» Current | 2018 | 2017 | 2016 | Subscribe

Linley Newsletter

Turing Accelerates Ray Tracing

October 9, 2018

Author: Linley Gwennap

An upgrade to the two-year-old Pascal, Nvidia’s new Turing architecture delivers a big shader-performance boost. It also enables “neural graphics functions,” or routines that improve image quality using neural networks (AI). But the pièce de résistance is Turing’s new ray-tracing (RT) engine, which allows game developers to create complex lighting and reflection effects unachievable with previous graphics cards. Be ready to pay up, though: the new PC graphics cards start at $499 and go beyond $1,000.

A shrink from 16nm to 12nm transistors allowed Nvidia to increase the number of ALUs (which it dubs Cuda cores) by 20% while raising the clock speed by 7%. These modest changes improve Turing’s peak floating-point performance by 29% over Pascal’s. In addition, the new architecture enables each shader core (SM in Nvidia’s parlance) to perform integer and floating-point operations at the same time. This new parallelism further increases performance by about 35% on many common workloads.

Implementing ray tracing entirely in the shader cores, Pascal can cast 1.1 billion rays per second (GRay/s), whereas Turing exceeds 10GRay/s on many programs. The new architecture also includes tensor units that can accelerate integer operations for AI inference. In this fashion, it can perform 130 trillion integer (INT8) multiply-accumulate (MAC) operations per second.

Nvidia is already shipping three PC graphics cards based on the Turing design. The GeForce RTX 2080 Ti features the TU102 die to deliver the maximum Turing performance at 250W. The RTX 2080 and RTX 2070 use smaller die configurations that reduce performance but enable lower prices. The company is shipping three additional Turing-based cards for professional users. The new graphics cards leave AMD’s Radeon GPUs in the dust, at least until that company’s 7nm models arrive, and are best for 4K gaming.

Subscribers can view the full article in the Microprocessor Report.

Subscribe to the Microprocessor Report and always get the full story!

Purchase the full article

Events

Linley Fall Processor Conference 2018
Covers processors and IP cores used in embedded, communications, automotive, IoT, and server designs.
October 31 - November 1, 2018
Hyatt Regency, Santa Clara, CA
Register Now!
More Events »

Newsletter

Linley Newsletter
Analysis of new developments in microprocessors and other semiconductor products
Subscribe to our Newsletter »