Talk:Dynamic random-access memory

From Wikipedia, the free encyclopedia
Jump to: navigation, search
WikiProject Computing / Hardware (Rated C-class, Mid-importance)
WikiProject icon 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.
C-Class article C  This article has been rated as C-Class on the project's quality scale.
 Mid  This article has been rated as Mid-importance on the project's importance scale.
Taskforce icon
This article is supported by Computer hardware task force (marked as High-importance).

ECC memory protection rates[edit]

How can anyone deduce such error rates ("from 10-10 to 10-17") for ECC modules based on the sigmetrics'09 paper [1]? I don't like statistics, but whichever the way I try, I obtain results orders of magnitude larger than this, more like 10-5 per bit and per hour. Would anyone with enough confidence in their calculations point me to the correct line of reasoning? —Preceding unsigned comment added by (talk) 16:05, 24 February 2010 (UTC) i am curious as to what is the current thinking is concerning soft-errors. first we hear it's alphas, then there's a comment about neutrons, then another comment about alphas again. this should be easy to test. Bob Emmett (talk) 07:17, 19 March 2010 (UTC)


Will someone please explain "unbuffered", "registered" and "fully buffered" RAM, particularly as it relates to ECC?

This last one is explained but not linked in: —Preceding unsigned comment added by Solinym (talkcontribs) 21:30, 29 January 2008 (UTC)

Isn't "registered" and "fully buffered" the same? Is the information about ECC memory being gone in modern computers accurate? It mentions mother boards, but AMD's memory controller is on the chip, so every AM2 motherboard supports unbuffered ECC with all 64 X2 processor. That would seem to be much more than none. I could be wrong. —Preceding unsigned comment added by (talk) 16:04, 10 February 2009 (UTC)

Removal of Spelling and pronunciation section[edit]

The spelling and pronunciation section was summarily removed (at 14:33, 23 November 2005) a few hours after I added it. I fail to see what is 'collaborative' about this approach to Wikipedia. It's simply a case of two people with differing opinions, and the one who shouts loudest wins.

Nobody would deny that some guidance on spelling and pronunciation of 'DRAM' would be very useful. So wouldn't it be sensible to leave the section in place for a while, even if you personally believe parts of it are incorrect, and so allow people other than yourself to have some input? That way we might actually see some collaboration, and end up with a useful section that's acceptable to everyone.

I would deny it; if I didn't I wouldn't have removed the text. I didn't remove it so much because I believed it was inaccurate, but because it is irrelevant. DRAM isn't a trademark or a copyrighted name, but a general industry acronym. There is no 'official' way to pronounce the acronym, and I've heard it pronounced both "Dee-RAM and Dram" by people who are far more experienced than myself. The point is, being that there IS no correct or incorrect way to pronounce the acronym, an encyclopedic article on the technology shouldn't attempt to endorse one over the other. Furthermore, capitalizing all the letters of an acronym is basic English grammar; we can assume that our readers are familiar with English grammar.
Please see the links below demonstrating widespread incorrect capitalization.
I gave a brief reason as to why I removed the text in my edit summary, and am glad to explain it further here upon request. That is collaborative editing, not "never remove text that authors add without a lengthy explanation." Most of the time a lengthy explanation isn't necessary, and omitting it where I don't feel its needed saves me valuable time. -- uberpenguin 18:04, 24 November 2005 (UTC)
Incidentally, Merriam-Webster gives BOTH pronounciations for the acronym; suggesting that either is correct by accepted English standards. -- uberpenguin 18:07, 24 November 2005 (UTC)
And this wouldn't be the first time that Merriam-Websters is totally out of touch on technical language. It's old media, remember. See below for some other dictionaries' point of view.
It's very silly to simply discount a respected dictionary because it doesn't back up your point of view... Fine, if you want to throw out M-W, here's another that gives "Dram" as the pronunciation:
Actually, you are mistaken, that page does not give "Dram" as a pronunciation for memory, only "Dee Ram".
Uberpenguin, As I already said, I think that in removing an imperfect but useful section based purely on your own opinion and experience you are making a mistake, and I strongly suggest you allow other people to offer their opinions on this. Try to consider your attitude to this and examine your own statements above, particularly 'saves me valuable time' - why is saving your time so much more important than my time or any other Wikipedia user's time? Why is your opinion more important? But I suspect I'm wasting my 'valuable time' here, so I'll just provide some links and leave your page to you.
I thought I made it clear that the primary reason I removed the section was because it is irrelevant. The fact that the material added is highly debatable is an aside. Dictionaries comment on correct pronunciation and spelling, not encyclopedias. -- uberpenguin 14:31, 25 November 2005 (UTC)
In my opinion, the fact that this matter is literally 'debatable' makes it worthy of inclusion in an encyclopedia.
And I reiterate, this is not a page on correct pronunciation in the English language. The pronounciation is hardly debatable since there is no grounds on which to argue one pronunciation over the other; it's totally a matter of preference. -- uberpenguin 13:28, 30 November 2005 (UTC)
References demonstrating 'Dee Ram' is generally accepted as the 'correct' pronunciation of DRAM:
References demonstrating non-standard use of capital letters when writing the acronym DRAM (e.g. Dram, DRam). BTW, pronouncing it incorrectly (as the single word: 'Dram') tends to lead people into using the incorrect capitalization.

I won't reinstate the section myself, but the text is below in case anyone wants to use it as the basis for further work.

Spelling and pronunciation

DRAM is correctly pronounced 'Dee Ram'. The word should always be spelt in all caps. Both spelling and pronunciation reflect the fact that the word is a partial acronym (or perhaps more correctly, an initialism), which was derived from the older acronym 'RAM' by the addition of the prefix 'D' to distinguish it from SRAM, VRAM and so on.

For other words derived from RAM, a more general pronuciation rule is that 'RAM' is always pronounced as single word, while any letters preceeding it are spelled out. So, for example, DDR SDRAM is pronounced 'Dee Dee Ar Es Dee Ram'

Both 'Dram' and 'DRam' are incorrect spellings, but are occasionally used by writers or editors unfamiliar with the term. This error, combined with mispronunciation, sometimes results in DRAM-related newspaper headlines that attempt punning references to whisky.

Thats a shame, to take this out of the page, I was just looking for a pronunciation guide to show people. So many people are getting this wrong. 'dram', you gotta be joking... Paulo

Spelling and Pronunciation[edit]

what about D-RAM, SD-RAM, DDR SD-RAM? probably a hint for correct spelling, writing out the speeling does not occur very much... I do not know if it is correct to suggest it here as "the" scientific correct spelling (however it looks like to me). alex 12:41, 1 May 2006 (UTC)

Agreed. This article shouldn't be a lecture on how to pronounce the English language, nor should it advocate what amount to various colloquial pronunciations. There really is not an "official pronunciation" because nobody officially owns the term DRAM. I brought this up awhile ago on this talk page, removed the section, and somehow it was again added. I've removed it again pending a good reason why it should stay. -- uberpenguin @ 2006-05-01 22:28Z

Anybody know read and write speeds for SRAM, DRAM, FPM, EDO, SDRAM, DDR SDRAM, DIRECT RAMBUS, AND VRAM? Cheers.

External references[edit]

A reference was just added: Links to free Interent resources and reliability data ( This seems inappropriate:

  • the resource is not free. It is free-of-charge for non-commercial use.
  • it does not appear to be relevant: it is about soft errors in ASICs and FPGAs (not DRAM)
  • the link description is neither precise nor accurate

For these reasons, I will remove the link. DHR 19:11, 4 May 2006 (UTC)

Removed eDRAM section[edit]

I removed this section for a number of reasons. First, historically EDRAM means something very different; usually hybrid chips that use SRAM cells to help decrease the latency of an SDRAM module. Second, after several minutes of searching online I can only find a few references to NEC's eDRAM, all of which lack details and have popped up within the last few days. The articles that do contain details describe eDRAM as a separate LSI chip to be used as high bandwidth video-related RAM (presumably for texture storage), not integrated with the XBox 2's PPC970-derivative CPU (which is designed by IBM anyway). [3] Furthermore, the Playstation 2 uses Rambus RDRAM, the Game Cube doesn't even use DRAM for the main system memory (uses SRAM), and all that is known about the Playstation 3's memory technology is that it will use the Rambus interface. Please cite some sources for your information, since I cannot find anything to support the section, but can find plenty of counter-examples. -- uberpenguin 00:50, 2005 May 3 (UTC)

This is a really good article, but in the leader index it includes:

  • fast page mode DRAM
  • video DRAM or VRAM
  • extended data out (EDO) DRAM

Although FPM memory and VRAM are discussed, there's nothing about EDO memory. Can't someone put something useful in here?
Craig Landes 5/03

Done it, though it's not a lot. Hope it is useful anyway.

The FPM description states that there's an address counter for accessing successive locations in a column. I think this is wrong. I need to check this, but I believe that there's no such counter and a new column address needs to be provided each time.

There was a type called nibble mode DRAM that allowed clocking out 4 successive locations without giving a new column address, so it contained a crippled 2-bit counter, but this did not prove very useful and was abandoned. 2003-11-22

explaination of 5-1-1-1[edit]

Could someone write up an explaination of what the above means? Or at least a link to an explaination...

This link shows the explaination, but in french.

At url article says "CAS latency, RAS-to-CAS delay, RAS precharge, and active-to-precharge delay" However a certain user here will not let you know. It because it was deemed too "fairly PC-centric terminology... datasheets for (S)DRAM modules usually have large tables of timing information". So anyone with a suggestion how to include this information ..? I DO find those numbers cryptic without explanation. Electron9 15:58, 23 October 2006 (UTC)

This is the wrong article for that. As I said, it's pretty PC-centric. The information should probably go into an SDRAM article (since we've split up the *DRAM articles along major lines). Regardless, I don't see how acronym-dropping makes the numbers any less cryptic; you're just substituting jargon with more jargon. Heck, I've worked with embedded computer design a good bit and I don't even know what those timings are without looking them up and preferably looking at a timing diagram. I don't find the explanation enlightening at all without some sort of diagram. -- mattb @ 2006-10-23T18:18Z

I find these 'x86-hyphenated-dram-timeing-numbers' (better word?) already at the mention of 'Burst EDO (BEDO) DRAM' so it's not only valid for SDRAM. I don't know where this number mess was introduced on the timeline but I would like to know :). Anyway I understand the mess with CAS/RAS/precharge etc.. And still found the explenation of timeing numbers to be an 'aha!' experience. Not even most BIOS manuals explain it.. (nor explaining other cryptic settings). This information is useful to determine say if a certain dram is alright for an project or not. Not knowing means not being able to evaluate. So where do you suggest this piece of information is added ..?

Another thought, Timeline for introduction of various memory technologies would be good. Like the ones for Incandescent light bulb
Electron9 19:22, 23 October 2006 (UTC)

"Useful to determine if a certain DRAM is alright for a project"? If you're doing a project that requires you to know DRAM timings, you will be looking at a full timing diagram and datasheet, not these abbreviated numbers. Inasmuch as I can tell, the numbers are only useful to the sort of people who buy the most expensive and cutting edge PC components possible and want to make sure they are getting the lowest latency DRAM. I still don't believe this information belongs in this article unless we include it in a larger section about timing measurements. This I'm also skeptical about since it has limited value for the average reader and is (in my opinion) unencyclopedic for a general overview article about DRAM. -- mattb @ 2006-10-23T19:54Z


There should be an explanation of the speed measurments of RAM and the range of speeds of the different types.

No DDR2 and DDR3 is now.

Could someone explain SD-PC133R?[edit]

I've encountered a type of ram described as: SDRAM forPC133 w ECC & Register. Seems to be a distinct type it certainly is much taller then the standard types of SDram. Maybe someone could explain this here, i can't find much information on. Pulled from an older compaq server.

VRAM is almost obsolete?[edit]

I may be wrong, but isn't VRAM standard for video or graphics cards?

Synchronous Dynamic RAM (SDRAM)[edit]

Just something I noticed in this section, but could apply to the whole article. Isn't the object of an encyclopedia to provide information in a way that a diverse audience can understand and appreciate, rather than making it's author sound impressive? Using archaic language, like the word "whilst," doesn't exactly serve this effort. In that case, why not just use "while?" Strunk and White are behind me on this, I'm sure.

Probably the author was British. Mirror Vax 03:26, 10 January 2006 (UTC)

RDRAM needs to be combined with its orphaned separate entry[edit] --Swaaye 22:58, 17 January 2006 (UTC)

Introduction's Transistor Count[edit]

I think that the introductory paragraph's transistor count is wrong. The paragraph claims, "only one transistor and a capacitor are required per bit." HOWEVER, this would imply that over one *billion* transistors are necessary for a gigabyte stick of ram. (yes, I recognize that one billion is low, but in the interest of simplicity and since it makes the point, please just play along) That is a huge transistor count insanely greater than is available on a chip using current technology. I suspect that the actual transistor count should be in the neighborhood of, say, log2(# of capacitors). Can someone verify/correct this please?

Note that there are around 8 chips, maybe 16, on a stick of RAM, so the density isn't quite that high per chip. Even so, memory can be much much denser than say, processor logic. Now, I haven't studied memory design in detail, but I don't see how you could get away with less than one transistor per capacitor, since the transistor is the switch that determines access to the storage element. You can ditch the separate capacitor, which they do in Flash, but you can't get rid of the switch. - mako 20:36, 27 January 2006 (UTC)
Actually, you'd need about 8.6 billion (109) FETs for a gigabyte stick of RAM, using the common connotation of "gigabyte" in this context (230 bytes). This isn't incorrect, and the IC density is high, but not incredible. Modern manufacturing methods "trench" each bit's capacitor under the transistor, so the capacitor portion of the cell doesn't contribute much to the die area. As mako pointed out, traditional digital design methods don't allow less than one transistor per storage bit. -- uberpenguin 22:57, 29 January 2006 (UTC)
From what i can gather the simple structure of dram and the fact that they can turn off bad blocks (they always put more blocks than strictly needed on a ram chip for this reason) means they can pack the transistors far closer than in more active IC designs. Plugwash 00:14, 25 February 2006 (UTC)

Yes, DRAM uses one capacitor and one transistor for each bit. (Plus a few more transistors for row select and column decode). A billion bits requires a billion transistors and a billion capacitors. (I've heard that MLC flash memory tries to store 2 bits on 1 capacitor with 1 transistor, but my understanding is that it was not commercially successful).

Plugwash is correct -- since RAM has exactly the same structure repeated millions (by Moore's Law, soon billions) of times over the majority of the chip, DRAM has a higher transistor density than other chips (except for flash memory) for 3 reasons:

  • designers can spend weeks figuring out ways to pack RAM bits together as tightly as possible. With other kinds of chips (CPUs, DSPs, FPGAs, etc.), where you have transistors connected in thousands of different ways, designers have much less time to figure out how to pack them together.
  • The irregular connections of so-called "random logic" causes lower transistor densities, even if you had enough time to come up with the maximum-density layout. (The wiring takes up more area than the transistors).
  • Because RAM has thousands of identical rows, the small amount of "extra" space used by designing 1 or more "extra" spare rows is more than compensated by the fact that the entire design can be shrunk much smaller. Even though the size is so small that the defect rate goes way up, most defects can be corrected by "locking out" the bad row and programming the spare row to replace it.

-- 15:37, 8 May 2007 (UTC)

Multiple gigabit densities have been announced, but I see no references for the hundreds per chip the introduction claims. A 16 GB DIMM will have over a hundred gigabits, but that's split between several chips on the module. — Preceding unsigned comment added by (talk) 20:18, 29 May 2011 (UTC)

spare cells[edit]

  • Because RAM has thousands of identical rows, the small amount of "extra" space used by designing 1 or more "extra" spare rows is more than compensated by the fact that the entire design can be shrunk much smaller. Even though the size is so small that the defect rate goes way up, most defects can be corrected by "locking out" the bad row and programming the spare row to replace it.

This sounds like something that needs to go into the article -- perhaps in the "Errors and error correction" section.

This seems like a reasonable thing to do, but how many DRAM manufacturers actually do this? References?

Sometimes I hear people explaining this using the term "bad blocks", which makes me think they are confusing DRAM with Flash memory (or perhaps hard drives). -- 15:37, 8 May 2007 (UTC)

Memory repair schemes are fundamental to the cost structure of DRAMs and all DRAM manufacturers utilize memory repair schemes. As the wafers leave the fab (but before they are cut into individual die), each chip is tested and a map of defective bits is generated. The memory design includes spare rows and columns along the edges of the memory blocks. Using a pin-point laser, a series of fuse links is blown to disable the bad rows or columns and replacem them with spares. If someone could go to an engineering library, any textbook on memory design will cover this & we'd have a proper reference. The memory manufacturers, while not exactly hiding the information, are understandably reluctant to focus on the fact that many (most?) of the DRAMs they are selling have been "patched up". (talk) 07:50, 9 January 2009 (UTC)

Removed double image.[edit]

I removed one of the images which was a double-- there was two on the top of the page. 20:03, 3 April 2006 (UTC)

They're different -- one's a read, and one's a write. - mako 20:46, 3 April 2006 (UTC)

Suggestion to refer also the RLDRAM[edit]

I suggest to mention and refer RLDRAM similary to the mentioning of the other DRAM types. User:elcha, 08 October 2006

Looks like RLDRAM is just a low(-er) latency SDRAM. It seems to me that it's more a brand name than an actual novel DRAM arrangement. -- mattb @ 2006-10-08T16:08Z

Define 'Precharge'[edit]

Can someone define 'precharge' before or along with its first use and perhaps give the origin of the term. In the introduction there is the phrase "row values must be restored to the capacitors" which sounds like the old 'write back' for core memories.

Is there a conventional name for state of a bank just after having been 'precharged'?

I suspect now that I know the answer. After a read or write the sense lines carry a charge which could pollute a subsequent read operation. This left over charge must be removed to detect the small signal that the sense line must detect on a read. I think steps to do this are called the ‘precharge’.

Upon write the sense lines carry the charge to the bit cell. Even a read operation includes a write-back phase. The data written will remain on the sense lines as noise unless much time passes or some sort of precharge is performed.

If someone knows this is wrong I hope they will delete it or improve it.

(This article has been very useful to me.)

NormHardy 05:29, 15 July 2007 (UTC)

The way the memory cell and bitlines work, when the pass transistor is opened up the charge from the cell capacitor spills out onto the bitline. It must be sensed as a low or high voltage by the sense amplifier at the distant end of the bitline. Of course the bitline node has a huge capacitance compared with the tiny cell capacitance, so there's not much of a signal. A trick that was developed in the early days, and is still used, is to make use of a dummy reference bitline that is not connected to an opened cell. The sense amp compares the voltage on the two bitlines (differential signaling) which is much more sensitive than trying to measure the absolute voltage on a single bitline. In any case, just prior to the cell pass transistor being opened, the two bitlines are briefly clamped together so as to have an identical voltage. This is precharge and it greatly increases the sensitivity of the detection scheme. (talk) 08:04, 9 January 2009 (UTC)

NEC VCM (Virtual Channel Memory)[edit]

I would like to see some information on how this type of RAM works. I understand it involves some sort of buffer, making precharge and refresh cycles obsolete. Maybe it should also be said why it never caught on, although major chipset maufacturers (e.g. VIA Technologies) included support in their chipsets even before these modules hit the shelves. —The preceding unsigned comment was added by (talk)

Yes, I have also heard about this memory and I know there is an article about it in a IT magazine I used to collect but I can't find that number. —Preceding unsigned comment added by Mariushm (talkcontribs) 13:01, 10 November 2007 (UTC)

Proposed merge[edit]

I think the section from Random access memory on DRAM packaging should be in the Dynamic random access memory article instead. Please add comments below. HeirloomGardener 11:56, 23 May 2007 (UTC)

There seems little point in retaining the information there if there's an entire article devoted to DRAM. A quick synopsis and link to the DRAM article from RAM would suffice. Especially the lists add little to the RAM article. Steevm 04:40, 11 July 2007 (UTC)

VRAM in IBM PC/RT graphics system[edit]

Under section "Video DRAM (VRAM)":

"high resolution graphics monitor introduced in 1986 by IBM with the PC/RT system"

Was it the graphics adapter or monitor that contained the vram ..?
I noticed the edit on another sentence replaced 'monitor' with 'display adapter'.
Electron9 23:13, 5 August 2007 (UTC)


Found this memory being mentioned in a IT magazine edited in October 1997.

According to the article, a consortium called Sync-Link or SLDRAM (made by Mitsubishi, NEC and Siemens) was established to compete against Intel-Rambus.

Memory chips were supposed to appear the next year (1998) and were following the concepts Rambus. As Direct RDRAM, SLDRAM would have worked on a 16bit bus at a high frequency. The consortium estimated the performances would be comparable to SDRAM while some members estimated 3.2GB/s by interconnecting several independent SLDRAM subsystems.

The article mentions two websites as further reference SciZZL and SLDRAM, the later is currently just a parked domain. I don't know if the memory chips were in the end actually made.

If anyone is interested in a copy of the article, I can scan the page from the magazine and publish it somewhere.

Stacked RAM[edit]

The Dynamic random access memory article states, without any reference:

" * Stacked RAM chips use two RAM wafers that are stacked on top of each other. This allows large module (like a 512mb or 1Gig SO-DIMM) to be manufactured using cheaper low density wafers. Stacked chip modules draw more power."

I am not aware of any stacked RAM chips that are currently made by stacking wafers on top of each other. If no reference can be found, I suggest that this should be removed.

On the other hand, there are stacked RAM modules that are currently made by packaging separate RAM chips in a vertical stack. If this is what the author meant, the entry should be corrected.

Wafer-stacked RAM chips are now on the drawing board, but they are predicted to consume less power, not more. See discussion here: [4]

Gretchenpatti 23:18, 15 November 2007 (UTC)

Having received no opposition, I made the correction. --Gretchenpatti (talk) 16:16, 23 January 2008 (UTC).... this also the first unoque

VRAM section[edit]

I've done some significant cleanup and pruning of the VRAM section. I do not believe some of the past editors fully understand the difference between video RAM (which may or may not be VRAM) and actual use of dual-ported VRAM. Although somewhat common for higher-end equipment at one point, it was never present on "most systems", and really had no impact on the consumer uptake of personal computers as suggested.

Video RAM provided the necessary ingredient to advance graphics from the poor quality, low resolution single green color screens of computers at that time, to the high resolution, multicolor displays of SVGA and above. It was a key element in bringing personal computers to the masses, allowing windows, spreadsheets, and high quality graphics to be introduced at very affordable prices.

This set the new standard for graphics, and was quickly adopted as an industry standard. For over 10 years. VRAM was used in nearly every display adapter sold, over the entire range of systems. The number of megabytes of VRAM was advertised as a feature in most PC sales in the late 1980s and early to mid 1990s.

I do not think there is really any usable content in either of those paragraphs, They are both not only inaccurate, but read like cheerleading.

The section still needs work, but I think it's better than it was.

Oregonerik (talk) 02:13, 4 February 2008 (UTC)

But what does it do?[edit]

This is all wonderful, but what does it do? (talk) 07:30, 4 February 2008 (UTC)

As a type of RAM, it stores information. Namely: it is a fast-access, volatile data storage.--portugal (talk) 10:10, 22 February 2008 (UTC)


How is the general feeling about the relevance of the addition of a section on security, based on this study?--portugal (talk) 10:12, 22 February 2008 (UTC)

Done. Electron9 (talk) 11:22, 22 February 2008 (UTC)


WRAM (Window Random Access Memory) redirects here, but there's no explanation of what it means. Can someone in the know help? Gypsum Fantastic (talk) 21:54, 6 June 2008 (UTC)

Loopy Link[edit]

If you look at Dynamic random access memory#Errors and error correction, it points to ECC_memory#Error-correcting_memory as the main article, which itself points back to Dynamic random access memory#Errors and error correction as the main article. =P (This has been crossposted to Talk:Error detection and correction and Talk:Dynamic random access memory) Copysan (talk) 04:05, 2 September 2008 (UTC)

Do we have enough information to make a stand-alone ECC_memory article? -- (talk) 20:41, 7 March 2009 (UTC)


Overall, this is an excellent article that is highly useful and understandable. Thanks to all those who have contributed. Todd (talk) 22:19, 29 November 2008 (UTC)

Out of Date template[edit]

I added the Out of Date template to the Memory Timing section because the memory numbers quoted in the second half of the section are badly out of date. PC3200 hasn't really been standard for a couple of years; the current standard is probably PC6400 (800mhz) for DDR2 ram. The typical latencies for this are around 5-5-5-16. Premium typically is more like 4-4-4-12. DDR3, though, is completely different. Its latencies are typically higher, but so are its clock speeds. I don't have any sources for this info, though, so I didn't want to change it. —Preceding unsigned comment added by (talk) 08:08, 29 December 2008 (UTC)

Feeling superior[edit]

"superior to SDRAM in some ways" Such as? Also, I recall a 64Mb DRAM proposal, but don't see it mentioned. (Or did I miss it...?) TREKphiler hit me ♠ 15:56, 31 October 2009 (UTC)

DRAM operation figures[edit]

First wikipedia post, so please excuse any poor etiquette.

It seems the read and write figures would be more easily understood if they were split into three, Row Activate, Col Read, and Col Write. The Col Write figure could then show the CAS mux position pointed in the correct direction to charge the cells.

What format are the figures in? I could certainly help with the dimple edits if needed.

Thanks, Jeff Stuecheli — Preceding unsigned comment added by Jeffastue (talkcontribs) 02:15, 8 April 2011 (UTC)


I made some changes to the article. I tried to keep consistency when naming and reorganized some sections while renaming some of them. Here are a few comments:

1. "Common DRAM modules" under "DRAM Packaging" seems somewhat redundant with the rest of the section. I would merge it or eliminate it.

2. "Given support of CAS-before-RAS refresh, it is possible to deassert /RAS while holding /CAS low to maintain data output."

I believe CAS and RAS should be marked as active-low signals in the first part of the sentence.

3. I moved two subsections under "Asynchronous DRAM". The two sections do not describe DRAM versions. They appeared to describe issues related to asynchronous memory.

4. The "Security" section does not explain what it's meant by "attack". Some explanation should be added.

ICE77 (talk) 22:28, 7 July 2011 (UTC)


This text

In 1964 Arnold Farber and Eugene Schlig, working for IBM, created a hard-wired memory cell, using a transistor gate and tunnel diode latch. They replaced the latch with two transistors and two resistors, a configuration which became known as the Farber-Schlig cell. In 1965 Benjamin Agusta and his team at IBM created a 16-bit silicon memory chip based on the Farber-Schlig cell, with 80 transistors, 64 resistors, and four diodes.

describes static RAM built on latches. It should be moved to appropriate article.Avivanov76 (talk) 07:59, 26 June 2013 (UTC)

Hybrid Memory Cube and High Bandwidth Memory[edit]

Hi. There are two proposed standards for 3D-packages DRAM memory with TSV connections: Hybrid Memory Cube (Micron, Samsung) and High Bandwidth Memory (AMD, Hynix, Nvidia; JEDEC JESD235). Where can we list them in this article? `a5b (talk) 13:33, 8 December 2014 (UTC)

Hello! To me, putting them as links into the "See also" section would fit nicely. — Dsimic (talk | contribs) 13:47, 8 December 2014 (UTC)
Both articles, HMC and HBM links to section 'General DRAM formats' as "* [[Dynamic random-access memory#General DRAM formats|Stacked DRAM]]". Can we have some in-text section (or in template:DRAM) about future directions of DRAM memory, especially stacked and w/TSV? (possible refs: ...Roadmap of DRAM from three major manufactures 2013; Memory: A System Optimization Challenge Cadence, 2013; The Future of DRAM, Synopsis 2014). PS Where is information about 2.5D stack integration of memory dies with wire bonding? `a5b (talk) 14:12, 8 December 2014 (UTC)
Sorry, I might have gotten something wrong, but both Hybrid Memory Cube and High Bandwidth Memory are standalone articles? — Dsimic (talk | contribs) 14:17, 8 December 2014 (UTC)
Yes. Do you think that they should have no separate articles? HMC has 3k visitors per month. `a5b (talk) 14:23, 8 December 2014 (UTC) (they are different standards based on similar ideas: TSV for true 3D stacking of memory dies. HMC has integrated memory controller on bottom die; and HMB has no controller on its bottom logic die).
I'd leave them as-is, there seem to be no reasons for merging them. — Dsimic (talk | contribs) 14:26, 8 December 2014 (UTC)
Regarding your edit, yes, that place would be good to briefly mention them, with a simple "Stacked RAM module technologies include Hybrid Memory Cube and High Bandwidth Memory." addition to the last bullet point. Hope you agree. — Dsimic (talk | contribs) 14:59, 8 December 2014 (UTC)


The article claims DRAM is packaged in creosol, which is a liquid disinfectant. There's cresol novolac epoxy resin which is made from cresol (a totally different chemical with one 'o' instead of two), which is probably what is intended.Billgordon1099 (talk) 16:37, 22 December 2014 (UTC)

Nice catch, thanks! Went ahead and cleaned it up further. — Dsimic (talk | contribs) 04:14, 25 December 2014 (UTC)


I was thinking of re-organizing the article something along these lines. Any comments?

Asynchronous DRAM *

   FPM RAM *
   EDO RAM *
       Burst EDO DRAM (BEDO DRAM)*

Synchronous DRAM *

   Reduced Latency DRAM (RLDRAM) (SDRAM)

Graphics DRAM

   VRAM *
   WRAM *
   MDRAM *

Rambus DRAM

   RDRAM (Direct Rambus DRAM) *

1T (capacitorless) DRAM

       1T-SRAM (Pseudostatic RAM/PSRAM)

--Jules (Mrjulesd) 19:04, 12 December 2015 (UTC)

Content: theory vs. instances[edit]

This article excessively focuses on the types of DRAM IC architectures and interfaces that have appeared in the 45 or so years DRAM ICs have existed, whilst mostly ignoring the actual DRAM technology itself. What do I mean by "DRAM technology"? I mean the aspects of DRAM pertaining to the basic circuits (cell structure, access transistor, capacitor, bitline, and sense amplifier design), the universal low-level array organization of DRAMs, their fabrication processes (which have greatly diverged from those used to fabricate logic-heavy ICs), the various motivations and resultant debates over alternative designs, and most importantly, technological trends. While this article may not appear to be particularly flawed to most, I think it reveals a core weakness of Wikipedia regarding theory vs. instances (products)—there is nearly always a strong bias towards describing products as individual artifacts whilst omitting the theory upon which they are based on or any great overview which ties disjoint topics together. Two articles should be split from this one; to provide overviews of DRAM architectures (eg. FPM, EDO, SDRAM, DDRx, etc.); and DRAM packaging as it appears to end-users (nicely packaged in a memory module). This is to prevent this article from becoming too redundant to those which provide overviews of SDRAM and DDR SDRAM; and crowded to provide adequete coverage of the omitted topics. Or am I the only one seeing the above shortcomings in this article? AZ1199 (talk) 06:55, 4 January 2016 (UTC)

If you really want to expand on the technological aspects of DRAM, how about creating a new article: Dynamic random-access memory technologies? You could copy text from this article and then expand. Note that the referencing of the present section is inadequate. I think to split DRAM architectures and DRAM packaging would be a mistakes, as they are the two most important subjects for the majority of readers.
I hardly think this article is redundant to say DDR articles, as its purpose is to provide an overview to all of DRAM topics, with DDR articles providing in depth descriptions of these technologies.--Jules (Mrjulesd) 15:16, 4 January 2016 (UTC)

Memory timing[edit]

The Dynamic random-access memory#Memory timing section has a number of serious problems:

  • It presents a subset of the full set of timing parameters for a type of asynchronous DRAM as its introduction to timing parameters.
  • It confuses the timing parameters present on SDRAM and DDRn DIMMs with the set of timing parameters of the DRAM IC.
  • It presents a section detailing a subset of abbreviations for asynchronous DRAM as if these were independent of the DRAM interface.
  • It presents timing parameters for asynchronous DRAM by their by their informal, non-standardized names, which have varied over time and manufacturer instead of the canonical, formally defined JEDEC names.

My opinion on this section is that it shouldn't exist. The timing parameters are specific to the DRAM IC interface and should be described at whichever section or article where the DRAM IC interface is primarily covered. The set of timing parameters that are applicable to DRAM core timing should be covered in the sections in this article that introduces and explains the hardware that the timing parameters measures. Would there be any objections to section being deemed problematic or the proposed solution? AZ1199 (talk) 06:58, 23 January 2016 (UTC)

Coverage of memory modules tangential to DRAM?[edit]

Most of Dynamic random-access memory#Packaging section is about memory modules, and this coverage dwarfs what has been written about DRAM integrated circuit packaging. Although many DRAM standards also define memory modules, I believe this kind of coverage is undue for this article. Such coverage better belongs at memory module. DRAM is fundamentally an integrated circuit product. As such, whether or not DRAM integrated circuits are mounted on modules for use by end-users is not particularly relevant. At most, little more should be said about memory modules other than this is how the average end-user would encounter DRAM; and that from the perspective of the DRAM interface, memory modules introduce its unique transmission and signal issues. On the other hand, the kind of integrated circuit packaging DRAMs use is of far greater relevance to the topic (DRAM). The kind of DRAM integrated circuit packaging has a multitude of effects; it relates to the cost of DRAM, the physical size limit of the die, routing of on-die signals (which affects latency and die layout), ability to dissipate power, signal integrity of DRAM transmitters and receivers, the PCB area occupied, the design of the DRAM protocol, among other things. I propose the content on memory modules be merged into to memory module. AZ1199 (talk) 04:58, 25 January 2016 (UTC)

External links modified[edit]

Hello fellow Wikipedians,

I have just added archive links to one external link on Dynamic random-access memory. Please take a moment to review my edit. You may add {{cbignore}} after the link to keep me from modifying it, if I keep adding bad data, but formatting bugs should be reported instead. Alternatively, you can add {{nobots|deny=InternetArchiveBot}} to keep me off the page altogether, but should be used as a last resort. I made the following changes:

When you have finished reviewing my changes, please set the checked parameter below to true or failed to let others know (documentation at {{Sourcecheck}}).

You may set the |checked=, on this template, to true or failed to let other editors know you reviewed the change. If you find any errors, please use the tools below to fix them or call an editor by setting |needhelp= to your help request.

  • If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
  • If you found an error with any archives or the URLs themselves, you can fix them with this tool.

If you are unable to use these tools, you may set |needhelp=<your help request> on this template to request help from an experienced user. Please include details about your problem, to help other editors.

Cheers.—cyberbot IITalk to my owner:Online 09:52, 30 March 2016 (UTC)

Kenner reference missing[edit]

The reference "Kenner", which appears to be the primary reference for this article, is not listed anywhere in the article. It's only referred to by the abbreviated name "Kenner". Perhaps there was an earlier citation, which provided the full reference, that was deleted? Can someone who has the full reference please add it? (talk) 05:42, 28 September 2016 (UTC)

External links modified[edit]

Hello fellow Wikipedians,

I have just modified one external link on Dynamic random-access memory. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

You may set the |checked=, on this template, to true or failed to let other editors know you reviewed the change. If you find any errors, please use the tools below to fix them or call an editor by setting |needhelp= to your help request.

  • If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
  • If you found an error with any archives or the URLs themselves, you can fix them with this tool.

If you are unable to use these tools, you may set |needhelp=<your help request> on this template to request help from an experienced user. Please include details about your problem, to help other editors.

Cheers.—InternetArchiveBot (Report bug) 07:15, 15 September 2017 (UTC)