A Guide to Processors for Deep Learning
Covers processors for accelerating deep learning, neural networks, and vision processing for AI training and inference in data centers, autonomous vehicles, and client devices.
Most Recent Guides & Forecasts
Communications Semiconductor Market Forecast 2018–2023
Provides five-year revenue forecasts for more than 20 categories of communications semiconductors, including Ethernet products, broadband interface chips, embedded processors, server processors and FPGAs.
Provides five-year revenue forecasts for more than 20 categories of communications semiconductors, including Ethernet products, broadband interface chips, embedded processors, server processors and FPGAs.
Communications Semiconductor Market Share 2018
Provides market share data for more than 20 categories of communications semiconductors, including Ethernet products, broadband interface chips, embedded processors, server processors and FPGAs.
Provides market share data for more than 20 categories of communications semiconductors, including Ethernet products, broadband interface chips, embedded processors, server processors and FPGAs.
A Guide to Processors for Deep Learning
Covers processors and IP cores for accelerating deep learning, neural networks, and vision processing for AI training and inference in data centers, autonomous vehicles, and client devices.
Covers processors and IP cores for accelerating deep learning, neural networks, and vision processing for AI training and inference in data centers, autonomous vehicles, and client devices.
A Guide to Multicore Processors
Covers 32- and 64-bit embedded processors with four or more CPU cores that are used for wires and wireless communications, storage, security, and other applications.
Covers 32- and 64-bit embedded processors with four or more CPU cores that are used for wires and wireless communications, storage, security, and other applications.
A Guide to Ethernet Switch and PHY Chips
Covers data-center switch chips for 10G, 25G, 40G, 50G, and 100G Ethernet. Also includes physical-layer (PHY) chips for 10GBase-T and 100G Ethernet.
Covers data-center switch chips for 10G, 25G, 40G, 50G, and 100G Ethernet. Also includes physical-layer (PHY) chips for 10GBase-T and 100G Ethernet.
White Papers
Accelerating AI Performance in an Open Software Ecosystem
Nvidia's software stack, based on its proprietary CUDA, creates a high barrier to entry for challengers. SYCL, the only open alternative to CUDA that has multivendor support, enables customers to move to higher performance hardware while retaining software flexibility. A commercial version of SYCL, branded ComputeCpp, is offered by Codeplay.
Nvidia's software stack, based on its proprietary CUDA, creates a high barrier to entry for challengers. SYCL, the only open alternative to CUDA that has multivendor support, enables customers to move to higher performance hardware while retaining software flexibility. A commercial version of SYCL, branded ComputeCpp, is offered by Codeplay.
Arm Ecosystem Reduces SoC Design Cost and Time to Market
This paper discusses three critical facets of the Arm ecosystem: design verification, physical design, and software development. Each one directly affects the time required to develop a complete SoC and software stack.
This paper discusses three critical facets of the Arm ecosystem: design verification, physical design, and software development. Each one directly affects the time required to develop a complete SoC and software stack.
AI Requires Many Approaches
Artificial intelligence (AI) is being applied to a wide range of problems, so no single processor can support them all. This paper describes and compares the various approaches needed for data center, autonomous driving, and consumer/IoT applications.
Artificial intelligence (AI) is being applied to a wide range of problems, so no single processor can support them all. This paper describes and compares the various approaches needed for data center, autonomous driving, and consumer/IoT applications.
Universal RDMA: A Unique Approach for Heterogeneous Data-Center Acceleration
Now available in 10G, 25G, 50G, and 100G Ethernet server adapters, RDMA reduces network processing overhead and minimizes latency. Cavium's FastLinQ stands out by handling both RoCE and iWARP, which are alternative protocols for RDMA over Ethernet.
Now available in 10G, 25G, 50G, and 100G Ethernet server adapters, RDMA reduces network processing overhead and minimizes latency. Cavium's FastLinQ stands out by handling both RoCE and iWARP, which are alternative protocols for RDMA over Ethernet.
