設計工具
High-bandwidth memory

HBM3E

The industry's fastest, highest-capacity high-bandwidth memory (HBM) to advance generative AI innovation

pause

Micron commences volume production of industry-leading HBM3E to accelerate growth of AI

Micron's HBM3E helps reduce data center operating costs by consuming about 30% less power than competing offerings. The 24GB 8-high solution will be part of NVIDIA H200 Tensor Core GPUs which will begin shipping in the second calendar quarter of 2024.

Micron's HBM3E, now in volume production. Fueling the AI revolution.

Advancing the rate of AI innovation

Generative AI

Generative AI opens a world for new forms of creativity and expression, like the image above, by using large language model (LLM) for training and inference. Utilization of compute and memory resources make the difference in time to deploy and response time. Micron HBM3E provides higher memory capacity that improves performance and reduces CPU offload for faster training and more responsive queries when inferencing LLMs such as ChatGPT.™

3D illustration of an astronaut made of crystals

Deep learning

AI unlocks new possibilities for businesses, IT, engineering, science, medicine and more. As larger AI models are deployed to accelerate deep learning, maintaining compute and memory efficiency is important to address performance, costs and power to ensure benefits for all. Micron HBM3E improves memory performance while focusing on energy efficiency that increases performance per watt resulting in lower time to train LLMs such as GPT-4 and beyond.

Light blue, pink and yellow data streams combining to form a full variation of datasets on the right

High-performance computing

Scientists, researchers, and engineers are challenged to discover solutions for climate modeling, curing cancer and renewable and sustainable energy resources. High-performance computing (HPC) propels time to discovery by executing very complex algorithms and advanced simulations that use large datasets. Micron HBM3E provides higher memory capacity and improves performance by reducing the need to distribute data across multiple nodes, accelerating the pace of innovation.

An artist in a dimly lit room using a stylus to create a colorful abstract design on a tablet

HBM3E built for AI and supercomputing with industry-leading process technology

Micron extends industry-leading performance across our data center product portfolio with HBM3E. Delivering faster data rates, improved thermal response, and 50% higher monolithic die density within same package footprint as previous generation.

Micron's HBM3 Gen3

HBM3E built for AI and supercomputing with industry-leading process technology

HBM3E provides the memory bandwidth to fuel AI compute cores

With advanced CMOS innovations and industry-leading 1β process technology. Micron HBM3E provides higher memory bandwidth that exceeds 1.2 TB/s.1

Blurred image of female wearing AI goggles

HBM3E unlocks the world of generative AI

With 50% more memory capacity2 per 8-high, 24GB cube, HBM3E enables training at higher precision and accuracy.

Splash of water on neon background

HBM3E delivers increased performance per watt for AI and HPC workloads

Micron designed an energy-efficient data path that reduces thermal impedance and enables greater than 2.5X improvement in performance/watt3 compared to the previous generation.

Generative AI illustration of a modern high technology server room in purple neon colors

HBM3E pioneers training of multimodal, multitrillion-parameter AI models

With increased memory bandwidth that improves system-level performance, HBM3E reduces training time by more than 30%4 and allows >50% more queries per day.5,6

Chat Bot Chat with AI or Artificial Intelligence technology. Woman using a laptop computer chatting with an intelligent artificial intelligence asks for the answers he wants

Micron HBM3E: The foundation for unlocking unprecedented compute possibilities

Micron HBM3E is the fastest, highest-capacity high-bandwidth memory to advance AI innovation — an 8-high, 24GB cube that delivers over 1.2 TB/s bandwidth and superior power efficiency. Micron is your trusted partner for memory and storage innovation. 

Cropped shot of African American data engineer holding laptop while working with supercomputer in server room lit by blue light, copy space

Frequently asked questions

There is no difference. It is just a name change.

Micron’s HBM3E delivers an industry leading pin speed of > 9.2Gbps and can support backward compatible data rates of HBM2 first generation devices.

Micron’s HBM3E delivers an industry leading Bandwidth of >1.2 TB/s per placement. HBM3E has 1024 IO pins and Micron’s pin speed > 9.2Gbps achieves > 1.2TB/s​.

Micron’s industry leading HBM3E provides 24GB capacity per placement with an 8-high HBM3E, Micron plans to announce a 36GB 12-high HBM3E device in the future.​​

Micron’s HBM3E delivers an industry leading Bandwidth of >1.2 TB/s per placement. HBM3E has 1024 IO pins and Micron’s pin speed > 9.2Gbps achieves > 1.2TB/s.​

HBM2 offers 8 independent channels running at 3.6Gbps per pin and delivering up to 410GB/s bandwidth. HBM2 offers 4GB, 8GB and 16GB of capacity. HBM3E offers 16 independent channels and 32 psuedo channels. Micron’s HBM3E delivers pin speed > 9.2Gbps at an industry leading Bandwidth of >1.2 TB/s per placement. Micron’s HBM3E offers 24GB using 8-high stack, and a 36GB using 12-high stack is planned for the future.

Please see our Product Brief.

Featured resources

1.  Data rate testing estimates based on shmoo plot of pin speed performed in manufacturing test environment.
2.  50% more capacity for same stack height.
3.  Power and performance estimates based on simulation results of workload uses cases.
4.  Based on internal Micron model referencing an ACM Publication, as compared to the current shipping platform (H100).
5.  Based on internal Micron model referencing Bernstein’s research report, NVIDIA (NVDA): A bottoms-up approach to sizing the ChatGPT opportunity, February 27, 2023, as compared to the current shipping platform (H100).
6.  Based on system measurements using commercially available H100 platform and linear extrapolation.

Customer support

Need to get a hold of us? Contact our support teams as well as get contact information for our individual locations.

Sales

Contact our sales support team directly by completing the Sales Support form on our Contact Us page.

Order a Micron sample

Your online source for placing and tracking orders for Micron memory samples.

Downloads & technical documentation

Dive deeper into product features or functionality and get design guidance.