Skip to content

What is Shared GPU Memory and How Does it Work?

Shared GPU memory refers to system memory (RAM) that is shared between the central processing unit (CPU) and the graphics processing unit (GPU). Instead of having its own dedicated video memory, the GPU can access part of the main system RAM to use for graphical tasks.

A Brief History

The concept of shared graphics memory has been around since the early days of personal computing. Back in 1984, the groundbreaking Apple IIe made use of this technique to reduce overall system costs while enabling basic gaming and graphics capabilities. The Commodore Amiga 1000 used chip RAM shared between its Motorola 68000 CPU and custom chips like Denise, which handled video output.

Over time, the approach improved with techniques like Intel‘s Dynamic Video Memory Technology (DVMT) first introduced in 2008. DVMT allowed for dynamically adjustable memory allotted to the integrated GPU based on workload, whereas early shared memory was static.

Integrated graphics and shared memory remain a key part of Intel, AMD, and Arm‘s system-on-chip (SoC) processors targeting mainstream laptops. As memory speeds and graphics technology continue evolving, shared GPU memory serves as an important stepping stone between basic display outputs and dedicated graphics.

How Shared GPU Memory Works

Motherboards have channels for installing specialized graphical RAM called GDDR (Graphics DDR) memory. GPUs designed for gaming and professional workloads have onboard GDDR memory amounts from 4GB up to 24GB on high-end cards.

Integrated graphics lack dedicated VRAM, but still need memory to process graphics. Shared GPU solutions connect to standard DDR or LPDDR system memory in the RAM slots. When running graphics workloads, a portion of this RAM is made available to the GPU.

For example, a laptop with an integrated Intel graphics chip and 8GB of DDR4-3200 system memory could allocate 4GB for the integrated GPU to use. This pool of shared memory is much slower than GDDR RAM, with 68 GB/s bandwidth for DDR4-3200 compared to 400 GB/s bandwidth on 16Gbps GDDR6 chips. But it allows the GPU to process light graphical workloads.

Leading Platforms Leveraging Shared GPU Memory

Here are some examples of integrated GPUs using shared system memory configurations:

  • Intel UHD Graphics on 11th & 12th Gen Core CPUs
  • AMD Radeon Graphics on Ryzen 5000U / 6000U mobile processors
  • Apple M1 / M2 GPUs in MacBooks and Mac minis
  • Qualcomm Adreno integrated GPUs on Snapdragon smartphone chips
  • Nvidia RTX A2000 Laptop GPUs for creative workflows

While fine for casual graphics, these solutions sacrifice performance versus their counterparts with dedicated VRAM. Discrete GPUs with onboard high-speed GDDR6 memory include:

  • Nvidia GeForce RTX 3000 Founders Edition cards
  • AMD Radeon RX 6000 RDNA 2 graphics cards
  • Nvidia RTX A5000 workstation GPUs

Diving Into the Pros and Cons

Shared GPU memory comes with a mix of benefits and limitations compared to dedicated graphics solutions:


  • Allows integrated graphics to perform basic GPU tasks
  • More memory efficient than isolating CPU and GPU pools
  • Saves costs for manufacturers and consumers
  • Enables slimmer device designs over discreet graphics


  • Shared memory bandwidth can severely bottleneck GPU
  • Graphics-intensive programs may crash or have errors
  • Cannot run most modern AAA games smoothly
  • Multitasking suffers with less total available system memory

Let‘s analyze the pros and cons in more detail…

Enabling Integrated Graphics

The key advantage of shared memory is expanding the capabilities of integrated GPUs. Entry-level laptops like the Lenovo IdeaPad 1 with basic AMD Radeon graphics would not functionally work without shared memory access. GPUs rely on memory bandwidth to process textures, polygons, pixels, and complex graphics pipelines. While no replacement for dedicated VRAM, shared system memory unlocks a base level of graphical functionality from integrated chips.

More Efficient Use of Total RAM

By pooling memory between the CPU and iGPU, shared configurations make better use out available RAM capacity. Separate memory pools for computing and graphics tasks can sit unused if workload is heavily skewed towards one component. With dynamic shared allocation, RAM shifts to where it‘s needed most transparently.

Cost and Design Wins

Using system memory sidesteps the expense of adding dedicated graphics memory. For laptop manufacturers, relying on shared memory from baseline DDR4/LPDDR4 allows them to hit aggressive price points for entry-level models. Shared graphics also enable sleek, lightweight designs not possible while including a bulky discrete GPU. Tablets, Ultrabooks, and productivity laptops emphasize portability where shared memory permits basic graphics in slim unibody chassis.

However, the performance sacrifices lead to some significant drawbacks…

Bottlenecked Bandwidth

Even with advanced memory like LPDDR5 or DDR5, system RAM cannot match the throughput of dedicated GDDR6 memory. Copying textures or geometry data to the GPU becomes massively bottlenecked. This leads to choppy frame pacing, input lag, and crashing in graphics-intensive programs.

While bandwidth continues improving, discrete VRAM maintains a large lead. Nvidia‘s flagship RTX 4090 graphics card boasts up to 1.8 TB/s bandwidth from 24GB of 21Gbps GDDR6X memory. DDR5-6400 offers 51 GB/s in comparison – not nearly enough for smooth 4K gameplay.

Gaming and Graphics Pitfalls

Between bandwidth ceilings and insufficient capacity, integrated GPUs with shared memory choke on gaming. Playing Counter-Strike: Global Offensive dips below 60 FPS. Resource-heavy AAA titles like Elden Ring or Cyberpunk 2077 become unplayable slideshows. Performance issues plague creative software like Blender and Unreal Engine as well.

Graphics workarounds like lowering resolutions and settings help a bit, but often still result in crashes or glitches. Shared memory leaves no headroom for high-resolution textures, geometry, post-processing effects that dedicated VRAM excels at powering.

Multitasking Suffers

With a large portion of RAM allocated to graphics, less remains for regular CPU/system memory. This directly cuts into multitasking capability and exacerbates the limited memory bandwidth. Quickly alt-tabbing between Chrome browser tabs while running a Photoshop project causes lag across both programs, where 16GB+ dedicated to the CPU avoids slowdown.

In today‘s multi-app world, 4GB shared between integrated graphics and the OS strains workflows. Photographers batch editing RAW images in Lightroom will constantly pinch memory bandwidth already throttling the entry-level GPU.

Based on these pros and cons, how does shared GPU memory ultimately compare against dedicated graphics solutions?

Performance and Capability Comparison

Here we will dive into some real-world benchmarks and program tests that reveal key differences between shared integrated graphics and dedicated discreet video cards.

Gaming Frame Rates

One area where limitations become extremely apparent is gaming performance. This table shows average FPS benchmarks across six popular game titles:

Game/Application Integrated: Intel Iris Xe DG1
(Shared memory)
Entry-Level: Nvidia GTX 1050 Ti
Mid-Range: AMD RX 6600
DOTA 2 55fps 130fps 260fps
GTA V 24fps 73fps 144fps
Call of Duty: Warzone Unplayable 68fps 150fps
Red Dead Redemption 2 Unplayable 51fps 120fps
Cyberpunk 2077 Unplayable 33fps 60fps

With shared memory struggling in DOTA 2 and GTA V, more intensive titles become completely unplayable. It takes at least an entry-level discrete GPU like the GTX 1050 Ti to hit 30+ FPS, while mid-range offerings achieve buttery smoothness.

Content Creation Benchmarks

| Program | Integrated: AMD Ryzen 7 6800U
(Shared memory) | Nvidia RTX A2000 Laptop GPU | Nvidia RTX A5000 Desktop GPU |
|Blender BMW Render| 18min 22sec | 4min 33sec | 1min 51sec |
|DaVinci Resolve 4K Export| 42min 17sec | 16min 51sec | 6min 22sec |
|Unity Terrain Editing Avg FPS| 31fps | 92fps | 258fps|

For creative workflows like 3D/video editing and game development, dedicated VRAM also massively outperforms shared configurations. Complex production tasks grind to a halt on integrated graphics.

While the benchmarks reveal poor performance for gaming and graphics workloads, shared memory does fit the bill for less intensive tasks.

Ideal Use Cases

Application Integrated Graphics w/ Shared Memory
Office productivity ✅ Yes
Web browsing, video streaming up to 4K ✅ Yes
Photo editing in Adobe Lightroom ✅ Yes
1080p Video editing in Premiere Pro ✅ Yes
Casual gaming (DOTA 2, CS:GO, Minecraft) ✅ Yes

For general computer use like writing documents, Web browsing, basic photo and video editing, and non-demanding games, integrated GPUs with shared memory function adequately. Just don‘t expect to run new AAA titles or smoothly render 3D/4K content.

Cost-Benefit Analysis

At under $800, budget laptops like the HP ENVY featuring Intel Iris Xe graphics with shared memory comfortably handle everyday tasks for most mainstream users. The largely capable system costs a fraction compared to a $2000+ high-end gaming notebook required to power modern games smoothly.

However, there are some long term advantages to spending more upfront on a mid-range discreet GPU:

  • Play modern games on high settings for 3+ years
  • Edit 4K video projects and complex 3D models faster
  • More future proof solution as applications demand more GPU acceleration
  • Avoid crashes and instability from overloaded shared memory
  • Cost per frame, render speed in GPU-accelerated software is lower

For users without graphics-intensive needs, shared memory integrated GPUs offer solid value. But content creators, hardcore gamers, and power users gain tangible productivity returns from discrete VRAM.

The Road Ahead for Shared and Dedicated GPU Memory

As integrated graphics continue evolving, we will likely see high-bandwidth LPDDR5 system memory combined with new mixed precision and upscaling techniques to better optimize performance from shared memory configurations.

Dedicated graphics memory will ramp up capacities and throughputs, reaching up to 48GB on upcoming AMD GPUs based on leaks. Nvidia‘s Hopper data center architecture implementing HBM3 indicates extremely high bandwidth video memory will downscale to prosumer desktop cards eventually.

Exciting hybrid solutions like AMD‘s Rembrandt Ryzen 6000 mobile processors packaged with RDNA2 graphical chiplets point towards a multi-chip future fusing integrated and dedicated graphics. Shared system memory could tunnel content to help manage the multi-GPU workloads.


Shared GPU memory enables compelling form factors with basic graphical capabilities perfect for mainstream laptops and pre-built desktops doing light gaming, content consumption and acceleration of non-intensive workflows. Dedicated VRAM still easily defeats shared configurations for AAA gaming, 3D modeling and GPU compute programs. But shared memory sustains the concept of good enough performance for many general computing needs and entry-level graphics use cases.

Advances in memory like DDR5 and LPDDR5 combined with software and hardware optimizations provide room for shared graphics solutions to slightly improve. However, discrete GPUs with abundant, blazing-fast GDDR6/X VRAM aren‘t going anywhere. Especially as immersive metaverse applications drive graphics hardware to new levels over the coming decade.