MacNN Forums (
-   Tech News (
-   -   Nvidia outlines Tegra, GPU roadmaps at GPU Tech Conference (

NewsPoster Mar 19, 2013 04:52 PM
Nvidia outlines Tegra, GPU roadmaps at GPU Tech Conference
Nvidia CEO Jen-Hsun Huang <a href="" rel='nofollow'>speaking</a> at the <a href="" rel='nofollow'>2013 GPU Technology Conference</a> in San Jose, CA unveiled the company's mobile Tegra chipset roadmap, and discussed future GPU technology in the pipeline for the company. Also shown was the rack mountable GRID Visual Computing Appliance (VCA), offering virtualized GPU computing and remote graphics capabilities to allow a server to handle processor-intensive rendering jobs, with the output streamed to lower powered clients.<br />
<br />
While light on technical details, the CEO revealed that following the <a href="" rel='nofollow'>Tegra 4 release</a> this year, two new system-on-a-chip products were imminent -- the Logan and Parker chipsets. Logan will feature a Kepler-based GPU and support CUDA 5 and OpenGL 4.3. The chip is currently the size of a dime, and allegedly offers up to three times the computing performance of existing products. The first Logan demoes will happen later this year, with full production in the first quarter of 2014.<br />
<br />
The Parker-based SoC is a 63-bit ARM with Denver-class CPU cores, and Maxwell GPU technology. The CEO claims that in five years, which is the expected rollout of Parker, Nvidia will have improved the performance of Tegra by 100 times over the Tegra 2. Parker is scheduled to arrive in 2015.<br />
<br />
The CEO briefly reiterated the upcoming <a href="" rel='nofollow'>Maxwell</a> technology expected in 2014. Maxwell gives double the performance per watt that the existing Kepler does. The incoming Volta architecture doubles the performance of Kepler, and implements the new Stacked DRAM technology, which keeps the signal closer to th processor, and resides on the same die as the CPU. Volta is expected to offer 1TB per second of memory bandwidth to the stacked DRAM, more than three times the bandwidth of the currently shipping GTX Titan.<br />
<br />
The <a href=" vice/" rel='nofollow'>Nvidia GRID VCA</a> is equipped with one or two eight-core Xeon processors, 192 to 384GB of RAM, and four or eight GRID boards each with two Kepler GPUs. During the speech, a MacBook Pro was connected to a demo GRID VCU, and executed several software packages on the VCA that aren't native to OS X -- namely 3D StudioMax and Solidworks. Pricing on the GRID VCA system ranges from $24,900 for a single CPU/eight GPU configuration, or $39.900 for the fully equipped dual CPU, 16 GPU loadout, with an additional licensing fee applicable annually.
All times are GMT -4. The time now is 06:21 AM.

Copyright © 2005-2007 MacNN. All rights reserved.
Powered by vBulletin® Version 3.8.8
Copyright ©2000 - 2017, vBulletin Solutions, Inc.

Content Relevant URLs by vBSEO 3.3.2