Can we use computer RAM for GPU ?


👤 Diwas Poudel    🕒 02 Sep 2023    📁 TECH

The answer to this question is not simple. Thus, let's explore them.
 
The CPU of our PC is a general-purpose processor that can do a variety of tasks. They are designed for processing serial instructions and are not compatible much with parallel instruction processing. Whereas a GPU is a special-purpose processor that performs certain tasks such as graphics rendering in parallel. They are designed for processing rapid parallel instruction and are not suitable for serial instruction processing. And the major aim of GPU is fast throughput, and when the speed of GPU and CPU is compared, GPU wins.
 
If you are not familiar with GPU you may not have heard that there are two types of GPU. One is an integrated GPU and the next is a Dedicated/Discrete GPU.
 
There are two types of GPU that can feed into your system. One is an integrated GPU and the next is a Dedicated/Discrete GPU.

Integrated GPU 

Integrated GPUs are GPUs that are attached to the CPU on a single die internally. They shared the memory used by the CPU. One of the popular Integrated GPUs is from Intel.

Dedicated GPU

The dedicated GPU is connected to the motherboard via the PCI Express Slot. These GPUs have their own RAM called VRAM, which stores the data and information that will be used by this Dedicated GPU. The GPU communicates directly with the VRAM. And if GPU has to communicate with the CPU then it happens via PCI Express Slot. Nvidia Graphics are mostly popular for Dedicated GPU.

Some Nvidia dedicated graphics cards are available in a variety of series such as the RTX 40 Series, RTX 4090, RTX 4080, RTX 30 Series, RTX 3090 Ti / 3090, RTX 3050, RTX 20 Series, GTX 16 Series, and so on.

You may also hear about System RAM and this RAM is attached to the motherboard in a RAM slot. The CPU communicates directly with the Main Memory (RAM) for providing needed information more quickly.

Now the question Can we use RAM as VRAM for GPU may come to your mind. If so let's discuss this.

If we take a look at laptops and some desktops, we can see that it has two types of memory: one is Dedicated RAM, also known as VRAM, and the other is Shared RAM.

What is Shared RAM?

Actually, the capacity of Shared RAM is only half that of Actual RAM. If you have 16 gigabytes of RAM, then your shared RAM will only be 8 gigabytes. If you have dedicated graphics of 4 gigabytes, then you will have a total of 8 gigabytes plus 4 gigabytes, which is 12 gigabytes of RAM. However, you should be aware that the bandwidth of shared RAM is significantly lower than that of dedicated VRAM. Therefore, it is imperative that you do not anticipate improved performance from shared RAM.

You can clearly see that I have 16GB system RAM, with 8GB reserved for Shared GPU Memory.

Shared GPU memory is half of System Memory
fig. Shared GPU memory is half of System Memory

What is GPU Dedicated RAM?

This RAM is known as VRAM, or video RAM, because it is specifically designed to store video, images, and graphics-related data and the Graphics Processor communicates directly with it for data and information.

Can we use RAM as VRAM for GPU?

It is speculated that shared memory will be utilized in the event that your graphics memory (VRAM) runs out of memory space. Therefore, the more random access memory (RAM) you have, the more shared memory you will have. But if Shared RAM is full then dedicated RAM is never used.

If you have enough dedicated VRAM, your system RAM (shared RAM) will least be used for high graphics work because System RAM is not a key source of memory.

In addition, the path/distance between VRAM and GPU is much less than the path between GPU and System RAM (via PCI slot).As a result, if System RAM is used, performance will be affected.

Furthermore, because GPU memory (VRAM) is larger than CPU RAM, the interface between the Graphics Processor and VRAM has a higher bandwidth.

Also, there is another perspective, During the execution process, GPUs do not have direct access to CPU RAM. Instead, data is directly exchanged between the GPU RAM and the CPU RAM both before and after the GPU executes its instructions. Because GPU execution requires manual virtual memory as GPU does not have a paging mechanism like CPU /RAM jas and because GPUs cannot exchange data with the CPU while it is being executed, programmers need to modify their applications and possibly segment them so that all relevant data can be found in the GPU when it is required. When dealing with data sets that are too vast to be stored on a single GPU, several memory transfers must take place between the CPU and GPU.

When Integrated Graphics is enough?

General activities like as browsing the web or watching videos on YouTube do not require the use of specialized graphics hardware or software. The graphical capabilities included in Intel are sufficient. Additionally, if you do not have enough money to get integrated graphics, then it is sufficient for you.

However, if you want to do animations, play games, or conduct video editing on a large scale, dedicated RAM is required.

Factors that affects how GPU uses System RAM 

Lets look some of the factors that can affects how GPU uses System RAM:

1. Amount of System RAM do you have.
2. Which GPU are you using.
3. Amount of VRAM available on GPU.
4. PCIe bus Speed 

Its important to note that GPU uses system RAM using PCIe bus. GPU uses this PCIe to communicate with CPU and other components of computer.So, speed of PCIe bus can affect performance of GPU , when you are using system RAM.

In this case GPU uses Shared Memory.Size of Shared memory depends on amount of physical RAM you have intalled in your system.

Conclusion:

In this manner, we successfully discussed whether the RAM may be used as VRAM for the GPU.