The Bottom Line
There’s a bunch of analysis of costs and such floating around since the leak, but here’s the distilled version of it. The die size is roughly the same, and the wafers haven’t changed diameter, so with double the capacity, we’re talking a neat 50% drop in cost per GB. Keep in mind this change won’t happen overnight. The fab has to make its money back from all of the R&D and retooling costs needed to make the transition. Because of this, we won’t see the price slashing that took place with the move to 34nm, but over time prices will drop as production ramps up and the old process is phased out.
Same size wafer + roughly the same die count + twice the capacity per die = easy math.
There is also some buzz about this flash supporting ONFi 2.2. This is not as big a deal in reality, as even the soon to be discontinued 50nm flash present in first gen Intel and early Micron drives supported ONFi 2.1, which is where the 200MB/sec speed mode was introduced. ONFi 2.2 did not add any groundbreaking speed boosts as compared to the older spec. ONFi 3.0, once ratified, should implement Double Data Rate transfers, which will double speeds to 400 MB/sec (as the name implies). We’ll have to wait for a new revision to this process to see those faster speeds.
ONFi 2.2, while not on this chart, maintains the same 200 MB/s speed mode.
Same size wafer + roughly the same die count + twice the capacity per die = easy math.
There is also some buzz about this flash supporting ONFi 2.2. This is not as big a deal in reality, as even the soon to be discontinued 50nm flash present in first gen Intel and early Micron drives supported ONFi 2.1, which is where the 200MB/sec speed mode was introduced. ONFi 2.2 did not add any groundbreaking speed boosts as compared to the older spec. ONFi 3.0, once ratified, should implement Double Data Rate transfers, which will double speeds to 400 MB/sec (as the name implies). We’ll have to wait for a new revision to this process to see those faster speeds.
ONFi 2.2, while not on this chart, maintains the same 200 MB/s speed mode.
I know, don’t tell me, I’m spoiling all the fun. What good is this stuff if it’s not (immediately) cheaper and faster than the older stuff? Simple. This new flash will push the SSD makers into another revision of their SSD controllers. Controller tech is still lagging *way* behind the full speed of flash memory. Most ‘modern’ controllers implement ONFi 1.0 (!), relying on sheer parallelism to boost transfer rates on the host side of the controller. Intel is planning a G3 line to be released later this year. Micron also plans to implement the new flash in their next generation controllers. These new revisions will boast further enhancements and optimizations, taking advantage of some of the array addressing time reductions that came with ONFi 2.2, along with a *long overdue* boost to the write speeds of Intel models. Intel is planning a coordinated effort among their various groups, where SATA 6Gb/sec will be rolled out in their chipsets and SSD’s simultaneously. And remember, over time the prices will come down as volume scales up. We hope to see better drops over time than seen with 34nm last year, which is still struggling to catch up to it’s price reduction target.
25nm flash is here to stay. Get it while it’s hot!
I’d like to thank Intel and Micron for having us out for this unique experience. It was great being able to meet with the IMFT staff as well as members from both companies. Getting a first-hand look at the bleeding edge of chip fabrication is something I know I’ll be talking about for years to come.
