How 3D XPoint Phase-Change Memory Works
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 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.
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:
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!
A 3D XPoint die, submitted for your viewing pleasure (click for larger version).
How PCM Works
To understand how XPoint reads and writes bits, let’s start with how phase change materials work, and to do that we need to know what makes a material PCM capable in the first place:
Periodic Table. Metalloids in yellow. Chalcogens blue boxed. source
Phase change materials are generally alloys of metalloids. Metalloids are elements that share properties with metals and non-metals. They act as insulators at room temperature and as conductors when heated (or when doped). Alloys of varying mixtures of the semimetals have been experimented with for decades. Boron is mainly used for doping, and Polonium is unstable and radioactive, so we won’t be seeing much of that one :). Silicon is great for standard transistors and other semiconductors, but less than optimal as a phase change material. That leaves Germanium (Ge), Arsenic (As), Antimony (Sb), and Tellurium (Te). Alloying these together results in a chalcogenide (kal-kuh-juh-nahyd), which in the context of this article is a compound containing a Tellurium anion (Te is the only stable metalloid belonging to the chalcogen group of the periodic table). Once mixed in the proper proportions, these materials offer some rather unique properties:
Specifically, metalloid alloys have multiple stable states that each come with their own distinct resistance characteristics. These can be manipulated by heating and cooling the material in various ways. The amorphous state resembles a glass, while the crystalline state more closely resembles a metal.
Reading and Writing Phase Change Memory
Ok, let's explain what is going on here. Voltage is applied across a section of chalcogenide material. If the material is in an amorphous (mixed) state, it does not begin to conduct until the threshold voltage (Vth) is exceeded. Once conducting, as voltage increases further, so does the current. Since the material is now acting resistive, it dissipates heat and therefore increases in temperature. If held at the ‘set’ (1) voltage, the material reaches ~350C, which is not hot enough to become molten, but *is* warm enough for its molecules to realign into a crystalline structure if the temperature is maintained for ~100 nanoseconds. Once formed, the crystalline structure behaves like a resistor and remains even after the voltage is removed and the material cools. Once in the set state, applying 0.5V would result in ~0.5mA (using the above example). The voltage no longer needs to meet a threshold in order for the material to conduct, and its response follows the plot line marked ‘crystalline’.
To reset the cell, we apply a much higher voltage, pushing currents and temperatures high enough (~600C) to heat the material to a molten state. This melts down the crystalline structure. The voltage is then removed and the material rapidly cools, passing through the crystallization temperature region too quickly to form any crystal structure, ‘freezing’ it in the amorphous state. It is now ‘reset’ (0), and applying that same 0.5V will result in near zero current. I should point out that we don’t need nearly that high of a voltage to perform a read, as even 0.1V would produce a readable difference in current between the two states used in our example.
An interesting thing to note about the above is that there is no ‘erase’ required before programming a cell as is the case with NAND flash. With PCM cells, we can perform a set or reset operation by simply applying the associated voltage/time profile without regard for the previously set state. Unlike NAND which must be written in pages (KB) and erased in even larger blocks (MB), PCM data can be overwritten ‘in-place’, and single bit overwrites are possible without disturbing adjacent cells.
Tweaking the Formula
Intel and Micron would have you believe that the stuff that makes up XPoint is an ancient Chinese secret. Well, it’s not. The common phase change alloy is a 2:2:5 stoichiometric ratio of Germanium, Antimony, and Tellurium. Ge2Sb2Te5, dubbed ‘GST’ for short. As is with most alloys, there are many slight variations possible to the recipe, and that is where the manufacturer-specific secrets come into play. That said, we do have some clues as to what might have been tweaked from a 2010 Micron presentation:
Those developing PCM technology will naturally finely tune the mixture to try and improve performance. Above we see an excerpt from a Micron brief showing how slightly increasing the concentration of Antimony (Sb) helped reduce the reset resistance (reducing the voltage needed), as well as reducing the time needed for a set operation. There are also external factors related to cell selection that might require tweaking the ratios further, which we will touch on shortly.
You might think phase change alloys are so exotic that you have never seen or held them, but you are likely wrong. Rewritable optical discs (CD-RW/DVD-RW) are extremely close cousins to the materials found in XPoint. Optical discs used Silver and Indium in place of the Germanium found in GST, which naturally changed the properties of the alloy. ‘Blank’ media was crystalline, and pits were written by heating spots by pulsing the write laser. The spots then quickly cooled without a chance to recrystallize, forming darker areas that could later be read as differences in reflectivity. Discs were erased by applying a lower power laser which started the recrystallization process (these alloys could continue crystalizing after the laser passed the area). Other metalloid alloy variants were used in various optical media technologies, aiming to improve the number of erase cycles and other performance characteristics. DVD-RAM actually used GST compounds but relied on its changing optical properties as opposed to electrical conductivity.