|Image: Micron website/video|
As anyone with a television or a DVR, let alone a smartphone, knows, technology evolves quickly, and, for the consumer, expensively. The record-player turned CD-player turned MP3-player, and the requisite refresh of all your content with each new version, is a costly and redundant proposition. Sure, these new technologies bring great benefits, but. like me and my iPhone4, some consumers stay behind the curve to stay solvent.
So, with the recent announcement by Intel and Micron of a new data storage technology called 3D Xpoint (called three-dee cross-point), will we all be trashing our last-century magnetic hard drives, and even our flash cards and SSD drives, and migrating all that content and data to this new technology? Is it, as proclaimed by Intel GM of Technology Manufacturing, William Holt, "the Holy Grail of memory", or is it a replay of Intel's doomed partnership with Rambus, which created Direct RDRAM, a previous Holy Grail that the market rejected, and that caused Intel to lose credibility with both their PC-maker partners and the consumer?
Let's take a quick look at the state of data storage today. The hard drive, which most of still use for our primary storage device, dates all the way back to 1956, when IBM introduced a drive with fifty 24-inch platter, two heads, and a total capacity of less than 4 megabytes. Obviously, we've come a long way in size, speed, and capacity, but the magnetic read-write function on a set of spinning disks is still the basis for our current drives. With the invention of NAND memory, commonly called Flash, we've become accustomed to the tiny "thumb-drives", and the Solid State Drives (SSD's), that contain our photos and videos, and that make iPhones and tablets possible. And inside the computer itself, we have register and cache memory on the processing chip itself, plus DRAM storing the data that the computer "has in memory" while it computes.
So we've already identified four major challenges. The first is proximity - as data gets farther away from the processor, access slows. Closer is not necessarily faster - DRAM is about 10 times slower than cache, Flash memory 10,000 slower, and data on our hard disk is over ten million times slower to access. The processor is often in a wait state, reducing the benefit of creating faster chips, and requiring us to go get a cup of coffee while our computer 'spins'. The fourth challenge is complexity - because of the current state of the technology, computer manufacturers must include all of these memory types - cache, DRAM, NAND, and hard disk, in order to serve the CPU with the volatile and non-volatile data it needs.
So a memory technology that could be cheap, fast, non-volatile (so it retains its data without power), and can be installed on the motherboard next to the processor, would indeed be the Golden Fleece. This is what Intel and Micron have proclaimed with 3D XPoint. This technology provides the potential to replace all the onboard system memory, plus external storage, either SSD or hard disk, with one type of super-dense, super-fast memory that lasts much longer and ultimately costs much less than current memory technology.
So haw does it work? According to technical specifications released by Intel and Micron, 3D XPoint is an array of 128 billion cells, each of which stores one bit. These cells are stacked in multiple layers, with a 'Selector' technology that varies voltage to these memory locations, eliminating the need for transistors. The key to the speed is the ability to write to each selected cell, rather than in blocks or pages as with current technologies, and to "flip bits' at a much higher rate than other memory configurations.
Sounds all good, but there are a few dark clouds. First, the connection point for this technology is currently expected to be the old standby PCIe bus, which is the fastest bus available but way too slow to get the full advantage of 3D XPoint. This means that, to get the benefits, PC-makers will have to revamp the bus structure of the PC, thus costing them much development time and making current systems obsolete. PC vendors also complain that Intel and Micron are being coy about how and when the technology will be available, thus foiling their planning cycles.
On the positive side, the idea of self-contained computers with no spinning disks, cache-like speed, and much greater density and longevity, should spur a new round of innovation that powers both hardware and software makers to rethink the capabilities of the PC, tablet, and phone. Get set for another round of innovation, and another pile of obsolete components.