Cloud Computing AI Servers

AI servers process tremendous amounts of data for deep learning workloads including training and inference which require a large memory capacity, high bandwidths, and cache coherency within the overall system. AI server system-on-chip (SoC) designs involve tradeoffs: high performance; low power; need for cache coherency; integrating high bandwidth interfaces that are scalable to many cores; implementing heterogeneous processing HW accelerators; addressing Reliability-Availability-Serviceability (RAS); massively parallel deep learning neural network processing; advanced FinFET technologies. Synopsys offers a portfolio of DesignWare IP in advanced FinFET processes that address the specialized processing, acceleration, and memory performance requirements of AI servers. For more information visit the DesignWare IP for AI web page.

Mouse over or tap to learn more about DesignWare IP solutions: 

DDR5/4 Controller and PHY

  • Scalable DDR architecture supporting up to 4800 Mbps with single, dual channels
  • High-performance multi-standard PHY
  • High-bandwidth and low latency controller
  • Learn More >>

USB Controller and PHY

  • Interfaces to USB 2.0 and USB 3.0 devices
  • 5 Gbps USB 3.0 read/write performance
  • Supports both host and device operation
  • Learn More >>

PCIe Controller and PHY

  • High-performance, multi-channel PCIe PHY supporting PCIe 5.0, 4.0, 3.1, 2.1, 1.1
  • Configurable PCIe controller supporting PCIe 5.0, 4.0, 3.1, 2.1, 1.1
  • Up to 16 lanes PCIe Dual Mode/Switch Port controller with SR-IOV and RAS features
  • Learn More >>

CCIX Controller & PHY

  • Supports data transfer rates up to 25 Gbps, ready for 32 Gbps
  • Enables faster data access with support for cache coherency
  • Supports Synopsys native interface with extensions and Arm AMBA AXI + CXS interface
  • Learn More >>

Ethernet Controller and PHY

  • Complete solution for up to 400G Ethernet applications
  • Configurable Ethernet controllers compliant with IEEE standards
  • High-performance, multi-standard NRZ and PAM-4 PHYs enable required reach and performance
  • Learn More >>

HBM2 PHY

  • High memory throughput with low power and latency
  • Supports pin bandwidth up to 2400Mbps
  • 10X better energy efficiency than DDR4
  • Learn More >>

AMBA Interconnect

  • ARM® AMBA® 4 AXI, ACE-Lite™, AXI3™, AHB™, APB™ interconnect IP
  • Automated AMBA subsystem assembly
  • Learn More >>

Security IP

  • Self-contained hardware secure modules with Root of Trust
  • Content protection embedded security modules for high-value content
  • Configurable cryptography IP
  • High-performance security protocol accelerators
  • Learn More >>

I2C, GIPI, UART …

  • Functions for on-chip and off-chip connectivity
  • Synthesizable to meet specific design requirements
  • Learn More >>

Embedded Memories and Logic Libraries

  • High-speed and high-density memory compilers and logic libraries
  • Optimized for advanced CPUs, GPUs, and DSPs
  • High Performance Core Design Kit for optimal processor speed, area and power
  • Learn More >>

Highlights: