AMD, Intel, ARM: for years, their respective CPU architectures required separate sockets, separate motherboards, and in effect, separate servers. But no longer: Facebook and the Open Compute Summit have announced a common daughtercard specification that can link virtually any processor to the motherboard.
AMD, Applied Micro, Intel, and Calxeda have already embraced the new board, dubbed “Group Hug.” Hardware designs based on the technology will reportedly appear at the show. The Group Hug card will be connected via a simple x8 PCI Express connector to the main motherboard.
It’s hard to overstate the potential of the technology. Although many other components within a server—including power supplies, hard drives, memory and I/O cards—have long been replaceable, processors have not. Frank Frankovsky, director of hardware design and supply chain operations at Facebook, told an audience at the Summit that, while a standard has been provided, it may be some time before the real-world appearance of servers built on the technology.
Nonetheless, with Group Hug, the architecture of the server seems ready to further disaggregate. “It’s all come together,” Frankovsky said.
Bob Ogrey, a fellow at AMD and one of the company’s cloud evangelists, previously suggested that the topic of swappable CPU designs was under discussion by some big firms, with Facebook and other companies requesting that CPU providers take that step.
“I do believe we’re going to have highly customized configurations,” Frankovsky added. “So the days of having to predict how the software is going to affect the hardware—we’re not going to have to predict as accurately as we did today, because we’re going to be able to have very malleable hardware provisions. We’re going to be able to say, ‘Oh, if I under-provision CPU for a given workload if the workload changes, I’m going to be able to add more CPU capacity.’ If I need a RAM sled, let me put a RAM sled in place. Or hey, I need more flash, so I need more hosts sharing that.”
Customization of servers will increase exponentially as a result; bottlenecks related to the placement of components on a traditional motherboard will disappear. “I’m going to be able to take the older [components] out,” he said, “and leave the ones still of value, still of use, in the rack.”
The announcement was a highlight of the Open Compute Summit, founded by Facebook just a few years ago with the intention of unlocking the value in the server space by sharing design information. Now, the Open Compute Group touts 50 corporate members.
It remains to be seen how the Group Hug board will change the component industry. Will AMD and Intel make all of the Opteron and Xeon chips Group Hug compatible, or just a few select models? Will the ability to swap one CPU for another commoditize the hardware to unsustainable levels? One thing is clear: for years, PC and server makers were virtually vassals to component makers such as Intel; servers added little hardware value outside of the core components. With Group Hug, server makers may have seized some of that power back.
Image: Olga Miltsova/Shutterstock.com