by Anton Shilov
09/21/2012 | 04:11 AM
Both leading suppliers of server microprocessors - Advanced Micro Devices and Intel Corp. - customize their chips for select customers in a bid to win large-scale contracts to sell their Opteron and Xeon central processing units. In some cases the companies just tweak clock-speeds or power consumption, in other cases the changes are much deeper.
Just three years ago three server giants - HP, Dell and IBM - basically commanded the market with industry-standard offerings that suited most businesses. Nowadays there are companies like Apple, Facebook, Google or Microsoft, who run massive cloud datacenters and require customized equipment. Some companies - Facebook and Google - even assemble their custom servers in-house. In a bid to solve specific problems, microprocessors inside servers for these large customers are specifically tailored for their needs. In fact, both Facebook and Google acknowledged in an interview with Wired that they request "specific silicon" from chipmakers.
"We want to give [our biggest customers] a way of differentiating their machines," said Diane Bryant, corporate vice president and general manager of the datacenter and connected systems group (DCSG) at Intel.
Large server buyers need to reduce power consumption, lower total cost of ownership as well as improve performance and add specific functionality of their datacenters in general. As a result, they get chips that regular customers do not have access to. In many cases, this is limited to tailoring clock-speeds and/or power ratings, but in some cases AMD and Intel may even modify their chips to support certain feature or even implement undocumented instructions that only work with certain software.
"There are situations where the company has turned off certain parts of a processor at the request of customers, or even added [new] instructions," said John Williams, vice president of server marketing and business development at AMD.
Neither chipmakers nor their customers talk about specifics, e.g., what modifications could be made to solve a specific issue. Nonetheless, over-time certain functionality most probably migrates into industry-standard offerings to solve mainstream problems.