Introduction, How PCM Works, Reading, Writing, and Tweaks

I’ve seen a bit of flawed logic floating around related to discussions about 3D XPoint technology. Some are directly comparing the cost per die to NAND flash (you can’t - 3D XPoint likely has fewer fab steps than NAND - especially when compared with 3D NAND). Others are repeating a bunch of terminology and element names without taking the time to actually explain how it works, and far too many folks out there can't even pronounce it correctly (it's spoken 'cross-point'). My plan is to address as much of the confusion as I can with this article, and I hope you walk away understanding how XPoint and its underlying technologies (most likely) work. While we do not have absolute confirmation of the precise material compositions, there is a significant amount of evidence pointing to one particular set of technologies. With Optane Memory now out in the wild and purchasable by folks wielding electron microscopes and mass spectrometers, I have seen enough additional information come across to assume XPoint is, in fact, PCM based.

XPoint.png

XPoint memory. Note the shape of the cell/selector structure. This will be significant later.

While we were initially told at the XPoint announcement event Q&A that the technology was not phase change based, there is overwhelming evidence to the contrary, and it is likely that Intel did not want to let the cat out of the bag too early. The funny thing about that is that both Intel and Micron were briefing on PCM-based memory developments five years earlier, and nearly everything about those briefings lines up perfectly with what appears to have ended up in the XPoint that we have today.

comparison.png

Some die-level performance characteristics of various memory types. source

The above figures were sourced from a 2011 paper and may be a bit dated, but they do a good job putting some actual numbers with the die-level performance of the various solid state memory technologies. We can also see where the ~1000x speed and ~1000x endurance comparisons with XPoint to NAND Flash came from. Now, of course, those performance characteristics do not directly translate to the performance of a complete SSD package containing those dies. Controller overhead and management must take their respective cuts, as is shown with the performance of the first generation XPoint SSD we saw come out of Intel:

gap.png

The ‘bridging the gap’ Latency Percentile graph from our Intel SSD DC P4800X review.
(The P4800X comes in at 10us above).

There have been a few very vocal folks out there chanting 'not good enough', without the basic understanding that the first publicly available iteration of a new technology never represents its ultimate performance capabilities. It took NAND flash decades to make it into usable SSDs, and another decade before climbing to the performance levels we enjoy today. Time will tell if this holds true for XPoint, but given Micron's demos and our own observed performance of Intel's P4800X and Optane Memory SSDs, I'd argue that it is most certainly off to a good start!

XPoint Die.jpg

A 3D XPoint die, submitted for your viewing pleasure (click for larger version).

You want to know how this stuff works, right? Read on to find out!

FMS 2014: HGST Claims 3 Million IOPS and 1.5us Access Time SSD - updated with pics

Subject: General Tech, Storage, Shows and Expos | August 7, 2014 - 02:17 PM |
Tagged: ssd, phase change memory, PCM, hgst, FMS 2014, FMS

According to an HGST press release, the company will bring an SSD based on phase change memory to the 2014 Flash Memory Summit in Santa Clara, California. They claim that it will actually be at their booth, on the show floor, for two days (August 6th and 7th).

The device, which is not branded, connects via PCIe 2.0 x4. It is designed for speed. It is allegedly capable of 3 million IOPS, with just 1.5 microseconds required for a single access. For comparison, the 800GB Intel SSD DC P3700, recently reviewed by Allyn, had a dominating lead over the competitors that he tested. It was just shy of 250 thousand IOPS. This is, supposedly, about twelve times faster.

HGST_CompanyLogo.png

While it is based on a different technology than NAND, and thus not directly comparable, the PCM chips are apparently manufactured at 45nm. Regardless, that is significantly larger lithography than competing products. Intel is manufacturing their flash at 20nm, while Samsung managed to use a 30nm process for their recent V-NAND launch.

What does concern me is the capacity per chip. According to the press release, it is 1Gb per chip. That is about two orders of magnitude smaller than what NAND is pushing. That is, also, the only reference to capacity in the entire press release. It makes me wonder how small the total drive capacity will be, especially compared to RAM drives.

Of course, because it does not seem to be a marketed product yet, nothing about pricing or availability. It will almost definitely be aimed at the enterprise market, though (especially given HGST's track record).

*** Update from Allyn ***

I'm hijacking Scott's news post with photos of the actual PCM SSD, from the FMS show floor:

DSC04122.JPG

DSC04124.JPG

In case you all are wondering, yes, it does in fact work:

DSC04125.JPG

DSC04126.JPG

DSC04127.JPG

One of the advantages of PCM is that it is addressed at smaller sections as compared to typical flash memory. This means you can see ~700k *single sector* random IOPS at QD=1. You can only pull off that sort of figure with extremely low IO latency. They only showed this output at their display, but ramping up QD > 1 should reasonably lead to the 3 million figure claimed in their release.

Source: HGST

Phase Change Memory might be on the market next year

Subject: General Tech | November 30, 2012 - 01:38 PM |
Tagged: nand, EMC, phase change memory

SSDs are not that old but already there is a challenge that must be overcome if it is to remain a viable storage medium.  As Allyn has discussed many times in articles and on the podcast, as NAND process shrinks continue, the number of write cycles before failure drops which lowers the life expectancy of the drive even while it allows for high capacity chips and lower power consumption.  Zahid Hussain is EMC's flash product division general manager and he is confident that his company will be able to do what Hynix, Samsung and others have so far been unable to do; work with Micron to replace the NAND chips with Phase Change Memory based chips.  This type of chip is non-volatile and could also find its way into DIMMs as well.  Read more at The Register.

ElReg_pcm_schematic.jpg

"It is anticipated that, as NAND process geometries shrink beyond 15nm or so, the working life will fall off drastically, speed will slacken and the error checking and correction logic will become much more complicated. At that point, roughly, it is hoped, a post-NAND technology will be productised and deliver chips that are denser than flash, faster than flash, approaching DRAM speed, byte-addressable instead of block-addressable, and with a longer working life. That seems like a real big ask."

Here is some more Tech News from around the web:

Tech Talk

Source: The Register

The possibilty of rising DRAM prices got you down? Try PRAM instead!

Subject: General Tech | July 18, 2012 - 02:39 PM |
Tagged: phase, phase change memory, micron, 45nm

Phase Change Memory is not new, Allyn listened to Intel talk about a breakthrough in this technology almost three years ago, but it is not common on the market.  It offers two major benefits over the current RAM on the market, the first being its lack of volatility as the crystals it forms will remain even after power is cut off, the second is more dear to computer enthusiasts as it should be faster than DRAM.  It may seem odd that a technology which requires the formation of crystals would be faster than the electronic flipping of bits but Micron claims that the trickle of voltage supplied creates seed crystals which speed the formation process during write cycles. The good news is that we should see real world testing soon as The Inquirer has heard that Micron has a good supply of PRAM to sell which means benchmarks are not far behind.

Silver_Cross_Balmoral_Pram.jpg

"MEMORY MAKER Micron has announced high volume availability of its 45nm phase change memory (PCM) chips.

Micron has been pushing the development of PCM chips with Intel for a number of years and is finally at a stage where it can offer chips to its customers. The firm announced that its 45nm PCM chips are available in a 1Gb PCM plus a 512Mb LPDDR2 package for mobile devices."

Here is some more Tech News from around the web:

Tech Talk

Source: The Inquirer