0
Rambus Unveils GDDR7 Memory Controller IP: PAM3 Signaling, Up To 48 Gbps Data Rates, 192 GB/s Bandwidth

Rambus Unveils GDDR7 Memory Controller IP: PAM3 Signaling, Up To 48 Gbps Data Rates, 192 GB/s Bandwidth

Rambus has it. The unveiling of that Next generation GDDR7 memory The controller features IP, PAM3 signaling, and transfer speeds of up to 48 Gbps.

Rambus Ready for Next-Gen Gaming and AI GPUs with GDDR7 Memory Controller IP: Delivering substantial improvements over GDDR6.

News for the newspaper: As the latest addition to the Rambus portfolio of industry-leading interfaces and security digital IP for AI 2.0, the GDDR7 memory controller will deliver the breakthrough memory throughput needed by servers and clients in the next wave of AI inference.

Supercharging AI Inference with GDDR7

The output of the AI ​​2.0 training process is an inference model that can be used to generate new multimodal content from user input. Because accuracy and fidelity increase with model size, there is a constant push for larger and larger inference models. As AI computing rapidly expands and moves out of the data center to the edge and endpoints, it requires more powerful processing engines with high-performance memory solutions across the computing landscape.

Image Source: Rambus

GPUs have been the inference engine of choice, and in the case of edge and endpoint applications, such as servers and desktops, it has been GPUs using GDDR6 memory. GDDR6, however, reaches the practical limit of standard NRZ signaling at a data rate of 24 gigabits per second (Gbps). To meet the bandwidth requirements of future GPUs, a new generation of GDDR using a new signaling scheme is required. Enter GDDR7 memory that can boost data rates to 40 Gbps and more using PAM3 signaling.

GDDR, now at the GDDR7 specification level, is the most advanced graphics memory solution today with a performance roadmap of up to 48 gigabits per second (Gbps) and memory throughput of 192 GB/s per GDDR7 memory device.

On the crucial parameter of bandwidth, GDDR7 memory really shines. At a data rate of 32 Gbps, and a 32-bit-wide interface, a GDDR7 device can deliver 128 GB/s of memory bandwidth, double that of any alternative solution. GDDR7 memory offers excellent speed, bandwidth and latency performance for AI estimation.

Rambus

Rambus Silicon IP for AI 2.0

As the preferred silicon IP provider for AI 2.0, Rambus offers industry-leading HBM, PCIe, and CXL controller IP and is now an industry first. GDDR7 Memory Controller IP The Rambus GDDR7 controller provides a full-featured, bandwidth-efficient solution for GDDR7 implementations. It supports 40 Gbps operation providing 160 gigabytes per second (GB/s) throughput for a GDDR7 memory device, a 67 percent increase over the industry's highest throughput GDDR6 controller (also from Rambus). Is.

Image Source: Rambus

The Rambus GDDR7 controller enables a new generation of GDDR memory deployments for advanced AI accelerators, graphics, and high-performance computing (HPC) applications.

“GDDR7 memory offers significant performance gains over GDDR6,” said Soo Kyum Kim, vice president of memory semiconductors at IDC. “The Rambus GDDR7 Controller IP solution will be an important tool for anyone looking to take advantage of the improved speed and latency features offered by GDDR7.”

Key Features of Rambus GDDR7 Controller:

  • Supports all GDDR7 link features including PAM3 and NRZ signaling.
  • GDDR7 supports a wide range of device sizes and speeds.
  • Optimized for high performance and low latency in various traffic scenarios.
  • Flexible AXI interface support
  • Low power support (self-refresh, hibernate self-refresh, dynamic frequency scaling, etc.)
  • Reliability, Availability and Serviceability (RAS) features – such as end-to-end data path parity, parity protection for stored registers, etc.
  • Comprehensive memory test support
  • Integration support is available for third-party PHY.
  • Verified using the latest GDDR7 VIP and memory vendor memory models

Share this story.

Facebook

Twitter

About the Author

Leave a Reply