VRAM: What is it and what is it used for?

Hardware Guide » Tutorials » VRAM: What is it and what is it used for?

La VRAM or Video RAM memory It's one of the most important elements when it comes to a computer that will be used for graphics tasks. The performance that can be extracted from the GPU will largely depend on it. For this reason, we're going to dedicate this article to exploring the secrets this memory holds and its nature.

You may also be interested in reading:

• GPU Architectures: The Differences Between Intel, AMD, and NVIDIA
• GPU Virtualization: What It Is

GPU: a "CPU" for graphics

When it comes to a graphics card, as you may well know, it is almost like a motherboard in itself, with a central processing unit, in this case a GPU dedicated to graphics, it also has a RAM, in this case the VRAM that we will see in more detail in this article, a firmware or BIOS, the PCB itself manufactured by manufacturers who also have experience in motherboards such as ASUS, Gigabyte, ASRock, MSI, etc.

In case of being a iGPU or integrated GPUIn this case, it will share a portion of RAM reserved for the GPU, and the rest will be used by the CPU. Therefore, this portion of main memory could be considered VRAM for iGPUs, even though it doesn't have exactly the same characteristics as GDDR.

VRAM vs RAM

La VRAM is a type of physical memory similar to RAMIn some cases, RAM is also used for graphics purposes, such as in iGPUs, where it is shared with the CPU. However, the needs of the GPU and CPU are different. The GPU requires higher latency, bandwidth, and speed due to the number of cores it manages and the parallelism required to process massive amounts of data, as is the case with graphics processing with SIMD streams. In these cases, a portion of main memory is reserved for exclusive use by the GPU, which can limit the CPU's capabilities.

To meet the needs of the GPU, it has been developed a version of RAM specially optimized for GPUs, called GDDR. Although based on conventional DDR RAM, GDDR has significant improvements in bus and speed to meet the GPU's bandwidth, speed, and latency demands.

In some systems, such as mobile devices, modern game consoles, and Apple M-Series processors, a unified memory system called Unified Memory is used. In these systems, the CPU and GPU share the same RAM address space, meaning that there is no need to have identical copies of data sets for the CPU and GPU. Although the RAM It's not optimized to meet the GPU's bandwidth and speed needs, so some systems use GDDR5 directly instead of unified RAM, as is the case with the AMD APU in the Sony PlayStation XNUMX. This allows for memory optimized specifically for the GPU.

Differences between GDDR and DDR

While DDR is used by the CPU, and in some cases by the iGPU as I mentioned earlier, GDDR is a memory specifically designed for the GPU:

• DDR (Dual Data Rate): It is a type of SDRAM (Synchronous Dynamic Random Access Memory) used to store all the data and instructions for the programs that the CPU must execute. In the case of an iGPU, it will also have a portion of this memory reserved for the GPU; in some architectures it is inaccessible to the CPU, while in others it is a unified memory, as in HSA systems. Furthermore, the main memory, or RAM for the CPU, can be implemented in modules, such as DIMMs, or also soldered onto the PCB itself.
• GDDR (Graphics DDR): These are also memory chips with SDRAM cells, and are even based on DDR specifications. They are also usually manufactured by the same manufacturers such as SK Hynix, Micron, Samsung, etc. However, in this case, it will only be accessible by the GPU, and will have some differences with DDR to enhance the needs of the GPU, which is a processing unit with a higher degree of parallelism compared to the CPU. On the other hand, in this case, it will always be soldered onto the PCB.

In summary, the Main differences between the two types of memory are:

Specificationss	DDR	GDDR
Bandwidth	Low	High
Energy consumption	High	low
Heating	Higher	L
Unlimited	low	High
Capacity	low	High
Clock frequency	low	High
Ribs	Less	More

What is VRAM memory?

Before continuing to talk about the required capacity, it is essential that you understand what VRAM is and its function in the system. VRAM (Video RAM) It's a random-access memory, similar to CPU RAM, but specifically dedicated to the GPU. This memory stores the data needed to render pixels, textures, meshes, and other framebuffer elements on the monitor. It's important to mention that, on many systems, VRAM can't be accessed in the same way as conventional RAM, as it isn't mapped or visible to the programmer.

It is essential to keep in mind that the GPU also needs receive information from the CPU, such as the position of objects and physics data for processing graphics. Therefore, this necessary data is also stored in VRAM, not just the results.

History of the VRAM

Frederick Dill, Daniel Ling and Richard Matick They were the ones who would invent VRAM while working at the IBM research center. However, it wasn't patented at the time, until five years later. And it wasn't until 5 that it was used for the first time in a commercial PC, specifically for the graphics adapter of the IBM PC RT.

After the release of this PC, the production costs of dual-port VRAM memory decreased, allowing other manufacturers to use it in their graphics acceleration products or graphics cards. This fact marked a inflection point in the world of graphics, as speed and quality have improved to this day. However, there have been some attempts to replace it with HBM, but this memory is more expensive and its capacity is more limited, which has meant it hasn't been a success in the PC world...

Types of VRAM

Throughout history, there have been developed a multitude of different types of VRAM, Although many people today only remember the most recent versions such as GDDR4, GDDR5 or GDDR6.

• SONIC: SAM memory was a type of linear register used in some older graphics cards to store the data needed for the digital-to-analog signal conversion required for connection to older monitors that only had analog ports.
• CDRAM (Cache DRAM): This is a graphics memory that combines features of SRAM static memory and DRAM dynamic memory, and is used as a cache to accelerate graphics. However, this memory was expensive and its capacity was limited.
• MDRAM or Multibank DRAM: is a type of high-speed RAM created by MoSys Inc. This memory is divided into 32KB DRAM banks, allowing individual access to each bank. Although this memory is no longer used today, it was once used in some devices such as the Hercules Dynamite 128, which had 4MB of MDRAM.
• Rambus VRAM: The Rambus company also patented a new bus to accelerate the flow of data between the VRAM and the frame buffer.
• 3D RAMMitsubishi created this graphics card, which included additional logic, such as an ALU, that allowed for calculations to be performed for Z-Buffer testing. These graphics cards were used for Sun Microsystems workstations based on UltraSPARC processors.
• SGRAM (Synchronous Graphics RAM): It is basically a VRAM memory based on SDRAM cells with a relatively low production price, but with only one access port, although it could act as a dual-port memory by opening two memory pages at the same time.
• WRAM (Windows RAM): It has nothing to do with the operating system; it's just another high-performance VRAM with two ports and high bandwidth. It was used, for example, in graphics cards like the Number Nine Revolution 3D, or the Matrox MGA Millennium.
• Extended or shared RAMShared memory: Shared memory is used when the GPU is integrated, as I mentioned earlier. It's not dedicated VRAM per se, but rather a DDR card that the GPU can also use, along with the CPU. In many cases, the size of the space reserved for the GPU can be modified from the BIOS/UEFI.
• HBM (High Bandwidth Memory)HBM is a high-speed memory that was implemented in AMD Radeon Fury graphics cards. Although it offered better performance than GDDR, it was more expensive and had limited capacity. Due to these disadvantages, AMD eventually opted to move to GDDR, although HBM is still used in the high-performance computing (HPC) sector.
• GDDR (Graphics DDR)GDDR (Graphics Double Data Rate) is a type of DDR SDRAM designed specifically for graphics cards, which differs from conventional DDR in that it has shorter access times and higher clock frequencies. Over time, several generations of GDDR have emerged, such as GDDR3, GDDR4, GDDR5, GDDR6, and others, each with improvements in speed and capacity.
• GDDRXAn example is the GDDR6X used by NVIDIA for its latest graphics cards. This is an improved version of the GDDR6 to gain more performance and was developed jointly with Micron. One of the most notable features is that it uses PAM4 encoding for the data bus instead of NRZ like conventional GDDR6.

Now you have a little more understanding of what VRAM is, the differences with RAM, etc. Let's continue exploring some interesting related topics...

How to know how much VRAM I have

If you want to know how much VRAM memory you have on your graphics card, you can follow the steps below that I explain for the different operating systems:

• Microsoft Windows Hereinafter 10:
  1. Open Settings.
  2. Inside the window go to System.
  3. Go to Screen.
  4. Then to Advanced Display.
  5. In the adapter properties you will see a line with the dedicated video memory.
• GNU/Linux (Ubuntu):
  1. Go to your distro's System Preferences.
  2. Click About this system.
  3. Then look at the information in the Hardware section, and there you will be able to see the amount of memory.
• Apple macOS:
  1. Click on the apple icon in the top left.
  2. Then in About This Mac in the drop-down menu.
  3. Check the Graphics section. If it's an Apple M-Series, remember that it will be unified and shared with the rest of the SoCs.

Also, use third party tools to know the details and the amount of VRAM memory you have on your computer. For example, I recommend using GPU-Z by TechPowerUp Available for Microsoft Windows. Once you've downloaded the app, run it and look at the Memory Size section of the Graphics Card tab.

Can VRAM be increased?

Answering this question is not so simple, Each particular case will have to be analyzedFor example, you can be guided by these points:

• You have a PC with a dedicated graphics card or dGPU:
  • If it's a desktop, you'll need to upgrade your graphics card to a higher-end one to expand the VRAM.
  • If it's a laptop with dedicated graphics, you won't be able to change the VRAM capacity either, as the chips will be soldered onto the motherboard. Some laptops have MXM-type modules, in which case, you'd also have to replace the card's module with a higher-quality one, although there aren't usually many options to choose from.
• If you have a PC or laptop with an iGPU If you're using an Intel or AMD APU, or any other integrated iGPU, such as those found in some ARM SoCs, etc., then you could change the amount of RAM shared with the GPU in some cases. To do this, you could use several methods:
  • Some systems support modifying the shared RAM capacity allocated to the GPU from the BIOS/UEFI:
    1. Accessing the Setup menu.
    2. Then in tabs like Advanced Features or Chipset or similar, depending on the firmware brand.
    3. Within the Graphics Configuration or Video Configuration or similar.
    4. And there, look for options like VGA memory size or VRAM size to modify them. This will depend on the amount of RAM you have installed. For example, with 4 GB of RAM you could use 512 MB, with 8 GB you could use 1 GB, with 16 GB you could use 2 GB for graphics, etc.
  • It can also be modified from the Windows Registry. To do so, follow these steps:
    1. Press Windows + R.
    2. Type regedit and press ENTER.
    3. Look for the entry HKEY_LOCAL_MACHINE\Software\Intel or AMD (depending on your case).
    4. Right click on Intel or AMD.
    5. Select New.
    6. Then Key.
    7. Name it GMM.
    8. You need to use DedicatedSegmentSize and assign it a Decimal value. This specifies the amount of memory you want to dedicate.
    9. Finally, reboot the system.
• If you have a Mac with dedicated GPU:
  • If it's an AMD or NVIDIA (on older models), you can always swap the graphics card for a supported model that has more VRAM, if available.
  • If it's a MacBook with a dedicated GPU, it can't be replaced as it's soldered onto the motherboard, as is the case with other laptops.
• If you have a Mac with iGPU
  • If it is an Intel iGPU, the memory allocated to graphics cannot be increased.
  • If you have an Apple M-Series, you won't be able to change it either, since it has unified memory and is the one for the SoC model you purchased.
• You could also make use of a External GPU, as some laptop systems support, and that this eGPU has a higher VRAM, like those that connect via Thunderbolt.
• For desktop computers, there is the alternative of using a multi-GPU configuration, if you have a PCIe x16 slot available to add a second graphics card and have them work together.

How is the memory required calculated based on resolution?

You may be aware of the amount of VRAM memory required for each screen resolution, but do you know how it's calculated? A common example is calculating the framebuffer size required for a FullHD resolution of 1920x1080 pixels and 32-bit color:

• 1920×1080 = 2073600 total pixels on the screen
• 2073600 x 32-bit = 66355200 bits
• 66355200 / 8 = 8294400 Bytes
• 8294400 / 1024 = 8100 KB
• 8100 / 1024 = 7,9 MB

It is true that with 8MB VRAM A single image could be displayed in Full HD resolution, but that's not enough for dynamic graphics in the real world. Larger capacities are needed because many other values ​​and copies of rendering textures must be stored. Furthermore, double or triple copies of data can also require even more capacity. Therefore, current VRAM capacities are necessary to ensure optimal graphics performance and a smooth gaming experience. The GPU can also store internal and transformed copies of rendering textures to avoid repeating texture dispatches from the CPU.

The amount of VRAM memory that you have on your graphics card is essential to achieve good performance in video games and graphics applicationsThis is because the amount of available memory determines the image quality, loading times, and resolutions that can be handled. If the graphics card doesn't have enough VRAM, it will fill up quickly, causing a significant drop in performance. Many claim that 4GB of VRAM is enough to run AAA games at 1080p, while 8GB is required for UHD resolutions. However, this claim has proven to be less true these days.

So: how much VRAM do I need?

Generally, you should follow these guidelines for how much VRAM you need for gaming and personal use:

• 2 GB for titles running at 720p.
• 4 GB for titles running at 1080p resolution.
• 6GB for 1440p gaming.
• 8 – 12 GB for 2160p or 4K.

If you want a graphics card for professional work, such as rendering, editing, etc., then the recommended capacities are:

• 8GB for 720p to 1080p.
• 16 GB for 4K.
• 32 – 64 GB for 8K, for commercial use.

Are these exact figures?

As you may know, in the new generations of graphics cards, it has been decided to put a minimum of 8GB of VRAM In some cases, however, this is causing a stir among some online users who complain that it's not enough to play current AAA titles at high graphics settings, even at 1080p or Full HD resolutions.

The reason for these concerns and why the 8GB minimum is a confusing number is that the capacity needed It will not only depend on the resolution as you thought. Other aspects also influence, such as

• Type of video game: Each AAA title may require more or less VRAM capacity depending on factors such as its optimization, the realism of its graphics, etc.
• Graphical configuration: since depending on the configuration or quality chosen by the user, the amount of VRAM used may increase or decrease.
• Graphics engines or Game Engines: As games become more advanced and offer increasingly realistic graphics, they're increasing the amount of VRAM required. An example of this is the famous Unreal Engine 5, which requires slightly more than 8 GB to run smoothly.

Currently, 12GB is arguably the new 8GB. So, it's best to get a graphics card with at least 12GB.

In the future, the video game developers will face a major challenge if they want to reach a wider audience with graphics cards with only 8 GB of VRAM, as it will be increasingly difficult for new titles to run on these cards, even if they are optimized. This is because current graphics quality and new graphics technologies require more and more data stored in VRAM, which increases its space consumption. VRAM not only stores textures, but also meshes, shaders, framebuffers, and other data necessary to render the scene, such as objects, shader values, and lighting information.

I hope it's now a bit clearer to you what VRAM is, what it's used for, and how much you should have...