MetaRAM, 8GB DIMMs and beyond

Memory capacity is one of those things that developers seem to have an infinite appetite for. How programs worked in the old days with a few Kbytes of memory is almost a mystery, but then again the type of information that is being crammed into memory for the sake of performance today is significantly different than even a decades ago. Yet, one has to wonder if programmers have gotten a ted lazy after trading in assembly skills for compiler optimizations. Regardless, MetaRAM, a two year old startup out of San Jose, California that emerged from stealth-mode a week ago is poised to make many new friends with their recent announcement of being able to quadruple the DRAM capacity of existing systems using existing DIMMs. Backed by several prominent venture capital firms including Kleiner Perkins Caulfield & Byers, Khosla Ventures, Storm Ventures, and Intel Capital, and led by Fred Weber who led the development of the Opteron processor at AMD, MetaRAM developed what they call MetaSDRAM technology.

As shown above, MetaRAM’s trick is a chipset that sits between the system memory controller and the actual DRAM. This chipset, which consists of an access manager chip, either the AM150 or AM160, and several flow controller chips, FC540, enables support for up to 16GB DIMMs without the need for any other hardware or software changes within the system. These chips work in tandem at speeds of up to 667 MT/s while being transparent to the host memory controller as well as the DRAMs.  This is definitely a clever trick, and as long as it does not inhibit the operating frequency it is a great way for upgrading current systems without having to re-work any additional hardware. Two questions emerge however that the company will have to address in the future: First, as the operating frequency of DRAMs increases, how long will MetaRAM be able to hide the latency of their chipset via clever buffering of reads and writes? Second, it is inevitable that memory controllers in the future will enable support for ever larger amounts of memory, is it possible therefore that at some point the amount of memory on a memory module will simply be limited by the physical integration limit, rather than by the controller capability? But for now things seem to be well at MetaRAM; several vendors have announced products based on MetaRAM’s chipsets and the company has also several open positions for those looking for something new to work on. If you want to find out a little bit more about the company’s co-founder and CEO head on over to The Register, where you can find a very interesting interview with Fred Weber which covers his career from the early days at Harvard, through NextGen, AMD, and finally MetaRAM.