Choose the Architecture Proven to Scale for Data Center Workloads
In real-world data centers, Arm processor-based platforms struggle to keep pace with the performance, efficiency, and value of modern AMD EPYC™ CPUs.
The Cloud Should Provide Seamless Computing — Hybrid and Multi-Cloud Included
Enterprises choose hybrid clouds to combine private, on-premises infrastructure with one or more public cloud instances. With x86 infrastructure across your hybrid cloud, workloads can move freely, scale up, and scale out with consistent performance, minimal development work, and low administrative overhead. Arm is incompatible with x86. Adding Arm to your cloud means refactoring, recompiling, and managing multiple code bases in exchange for uncertain gains.
Build In Performance and Savings with AMD EPYC Processors
With AMD EPYC processors you can get better performance per dollar in the cloud and more performance per watt on premises, plus you don’t have to deal with compatibility issues.
Get Up to 75% Better Performance per Dollar Than AWS Graviton4
4th Generation AMD EPYC 9004-based instances on AWS (R7a) outpace AWS Graviton4 instances (R8g), delivering more work for less money.1
Performance per Dollar1 (higher is better)
416.5
238
Use Less Energy On-Premises Than AmpereOne
5th Generation AMD EPYC 9965 CPUs deliver up to 1.3X higher performance per watt than AmpereOne.2
Performance per Watt2 (higher is better)
1225.64
764.75
x86 Compatibility. Zero Cloud Silos.
x86 Delivers Seamless Compatibility and Effortless Integration
Most enterprise software on-premises and in the cloud runs natively on x86 platforms from AMD or Intel. All x86 chips work together seamlessly for maximum flexibility and true workload portability across environments.
Arm processors use instruction sets that are incompatible with the millions of lines of software written for the x86 architecture. The result: siloed clouds that increase operational complexity and reduce compatibility, portability, and flexibility.
Build Unified Hybrid and Multi-Clouds with AMD EPYC™ Processors
With x86, every instance is compatible from on-premises to every x86 cloud provider, enabling truly seamless hybrid and multi-cloud x86 architectures. This type of universal compatibility doesn’t exist across Arm instances.
AWS Graviton, Azure Cobalt, and Google Axion each use a custom version of the Arm architecture. In many cases, workloads can’t travel freely between them, which is how Arm can turn your seamless multi-cloud into a costly, multi-code-base to-do list.
Stick with x86
It’s What Makes the Cloud, the Cloud
The cloud—including hybrid cloud and multi-cloud—is core to modern IT. All these clouds can work seamlessly thanks to the millions of x86 CPUs and the vast array of software available. Choose AMD EPYC processor-based x86 to get leadership performance, energy efficiency, and affordability without the challenges of incompatible Arm instances.
Simplify Hybrid and Multi-Cloud with AMD EPYC CPUs and Skip Arm Migration Hassles
AMD EPYC CPUs provide consistency that Arm simply can’t. When you run all your instances on AMD EPYC CPUs, you don’t have to retrain your people, and you don’t have to refactor, recompile, or manage multiple code bases. In a world of difficult choices, choosing AMD EPYC can be an easy one.
AMD EPYC Portfolio
AMD EPYC x86 CPUs Hybrid Cloud News
Resources
Footnotes
- Phoronix, “AWS Graviton4 Benchmarks Prove to Deliver the Best ARM Cloud Server Performance ,” page 7, July 12, 2024
Performance per dollar calculated as geometric mean performance divided by total cost to complete workloads. On-demand pricing shown is for general-purpose cloud compute instances in the US East region, based on rates from July 2024 and last checked in June 2025. No changes were observed during this period. Pricing may change at any time.
- Phoronix, “AMD EPYC 9965 "Turin Dense" Delivers Better Performance/Power Efficiency vs. AmpereOne 192-Core ARM CPU ,” page 5, October 10, 2024
Performance-per-watt based on the geometric mean of server benchmarks (1P, 192 cores each). Up to 1.3x reflects EPYC 9965’s 1.6x higher performance divided by 1.2x greater average power use.
EPYC 9965 average Watts 275, minimum Watts 19, AmpereOne average Watts 230, minimum Watts 101. Bar values show absolute performance per watt; header shows relative gain.
© 2025 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, EPYC, and combinations thereof are trademarks of Advanced Micro Devices, Inc. in the United States and other countries. Arm is a registered trademark of Arm Limited (or its subsidiaries or affiliates) in the US and/or elsewhere. Other product names used in this publication are for identification purposes only and may be trademarks of their respective owners.
Footnotes
- Phoronix, “AWS Graviton4 Benchmarks Prove to Deliver the Best ARM Cloud Server Performance ,” page 7, July 12, 2024
Performance per dollar calculated as geometric mean performance divided by total cost to complete workloads. On-demand pricing shown is for general-purpose cloud compute instances in the US East region, based on rates from July 2024 and last checked in June 2025. No changes were observed during this period. Pricing may change at any time. - Phoronix, “AMD EPYC 9965 "Turin Dense" Delivers Better Performance/Power Efficiency vs. AmpereOne 192-Core ARM CPU ,” page 5, October 10, 2024
Performance-per-watt based on the geometric mean of server benchmarks (1P, 192 cores each). Up to 1.3x reflects EPYC 9965’s 1.6x higher performance divided by 1.2x greater average power use.
EPYC 9965 average Watts 275, minimum Watts 19, AmpereOne average Watts 230, minimum Watts 101. Bar values show absolute performance per watt; header shows relative gain.
© 2025 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, EPYC, and combinations thereof are trademarks of Advanced Micro Devices, Inc. in the United States and other countries. Arm is a registered trademark of Arm Limited (or its subsidiaries or affiliates) in the US and/or elsewhere. Other product names used in this publication are for identification purposes only and may be trademarks of their respective owners.