Nvidia to enter server chip Market Challenge Intel

Lead: U.S. It website pc today, in addition to continuing to consolidate the advantages of the graphics processor market, Nvidia also plans to use the next generation of Tegra chips to enter the server market, challenging Intel's dominance in this area. The following is the full text of the article: NVIDIA is currently developing the first PC and server CPU code-named "Denver Plan" (Project Denver). This CPU uses ARM architecture and is also geared toward mobile devices.　　Steve Scott, chief technology officer at the nvidia Tesla product line, said that the "Denver" kernel would be embedded in future Tegra chips, with special optimizations for server chips. "We have done a lot of work on this," said Scott. It may join the Tegra product line sometime in the future. "Nvidia's current business in the server area is primarily related to its Tesla graphics processor. Some of the world's fastest supercomputers are using Tesla processors to perform complex scientific and mathematical operations.　　Oak Ridge National Laboratory is building a supercomputer called Titan (Titan). This computer will use the NVIDIA Tesla processor and the AMD 16 core Opteron CPU to achieve 20000 trillion floating-point peak computing capabilities. Scott did not disclose how Nvidia will be targeting servers to optimize the details of future generations of Tegra chips. But Nvidia said the "Denver" kernel will set the NVIDIA graphics processor parallel processing capabilities and ARM CPU Low-power features.　　The ARM CPU also improves server performance. Most servers currently use Intel (micro-blogging) Xeon and AMD Opteron processors, but as companies focus on reducing their electricity costs, low-power arm processors are also becoming popular.　　Although ARM chips may be less performance and reliability than traditional server chips in performing critical tasks, they can handle large amounts of online transactions at lower power consumption, analysts say. According to Scott, using CPU and graphics processor parallel processing capabilities for complex computing, while increasing the speed of the server, but also reduce energy consumption and computing costs. Nvidia has developed the graphics kernel based on the current Tegra processor. "The ARM command group is more energy efficient than x86, which is why people are looking to develop an arm-based server because ARM chips are more efficient," says Scott. Dean McCarron, Dien Mccarren chief analyst at market research firm Mercury, said Nvidia's push Tegra chip into the server market was significant, and server chips had a higher profit margin than mobile device chips. He said: "Nvidia has interesting parallel processing technology, and arm is very good, two companies to enter the server chip market significanceUnusual. "The NVIDIA server chip target market may be a system that relies on GPU for graphic processing or cloud computing," says McCullen. The ARM processor does not have the GPU proficiency in performing complex operations, so Nvidia may end up having to make trade-offs with its arm CPU design so that product performance can be further improved. "When you enter the server market, there is evolutionary pressure to push you forward," McCullen said. "Companies such as Seamicro and Dell are currently developing servers based on Intel Atom Low-power processors, but Nvidia's involvement in this area may make markets more interested in arm servers," he said.　　Nvidia will have many rivals: Marvell last year announced the launch of a 1.6GHz four-core ARM server processor, while Calxeda is also developing a quad-core ARM server processor. One major obstacle to arm's foray into the server market is software compatibility, since most data center software code is written for x86 servers. But Mr Scott says x86 binary compatibility is not a big issue for future generations of Nvidia server chips to meet cloud computing services. "The importance of binary compatibility is greatly reduced in a cloud computing environment," he said. They can provide online services and can rely on arm services. That means higher energy efficiency. "(Xuan Chen)