Enterprise Solutions for AmpereOne^® Family

Delivering Power-Efficient, High-Performance AI Compute with the New AmpereOne^® M Processors

Discovering New Levels of Core Dense Systems for AI Compute

Expanding the already core-dense CPU lineup of the AmpereOne^® family, the new AmpereOne^® M processors further extend memory bandwidth with support for 12 DDR5 memory channels, while maintaining an impressive core density that tops at 192 cores. Built on the same single-threaded Arm^®-based custom core design, AmpereOne^® M integrates large L2 caches and 2×128b vector units per core, ensuring highly efficient execution across diverse workloads.

The AmpereOne^® M processors are purpose-built for AI inference, analytics, and other memory bandwidth-sensitive applications where efficiency and scalability are critical. With leading power efficiency, robust RAS capabilities, and cloud-native features at its core, AmpereOne^® M provides the performance foundation demanded by modern data centers. By balancing compute density, bandwidth, and reliability, AmpereOne^® M enables operators to maximize performance per rack while meeting the growing needs of AI-driven services.

Sustainable AI Compute from the Edge to the Cloud

Redefine Performance with AmpereOne^® M Processors

5nm

process

Ampere Computing designed an in-house Arm processor using chiplets

96-192

custom cores

More tasks can be done simultaneously and faster with dedicated custom cores

PCIe 5.0

lanes

Increased I/O throughput achieving 128GB/s bandwidth in x16 lanes

2MB L2

cache/core

Doubling of L2 private cache compared to prior gen Ampere CPU

DDR5

memory

Increased memory throughput and higher DDR5 capacity per DIMM

BFloat16

support

Added to list of data formats that includes: FP16, Int8, and Int16

Advantages of AmpereOne Family

Predictable High Performance

192 single-threaded cores at a consistent 3.2 GHz and increased memory bandwidth are ideal for AI compute. Also, large private caching and memory management features for consistency.

Energy Efficiency

A high CPU core count enables dense computing and increases the number of cores per rack, delivering efficiency in GIGABYTE servers that are VM-dense.

Power Efficient

192 Ampere cores run below 425W with multiple power management features, including dynamic power optimization. This reduces the carbon footprint and stays within rack power constraints.

GIGABYTE Servers for AmpereOne Family

Explore AmpereOne^® M based servers by GIGABYTE with Ampere's Jeff Wittich

Partnering with Ampere Computing since the launch of the Ampere Altra Family processors, GIGABYTE has continued to deliver innovative solutions powered by Arm architecture. Join us at COMPUTEX 2025 to hear from Jeff Wittich, Chief Product Officer at Ampere Computing, and discover GIGABYTE’s latest servers built for AmpereOne^® M—engineered to excel in both AI inference and modern cloud workloads.

Dive into GIGABYTE's Arm-based servers with the AmpereOne^® platform

With Ampere Computing releasing its new AmpereOne Family processors with a maximum of 192 cores, GIGABYTE introduces a brand new server lineup with numerous features that benefit cloud applications on the new platform. Expanding the GIGABYTE server portfolio for the scalable and highly efficient Ampere Arm-based platform, which has already been proven successful with the Ampere Altra product line.

Applications

Cloud

Cloud native operations prefer high core count CPUs because they are utilized in an environment where many requests are needed to be handled concurrently, and they use multi-tenancy architecture with multiple instances for users. It becomes even more important when VM is utilized, making traditional multi-threading technology less critical than having more physical cores. With the lower Watt/core, AmpereOne Family processors make it possible to achieve up to 3 times the physical core count in a single rack under the same power limitations.

AI Inference

AI Inference has started to take place in every aspect of our lives, from heavy computing data centers to the daily use of smartphones. While inference can be run on traditional servers with GPUs installed, the hardware requirements are much lower than those for AI training. For traditional AI models and SLMs <15B parameters, AI optimized CPUs excel. The AmpereOne^® M processors, which feature high memory bandwidth and support memory capacities of up to 1.5 TB, allow systems to easily perform AI inference in a GPU-less environment, leading to a significant reduction in costs.