<div class="custm_img_blk"><img src="http://www.techgig.com/files/photo_1337167504_temp.jpg" border="0" width="300" height="200" align="left" /></div><p style="margin: 0px 0px 15px; outline: 0px; vertical-align: baseline; font-family: arial; font-size: 15px; font-style: normal; font-variant: normal; font-weight: normal; line-height: 21px; border-width: 0px; padding: 0px">"Computer graphics is at the cusp of being revolutionized again," thanks to the ability to do fluid and light simulations on graphics chips, Nvidia CEO Jen-Hsun Huang said at the opening of the company's GPU Technology Conference (GTC) today. Huang showed new applications of GPUs in gaming, announced two new Tesla GPU boards aimed at high-performance computing, and demonstrated GPU cores in the cloud.</p><p style="margin: 0px 0px 15px; outline: 0px; vertical-align: baseline; font-family: arial; font-size: 15px; font-style: normal; font-variant: normal; font-weight: normal; line-height: 21px; border-width: 0px; padding: 0px">Huang started by talking about computer graphics, and the company's recently announced 28nm Kepler architecture. He called the latest<span>GeForce GTX 690</span> "the world's most advanced graphics board, based on two Kepler GPUs" (technically the GK104).</p><p style="margin: 0px 0px 15px; outline: 0px; vertical-align: baseline; font-family: arial; font-size: 15px; font-style: normal; font-variant: normal; font-weight: normal; line-height: 21px; border-width: 0px; padding: 0px">He showed a number of traditional gaming demos, but I was more interested in the demo of real-time ray tracing on the board.</p><p style="margin: 0px 0px 15px; outline: 0px; vertical-align: baseline; font-family: arial; font-size: 15px; font-style: normal; font-variant: normal; font-weight: normal; line-height: 21px; border-width: 0px; padding: 0px">This technology will expand over the next few years, and as a result, computer graphics will look nothing like the simple shading we see in console and PC games today.</p><p style="margin: 0px 0px 15px; outline: 0px; vertical-align: baseline; font-family: arial; font-size: 15px; font-style: normal; font-variant: normal; font-weight: normal; line-height: 21px; border-width: 0px; padding: 0px">"Kepler is a big deal for computer graphics, but an even bigger deal for high-performance computing," Huang said. It has major advances, including a streaming multiprocessor core called SM; and the new SMX architecture, which has 192 cores and is three times more energy efficient than the architecture used in the company's previous Fermi chip. Another feature is Hyper-Q, which allows 32 work queues, which lets multiple CPU cores send much more work to the graphics processor, and thus keep the graphics cores busy all the time. Finally, it has "dynamic parallelism," where the GPU cores themselves can create work within their threads, based on their own results.</p><p style="margin: 0px 0px 15px; outline: 0px; vertical-align: baseline; font-family: arial; font-size: 15px; font-style: normal; font-variant: normal; font-weight: normal; line-height: 21px; border-width: 0px; padding: 0px">These features will make Kepler particularly suited for high-performance computing. As an example, he introduced a demo of an n-body simulation. With Fermi, the company had shown a demo of 20,000 bodies interacting with each other, but with Kepler, it showed a demo of 280,000 bodies interacting, revealing a very fast simulation of the interactions between the Milky Way and the Andromeda Galaxy. On Fermi, the demo could show one million particles per second, while on Kepler the demo could handle 10 million particles per second.</p><p style="margin: 0px 0px 15px; outline: 0px; vertical-align: baseline; font-family: arial; font-size: 15px; font-style: normal; font-variant: normal; font-weight: normal; line-height: 21px; border-width: 0px; padding: 0px">For this market, Huang announced two new boards. First up is the single-precision Tesla K10, out now and aimed at imaging, signal and seismic applications, with three times the performance and 1.8 times the memory Bandwidth of the previous boards. This is slated to be followed in the fourth quarter by the double-precision Tesla K20—aimed at fluid dynamics, finite element analysis, and financial and physics applications—which adds the Hyper-Q and dynamic parallelism features.</p><p style="margin: 0px 0px 15px; outline: 0px; vertical-align: baseline; font-family: arial; font-size: 15px; font-style: normal; font-variant: normal; font-weight: normal; line-height: 21px; border-width: 0px; padding: 0px">"We're going to put the GPU into the cloud," Huang said. Kepler is the first GPU designed for cloud computing, he said. That's because it offers a virtualized GPU, low-latency remote display features in that it no longer needs to be physically connected to a display, and low power so it can run in distributed data centers. The virtualized GPU is the core of this, as until now GPUs had to be dedicated to a particular application.</p><p style="margin: 0px 0px 15px; outline: 0px; vertical-align: baseline; font-family: arial; font-size: 15px; font-style: normal; font-variant: normal; font-weight: normal; line-height: 21px; border-width: 0px; padding: 0px">One application for this is the "BYOD" movement where companies allow users to bring their own devices to work, but allow them access to virtual machines in a private cloud with streaming applications.</p><p style="margin: 0px 0px 15px; outline: 0px; vertical-align: baseline; font-family: arial; font-size: 15px; font-style: normal; font-variant: normal; font-weight: normal; line-height: 21px; border-width: 0px; padding: 0px">Citrix and other companies today allow for virtual desktop infrastructure (VDI) with products such as Xen Desktop. But until now, such devices have used software-based GPUs because GPUs cannot be virtualized. Now, the company has introduced the virtual GPU hardware in Kepler, along with driver and software that works with Citrix's solutions, allowing power users and designers to have a remote GPU dedicated to their environment. Huang demoed this with Citrix Xen Desktop running Windows on an iPad with GPU acceleration.</p><p style="margin: 0px 0px 15px; outline: 0px; vertical-align: baseline; font-family: arial; font-size: 15px; font-style: normal; font-variant: normal; font-weight: normal; line-height: 21px; border-width: 0px; padding: 0px">In another demo, a computer graphics artist from Industrial Light and Magic used a MacBook Air to access a Windows machine on a server running Autodesk's Maya modeling software.</p><p style="margin: 0px 0px 15px; outline: 0px; vertical-align: baseline; font-family: arial; font-size: 15px; font-style: normal; font-variant: normal; font-weight: normal; line-height: 21px; border-width: 0px; padding: 0px">He then showed a wall of 100 different machines all accessing the GPUs from a single server.</p><p style="margin: 0px 0px 15px; outline: 0px; vertical-align: baseline; font-family: arial; font-size: 15px; font-style: normal; font-variant: normal; font-weight: normal; line-height: 21px; border-width: 0px; padding: 0px">A Cisco representative talked about a new version of Cisco's Unified Computing System (UCS) that supports Nvidia's virtual GPUs technology. Huang added that the GeForce Grid, a method for streaming video games over the cloud, is more convenient. </p>
<a href='http://techgig.com/tech-news/editors-pick/Nvidia-CEO-We-re-Bringing-GPUs-into-the-Cloud-12437'>View More</a>
![http://www.techgig.com/files/photo_1337167504_temp.jpg"](http://www.techgig.com/files/photo_1337167504_temp.jpg"){kind=link}
No comments:
Post a Comment
If you have any question please let me know