Is it faster than flash, hence why to use instead of?
If it is faster and its cost is near that of flash, why bother using flash at all?
IBM has claimed a significant break through in phase-change memory which could potentially replace DRAM. Researchers at IBM say they have successfully managed to store 2 bits per cell in a chip employing a strip of 64,000 cells at some toasty temperatures and while under quite a bit of stress "for the first time ever". Multi- …
Because 2 bits per cell at 64,000 cells is 128kbit. Kilobit. We're not talking 128GiB DIMMs here. We're not talking 4TiB PCM drives.
What we're talking about is a potentially useful high endurance write buffer for flash chips. Think about writing log files. Lots of tiny little changes that are definitely "sub-cell" in size, when we lok at writing all of that to a flash device. Logs can wear out flash in a hurry.
But what if you front-ended that flash with this PCM? Absorb the writes until there is enough change to require a full cell's worth of writes, thus optimizing the flash?
Physically, flash will occupy less space for some time yet. And never believe IBM regarding price. They aren't talking about PCM as being the same cost as bulk 3D NAND, but competitive with the most expensive SLC flash you can find, I promise you.
So the initial applications will be PCM as a flash cache layer until some refinements are made to packaging, temperature reliability and so forth. Basically, taking the SLC cache out of today's flash drives and putting PCM in instead.
Ten Years? Twenty...?
In the meantime, everything currently being used will slowly evolve and surpass it - I'm still waiting for my "Cheap as Chips" (sorry!) Bubble Memory, Memristor-based storage, and the Sci-Fi stalwart that is always being touted as the Saviour of Humankind: Holographic Storage.
All of these things exist in one form or another, but just have never (not yet?) made the grade.
Great news, but not holding breath.
Firstly hats off to IBM for continuing Research into new Technology. They do have the money and the expertise to come up with new ideas - while most others just implement IBM's research or find an application for it.
The storage industry alone has spent the last 40 overcoming the limitations of spinning disk. (Raid, Snapshots, IOPS intensive workloads).
Toshiba invited Flash in the 80's and it is only mainstream now.
Therefore you also have to consider the commercial aspect and the greed of corporations.
Could we have perfected Flash in the 90's? Probably, but there was plenty of $$$ to be made on spinning disk.
Intel sells you CPU's that are incrementally faster than the last generation - because people keep buying.
Next is the electric car - vs - oil lobby.
Fossil energy lobby -vs - renewable...
That's what's wrong with Research. Researchers would happily push forward, but its the economic forces slow us down...
Yup. IBM pioneered the giant magnetoresistive effect back in the '80s that was a breakthrough for hard drives, and let you store megabytes on something smaller than a house brick.
If HP said it'd discovered this, I'd laugh... but IBM I believe, and I'm pretty sure they'll push it through to market if it's viable at all.
I would not have disbelieved it of HP even ten years go. HP Labs was mutated into a marketing gimmick but Carly, and then hollowed out by a timid board, more interested in their own short term riches than long term investment. I know, I worked with HPL for fifteen years, and what was lost through greed and apathy almost made me cry.
I mean phones typically have power 24/7 since they have a battery... plus it would make it easy and cheap to implement meaningful security measures, like making it erase its contents when it's opened or the temperature suddenly drops. Since those things can be done mostly in software you could sell "high security" phones for a decent premium without it costing you much more.
I could see using a smaller amount of DRAM and paging from a phase change backing store, thereby saving power. The performance cost would only be noticeable if you were switching apps constantly, or expecting them all to stay active with a lot of wired memory even when they weren't in the foreground.
"I could see using a smaller amount of DRAM and paging from a phase change backing store, thereby saving power. "
That missed the point that DRAM is a bigger bottleneck to performance than anything further down the chain.
CPUs spend most of their time waiting for ram to catch up. The fact that storage past that point is a few hundred/thousand times slower only matters when loading in data or apps.
PCM and friends are good as a rival to keep spurring flash along (even if they never are commercially viable, they keep development of existing tech going), but what's sorely needed _right now_ is affordable off-cpu memory with random access latency of less than 10ns, preferably under 5ns
Biting the hand that feeds IT © 1998–2020