Overview
The Synopsys Die-to-Die Controller IP, optimized for latency, bandwidth, power and area, enables efficient inter-die connectivity in server, AI accelerator, networking and high-performance computing SoCs. The controller interoperates with the Synopsys 112G XSR PHY to deliver a complete Die-to-Die solution for a seamless connection between the on-die interconnect fabrics in both dies via the standard CXS port. The Synopsys Die-to-Die Controller uses
a FLIT-based architecture to minimize latency. It implements an advanced error detection and correction mechanism including Cyclic Redundancy Check (CRC) and optional latency-optimized Forward Error Correction (FEC) to reduce Bit Error Rate (BER) to a very low level for PAM-4 or NRZ PHY
signaling. The embedded retry protocol enables very low latency, error free links between two dies.
The Synopsy Die-to-Die Controller optimizes system performance by supporting two configurations for coherent and non-coherent data traffic between the SoC bus and each die. The latency-optimized configuration interfaces with the SoC fabric via a FLIT-based interface (Arm® CXS). The Synopsys Die-to-Die Controller can be extended to support any aggregate bandwidth between the two dies using bifurcation into multiple parallel links.
Learn more about Custom IP core
Synopsys delivers standards-compliant 112G PHY IP for linear optics, helping reduce AI interconnect power, latency, and heat.
“We are in an era of pervasive intelligence,” said Shankar Krishnamoorthy, chief product development officer at Synopsys. Moderating a panel discussion at the inaugural Synopsys Executive Forum, he described the transformative impact of artificial intelligence (AI) and the challenges ahead.
As artificial intelligence (AI) reshapes industries and advances technological frontiers, its success hinges on advanced memory capabilities. Leading this transformation is High Bandwidth Memory (HBM), which offers unparalleled speeds and efficiencies.
With more than 25 years of experience in delivering field-proven silicon IP solutions, we are thrilled to partner with NVIDIA and the NVIDIA NVLink ecosystem to enable and accelerate the creation of custom AI silicon. This strategic collaboration will leverage Synopsys' expertise in silicon IPs to assist in the development of bespoke AI silicon, forming the foundation for advanced compute clusters aimed at delivering the next generation of transformative AI experiences.
The new ARC VPX6 processor introduces 1024-bit vector processing to the lineup, extending the existing VPX family, which includes VPX5 (512-bit), VPX3 (256-bit), and VPX2 (128-bit) variants. As such, VPX6 doubles the attainable peak performance while preserving full software compatibility with the other members of the VPX family. By simply upgrading hardware without any recoding, designers can instantly realize performance gains from day one.
For AI workloads to be processed reliably at a fast rate, the die-to-die interface in multi-die designs must be robust, low latency, and most importantly high bandwidth. This article outlines the need for 40G UCIe IP in AI data center chips leveraging multi-die designs.