The key to Watson’s success was its reliance on parallelism – it corralled the computing power of 2,880 CPUs and 15TB of DRAM. When considering an average desktop today will have around 4GB of DRAM, this is the equivalent of nearly 4,000 PC’s worth of DRAM. More interestingly, perhaps, is the DRAM/CPU – about 5.2GB. For a typical PC today the number is much closer to 2GB/CPU. Clearly Watson needed near instant access to vast quantities of data.
While consumer PCs are a long way from adopting 15TB of DRAM and it’s unlikely that IBM is going to sell enough Watsons to really impact the DRAM market, there are some important lessons to be learned from this exhibition. The first is that DRAM is still much, much faster than NAND technology. NAND is certainly an emerging technology that will find its way into more and more of our devices, but it is still a long way behind DRAM when it comes to raw speed. The second lesson is that while search has come a long ways in the past 15 years, largely in thanks to Google, computing horsepower is going to enable even better search. Of course, it will take more than one Watson to power the Google super-search engine of the future, it will take many thousands – and that means a LOT of DRAM.
Find Out More >Latest iSuppli DRAM Research