San Francisco – The emergence of dual-core processors will bring the parallel computing power necessary to virtualize enterprise IT environments and ease manageability issues for CIOs, executives told the Intel Developer Forum.
Speaking to a crowd of approximately 5,500 software firms and
industry partners, Intel president Paul Otellini promised dual-core additions to its product line that would span its high-end Itanium chips to the Xeon server products and Pentium desktop products.
Dual-core chips are designed to significantly increase performance without increasing power consumption, a key concern among original equipment manufacturers and developers. By 2006, Otellini said he expected more than 40 per cent of Intel desktop chips to contain two cores and 85 per cent of its server chips.
These products will also include hyperthreading, a technology that takes some of the unused circuitry (called “registers”) on the chip and, in a sense, fools the system into thinking it is running on a dual processor. Hyperthreaded or multi-threaded CPUs could either run two or more applications at once (by using both the floating-point integer and the graphics unit, for example), or simply make a single application run faster.
Otellini said dual-core processors and hyperthreading would usher in an era of “parallel personal computing” whereby computers will first recognize a piece of data, mine for more context in an archive and then synthesize it into meaningful information. For example, a financial services company could create an application that would lean to spot a hedge fund, search for a hedge fund investment opportunity and then factor in what would happen when interest rates change. Desktop applications of this nature would require 50 Gigaflops of processing speed, as opposed to the five to seven available on high-end PCs today.
“Instead of thinking about how many chips are on the computer, you’re going to start thinking about how many computers are on a chip,” Otellini said.
Intel is planning to embed features in future processors and chip sets that will facilitate parallelism, Otellini said.
These include Vanderpool, a technology that will allow developers to build systems that virtualize end user systems into four partitions. One of these might be devoted to corporate applications, another to personal computing applications like games, a Linux workstation and a manageability area that would look for the latest bug patches and virus protection updates.
Vanderpool is expected to be available when Microsoft launches the next version of Windows, code-named Longhorn, in 2006, but Intel co-CIO Stacy Smith demonstrated its capabilities with a desktop chassis painted with four colour bars, like a rainbow, to represent each partition. Smith said Vanderpool would eliminate thousands of calls to IT help desks and would remove the second system from many developers’ desks.
Perry Longinotti, a project manager with payroll and HRIS developer Aspen Avanti Inc. in Calgary, said dual-core processing wouldn’t necessarily have a huge impact for his customers because its software is not computationally-based. “We use a SQL database, so our product will scream just as fast on a 700 MHz system as a dual-core 300 MHz system,” said Longinotti, who is not attending IDF this year but spoke by telephone. “Our customers will get more from the development of faster buses, faster drives and the ability to address more memory.”
Vandepool, however, offers major potential, given that Aspen Avanti’s product would likely reside in the corporate applications partition. “That’s pretty interesting, and it means a lot of companies could farm out some of the manageability stuff that take up a lot of their time now,” Loginotti said.
Intel is not the only company working on a dual-core product. IBM unveiled a dual-core version of its Power chip earlier this year, while Advanced Micro Devices (AMD) has already started demonstrating a server-based version of a dual-core Opteron. Intel’s demonstration focused instead on a dual-core Itanium, which sells in lower volumes and has been in the works for some time. In a question-and-answer session following his keynote speech, however, Otellini told ITBusiness.ca he isn’t afraid of falling behind in the development process.
“This isn’t a race. This is a sea change in computing,” he said. “I’m happy to see our prime competitor is embracing this move because it shows we’re moving in the right direction.”
Intel highlighted the benefits of dual-core processing by highlighting NASA, which is building a supercomputer based on SGI Altix hardware and running more than 10,000 dual-core Itanium “Montecito” CPUs on Linux. Abhi Talwalkar, vice-president of Intel’s enterprise platforms group, said NASA will see a 1.5 to 2X performance boost from the dual-core chips. Walter Brooks, chief of the NASA Advanced Supercomputing Division, said approximately three thousand of those processors are already deployed today, and the rest should be in production within about four months.
“The main thing is not stringing all these processors together but getting our applications guys on board,” Brooks said. “In some areas, we’ll be using it for modeling the Sun and climate, and perhaps more importantly we’ll be using it to get the shuttles flying.”
Intel is also using IDF to launch the Cross-Platform Manageability Program, which will seek to create standard ways of supporting common and consistent manageability capabilities, interfaces, and protocols across all Intel platforms, including cell phones and servers. Otellini said he expects the company to release a public industry specification with input from developers at the spring IDF in 2005.