Talk:Dual-ported video RAM

Article name

This article was created at VRAM, moved to Video RAM, and then merged into Dynamic random access memory. I have relocated the content to its original location, VRAM. The DRAM article is too long as it is, and "VRAM" passes the usage test in MOS:ABBR, that is, references to this technology primarily appear in its fully abbreviated form. There is also some ambiguity with respect to Video RAM and Video memory, while "VRAM" is distinctly about this particular implementation of the technology. Ham Pastrami (talk) 06:33, 9 April 2009 (UTC)

The article should be called "Dual-ported VRAM" or something like "Special type of VRAM", but it should be really-really not called "VRAM".

VRAM, Video RAM, or Video Memory is just referring to memory that contains video data - which might be SRAM, DRAM, or (rarely) the special dual-ported stuff that's currently described as "VRAM".

Just look at https://en.wikipedia.org/wiki/Special:WhatLinksHere/VRAM - there are lots of articles referring to "VRAM", and I would assume that most of them are referring to general VRAM, not to dual-ported VRAM. For example, hardware from the 80ties and early 90ties is almost certainly not using using dual-ported VRAM, I am quite familar with schematics from that era, and I've never seen dual-ported chips being used in that schematics. For newer hardware: The Sony Playstation and PSone consoles are usually using normal DRAM as video memory (except very early Playstation models with PU-7 mainboards: These were actually using dual-ported DRAM, but Sony dropped using that technology on later mainboard revisions).

Concerning the official name(s) for dual-ported VRAM, I've checked some sources,

• http://www.google.com/patents/US4541075 - the patent is just calling it A semiconductor random access memory is provided having a second asynchronous input/output port. The patent doesn't use the terms "VRAM" or "Video" at all. Instead, it's using the term "display" in several places - but "display" is just an example application (explicitely stated in the sentence: Although the apparatus has been described with respect to mapped raster scan display devices, it is clear that the utility provided by the second I/O port may be used for other connected peripheral devices to enhance data flow to and from the memory.)
• http://pdf.datasheetarchive.com/indexerfiles/Datasheet-048/DSA0080861.pdf - this NEC datasheet is using the name Dual Port Graphics Buffer. It doesn't use the terms "VRAM" or "Video" at all.
• http://pdf.datasheetarchive.com/datasheetsmain/Datasheets-110/DSAP002477.pdf - this Samsung datasheet is equivalent (pin-compatible) to above NEC datasheet. Samsung is actually referring to their chip as a Video RAM chip, but I am sure that they are just meaning that their chip is intended or optimized for being used as Video RAM - they can't be possibly claiming that all past and future computers with video memory are having to use that special type of dual-ported RAM.

Concerning an artice with name "VRAM" - that should describe video memory in general (giving a quick run-down on terms like bitmap, framebuffer, tiles, bg maps, palettes, textures, color depths, etc, and mentioning normal DRAM and SRAM commonly being used as normal video memory, and dual-ported RAM being mentioned only as a special variant). I am surprised that wikipedia doesn't seem to have any such article yet - at least I can't find it anywhere - Video RAM, and Video memory are all redirected to VRAM (which is fine, aside from the big issue that VRAM should be described as general video memory, not as dual-ported memory). And Display memory is redirected to Framebuffer (which is rather nonsense since Display Memory isn't neccessarily referring to bitmaps).s

2OO.3OO.2OO.3OO (talk) 16:18, 5 January 2016 (UTC)

Video RAM shouldn't redirect here but should list all the different variants of RAM used for video. VRAM is a specific type of RAM, so it's not the same at all. --Zac67 (talk) 19:48, 5 January 2016 (UTC)

VRAM (the special dual-ported type of DRAM) was certainly not "rare" or unused in 1980s or 1990s hardware; and "Video RAM" ("VRAM") was exactly what it was called. I'm not sure what the patents and two data sheets prove. It is the nature of patents to avoid referring to technologies by their noun. Instead, technologies should always be described as ideas. That's how patents are written. As for the two data sheets, "VRAM" could have been a trademark of IBM at they time. NEC and Samsung may not have wished to or been able to call their equivalent products "VRAM". An example of such practice would be "Tri-state buffer". National Semiconductor (if my memory serves) trademarked this term. Everyone else had to call the equivalent functionality a "three-state buffer". Wikipedia is an encyclopedia that should name articles after what they are called, and this kind of DRAM was called VRAM.

In support of my remarks: In Yasunao Katayama's "Trends in Semiconductor Memories", a survey article, in IEEE Micro (Nov./Dec. 1993, pp. 10–17), he describes a class of DRAM devices that have added additional functionality to the basic DRAM core called "high-functionality devices". He describes this class: "Instead of making the DRAM faster, designers responded to the needs of special applications—particularly graphics—and followed a different trend in DRAM development." This proves that the needs of graphics applications were served with specialty DRAMs, contrary the claim that they were rare and unused. As further support to the contrary, the article continues: "The first commercially successful high-functionality DRAM product was VRAM (video RAM). Introduced in the mid-1980s, it realized concurrent dual-port access by multiplexing the massively parallel data in the DRAM array before the data were demultiplexed in the column data path." Katayama continues with a brief description of VRAM architectural differences, features, characteristics, and impact (as is proper for a survey article), and ends with "VRAM is JEDEC standard."

Another source is Kiyoo Itoh's VLSI Memory Chip Design, which states on page 340, briefly (since this book is about the latest technologies), "In the past, the dominant form of graphic (sic) memory was video RAM (VRAM), which has independent display and drawing terminals (i.e. a dual port)...". This doesn't contradict Katayama's article.

I think there may be some confusion as to what "dual-port" is referring to. It is not referring to a general-purpose dual-port RAM, it's referring to the fact VRAM has a wide parallel read/write port (like a conventional DRAM) and a serial data out port for outputting to the DAC.

Regarding the article title, I am in full agreement that the present title is ambiguous and the subject of much confusion and mistaken links from other articles. A better title would be "VRAM (JEDEC standard)". The "Video RAM" and "Video memory" redirects should be turned into disambiguation pages, per the comments regarding the confusion that they cause. AZ1199 (talk) 03:09, 6 January 2016 (UTC)

A three-state buffer is something completely different and much simpler. The "VRAM (JEDEC standard)" is the right direction but I'm not sure at all if it was ever standarized by JEDEC or if there was an industry standard predating it. How about "VRAM (memory type)"? --Zac67 (talk) 09:53, 6 January 2016 (UTC)

I never equated three-state logic to memory, let alone VRAM. I only stated that manufacturers do not always call a spade a spade. If a further example is required, a similar situation exists with the I2C bus. "I2C" is a trademark of Phillips (or whatever its semiconductor operations are called now). Some manufacturers that implement the I2C bus as an interface refer to it as a "industry standard serial bus" or some other similar descriptive term. That said, there are manufacturers that do refer to VRAM as VRAM (http://www.ti.com/product/tms55160).

As for whether JEDEC standardized VRAM, it did. Searching their site, JEDEC Committee JC-42 is responsible for VRAM, and it's a part of JEDEC Standard No. 21C (Section 3.10). Interestingly, JEDEC refers to VRAM as "multiport dynamic RAM" and includes Synchronous Graphics DRAM along with those that would be VRAM.

As for whether "VRAM (memory type)" is a better title or not; I've done a quick Google search concerning JEDEC standardization of VRAM, and I've found a few results (snippets in books) that say that the standardization occurred long after VRAM became successful and that later VRAMs exceeded the functionality and performance specified by JEDEC. The former would lend more weight to "VRAM (memory type)", since it is obvious that to characterize VRAM as a JEDEC standard would be giving undue weight. The latter, I am unsure, since manufacturers have offered (and still do) SDRAM and DDRx SDRAM that exceed JEDEC's specifications, but that seems to make no difference on Wikipedia. However, is there potential for "memory type" to be ambiguous? There is already some confusion concerning what VRAM means, as evidenced by its usage in various articles (and I've cleaned up a few). Referring to VRAM as a type of memory could be mistakenly interpreted to mean "type, as determined by use" (that is, any memory used for video), rather than "distinct architecture". AZ1199 (talk) 01:02, 7 January 2016 (UTC)

How about "VRAM (multiport video RAM)" as article name? That would match up with the name used at http://www.ti.com/product/tms55160 and would more or less match JEDEC's "multiport dynamic RAM" (plus pointing out that it's video-related), and it might be a bit more intuitive than "VRAM (memory type)" or "VRAM (JEDEC standard)".

For its popularity, it might be interesting to compile a list of hardware that used VRAM (and that didn't do so), for example:

• MCGA, but not consumer grade EGA/VGA/SVGA (as somebody pointed out in a comment below).
• X68000 (japan only), but not Commodore Amiga (although Commodore planned to use it in unreleased Amiga Ranger Chipset).
• Early Playstation mainboards, but not later Playstation/PSone mainboards (haven't checked how many years it was used here).

Going by that it would appear to be restricted to professional users, and to relative short-lived use in gaming hardware - or was it also used in mass-produced hardware for longer periods? The quoted sentences "designers responded to the needs" and "dominant form of graphic (sic) memory" could be read this or that way (saying that there was a "need" for special "graphic memory" doesn't neccessarily mean that that special memory was widely used, ie. using normal DRAM in video hardware might have been even more dominant than using dedicated "graphic memory").

For video memory in general, I would would recommend describing that in Video memory (aside from Video RAM, that page may also include Video ROM; as used by old Text displays, OSD chips, and NES gaming consoles), and redirecting Video RAM to that page. And maybe also redirecting VRAM to it, or at least adding some "For general Video RAM, see..." link on the VRAM page.2OO.3OO.2OO.3OO (talk) 02:44, 8 January 2016 (UTC)

I agree that "VRAM (multiport video RAM)" would be a much better title than any of the other proposed ones that solves all the problems with the current one. I'm not sure how much could be said about video memory in general, other than that video/graphics hardware may use different types of memory for all sorts purposes, since they aren't discussed as one topic, but always in the context of specific systems. Eg. the monochrome graphics adapter in the original IBM PC used SRAM to hold character and character option codes, and a character ROM to generate the actual pixels on-screen; whilst SGI 3D graphics accelerators of the 1990s had separate display list, working-set, texture, and framebuffer memories. Maybe "video memory" could serve as a disambiguation page, and direct readers to those articles where they are fully discussed in their contexts.

As for whether VRAM was less or more commonly used in video cards than DRAM, the following is a brief summary of the four most relevant results from the first ten results for a search for "VRAM" in full-view magazines on Google Books:

• The July 1993 issue of PC Magazine (https://books.google.com/books?id=eX8w8B-OhIIC&pg=PA168&#v=onepage&q&f=false) benchmarked seven video cards/graphics accelerators. Six of them use VRAM, and only one didn't.
• The 24 June 1994 issue of the same magazine (https://books.google.com/books?id=-Qil3sD8bgEC&pg=PA214#v=onepage&q&f=false) discuses eight graphics accelerators. Six of them use VRAM exclusively, and only two use can either VRAM or DRAM.
• The 25 July 1994 issue of InfoWorld (https://books.google.com/books?id=kDgEAAAAMBAJ&pg=PA81#v=onepage&q&f=false) talks about DRAM vs. VRAM in graphics cards. It does say that one of the most important factors in performance was which type of DRAM the card used. There's some discussion of benchmark tests between more numerous VRAM-based cards and DRAM-based cards.
• The 5 December 1995 issue of PC Magazine (https://books.google.com/books?id=FTbctntiaHgC&pg=RA1-PA178#v=onepage&q&f=false) compared eight models of moderately priced retail PCs. Six of them had VRAM-based video cards; only two had DRAM-based graphics, and they were integrated versions.

It would appear from these results, VRAM was indeed common. AZ1199 (talk) 03:45, 8 January 2016 (UTC)

Image?

The article have an image https://www.google.com/search?q=imagen+de+vram&rlz=1C1CHBD_esSV885SV885&sxsrf=ALeKk00gQ_g2y235Cfteg_2ea2F5Y_WUqQ:1604869157793&tbm=isch&source=iu&ictx=1&fir=4h_jmqThd7TT0M%252C2Uzr2xKYXIt9hM%252C_&vet=1&usg=AI4_-kSX8pPS3TUdRffiFCeE6ID2nHoUzw&sa=X&ved=2ahUKEwipor7A6_PsAhWJjVkKHbCrBmYQ9QF6BAgDEDo#imgrc=4h_jmqThd7TT0M —Preceding unsigned comment added by 201.196.172.14 (talk) 02:51, 23 July 2010 (UTC)

The article is exaggregated

While it is true that VRAM dramatically increased the available bandwidth for displaying video data in the 1980s, it is untrue that Microsoft Windows or other GUIs would have been impossible. It also is untrue that VRAM was exclusively used in the beginning of the 1990s. Indeed, it was used on high-performance graphics cards those days, but not on consumer-grade hardware. On the low-end scale, the IBM PS/2 Model 30 MCGA graphics adapter indeed does use VRAM (at 8 bits bus width), but the very popular EGA and VGA graphics cards use classic DRAM (with 32 bits bus width). Most consumer-grade SVGA card controllers (like the OAK OTI-077, the Tseng Labs ET4000, Trident TVGA TVGA8900/TVGA9000 and Cirrus-Logic GD520/522x chips) also use DRAM. — Preceding unsigned comment added by 82.113.121.203 (talk) 12:16, 11 June 2011 (UTC)

Merger Proposal

I propose this merger because the dual ported ram article is barely a stub, and this, the vram article actually describes what dual ported ram is and what it does.TeeTylerToe (talk) 03:54, 2 July 2013 (UTC)

Oppose – VRAM is a special kind of dual-ported RAM but the applications are/were quite different (dual-ported RAM is usually used in complex interfacing). Zac67 (talk) 17:35, 2 July 2013 (UTC)

Oppose. I disagree with the merger. Dual-ported RAM is more general concept which may be used for various purposes (e.g. in network cards, for interprocessor communication); VRAM is just one of its uses (although perhaps the most widespread one). Kolarp (talk) 11:06, 6 July 2013 (UTC)

Oppose - VRAM is a kind of dual port RAM, BUT dual port RAMs are used in other systems too, specially with some embedded systems, NIC, in some FPGA designs, etc. So merge "dual port" RAM into VRAM may induced some confusion. Dilshan (talk) 04:05, 29 August 2013 (UTC)

Bad Links

Can something be done about the links to this page? It's extremely confusing. This page is linked to by many others and I have yet to see a single link that actually involves dual-ported RAM. Every link coming to this page expects it to be about video memory in general. So you have the pages for individual computers claiming to have X amount of "VRAM" and linking here, when those computers didn't use this kind of dual-ported RAM. 99.183.184.100 (talk) 13:20, 17 April 2020 (UTC)

The whole article is inaccurate as to who invented what and when

Texas Instruments D.Redwine and G.R. Rao filed in 1979 for what became US Patent 4,440,852 that was issued May 18, 1982. (see: https://patentimages.storage.googleapis.com/57/a3/40/bf063176a7eeee/US4330852.pdf). The concept was further refined and improved at Texas Instruments including US 4,639,890 and 4,648,077 among many others.

Texas Instrument put the TMS4141 on sale in 1983 (Pinkham, R., Novak, M., and Guttag, K., "Video RAM Excels at Fast Graphics", Electronic Design, Vol. 31, No. 17, Aug. 18, 1983, pp. 161-182) and it was designed into many systems by 1986. See also: https://books.google.com/books?id=zZcm2V37uioC&pg=PA10&lpg=PA10&dq=TMS4161+1984&source=bl&ots=2EmVl8kx1H&sig=ACfU3U0w0ZSsMOy0afsVYuQdSk9Oi2-Aow&hl=en&sa=X&ved=2ahUKEwj8m4DKw-7yAhVck2oFHTAyAF0Q6AF6BAgWEAM#v=onepage&q=TMS4161%201984&f=false. This is 2 to 3 years before the 1986 IBM RT PC

Where exactly is that dual-ported? I can only see serial access. --Zac67 (talk) 06:40, 8 September 2021 (UTC)

According to this article, that is the whole point, one of the "ports" is serial. Dual random access is a different thing.Spitzak (talk) 16:58, 8 September 2021 (UTC)

@Spitzak: Looking over the patent you linked to, I see nothing like dual-ported RAM. Unless I am greatly mistaken, this is a way to use existing DRAM production methodologies to make a single-output serial RAM like a very large shift register. That is, after all, what the patent says it is.

The patent describes this as a way to interface with serial devices like CCDs and disk drives. At the time, this was normally accomplished with a DRAM and external circuits to latch the data and then feed it a bit at a time. Using this memory would eliminate that external circuitry and associated data lines on the circuit board.

The system discussed in this article is dual-ported and that is its key quality. Dual porting allows the CPU to read and write the data in parallel as if it is a conventional DRAM, while a second separate port is feeding the display at the same time in serial.

The patent you point to does not describe this. Unfortunately I can't find a data sheet for the TMS4141 online, so I can't cross-check, but I'm pretty sure if this patent is the basis for the TMS4141 when would be described as "single-ported serial DRAM". Maury Markowitz (talk) 13:16, 21 January 2024 (UTC)

I am not the one who posted the patent. I was just confused by Zac67's comment saying "I can only see serial access" when the Wikipedia article makes it pretty clear that one of the ports is serial. Spitzak (talk) 20:56, 21 January 2024 (UTC)