New solutions provide scalability, low resource utilization and high performances in managing large amounts of data for designers and architects in the high-end computing, AI, machine learning and IoT spaces
SAN JOSE, Calif. — (BUSINESS WIRE) — April 4, 2019 — PLDA, the industry leader in PCI Express® IP solutions, today announced two innovative DMA engine solutions designed to manage large and heterogeneous data traffic across a PCIe link or across an AMBA AXI fabric solution. PLDA’s new IP products are an elegant and streamlined solution that delivers high-end performance and is easy to integrate into design planning.
This press release features multimedia. View the full release here: https://www.businesswire.com/news/home/20190404005532/en/
The explosion in data traffic due to the advent of artificial intelligence, machine learning, automotive and IoT has prompted the development of new System-on-Chip architectures. These architectures allow system developers to build high-performance applications that can ingest and process huge amounts of heterogeneous data. In order to take advantage of these architectures, some SoCs integrate a large number of processing elements and require a scalable solution for moving data across computing elements or across a PCIe-attached host platform. In order to meet these needs, PLDA has developed a high-performance DMA Engine available in two versions: As a standalone vDMA engine IP with AMBA-AXI interconnect or as a PCIe 5.0 IP controller with integrated vDMA engine.
This standalone vDMA-AXI IP product is the ideal centralized AMBA AXI DMA for SoCs that integrate a large number of processing elements and require a scalable solution for moving data across these computing elements. vDMA-AXI is particularly suited for applications in AI, machine learning and automotive spaces and delivers:
- Up to 2048 DMA channels
- Up to 512 Virtual Machines or domains
- A configurable AMBA AXI3/4 interface with either a 256-bit or 512-bit data path
- Up to 128 outstanding read and write requests
This PCIe 5.0 controller IP with built-in vDMA engine is designed for applications that require the highest performance and flexibility for moving data across a PCIe-attached host platform, whether virtualized or not. XpressRICH5-AXI with vDMA is particularly suited for applications in cloud computing, HPC, AI, machine learning, storage, and networking. In addition to the features outlined in the standalone vDMA engine, XpressRICH5-AXI also delivers:
- Support for x16, x8, x4, x2, x1 at PCIe Gen5, Gen4, Gen3, Gen2, and Gen1 speeds
- Compliance with PCI-SIG Single-Root I/O Virtualization (SR-IOV) Specification
- Support for Endpoint, Root-Port, and Dual-mode configurations
- Support for up to 32 Physical Functions (PF) and 512 Virtual Functions (VF)
According to Arnaud Schleich, CEO of PLDA, “New applications and the ever-growing quest for bandwidth is driving a need for faster I/O combined with a flexible design. PLDA’s vDMA solutions provide the performance needed by higher end applications with an ease-of-integration that delivers silicon success.”
Detailed information on these products is available online via video demonstrations and product brochures. You may also contact PLDA directly to discuss your individual project at firstname.lastname@example.org.
- Video Series: Check out the video series which describes how vDMA works and demonstrates the performance and capabilities of PLDA’s vDMA engine, as a stand-alone IP or as an integrated feature in XpressRICH5-AXI.
- Product information and specifications: