.png)
Xilinx Unveils New ACAP to Enhance Adaptive Cloud
Published2020.03.11 10:37Up to 3 times higher throughput than current generation FPGAs
Provides PCIe Gen5 and built-in DMA, CCIX, and CXL
The Adaptive Computing Acceleration Platform (ACAP) has been unveiled, designed for high-bandwidth networks operating in environments with thermal or spatial constraints, and for cloud service providers requiring scalable and adaptable application acceleration.
On the 11th, Xilinx announced 'Versal™ Premium,' the third series of the Versal ACAP product family. Versal Premium integrates interconnected power-optimized cores on an adaptive platform and provides high bandwidth and computing density.
Developed with TSMC's 7nm process technology, Versal Premium features software programmability and dynamically configurable hardware acceleration, as well as pre-designed connectivity and security features, enabling faster time-to-market.
It provides up to three times higher throughput than current generation FPGAs by supporting integrated Ethernet, Interlaken, and encryption engines to enable the implementation of fast and secure networks.
Therefore, it not only doubles the computing density of currently deployed major FPGAs but also provides adaptive capabilities to handle increasingly evolving cloud and networking workloads.
Versal Premium is implemented based on the currently available Versal AI Core and Versal Prime ACAP series. It supports 112Gbps PAM4 transceivers, hundreds of gigabit Ethernet and Interlaken connectivity, high-speed encryption and PCIe® Gen5, as well as built-in DMA, CCIX, and CXL.
In addition, it increases productivity by providing a solution stack for hardware and software developers together with the Vitis™ integrated software platform and Vivado® design suite.
Improve network speed and security
With the emergence of 5G networks and the increase in network traffic, there is a growing demand for power-optimized throughput and computing density within existing installation space and power specifications.
The Versal Premium Series solves this problem by providing adaptive serial bandwidth scalable up to 9 Tb/s. By utilizing 112G PAM4 transceivers and integrated connectivity features, bandwidth density per port can be doubled and latency reduced by up to 50% in core, metro, and DCI (Data center interconnect) infrastructures.
Pre-designed connectivity features have the flexibility to support various data rates and protocols, enabling the implementation of several terabits of Ethernet.
The channeled Ethernet core provides throughput of up to 5 Tb/s while minimizing footprint, and the high-speed encryption engine supports encrypted line rate throughput of up to 1.6 Tb/s and AES-GCM-256/128, MACsec, IPsec, etc.
Adaptive Cloud Acceleration
Versal Premium is designed to achieve the highest level of acceleration for a wide range of hyperscale computing data center workloads.
With a customizable data hierarchy and on-chip data bandwidth of over 120 TB/s, major bottlenecks can be eliminated and data movement reduced, and drop-in integration into existing cloud infrastructure is possible through pre-designed connectivity features and cores.
Samples of the Versal Premium series are scheduled to be supplied to early adopters in the first half of 2021. Documentation is currently available, and customers can begin prototyping using the Versal Prime evaluation kit. Versal Prime devices are implemented with the same architectural blocks as Versal Premium devices, and pin migration to Versal Premium is supported.
관련뉴스
MATLAB and Simulink Support Universal Verification Methodology 'UVM' to Accelerate FPGA and ASIC Verification
HDL Verifier enables design verification engineers developing FPGA and ASIC designs to directly generate Universal Verification Methodology (UVM) components and testbenches through Simulink, and supports their use in simulators that support UVM, such as Synopsys, Cadence, and Mentor Graphics.
2020.01.15by 이수민 기자
MathWorks Supports Automation of FPGA/ASIC-Based Vision System Design with Vision HDL Toolbox R2019
MathWorks announced that Vision HDL Toolbox included in MATLAB and Simulink Release 2019b supports native multi-pixel streaming for HFR and high-resolution image processing on FPGA. Vision HDL Toolbox accelerates the work of video, image processing, and FPGA design engineers to explore or simulate trade-offs between system operation and implementation when processing high-resolution video such as 4K or 8K video at frame rates exceeding 240 frames per second.
2020.02.27by 이수민 기자
TI Launches Stacked DC/DC Buck Converter to Maximize FPGA Power Density
TI has launched the new 40A SWIFT DC/DC buck converter TPS546D24A PMBus, the industry's first buck converter capable of stacking up to four ICs. This power IC delivers a maximum output current of 160A, which is four times higher than competing products. It can reduce power loss by 1.5W in high-performance data center and enterprise computing, medical, wireless infrastructure, and wired networking applications.
2020.03.10by 최인영 기자
Intel Unveils Next-Generation 'Core Ultra Series 3' and vPro Platform for Enterprise PCs
Intel has officially unveiled the "Intel Core Ultra Series 3" for next-generation enterprise PCs and the latest vPro platform, addressing a wide range of demands from enterprise notebooks to high-performance workstations.
2026.03.26by 배종인 기자
ROHM Launches ESD Protection Diode Compatible with High-Speed Interfaces of 10Gbps or Higher
Semiconductor specialist ROHM has developed the 'RESDxVx series' of ESD (electrostatic discharge) protection diodes capable of responding to next-generation high-speed data communication environments exceeding 10Gbps, simultaneously enhancing both signal quality maintenance required in high-speed interfaces and IC protection performance.
2026.03.26by 배종인 기자
