Recently, while reading various forums, I realized that many people do not understand the connection between RAM and GPU. Unfortunately, some people still think RAM and VRAM are the same things. We decided to change that by writing this article.
Today, we’ll examine the connection between these two critical components and answer the question, “Does GPU use RAM?” After in-depth research, we concluded the following:
Whether or not GPUs use RAM depends on the type of GPU. Integrated GPUs always use RAM since they don’t have their own memory. Contrary, dedicated GPUs use their own memory (VRAM). However, in some rare circumstances, when the VRAM is full, they will use RAM.
The answer is even more complicated, so if you want to make the best possible decision, you will want to keep reading, trust me.
It depends on the type of GPU
The type of GPU will play a crucial role in determining whether it will use RAM or not. Now, we can divide them into two categories: dedicated GPUs and integrated GPUs.
Dedicated GPUs
These are separate components installed on your motherboard, which have their own memory, known as VRAM.
In almost all cases, they primarily use VRAM, which is way faster than RAM. However, in some cases, when VRAM is full, a dedicated GPU may use RAM to store additional data.
Please note that this is not preferred because RAM is much slower, which often causes performance drops.
Integrated GPUs
Unlike standard GPUs, integrated GPUs are built into the CPU and do not have their own memory. These are mostly found in lower-end computers that do not require high graphics.
Since they don’t have their own memory, integrated GPUs have to use systems RAM, which is typically around 50% of total RAM.
To learn more about this and how to improve your iGPU performance, make sure to read Does Integrated Graphics Use RAM? (All You Need to Know).
Does RAM affect GPU?
A dedicated GPU is like a separate little helper inside your computer that takes care of all the graphics and visuals. It has its own special memory called VRAM, which it uses to store and manage all the images, textures, and other visual stuff.
Now, your computer also has regular memory called RAM, which is used for various tasks and programs. Most of the time, the dedicated GPU doesn’t need to use this RAM because it has its own VRAM to handle everything. However, sometimes, if the GPU’s VRAM gets full or there’s just too much going on, the GPU might need a little extra help.
In those cases, GPU can use some of your computer’s RAM to store extra graphics data temporarily.
So, if you have more RAM or faster RAM, it can help your dedicated GPU when it needs that extra boost. But keep in mind that this doesn’t happen very often, and most of the time, the dedicated GPU can manage everything with its own VRAM.
Note that certain dedicated GPUs borrowed system RAM in older systems, but most modern ones do not because they use GDDR 5 or faster memory. In contrast, older dedicated GPUs use the same speed as the installed system RAM.
In conclusion, most excellent dedicated GPUs will have GDDR 5 or faster VRAM and will utilize it. But, integrated GPUs must consume system RAM. For instance, if you need 8 or 16 GB of RAM for an application and are using Intel HD graphics or an AMD APU-style processor like the Vega, plan on adding 4 GB of RAM to accommodate the integrated GPU’s potential needs.
On the other hand, if you have a dedicated GPU, the only way to increase VRAM is to get a new GPU with more VRAM.
Remember that insufficient RAM can cause performance bottlenecks for the GPU and CPU. For example, suppose the system is running out of RAM. In that case, it may use slower storage like the HDD or SSD for caching, which can lead to stuttering and reduced frame rates in games or slower performance in graphics-intensive applications. Additionally, the speed of the RAM can affect the overall system performance, which may indirectly impact the GPU’s performance.
Shared memory vs VRAM
Ok, we have a solid understanding of the RAM-GPU connection. Now we will explain the difference between shared memory and VRAM, which are two different ways of providing memory for a GPU.
VRAM
Most of you probably know this already, but we will explain it to those who might not. Namely, when you think of VRAM, think of a dedicated memory found on a dedicated GPU, which is designed for storing and managing graphics data such as images, textures, and rendering buffers.
In terms of performance, VRAM is noticeably faster, since it is solely dedicated to the GPU. It is built using high-speed memory techs like HBM or GDDR, which allows quick data transfer between memory and GPU.
Since VRAM is made for dedicated GPUs, you will commonly find it on gaming rigs, workstations and other systems that require high graphics.
Shared Memory
Conversely, shared memory is just an allocation of the system’s RAM for graphics processing. We often see this in integrated GPUs built into the CPU that doesn’t have their own VRAM.
However, shared memory is slower since it relies on RAM with a lower bandwidth than VRAM. Therefore, it often reduces overall performance. Mostly, you will use this memory in systems with iGPU (low-end desktops, laptops in which power efficiency and cost are more important than graphics performance).
To summarize, the major difference between shared memory and VRAM is in their design and performance. Shared memory is just a portion of the RAM (usually 50%) allocated for the GPUm, which is slower. In contrast, VRAM is dedicated, high-speed memory for dedicated GPUs, offering much better graphic performance.
Some people report having their shared memory used before VRAM is full. I know a guy who had 8GB of VRAM and 8GB of shared memory. His GPU was GTX 1070. He noticed that his GPU used six gigs of VRAM and some small percentage of shared memory (3-4%).
Should you be worried if that happens to you? Well, it depends.
While shared memory is way slower and can cause performance loss, if you are only using a small percentage of SM with normal VRAM usage, there is nothing to worry about. For instance, in the case of a previous guy, it is obvious that VRAM is doing most work.
The only reason why shared memory is in use may be either as a pass-through buffer or a shared area for your CPU. Note that using shared memory is not an immediate problem. It simply means that there are some processes that require CPU.
That GPU is likely not demanding an excessive amount of memory, and the shared memory usage you observe might be due to stored pages from other desktop applications running in the background while you’re in full-screen mode. Windows retains these pages to facilitate faster pre-loading.
A 6 GB VRAM capacity is quite typical for a game using an HD texture pack at 4K resolution with mid to high-quality settings, depending on your monitor size. However, if you’re experiencing sluggish performance in your game, consider overclocking your GPU before deciding to upgrade the card. Many graphics cards and CPUs in the gaming market are designed to support overclocking, as it is an attractive selling feature.
However, if shared memory was up significantly, you might have something to be concerned.