Linley Newsletter: September 1, 2017

 weSRCH's Best of the Internet Award
  7th-Sep-2017
 679

Linley Newsletter
(Formerly Processor Watch, Linley Wire, and Linley on Mobile)
Please feel free to forward this to your colleagues

Issue #563
September 1, 2017

Independent Analysis of Microprocessors and the Semiconductor Industry

Editor: Tom R. Halfhill
Contributors: Loyd Case, David Kanter

In This Issue:

- Pushing CMOS Beyond 7nm
- AMD Vega Shoots for GPU Stars

New Report: "A Guide to Processors for Deep Learning"

By Linley Gwennap, Mike Demler, and Loyd Case

Order by September 30 and save $300!

Deep learning, also known as artificial intelligence (AI), has seen rapid changes and improvements over the past few years and is now being applied to numerous applications. Typically implemented using neural networks, deep learning powers image recognition, voice processing, language translation, and many other web services in large data centers. It is an essential technology in self-driving cars, providing both object recognition and decision making. It is even starting to move into client devices such as smartphones and embedded (IoT) systems. Our new report is written for:

- Engineers designing chips or systems for deep learning or autonomous vehicles

- Marketing and engineering staff at companies that sell related chips who need more information on processors for deep learning or autonomous vehicles

- Technology professionals who wish an introduction to deep learning, vision processing, or autonomous-driving systems

- Financial analysts who desire a hype-free analysis of deep-learning processors and of which chip suppliers are most likely to succeed

- Press and public-relations professionals who need to get up to speed on this emerging technology

This market is developing rapidly -- don't be left behind! Order by September 30 to save $300 off the single-license price of $4,495. For more information, access:
http://www.linleygroup.com/report_detail.php?num=65

To order the report, access: https://www.linleygroup.com/report_order.php?num=65

Pushing CMOS Beyond 7nm
By David Kanter

Leading semiconductor companies agree that FinFETs are the best high-performance transistors for the next couple of nodes and that copper will deliver the right mix of low resistance and reliability for interconnect and via layers. Most manufacturers also agree that scaling these technologies beyond the 7nm foundry node (equivalent to Intel's 10nm node, which is expected later this year) will be difficult, if not impossible.

Semiconductor manufacturers are researching technologies to replace FinFETs and copper interconnects. The leading candidates for 5nm CMOS (equivalent to Intel's 7nm node) and beyond are gate-all-around (GAA) transistors and copper interconnects with a more complex and robust barrier. This combination should enable further scaling with good performance. Both of these technologies require big changes to chip manufacturing flows, but they're evolutionary extensions of the current approaches.

Whereas a FinFET wraps a gate around the transistor channel on three sides, a GAAFET surrounds the channel on all four, increasing the control of electron flow from source to drain. This tighter control should enable lower-voltage operation, albeit at higher manufacturing cost. Samsung has already committed to a variant of GAAFETs for its 5nm process, which is due to enter production in 2020. IBM's research alliance also identified potential replacements for copper interconnects, using tantalum-nitride (TaN) barriers. It concluded that copper wiring with a new barrier would offer the best performance, beating out new metals that don't need barriers.

These and other innovations will sustain scaling for the next five years, but not for all applications. The new transistors and wiring schemes are more expensive than their predecessors, further increasing the cost gap between cutting-edge and mainstream products. High-performance processors will be able to afford the greater performance, power efficiency, and density, but manufacturers must reduce costs to draw other customers.

Microprocessor Report subscribers can access the full article:
http://www.linleygroup.com/mpr/article.php?id=11852

AMD Vega Shoots for GPU Stars
By Loyd Case

AMD's CPU group has been on a new-product roll; now its GPU group -- called the Radeon Technology Group (RTG) -- is joining in with a new Radeon RX product. Code-named Vega, this offering is the most substantial revamp of an AMD GPU architecture since the company introduced Graphics Core Next in 2011. The Vega 64 and Vega 56 chips break a long drought at the high end that began after the company shipped its Fiji GPUs in 2014. Graphics cards based on Vega 64 are shipping now; Vega 56 cards are slated for a September release.

The first iteration, dubbed Vega 64, includes 64 compute units (CUs) operating at up to 1.67GHz. Each CU in turn consists of 64 shaders for a total of 4,096 ALUs in the chip. The GPU comprises 12.5 billion transistors on a 486mm^2 die. AMD builds Vega 64 in GlobalFoundries' 14nm FinFET process.

The high-end Fiji GPU has 64 earlier-generation CUs and is manufactured in a 28nm process with 8.9 billion transistors, requiring a 596mm^2 die. The company will also market a version called Vega 56 using the same die with 56 CUs enabled, reducing performance by about 10-15% relative to Vega 64.

Accompanying Vega's additional shader horsepower and high-speed memory are new graphics features. AMD says the GPU offers graphics capabilities beyond current APIs but declined to provide details; we assume it's been working closely with Microsoft and the Khronos Group. Any extended capabilities will likely appear in upcoming versions of DirectX, OpenGL, and Vulkan, and perhaps earlier as extensions to these APIs.

Microprocessor Report subscribers can access the full article:
http://www.linleygroup.com/mpr/article.php?id=11851

About Linley Newsletter

Linley Newsletter is a free electronic newsletter that reports and analyzes advances in microprocessors, networking chips, and mobile-communications chips. It is published by The Linley Group and consolidates our previous electronic newsletters: Processor Watch, Linley Wire, and Linley on Mobile. To subscribe, please visit:
http://www.linleygroup.com/newsletters/newsletter_subscribe.php

Copyright 2017, The Linley Group

Domain: Electronics
Category: Semiconductors
weSRCH App on Apple

Recent Newsletters

Linley Newsletter: August 8, 2019

Linley Newsletter Please feel free to forward this to your colleagues Issue #664 August 8, 2019 Independent Analysis of Microprocessors and the Semiconductor Industry E

08 August, 2019

Linley Newsletter: August 1, 2019

Linley Newsletter Please feel free to forward this to your colleagues Issue #663 August 1, 2019 Independent Analysis of Microprocessors and the Semiconductor Industry E

01 August, 2019

Linley Newsletter: July 25, 2019

Linley Newsletter Please feel free to forward this to your colleagues Issue #662 July 25, 2019 Independent Analysis of Microprocessors and the Semiconductor Industry

25 July, 2019