VRAM (Video RAM)

From Bauman National Library
This page was last modified on 3 November 2016, at 21:28.
Type Dual-ported dynamic random access memory
Release date since 1980
Website www.nvidia.co.uk/page/products.html

Video RAM, or VRAM, generally is the totality of all forms of RAM for a temporary storage of images which are formed by the display adapter and transmitted to the monitor. All types of VRAM are DRAM special mechanisms.


VRAM was invented by F. Dill, D. Ling and R. Matick at IBM Research in 1980, with a patent issued in 1985 (US Patent 4,541,075). The first commercial use of VRAM was in a high-resolution graphics adapter introduced in 1986 by IBM for the PC/RT system, which set a new standard for graphics displays. Prior to the development of VRAM, dual-ported memory was quite expensive, limiting higher resolution bitmapped graphics to high-end workstations. VRAM improved the overall framebuffer throughput, allowing low cost, high-resolution, high-speed, color graphics.[1]

VRAM purposes and specifics

VRAM is a buffer between computer processor and monitor and it’s often called a frame buffer. When the image is to be displayed on the display, it is, primarily, is read by the processor as the data from the RAM and then it is written in the video RAM. The data is converted from the video memory by RAMDAC (analog-to-digital converter) into analog signals that are transmitted by the mechanism of the display, a cathode-ray tube. Typically, VRAM is located on the video card. Most of VRAM types are a dual-ported memory. It means it can simultaneously write the data to modify the image while the video adapter continuously reads the content to draw it on the screen. Dual-ported design is the main difference of VRAM from RAM.[2]

Necessary volume of VRAM

The necessary volume of the video memory depends on several factors:

  • Display resolution
    The higher resolution requires more video memory.
  • anti-aliasing — smoothing.
    Basic anti-aliasing works so: each frame is rendered in the resolution, exceeding the resolution of the monitor several times, then it is shrinked back to the required size and displayed on the screen. Consequently, the greater smoothing requires the greater graphics memory.
However, it is valid not for all anti-aliasing algorithms. There are, for example, FXAA anti-aliasing which requires very few resources, but makes the picture a little blurred. In the contrary, the MSAA algorithm consumes a lot of resources, but provides a clearer picture. Thus, the memory consumption will strongly depend on the specific type of smoothing.
  • Texture & shadow resolution
    The higher texture / shadow resolution needs more memory.

In most cases, the necessary volume is determined by the specific user tasks. Using the PC in the office or watching movies needs about 1 or 2 GB of video memory. For demanding games, for installation modes with 8k textures, for using of multiple monitors or of one, but with a high resolution, it will require much more graphics memory - more than 4 GB.[3]

VRAM forms

VRAM is not the only video technology of its kind. Alongside with it a number of other variations exists, but all of them are the dynamic random access memory, and the only difference in their characteristics is the implementation of each one. The dynamic RAM has its own characteristics. They produced an access to the memory in large banks. Dynamic memory should quickly rewrites large amounts of data, without interrupting the reading procedure, because the image on the screen is continuously been reading from this memory with a frequency of the frame sweep of the monitor, and at the same time operating system writes data exactly there which leads to changes in the image. The dual-ported RAM performs this task with the maximum efficiency. High-priced VRAM itself and WRAM, which will be described below, relate to for this type of memory devices. In dual-ported memory systems, you can safely increase the sweep frequency without fear of lower video adapter performance, so for the work with large (19-inch and more) monitors this video memory is used basically. All is about the performance, so there are those variations of the video memory where access to memory is speeded up by the graphics accelerator (it is particularly important for 3D-accelerators). Examples of the described memory are MDRAM and SGRAM, analogue of SDRAM optimized for video memory tasks and ousted with two-ported WRAM in fact. Listed VRAM forms and analogues are described below:

  • Window RAM (WRAM) is a schematic development of VRAM, with an ~25% capacity increase and the support of some commonly used functions such as font rendering, moving imaging units etc. In this implementation of the memory the electronic logic circuits are added. They accelerate common video functions, such as the bit-block transfer and the pattern fill. It is produced by the accelerator firms Matrox (Millenium / Millenium II) and Number Nine mainly, because it requires special access methods and data processing. This type of memory is produced only by Samsung Electronics Co., Ltd — by the creator of WRAM, — that greatly reduces the possibility of its use. Video adapters built using this type of memory don’t tend to drop in performance when you install large resolutions and screen update frequencies, in such cases RAMDAC takes more time to the video memory bus access and video adapter performance can fall heavily on the same single-port memory.
The position of Samsung Electronics is based on hilghlighting of increased data transfer rate in WRAM (up to 640 MBps: in compare with the same characteristics of VRAM, this rate is 50% higher), while window RAMs cost slightly less than VRAM analogues. However, after several years of exploitation WRAM technology has given way to SDRAM-memory[4], which feature, as the name implies (SDRAM means Synchronous Dynamic Random Access Memory), is a synchronous data exchange.
  • Synchronous Graphics RAM (SGRAM) is a variation of RAM with a synchronous access using for video memory. SGRAM is considered as a relatively inexpensive type of graphics memory. The peculiarity of this type is the masking of recording block. Masking of recording allows you to select the data to be changed in one single operation. It should be noted that in the video cards block record filling of buffer with data for background and foreground images are processed with more efficiency in compare with the traditional sequence of read, write, update operations.
SGRAM is a moderately fast single-ported memory. It works at frequencies above 66 MHz synchronously with external frequency of the CPU bus. Later this decision allowed the use of SGRAM in the video cards with the AGP-interface.[5] The example of the video card using SGRAM, is Matrox Mystique.
  • Multibank Dynamic RAM (MDRAM) is a high-memory version developed by MoSyS Inc. This type is implemented as a plurality of independent banks with volume of 32 KB each one (now their number is larger) which can be accessed individually. This provides an inexpensive cost of this type of memory because, unlike other forms of video memory, card can be produced with the required for this resolution amount of access memory. From a performance standpoint MDRAM is no worse and no better than other types of video memory.
  • Rambus Dynamic RAM (RDRAM) is a video memory developed by Rambus in collaboration with Intel in 1996. Its feature is an accelerated data flow between the graphics memory and the frame buffer. This video memory is optimized for streaming video. RDRAM is used in the PlayStation game consoles. Originally it was intended as the only memory suitable for Pentium 4 CPUs, but it was found out that DDR developed a long time before fits none the worse. The high frequencies of memory provided 99% channel loading, while loading of the competing standards reaches maximum of 70%. The memory bandwidth is 1 GBps, later it reached 4 GBps. Intel was going to use RDRAM in the Intel 850 board, but two weeks before the presentation there have been reports in the Internet that there was a memory bit error in RDRAM. Now this type isn’t used due to the high prices on it.[6]


Cite error: Invalid <references> tag; parameter "group" is allowed only.

Use <references />, or <references group="..." />


  • Video RAM (dual-ported DRAM) from Wikipedia, the free encyclopedia: Version 745919293, saved at 04:40 UTC 24 October 2016 / Wikipedia editors // Wikipedia, the free encyclopedia. — San Francisco: Wikimedia Foundation (2016). — Available at: https://en.wikipedia.org/w/index.php?title=Video_RAM_(dual-ported_DRAM)&oldid=745919293
  • TechTarget (2016). What is video RAM? Available at: http://searchstorage.techtarget.com/definition/video-RAM (accessed 30 October 2016).
  • Pikabu (2015). Video memory and how much of it do we need. Available at: http://pikabu.ru/story/videopamyat_vram_i_skolko_eyo_nuzhno_3505611 (accessed 30 October 2016; in Russian).
  • WRAM from Wikipedia, the free encyclopedia: Version 68028263, saved at 17:28 UTC 18 January 2015 / Wikipedia editors // Wikipedia, the free encyclopedia. — San Francisco: Wikimedia Foundation (2016). — Available at: https://ru.wikipedia.org/w/index.php?title=WRAM&oldid=68028263 (in Russian).
  • SGRAM from Wikipedia, the free encyclopedia: Version 74698985, saved at 13:58 UTC 23 November 2015 / Wikipedia editors // Wikipedia, the free encyclopedia. — San Francisco: Wikimedia Foundation (2016). — Available at: https://ru.wikipedia.org/w/index.php?title=SGRAM&oldid=74698985 (in Russian).
  • RDRAM from Wikipedia, the free encyclopedia: Version 79560410, saved at 15:50 UTC 14 July 2016 / Wikipedia editors // Wikipedia, the free encyclopedia. — San Francisco: Wikimedia Foundation (2016). — Available at: https://ru.wikipedia.org/w/index.php?title=RDRAM&oldid=79560410 (in Russian).