Rambus launches RDRAM fault tolerance

Based on Big Blue's ideas, but it's IDM not IBM

Rambus today announced its version of fault-tolerant memory for high-end servers, as predicted here. Dubbed IDM (Interleaved Data Mode), the technology will be incorporated into the 256Mb version of Rambus Direct DRAM chips, due to sample at the end of 1999 and go into production next year. Essentially, the technology mirrors an approach laid down by IBM, which developed its Chipkill memory fault tolerance technology last year and now uses it in its high-end Netfinity servers. Rambus' announcement doesn't refer to the IBM effort beyond stating its "support for Chipkill high availability memory systems". As Rambus puts it, IDM "enables whole-chip and singular pin failure detection and correction using an approach similar to today's SDRAM-based systems" [our italics]. So IDM is essentially about allowing memory companies supplying RAM for Chipkill systems to use RDRAM, with its higher bandwidth and smaller chip size, instead of SDRAM. And let's not forget the naming of the technology. IDM sounds just like... well... IBM. Coincidence or a dig at Big Blue? We suspect the latter. ®

Sponsored: 10 ways wire data helps conquer IT complexity