Advanced Micro Devices is beginning to line up partners that will offer server systems built around the chip maker’s 45-nanometer “Shanghai” processor, which has been shipping since September and is scheduled for release in November.
The partners that have lined up behind AMD’s Shanghai processor include many of the smaller server and HPC (high-performance computing) vendors that have been longtime supporters of the chip maker’s technology. Companies with plans to release systems built around Shanghai include Penguin Computing, Colfax, Appro, Verari Systems and Rackable Systems.
By showcasing partners before the Shanghai processor officially launches, AMD is looking to erase some of the unease created when it failed to deliver its 65-nm, quad-core Opteron processor “Barcelona” on time in 2007. Since it began shipping Barcelona in earnest earlier in 2008, AMD has gained back some of its market share in the server and workstation market, according to a recent report by IDC.
In the case of Shanghai, AMD wanted to get the new parts into the hands of customers sooner in order to allow for more testing before the actual launch. Since Barcelona shipped late, many of the silicon problems went unnoticed at first. The new strategy also allowed AMD to push up the release date of Shanghai from the first or second quarter of 2008 to November.
These early of adopters of Shanghai discussed the benefits of the new chip at a panel discussion on Nov. 6. While the five vendors mentioned above and AMD talked about some of the benefits of the new processor, none of the participants offered details about the clock speeds and prices of the chips.
When Shanghai does come to market later in November, AMD is likely to offer processors for two- and four-socket servers first. Shanghai will use four processing cores and contain 6MB of Level 3 cache, compared with the 2MB of L3 cache in the company’s current crop of quad-core Opteron processors, which should add to the overall performance boost. AMD is also looking to boost the clock speed about 20 percent, which will get the clock speed closer to 3.0GHz.
During Thursday’s discussion, several of the vendors noted that since Shanghai uses the same socket-Socket F (1207)-as some previous Opteron processors, it made testing, evaluating and creating updated systems easier. AMD is emphasizing that at a time when a financial crisis is still looming, it’s better to work with technologies that are familiar.
AMD Changes Little to Keep Cash-Strapped Customers
“We have kept a very consistent architecture when it comes to our fifth generation of products,” said Burke Banda, an AMD marketing manager for the Server and Workstation division. “This is our second-generation, quad-core product and we have kept the same socket and worked with the same technology partners. The feedback we are getting from our customers is that it’s a time to be risk-averse and it’s a time to look at technologies they are familiar with, and while they want to move their business forward, they want to have familiar products.”
This is also a strategy AMD is using to try to point out that Intel’s upcoming line of processors, based on the “Nehalem” microarchitecture, will require new systems and IT departments will have to invest in learning about the new Intel architecture. This week, Intel has been making news with its first Nehalem chip, called the Intel Core i7, which is scheduled for release on Nov. 17.
When it comes to Shanghai, several of the vendors noted that the processor offers better energy efficiency and ability to handle virtual machine workloads than previous generations of AMD Opteron processors. Ed Holden, a manager of server products for Verari Systems, said his company was interested in building Shanghai-based systems for several specific areas, including Web 2.0, financial services, and video and media content creation.
“We have a lot of customers in the rendering space and media and the entertainment space, and those guys are extremely tied to memory bandwidth and I/O,” Holden said. “So any improvements that can be made to I/O and memory bandwidth help a render farm exponentially, so that’s been an area we have focused on.”
George Reitz, a vice president at Rackable Systems, said the small die size of Shanghai helps to make the motherboards smaller and allows the company to pack more boards onto servers, which allows for denser servers and ultimately a denser data center. “We are looking at getting up to 240 processors and getting many, many cores within the same rack and having some very interesting power profiles,” Reitz said.
The panel members and Banda agreed that the financial crisis on Wall Street and the credit crunch in the United States could have a significant impact on their businesses, especially for those that sell systems to financial services. Holden, the Verari representative, did note that many of his company’s customers were still interested in investing in R&D now in order to be ready for when the crisis ends and the business atmosphere in the United States rebounds.
The first systems based on Shanghai are not expected until later in December. So far, AMD has not announced support from larger vendors such as Dell, Hewlett-Packard, IBM or Sun Microsystems.