IBM doubles memory capacity

IBM has developed a technology which it claims will double server memory capacity without having to add extra physical memory.

Big Blue said the technology, called Memory eXpansion, will be used in its Netfinity line of servers. It can also be adapted to improve the memory power of PCs and handheld devices, said the company.

Memory eXpansion is based on a memory controller chip that automatically stores frequently accessed data and instructions close to the machine's microprocessors so that they can be accessed immediately.

The hardware implementation integrates a new cache specifically designed to handle data and instructions on a memory controller chip. By combining this new hardware-based compression algorithm with millions of tiny transistors, IBM claims it can double the server memory capacity for most applications.

Mark Dean, IBM vice president of systems research, said: "Adding memory is often the most effective way to improve system performance, but it's a costly proposition. Memory eXpansion improves system performance without adding costly physical memory."

The technology was initially designed for Intel-based servers, and IBM will use it in its own Netfinity server line. The company is also looking at ways to take advantage of the technology in its line of data transaction and web application servers and storage subsystems.

Mark Melenovsky, an analyst at researcher IDC, said Memory eXpansion would be a big boon to the low-end Intel server market. "Memory is one of the constraints that the Intel-based platform is facing in order to continue to expand," he said.

ServerWorks, which develops chips for Intel-based servers, has become the first licensee of the technology and will make it available to other server manufacturers.