Rambus circle

Artificial Intelligence & Machine Learning

Speed and Security for the Artificial Intelligence & Machine Learning Revolution

Artificial intelligence (AI) and machine learning (ML) are at the heart of the latest virtuous cycle of computing. Enormous gains in computing power have made practical the neural networks underpinning the revolutionary strides being made in AI and ML. The explosion of AI and ML applications drives the creation of application-focused processors which then take AI and ML to new levels of performance. Concurrently, the development of enormous digital data sets, also thanks to advances in computing and networking, provides the vast training data on which ML depends.

At Rambus, we develop products that move and protect the data critical to the development and performance of AI and ML. We provide high-speed interfaces, security cores, and chips that optimize the computing and networking devices at the foundation of the AI and ML revolution.

Real-time Insights with More, Faster Data icon

Speeding Neural Networks

Huge advances in parallel processing for neural networks power the great leaps realized in AI and ML. But applications made possible by these developments whet greater demands for even higher performance. At the hardware level, the bottleneck has moved from the processor core to the memory and chip-to-chip interfaces at the SoC boundary. At Rambus, we’re pushing the envelope of neural network performance with memory and SerDes IP cores, and memory interface chips, to unleash the performance of the next generation of AI and ML hardware.

SerDes PHYs arrow in blue circle
Memory PHYs arrow in blue circle
Memory Interface Chips arrow in blue circle

Going Beyond GPUs with GDDR6

Going Beyond GPUs with GDDR6

Supported by Micron Technology, SK Hynix and Samsung, GDDR6 SGRAM will feature a maximum data transfer rate of 16 Gbps, along with an operating voltage of 1.35V. GDDR6 offers higher densities compared to previous-generation graphics memory. In addition, GDDR6 doubles the speed (12–16 Gb/s) of GDDR5 and provides more than 5X the 3.2 Gb/s speed of DDR4. Although initially targeted at game consoles and PC graphics, the latest iteration of GDDR is expected to be deployed across multiple verticals, with Micron specifically highlighting the data center and automotive sector.

Download eBook
Safeguarding the Most Valuable Currency: Data icon

Securing Training Data and Algorithms

Given the immense value of the data and algorithms powering AI and ML, safeguarding these from malicious attack is of mission-critical importance. Doing so requires a multi-tiered approach built on a foundation of hardware-level secure silicon. At Rambus, we’ve developed robust hardware-based security solutions that safeguard the SoCs at the heart of AI and ML computing systems. With secure silicon IP and provisioning services, Rambus is at the forefront of protecting advanced AI and ML processing.

CryptoManager Platform arrow in blue circle
DPA Countermeasures arrow in blue circle

The CryptoManager Root of Trust

The CryptoManager Root of Trust

Built around a custom RISC-V CPU, the Rambus CryptoManager Root of Trust (CMRT) is at the forefront of a new category of programmable hardware-based security cores. Siloed from the primary processor, it is designed to securely run sensitive code, processes and algorithms. More specifically, the CMRT provides the primary processor with a full suite of security services, such as secure boot and runtime integrity, remote attestation and broad crypto acceleration for symmetric and asymmetric algorithms.
Download White Paper

FREE Webinar: Secure Silicon IP Series: Complexity vs. Security