Why Do Graphics Cards Have Memory?

If you’re shopping for a new graphics card, you must’ve noticed that there’s always a “Memory” spec on the box. Moreover, a single graphics card model often comes with multiple memory sizes. But why do graphics cards have memory if you already have plenty of RAM in your computer?

Graphics cards have memory to store rendered frames and provide the GPU with fast access to essential data. The image on the screen is always held on the frame buffer, which is a part of the memory. Textures, shadows, and depth buffers are also stored in the graphics card memory.

This article will explain why graphics cards have memory, how it affects performance, how much graphics card memory you need for gaming, what shared GPU memory is, and whether you can upgrade memory on your graphics card.

Why graphics cards have memory

a graphics card with the heatsink assembly removed showing the memory located around the GPU

There are a few good reasons why graphics cards don’t rely solely on your system RAM. 

VRAM is often faster, more easily accessible, has more bandwidth, is easier to optimize to work with, etc.

Let’s take a closer look to understand why graphics cards come with built-in memory.

Graphics cards use the memory to store graphics data

Your graphics processing unit (GPU) does thousands of calculations every second to display a crystal-clear image onto the screen.

The data required for the calculations must be stored somewhere. Unfortunately, your primary storage (SSD/HDD) is too far and way too slow to be useful.

All data required to render a frame goes from your primary storage disk to RAM. The CPU recognizes the data is for graphics, so it’s sent to your VRAM. That’s where it sits until it’s required.

Textures are the perfect example of vital graphics data. Without textures, your games would look like a bunch of random geometric shapes with no immersion quality to them.

Everything from grass and stones to various elements of your gun is a texture glued onto a 3D model.

You can think of video game textures as thousands of high-resolution images. Originally stored in the game folder, textures must stay in your VRAM for easy access when required.

Here are a few other examples of data stored in your VRAM:

  • Z-buffer (depth)
  • Shadow maps
  • Vector graphics
  • Mesh
  • Shaders

Rendered frames are stored in the memory

Your GPU takes the data from the Z-buffer, adds shadow maps to it, does all the required shading, etc.

Once a frame is completely finished, it has to go somewhere. Of course, you see it on your screen, but the monitor only displays the image. 

The frame is stored within a special part of the graphics card memory called the frame buffer. New frames are constantly overwriting the old ones to achieve a smooth image.

You can even change the number of pre-rendered frames in the graphics card drivers’ settings. 

It’s usually only one frame, but setting it to two or three may improve performance in some cases (or make it worse in others).

Graphics cards use a special type of fast memory (VRAM)

a 2d rendering of a gpu and graphics memory

Video random-access memory (VRAM), in its broadest sense, is similar to your system RAM. 

All data stored in VRAM is temporary, so it’s a type of cache used by your GPU to hold large amounts of data required for graphics.

VRAM is usually much faster than traditional RAM.

More importantly, it’s very close to the GPU physically. It’s soldered directly onto the graphics card PCB.

This allows for super-fast data transfer with minimum delays, making high-resolution graphics possible.

VRAM differs in size, speed, and bus width on modern graphics cards. 

More VRAM means it can store more data. Faster VRAM allows for faster data transfers and, consequently, better performance.

Lastly, the bus width determines the maximum available bandwidth. Think of it as a chokepoint between your graphics card memory and your GPU.

There’s also a particular type of VRAM called high bandwidth memory (HBM). A stacked layer of memory sits on a controller die that transfers data to the GPU through a layer of silicon.

HBM allows for small form factor graphics cards, faster data transfer rates, lower power usage, etc.

Does a graphics card memory size matter?

graphics card memory usage data in windows task manager

Graphics card memory size matters because it affects GPU’s performance. Not having enough memory on your graphics card limits the resolution size, textures, shadows, and other graphics settings.

Let’s use a simple analogy to help better understand how graphics cards work. Your GPU is like the engine in a car, and the graphics card memory is the passenger space.

You can only go as fast as the engine, but you can cram in more passengers if you have a larger car. This improves efficiency, meaning you can bring more people with you for the trip.

Similarly, the performance of your GPU affects how much FPS you can get at certain graphics settings. 

If you have plenty of VRAM, you can allow the GPU to render frames at a higher resolution, bump up texture quality, increase details, etc.

You’re still restricted by the GPU’s power, though. So, it’s best to look at the graphics card model first, then VRAM size.

More VRAM means you won’t experience constant stutters when the graphics card loads an asset from your RAM or storage disk.

Assuming your GPU can handle it, you need sufficient VRAM space to bump up the texture settings, shadows.

Even the most powerful GPU on the market can’t run games smoothly if it doesn’t have enough VRAM to store assets.

How much graphics memory do I need for gaming?

You need a minimum of 4 GB of graphics memory for 1080p gaming. 6 GB is ideal if you want to set texture quality to High or Ultra. For 1440p gaming, 8 GB – 12 GB is sufficient to run games at medium-high settings. 4K gaming requires at least 12 GB of VRAM for decent texture quality.

a young guy gaming on a gaming computer

In general, more VRAM is always better. If we’re talking about the exact GPU model, always aim for the model with more VRAM if the price is similar. 

You’ll probably get better FPS, less stutter, and you can set Textures to Ultra with a minimum performance loss.

If you’re only playing eSports titles, you could even get away with a 2 GB graphics card at 1080p

But any modern AAA title fills up 2 GB very quickly, and VRAM usage can even go over 6 GB in some instances. 

Gaming at 1440p is more demanding because there are more pixels on the screen. The higher-resolution frame takes up more space inside the frame buffer.

4K resolution is even more demanding. You already need a powerful graphics card to play games at 4K. 

If your graphics card came with less than 10-12 GB of VRAM, don’t even consider 4K gaming unless you want to upscale the resolution.

Upscaling allows you to preserve a lot of VRAM space because it uses special textures that are made to look like they’re high-resolution.

If you want to game at 8K, you’ll probably need over 20 GB of VRAM to store the enormous frame.

Can you upgrade graphics card memory?

Graphics card memory cannot be upgraded because it’s soldered onto the PCB. The circuit board design predetermines the maximum memory size on a graphics card. If a manufacturer decides to increase memory size on a graphics card, the memory chip dies are doubled.

Technically speaking, you could desolder all memory chips on your graphics card and replace them with chips that have twice the memory.

However, this is extremely risky, expensive, and difficult.

So, there is no practical way of upgrading the memory of your graphics card.

Soldering the memory onto the PCB allows manufacturers to optimize the data transfer rate between the GPU and the memory chips.

But you shouldn’t concern yourself too much about VRAM size in the first place. Both AMD and NVIDIA do a lot of careful planning to decide how much VRAM they’ll put onto their graphics cards.

It wouldn’t make any sense to combine a low-end GPU with a bucketload of VRAM.

How much VRAM they can put is predetermined by the memory bus size

For example, if a graphics card has eight 512 MB chips to achieve 4 GB of VRAM, the manufacturer must replace the 512 MB chips with 1 GB units to upgrade the VRAM to 8 GB.

They usually don’t mix and match memory sizes and speeds for this reason.

What is shared GPU memory?

Shared GPU memory is virtual memory that’s located on your system RAM. Your graphics card uses shared GPU memory when it runs out of VRAM space to store assets required to process graphics. Shared GPU memory is significantly slower than VRAM, so it’s only used when necessary.

a closeup rendering of a gpu and graphics memory onboard integration

Ideally, your graphics card will never use shared GPU memory. However, you can easily tell that your graphics card runs on shared GPU memory. As a result, you’ll get visible jitter, textures load very slowly, etc.

You can also turn on the performance metrics in some games or via your graphics card driver to track VRAM usage. 

If the VRAM usage is at 100%, your graphics card will start running on shared GPU memory.

In that case, lower the resolution, texture quality, shadow quality, or some other graphics settings. As a result, you’ll eliminate stutter and various graphical artifacts.

If you’re running on an integrated graphics card, your computer doesn’t have any VRAM. 

Instead, your GPU is running exclusively on shared GPU memory.

In that case, there isn’t anything that you can do to improve performance, aside from overclocking.

Does GPU memory matter?

GPU memory matters because more VRAM allows your graphics card to store high-resolution frames, more detailed textures, and other assets. More GPU memory doesn’t always mean higher FPS, but it allows for a smoother, more consistent experience, especially at high graphics settings.

When buying a new graphics card, pay attention to VRAM size. You don’t want to waste money on more VRAM when you don’t need it because your monitor is 1080p.

On the other hand, not having enough VRAM will seriously affect overall performance. This is because you need enough VRAM to store the outgoing frame, along with all graphics assets.

Ultimately, if the price between different VRAM sizes isn’t enormous, always go for the one with more VRAM. It’s a way to make your computer more future-proof.

Even if a modern game can’t occupy all the VRAM today, that won’t be the case in a couple of years. 

Assets, especially textures, are constantly increasing in size drastically.

GPU Memory vs. RAM

GPU memory, often called VRAM, and system RAM has many things in common. First, they’re both random-access memory, which means you can write and read any type of data to them.

However, VRAM is often split into several chunks that store different things. For example, frame buffers, depth buffers, shadow maps, textures, mesh, and other assets are all stored in VRAM.

Your system RAM temporarily stores all the data your CPU needs or may need soon. In addition, the programs and files that are being used are held in RAM.

For example, all Chrome tabs are stored in RAM, allowing lightning-fast switching between tabs.

VRAM speed is determined by GDDR (or HBM) version, memory bus width, and frequency. Speaking of frequency, VRAM easily reaches speeds of 10,000 – 15,000 MHz.

RAM is much slower. Clock speeds usually range between 2400 – 6400 MHz, depending on the DDR version and clock speed.

Moreover, the maximum data transfer rate is much higher for VRAM.

Final thoughts

Graphics cards have a special type of random-access memory called VRAM. VRAM comes in the form of memory chips soldered directly onto the graphics card to allow fast transfer rates.

Your graphics card needs VRAM to store the outgoing frame. It also holds textures, mesh, shadow maps, 3D depth information, and other assets that your GPU accesses frequently.

Having memory on the graphics card instead of using system RAM provides several advantages. 

VRAM is much closer to the GPU, it’s much faster, and it’s optimized to work with your GPU, improving overall performance.