo setup times: ~120–160 seconds.
With instance reservations, startup time can be reduced by ~60% (because capacity is pre-allocated).
Even in slower cases, startup usually stays under 200 seconds for g4dn instances.

Cost Comparison - Per Game Session (Per Hour)

Cost Component
g4ad.xlarge
g4dn.xlarge
Instance Price (On-Demand)
$0.473 / h
$0.658 / h
OS License
$0.184 / h
$0.184 / h
Storage (OS + Game, 150 GB @ $0.08/GB-month)
~$0.024 / h*
~$0.024 / h*
Snapshot Storage (110 GB @ $0.05/GB-month)
~$0.011 / h*
~$0.011 / h*
Total (with Windows license)
$0.692 / h
$0.877 / h
Total (BYOL/KMS)
$0.508 / h
$0.693 / h
There are no rows in this table
* Storage amortized: (150 GB × $0.08 = $12/month) ÷ 720h ≈ $0.016/h.
* Snapshot amortized: (110 GB × $0.05 = $5.5/month) ÷ 720h ≈ $0.008/h. Rounded slightly.

Infrastructure Comparison

Provider
Instance
Instance
AWS
g4ad.xlarge
$0.69 / h
AWS
g4dn.xlarge
$0.88 / h
Azure
NVIDIA A10 v5 (½ GPU)
$0.90 / h
Google Cloud
G2 (1× L4)
$0.92 / h
There are no rows in this table
Alternative Approach — Multi-GPU Dedicated Server
For production environments with consistent daily active users and near-full GPU utilization, a dedicated multi-GPU server can dramatically reduce costs.
Example:
3× RTX 3060 Ti server from
Monthly cost: $499.00
Can host: 3 simultaneous players
Effective cost: $0.23/h per GPU ($499.0 ÷ (3 GPUs × 720 hours))

Considerations

No on-demand scalability
Requires capacity management and uptime monitoring

Cloud Gaming Platforms That Use AWS

, Amazon’s own cloud gaming platform, is a real-world example that uses AWS EC2 G4 (g4dn.xlarge) instances for its backend.
In contrast, , , and operate on dedicated datacenter or hybrid cloud models with their own GPU clusters to achieve lower per-hour costs and more predictable performance at scale.
Their monthly plans are significantly cheaper because these platforms run on long-term reserved or owned infrastructure instead of hourly-billed cloud GPUs. For example, GeForce NOW offers 100 hours of playtime for just $9.99/month, which translates to roughly $0.10–$0.16 per hour — far below on-demand AWS or Azure GPU rates.
This cost efficiency comes from:
Full control over hardware utilization
No per-hour cloud billing overhead
Optimized encoding pipelines tuned to their datacenter setup

Subscription Per Month
GeforceNow
$9.99/monthly
Boosteroid
$8.74/monthly
Shadow PC
$32.99/monthly
Amazon Luna
$9.99/monthly
There are no rows in this table
As a result, dedicated or hybrid models remain the most cost-effective.
Summary:
g4ad.xlarge → ~$0.69/hour per session (Windows licensed) or ~$0.51/hour (BYOL)
g4dn.xlarge → ~$0.877/hour per session (Windows licensed) or ~$0.69/hour (BYOL)
Studies show the average teen plays ~2.13 hours/day, or about 60 hours/month.
If building a subscription model:
A 60-hour baseline can be assumed per user.
Light players (<60h) help subsidize heavy players (>60h).
This keeps pricing predictable and consumer-friendly, unlike pure pay-as-you-go billing.

Pricing Example

On-demand costs put g4ad.xlarge at ~$30–40/month per user for 60h.
This is too high compared to competitors like GeForce NOW or Xbox Cloud Gaming.
The realistic approach is capacity reservations (or savings plans):
Reserved instances cut costs by ~50% or more.
With this, subscription pricing drops to around:
g4ad.xlarge → ~$15/month per user
g4dn.xlarge → ~$20/month per user
⚠️ Note: these numbers exclude any platform/service commission (X%) that must be added on top.

Resources


 
Want to print your doc?
This is not the way.
Try clicking the ⋯ next to your doc name or using a keyboard shortcut (
CtrlP
) instead.