CASE STUDY
Nogales Lab Reduces Power Consumption by 60% and Increases Capacity by 40% with Ampere®
Nogales Lab Reduces Power Consumption by 60% and Increases Capacity by 40% with Ampere®
02 September 2023
02 September 2023
INTRODUCTION
Groundbreaking cryo-EM research at Nogales Lab was consistently impeded by critical infrastructure limitations. Their Intel x86 servers, tasked with demanding CPU-intensive workloads, suffered from chronic overheating and debilitating thermal throttling, resulting in significant research downtime. Further complicating matters, a constrained data center, insufficient cooling, and unpredictable budgets rendered conventional compute capacity expansion impractical. Facing these significant hurdles, Nogales Lab sought an innovative solution, ultimately deploying Ampere Altra processors to overcome bottlenecks and dramatically enhance their research capabilities.
COMPANY DESCRIPTION
Nogales Lab, a leading cryo-EM facility at UC Berkeley, pioneers the study of macromolecular assemblies. They utilize cryogenic electron microscopy to create detailed 2D and 3D images of single particle and microtubule biological structures. By directly visualizing architecture, functional states, and regulatory interactions, the lab drives critical advancements in understanding molecular biology, enabling breakthroughs in health and life sciences research.
Challenges
The lab’s Intel x86 servers couldn't handle CPU-intensive cryo-EM workloads, leading to overheating and thermal throttling, which halted research. A limited data center footprint, inadequate cooling, and fluctuating budgets made expanding compute capacity difficult, prompting an urgent need for a more efficient processing solution.
Solution
To overcome their limitations, Nogales Lab’s HPC Administrator deployed Ampere Altra Cloud Native Processors in Supermicro servers. The integration was surprisingly easy, functioning just like existing hardware. This strategic, entirely new CPU architecture efficiently addressed their escalating compute, power, and thermal challenges, marking a significant step forward for the lab.
Benefits
Ampere Altra processors delivered 60% less power consumption than their newer AMD EPYC system, providing 40% more overall compute capacity. This eliminated thermal throttling, allowing full utilization of all 128 cores. Researchers could complete jobs faster, benefiting from 40% better performance per watt, enhanced energy efficiency, and significant cost savings.
Performance Results
| Altra Max M128-30 | 2 AMD Epyc 7702 | 2 Intel Xeon 6238R | Power10 48-core, 384 thread | |
| Benchmark | 128 cores / CPU | 64 cores / CPU | 28 cores / CPU | 48 cores / CPU |
| 3D Classification | 58 hours, 22 min (128 core) | 42 hours, 12 min (256 threads) | 46 hours, 12 min (112 threads) | 26 hours, 13 min |
| AutoPicking | 1 hours, 17 min (64 core) | 1 hours, 2 min (64 core) | 1 hours, 3 min | 29 min, 28 sec |
| 2D Classification | 24 min, 12 sec (128 core) | 21 min, 4 sec (256 threads) | 52 min, 24 sec (112 threads) | 15 min, 23 sec |
| Notes | Power draw 515W | Power draw 1300W | Does not support GPUs, cost 12X vs Amper |
All tests performed with Relion 4.0.1 and CPU acceleration 0.603846154
“We found ourselves needing more and more CPU specific servers, and with the increase in the Intel and NVIDIA tax, we began looking for options that could deliver similar performance while lowering power consumption and price. What I discovered was that an entire architecture change was needed to lower the power consumption, lower the heat generation, and still be performant. That’s where we decided on Ampere.”
— Kurt Stine,
HPC Systems Administrator, Nogales Lab
“Performing his own benchmarking studies, Stine discovered that not only does the Ampere Altra processors deliver 40% better performance per watt, but he also saw equivalent and sometimes better performance than AMD EPYC. “Our initial results with Ampere looked really good—even without Relion being optimized— we’re getting similar performance numbers to EPYC. For the vast difference in price, it’s perfect for our compute requirements and our limited budget.”
— Kurt Stine,
HPC Systems Administrator, Nogales Lab
© 2025 Ampere® Computing LLC. All rights reserved. Ampere®, Ampere® Computing, Altra and the Ampere® logo are all trademarks of Ampere® Computing LLC or its affiliates. Other product names used in this publication are for identification purposes only and may be trademarks of their respective companies.
Ampere Computing LLC
4655 Great America Parkway Suite 601
Santa Clara, CA 95054
