What Is Shared GPU Memory – Is It Important?

If you’ve ever checked task manager, which is very likely, you’ve probably seen something called “shared GPU memory”, and it’s usually half the capacity of the available RAM you have. So, if you have 64GB of RAM, your system will make around 32GB available to your GPU, it doesn’t matter whether it’s dedicated or integrated.

With half your RAM being allocated as shared memory, you’re probably wondering if this has any noticeable impact on the performance of your computer. What would happen if you were to increase the shared GPU limit or potentially disable it? Well in this post, that’s exactly what we’re going to find out.

Answer: Shared GPU memory is normal RAM used as Video-RAM and it’s used by your integrated graphics card, this is because it is located on your CPU and doesn’t physically have its own RAM. For a dedicated graphics card, the shared GPU memory is a form of virtual memory used once you’ve used all of your VRAM buffer.

What Is Shared GPU Memory?

shared GPU memory is dram allocated to the integrated GPU or dedicated GPU

Shared GPU memory is just normal system memory or RAM that is used by either your integrated or dedicated graphics card. Your dedicated GPU will use RAM as virtual memory once the VRAM buffer is full. It’s very similar to swap memory, when your system uses all available RAM, it will use the secondary storage(HDD or SSD) as virtual memory. But the difference is that your dedicated GPU won’t store unused assets in the RAM.

RAM or system memory is shared by your integrated graphics card, this is because it doesn’t have its own dedicated RAM. It would be overkill if integrated graphics cards were to have their own super-fast GDDR6 VRAM, so it makes better sense to allocate RAM to the iGPU. The iGPU is pretty slow anyway, so the shared GPU memory is unlikely to get full.

Since the GPU can just use the shared GPU memory when it uses all of its available VRAM, does this have any performance implications? Well yes, RAM is many magnitudes slower than VRAM, so when your graphics card uses all of its available VRAM, you will notice this in the form of stuttering or loss of performance. So, the shared GPU memory will allow you to keep playing your video game but with diminished performance.

Can I Modify The Shared GPU Memory?

modifying the shared GPU memory

If someone is using a dedicated graphics card, modifying the shared GPU memory by decreasing or increasing it won’t result in a huge change in performance. The performance will severely diminish once the VRAM buffer is full, so increasing the shared GPU memory won’t really make a difference.

In the case of integrated graphics, increasing the shared GPU memory is again unideal, the iGPU is pretty slow anyway but very efficient, so they can operate with very limited shared GPU memory. Integrated graphics can work pretty fine with the limited bandwidth the shared GPU memory offers, but dedicated GPUs require faster, and more memory bandwidth than your standard system RAM.

Another reason to not modify the shared GPU memory is that it can result in performance loss, the operating system is pretty efficient at managing available RAM. So if your integrated graphics is idle, it pretty much won’t use any system memory, or it will use the minimum required. Increasing the shared GPU memory can result in less available system memory; if you increase the integrated graphics memory by 2GB, and you have a 16GB system, you will essentially have 14 GB available.

We can understand the frustration of seeing half your memory seemingly being needlessly allocated to your integrated graphics card even though it doesn’t need it. But this is due to the system automatically allocating shared GPU memory to the GPU. With DVMT(Intel), your integrated graphics driver can call windows to allocate and deallocate the required memory capacity.

The same thing applies with AMD’s UMA frame buffer, automatically your system should allocate and deallocate the necessary amount of memory to your integrated graphics, so there’s really no reason to modify it. In the BIOS, this feature is usually set to auto by default, but in very rare scenarios, you may be able to benefit from increasing the buffer if you know what you’re doing.

How Much VRAM Do I Need?

vram vs ram for gaming

For gaming scenarios, the VRAM plays a huge role when it comes to handling graphical data, and it’s supposed to work extremely quickly. The VRAM capacity has a huge role in what resolutions you can play, as well as what graphical settings. Messing with these values can increase or decrease the impact of the data being handled by the VRAM.

So, if you run your games at ultra settings at 4K, you can imagine how large these graphical files can be, this is including – textures, terrain data, and geometry data, in this case, you may need a large amount of VRAM. It’s also heavily game-dependent, for example, if you’re playing a really graphically intensive game, or an unoptimized game you will require more VRAM.

But in general, if you’re playing at 1440P, you should be able to get away with at least 6-8GB of VRAM, but if this capacity is ever exceeded, your graphics card will dip into the shared GPU memory. Remember, the goal is to never use the shared GPU memory as it’s slow and inefficient for the graphics card. If you’re gaming at 4K, you may need around 8-10GB of VRAM.

The stress on the VRAM can be caused by:

  • Video games resolution
  • In-game graphical settings – (low, medium, high ultra)
  • What video game you’re playing (is it graphically intensive)
  • Anti Aliasing(AA)

VRAM Vs RAM For Gaming?

In most gaming scenarios, you will require more RAM than VRAM, this is because video games nowadays require at least 16GB of VRAM to run smoothly. But, if you’re asking what is more impactful, we will have to say the VRAM is, this is because the VRAM handles so much graphical data, if you were to run out, your game may crash.

But capacity-wise, it’s pretty rare to find graphics cards that use upwards of 16GB of VRAM, the RTX 3090 which is unobtainable by most individuals due to its price has 24GB of VRAM. But the standard RTX 3080 has 10GB, and the RTX 3070 has around 8GB. The RTX 3080 and 3070 are no slouches of a graphics card, they’re pretty high-end, but they come nowhere close to 16GB.

But nonetheless, the dedicated graphics card will have RAM allocated towards it especially if it were to run out of VRAM, so in a way, the RAM is something the graphics card can fall back on. Using the shared GPU memory for gaming can be quite challenging, you will constantly stutter and the game will be at risk of crashing.

Verdict

In conclusion, shared GPU memory is allocated and deallocated accordingly, the system is pretty good at handling RAM so there’s really no need to modify this value. If your iGPU is idle, it will essentially have the minimum amount of shared GPU memory allocated towards it.

With dedicated graphics cards, they will use the shared GPU memory once they run out of VRAM, this is to keep the game or application running. In this case, you will notice diminished performance in the form of less FPS or stuttering, or maybe even crashes.