VRAM or Video Random Access Memory is a form of RAM used specifically for graphics processors. VRAM is used to store data that is actively being displayed on the screen or is likely to be required. Video games generally load the graphical textures for the entire map or scene into the VRAM when loading an area, this way all textures are easily and quickly available for rendering and you don’t have to wait while textures are slowly loaded in from slower storage or RAM.
VRAM works slightly differently on integrated and discrete graphics processors. With integrated graphics processors, VRAM is simply allocated from a small portion of the system RAM, this saves on cost but does come with a performance impact compared to how VRAM works on discrete graphics cards.
Tip: Integrated graphics is a graphical processing chip integrated into a CPU. Integrated graphics is generally relatively low power and designed primarily for day-to-day usage such as web browsing rather than intense graphical workloads such as video games. Discrete graphics processors are a completely separate processor chip from the CPU and are located on a graphics card that needs to be plugged into the motherboard. Discrete graphics processors are generally designed for higher graphical workloads such as video games – any modern PC will almost definitely have one.
On a discrete graphics card, VRAM uses a faster form of flash memory than standard RAM, as it needs to be accessible as fast as possible. Dedicated VRAM is physically located right next to the GPU on the graphics card to minimise transmission times between the VRAM and the GPU. Modern graphics cards generally use either GDDR6 (Graphics Double Data Rate type 6) or HBM2 (High Bandwidth Memory generation 2) memory which has both higher bandwidth (total amount of data that can be transferred per second) and lower latency (delay between making a request and receiving a response) than the DDR4 (Double Data Rate type 4) RAM used by the rest of the computer. Both the high bandwidth and low latency allows as much data to be accessible to the GPU as fast as possible. The primary downside of these faster flash memory types is that they are more expensive to make and have lower yields.
Tip: Yield, when used in the context of computer memory or processor silicon means the percentage of chips that work as intended. All processes have flaws with occasional failures that are identified and discarded during testing. Bleeding- and cutting-edge technologies generally cost more to make and have lower yields, meaning more product needs to be discarded during testing. As processes mature and are refined, yields generally increase.