# Talk:Phase-change memory

WikiProject Computing / Hardware
This article is within the scope of WikiProject Computing, a collaborative effort to improve the coverage of computers, computing, and information technology on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.

## Quality

Put bluntly; this is poorly written and hard to follow.

## title

Agreed, moving to "phase-change memory" Maury 20:25, 20 October 2005 (UTC)

## Diodes?

Does anyone know why PRAM doesn't need a transistor for a switching element, and can use a diode instead? This is interesting, and seems to deserve a real explaination in the article. Maury 22:02, 18 September 2006 (UTC)

I think most forms of RAM use a capcitor to store the state (0/1) as a variable charge. PRAM uses a variable resistor to store the state change. The challenge is how to access the cell, to store the data and then to read it out. How to do this as fast as possible, with minimum circuitry etc? A general row-column wiring matrix is usually used. It seems that dealing with resistors may be a little simpler than capacitors. Generally at least a diode per cell is needed, to isolate the cell being accessed, but a special kind of PRAM incorporates the diode function in the cell itself, minimizing additional circuitry needs. 69.87.200.92 19:02, 11 December 2006 (UTC)

Generally a memory requires two things: a storage mechanism (the floating gate in Flash, the GST in PRAM) and a way to address the storage mechanism (MOS transistors in Flash, MOS or BJT transistors in current PRAM technology). In the digital realm, transistors are generally on (conducting current) or off (not conducting). The voltage at which the device transitions is called the threshold voltage, Vt. To read a Flash bit, one applies a voltage to the transistor, which -- depending on its Vt -- either conducts or does not. The charge on the floating gate controls the Vt, and hence whether one sees current or not when the device is turned on. Again: apply a voltage and either see current, or not. Now consider a diode. It also has two states -- conducting or not. The response is more gradual than a transistor, but at some voltage Vt, it begins to conduct. Put a variable resistor in series with the diode and apply a voltage. If the resistance is low enough, the voltage at the diode will be above its threshold and it will conduct. If the resistance is too high, the voltage drop across the resistor is too great and the diode sees something less than Vt (and doesn't conduct current).

## Reverse bias, high voltage needed for diode selection PRAM

I think reverse bias is needed to ensure not every cell on an accessed line is inadvertently selected for reading or writing. Only the cell to be accessed should be forward biased. Also the high voltage needed to exceed the diode threshold voltage for the high-resistance state suggests that a high operating voltage is needed for PRAM. 219.70.70.32 02:31, 1 July 2007 (UTC)

## Macronix chalcogenide RAM

As originally described earlier, it utilizes only the amorphous state. Hence it is not a phase-change memory. It should go into another more appropriate article (maybe RRAM?).219.70.70.32 02:39, 1 July 2007 (UTC)

## Capacity units

There are several devices listed as having capacities in Mb. This is traditionally megabit (from Mbit in networking parlance), and it's also traditional to list capacities of memory chips in megabits. However, these devices are compared directly to others with capacities listed in GB for example. Which unit is intended? I'm assuming MB is correct; this should be changed in the article.--Ktims 07:10, 18 April 2007 (UTC)

Changed.

Should be removed, as they are in their current form mostly bullshit. Neither will flash rewritability kill your ipod (maybe in 5000 years...), nor does it make flash unsuitable for hardisk applications (see the tons of flash SSDs now starting to push into the marketplace. 134.105.82.206 12:56, 21 May 2007 (UTC)

Given that Flash HDs are available now to consumers, the "makes Flash unsuitable for hard drive replacements in desktop computers" comment is clearly bogus and was removed. Please find a source to support your claim that Flash write cycling degradation will not cause failure of an iPod, or other consumer electronics device for 5000 years.

## Phase-Change Memory vs Volatile Memories like DRAM

• Oppose, they seem to be way too different. The only thing they appear to have is that they both are competing fast non volatile memory. --soum talk 13:17, 9 July 2007 (UTC)
• Opposed, phase-change memory is a well-defined technology, while RRAM is not (yet). Guiding light 06:06, 17 July 2007 (UTC)

## Edit timeline?

The timeline says Stanford Ovshinsky filed the first patent on phase change technology. First, does anyone know if this application was granted as a patent and what number it is? Second, it says patent 3,448,302 was licensed to him claiming first reliable operation of phase change memory. Patent 3,448,302 was filed in June of 1966, 3 months before Stanford Ovshinsky's supposed first patent filing. Does anyone have more information on this? Carl 612 16:37, 16 August 2007 (UTC)

• The patent referenced in the time line is probably 3271591 (filed Sept 6 1966). I didn't review it it detail, but on page 4 it has the classic chalcogenide I-V curve drawings. Papers on the resistivity changes of chalcogenide materials date as far back as 1923. The biography on Stan Ovshinsky here says he discovered and developed switching materials between 1958 and 1961. There's also a list of publications that doesn't mention patents. Fhaigia 09:31, 15 September 2007 (UTC)

## Wrong Drawing

The phase change memory cell drawing looks very wrong. The bit line should be perpendicular to the word lines. That is not shown in the picture.218.168.141.188 14:38, 4 September 2007 (UTC)

• I actually deliberately didn't draw the bit line's top part, because it is not what this drawing should deal with. There could be many different configurations regarding wiring. In this drawing, I only want to show the two states of the charcogenide and leave the wiring part for other drawings. --Cyferz 21:07, 4 September 2007 (UTC)
• I agree. I think as positioned, the bit line is more confusing than helpful. Also, I'm not sure what "poly chalcogenide" is supposed to be. Chalcogenide is a metal alloy of germanium, antimony, and/or tellurium; poly as used in this context usually refers to heavily doped conductive silicon. The green area probably should be labeled crystalline chalcogenide. I clarified in the caption to indicate we're looking at a cross-section cut along the bit line. Fhaigia 08:46, 15 September 2007 (UTC)
• Drawing updated. --Cyferz 16:10, 15 September 2007 (UTC)

## Programming/Soldering

Maybe it's just me, but I don't follow this problem. Surely the manufacturer could just solder a socket in place and insert the chip after soldering - or not?? TinyMark 17:44, 17 September 2007 (UTC)

Yes, your suggestion is one work-around for the stated problem. However, sockets are expensive, likely costing more than the memory itself in this case. Particularly in embedded products, a bill of materials cannot absorb that much cost increase without affecting the price of the final product. Fhaigia 13:33, 4 November 2007 (UTC)

## RW optical disc - is it a phase change memory?

It just occured to me, that CD-RW, DVD-RW, DVD+RW, DVD-RAM, BD-RE, HD DVD-RW media use phase change between crystalline and amorphous form to store data too. Just like PRAM. Except they read it with electromagnetic field (laser), not with electrical current like PRAM, which is a minor difference really. Are those "phase change memories" too? Why not? Is it the "random access" part? Could someone comment on this? I have no reliable sources on PRAM, so I'm not altering the article itself. --Kubanczyk 21:44, 10 October 2007 (UTC)

• The minor difference you mentioned decided they are not. Nontheless they are related. Some time there are not clear borders between things. --Cyferz 02:08, 11 October 2007 (UTC)
• This is an article about a specific type of nonvolatile memory technology. It is not a general article about any memory storage system that relies on any type of material which changes state. Those media do use the same or similar phase-change material, but the mechanism of operation is entirely different, and hence belongs in articles specific to those media. CDs/DVDs/etc. rely on changes in reflectivity that accompany state changes in chalcogenide materials. PRAM relies on resistivity changes in chalcogenide materials. In a general article on chalcogenides, it would be appropriate to mention both behaviors, but except as incidental curiosity, reflectivity change is irrelevant to PRAM. Calling a laser an electromagnetic field may be technically correct, but its frequency is vastly different than those used in typical electronics. Most people think of electricity as something that comes out of a wall socket, and the stuff they see with their eyes as "light." Fhaigia 13:43, 4 November 2007 (UTC)
• They use the same materials. In CD-RWs, the phase change material $Ge_2Sb_2Te_5$ is used, in DVD-RWs $Ag_5In_5Sb_{60}Te_{30}$. The way, how the energy needed for the transformation, is deposited in the system differs. For optical discs the laser is used, for electrical PRAM current is used. Some of the most promising materials for PRAm are $Ge_2Sb_2Te_5$, $Ge_10Sb_65$ and $InSbTe$. —Preceding unsigned comment added by 88.77.134.108 (talk) 17:48, 13 April 2008 (UTC)

## HP Micromover?

@76.104.212.98: I just noticed the external link to HP probe storage, and the comment added about the HP Micromover technology, which I presume is the trade name for the technology discussed on the HP probe storage web page. What does this have to do with PRAM? As far as I can tell, the two are completely unrelated and -- for how PRAM is currently being developed -- incompatible. Can you please illuminate for me why that link and comment belong in this article? Fhaigia (talk) 04:37, 16 August 2008 (UTC)

• There has been no reply... I'm removing the Micromover reference. Fhaigia (talk) 16:05, 8 October 2008 (UTC)

## Flash vs SRAM

The article states it take Flash 10,000 times as long to write a block as it does for SRAM to write a byte. Nintendo Gamecube memory cards have a 512 KB block and block size has probably increased with increasing storage size. In 1 ms flash would do at least 5 MB while SRAM would only 1 KB. Why isn't flash the fastest memory available? — Preceding unsigned comment added by Sdmitch16 (talkcontribs) 17:49, 12 April 2011 (UTC)