Researchers See Gloom And Doom For SSDs–Shortsightedness Never Ceases To Amaze

Researchers at the University of California, San Diego (UCSD), have announced that SSDs only have about a decade of life left before they become impractical from a functional standpoint. The researchers have stated that as flash memory shrinks, read and write latency, not to mention cumulative bit errors increase.

Apparently, one of the team members wrote a paper which asserted that the future of SSDs looks “cloudy”, due to problems associated with using finer manufacturing processes.

The thing is, this isn’t really breaking news, as the industry has known about this issue for quite some time now. Indeed, this is precisely the reason why multiple major players in the field have been developing new technologies to replace flash when it does inevitably run out of steam.

One of these potential successors, MRAM, has actually been around for quite some time now, but has had the disadvantage of being relegated to older fab processes and hasn’t really gotten as much attention as it probably should have, until recently that is. Newer developments such as Spin-Transfer Torque MRAM and perpendicular magnetic domains have allowed densities that are an order of magnitude greater than what is possible with conventional MRAM products.

HDDs have actually been using perpendicular recording, which is similar in concept, for quite some time now and are indicative that this technique should make MRAM a viable option well into the future. Add to that the fact that MRAM performs similarly to SRAM, and it’s easy to see that this potential replacement is ripe for the market when flash finally does give up the ghost. With NAND no longer an option, I’m sure fabs will be able to find some extra room on those freshly vacated state of the art lithography lines.

Another alternative would have to be memristors, which are actually the fourth fundamental passive electronic component, complementing resistors, capacitors and inductors. It’s true, HP has stated that memristors (in the form of ReRAM) have the ability to replace all memory types and even have the potential to supplant transistors as the basic building block of the logic that is foundational to modern computing. Having such a wide range of capabilities, I think, at the very least, they will be up to the task of serving as the main component of future post-Flash SSDs, offering lower power consumption and higher performance to boot.

These aren’t the only two options either, (though they do seem to be the front-runners by a significant margin) other contenders include Racetrack Memory, Nano-RAM (NRAM), and programmable metallization cell (PMC) memory, all of which seem likely candidates to replace flash when the industry deems it necessary.

My point is that SSDs aren’t likely in any danger of going the way of the Dodo, now or in the future. The industry has known about flash’s limited shelf life for quite some time and has been diligently pursuing new paths in search of a replacement. To those who want to reiterate that flash will be approaching the end of its useful lifespan in the next decade or so, my answer will have to be a succinct, if perhaps a little bit surly ‘DUH!’

Oh well, I suppose the real proof will come in the form of SSDs that are even faster and approaching the densities of magnetic storage. In all honesty I can’t say I’ll be surprised if and when such a scenario comes to fruition, and I can’t imagine anyone reading this will be either.

In the meantime, I’ll be enjoying the blistering speed of my crochety, borderline obsolescent NAND based SSDs, patiently awaiting my deliverance from this high speed ghetto. Joking aside, it’s hard to imagine how anyone could view the future of SSDs as anything but positive. With a veritable cornucopia of future options available to the industry, the only realistic direction, in my opinion, appears to be ‘up’. In any case, see you on the other side.


  1. Great points, Josh! Isn’t it amazing how some researchers seem to think that they are prescient about technologies that are still in many ways in their infancy! Who can say what yet-to-be-discovered developments could change these technologies that they are all but writing off already. Must of had to come up with some ‘result’ (however unprovable) to justify their research grant.

  2. Very Shortsighted in the fact it quoted the end of ssd drives instead of the use of Nand flash memory.

  3. When I read this whitepaper a few weeks ago (Usenix released all of the whitepapers to us before the conference) I was amazed that they were quoting ECC as being a major inhibitor of progress with regards to latency. With solutions such as CLEARNand from Micron bringing ECC on-die with the NAND, these issues are already being addressed at the base level of the architecture itself. Removing ECC from the processing overhead yields tremendous results going forward. That was not taken into account for this study, yet this relatively ‘new’ tech is already being fielded. They even refer in the whitepaper as ECC being handled by the controller incurring latency, when ECC is already being removed from the controller!
    Another obviously overlooked aspect is signal processing. They make absolutely no mention of the increasing viability of signal processing and its effect on ‘cleaning up’ data to alleviate many of the issues that they are citing as detrimental to latency.
    Anobit and others are pioneering this work, and the early signs of success are already being seen simply by the fact that these companies are being snapped up right and left, as Anobit was by Apple. Apple specifically went after Anobit for the signal processing capabilities.
    Basically, they are inferring that if ABSOLUTELY NO further technological advancements are made, and if we ‘unthink’ and roll back a few already existing technologies that are ignored in the study, then NAND will cease to be viable.
    LOL. a joke of a study.
    Sure greater things will come along, but NAND is like HDD…it isnt going anywhere for a lonnnnggg time. it will have its place in the storage world for a much longer time than this incomplete study asserts!

  4. Seems they don’t have a realization of what ‘Solid State’ means. Also, the study name should have said ‘…for NAND based SSDs…’. I agree…huge waste of grant money.

  5. I think the point is that NAND is getting worse as it gets larger, so there is no where to go from here. The tech has existed since 1984, its not new, but it is getting worse as we create newer versions.

  6. An interview with Fujitsu CTO Dr Joseph Reger mentions Phase Change Memory (PCM) as a replacement for flash. Check out original URL:

Leave a Reply

Your email address will not be published. Required fields are marked *