Recommended Free Tools
An integrated GPU usually does not have a separate VRAM pool to free. Intel integrated graphics use system RAM, while a discrete GPU has its own video memory. You can test whether changing display routing leaves more room on the discrete GPU, but only a controlled before-and-after comparison on your PC can establish that. A large “shared GPU memory” figure in Windows is a limit, not proof that RAM is reserved or currently in use.
What “freeing VRAM” can mean
These are two different memory questions:
- Does the iGPU use memory? Yes. Intel integrated graphics draw on system memory rather than a separate VRAM pool. Increased graphics use can reduce RAM available to other work.
- Can using the iGPU leave more room on a discrete GPU? Potentially, if moving display work off the discrete GPU reduces its live memory allocations. That is not the iGPU freeing its own VRAM, and the result depends on the actual setup.
Windows figures can be misleading if treated as physical capacity or current use. Intel says the reported Shared System Memory is not an ongoing reservation. Intel drivers may also report 128 MB of fictitious Dedicated Video Memory for compatibility with applications that do not understand unified memory; that figure is not physical iGPU VRAM. See Intel’s graphics memory FAQ.
How to compare memory use on Windows
For a meaningful comparison, keep the inference workload and other conditions consistent. Record the adapter and workload phase beside each reading; model loading and generation may have different memory needs.
- Record the setup. Note the CPU and GPUs, Windows version, graphics driver, inference backend, model and quantization, context length, GPU offload settings, display connection or assignment, and background applications.
- Capture an idle baseline. In Task Manager, record each adapter’s dedicated and shared GPU memory, system memory in use and available, and the inference process if it is already running. Intel graphics may appear to have GPU memory even though the iGPU has no separate VRAM pool. Intel’s Graphics Command Center memory FAQ also discusses interpreting graphics-memory reporting.
- Run the same inference. Keep the model, quantization, context length, offload layers, display routing, and background processes unchanged. Record the same counters after the model loads and again during generation.
- Compare like with like. Compare readings from the same workload phase. Check live memory use or free budget for each adapter, system RAM available, whether the model loads and remains stable, and generation performance only if you measure it. Do not credit a change to the iGPU if other settings or applications changed at the same time.
How to interpret the numbers
| Windows figure | What it can tell you | What it does not establish |
|---|---|---|
| Dedicated GPU memory | On a discrete card, a memory segment can represent VRAM. On an integrated GPU, it can instead be memory reserved by firmware or the driver. | A reported iGPU dedicated-memory value does not necessarily mean physical VRAM is present. Microsoft explains the distinction between GPU memory segments in its GPU Segments documentation. |
| Shared GPU memory | A system-memory limit or segment that graphics may use. | The maximum is not the same as memory continuously reserved or currently used. Intel explicitly says its reported shared-memory value is not an ongoing reservation. |
| GPU memory in use | A live-use counter, interpreted with its adapter and reporting method identified. | A maximum-capacity or adapter-properties number does not prove current use or reclaimed RAM. |
| System memory available | How much RAM remains available to the system at the time of the reading. | A high shared-memory maximum does not mean that amount has been taken from RAM; actual iGPU activity can still consume system memory. |
For more detailed Intel-runtime accounting, Intel’s Compute Runtime Device Memory Accounting guide distinguishes OS/kernel-usable size, runtime-advertised total, and currently free or usable memory. Where supported, it identifies zesMemoryGetState().free and the optional currUsableMemSize extension as device-wide readings. Their availability and behavior depend on the driver and runtime; a reported maximum alone is not a substitute for a live reading.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstall#1 Best Overall
- Powered by Radeon RX 9070 XT
- WINDFORCE Cooling System
- Hawk Fan
- Server-grade Thermal Conductive Gel
- RGB Lighting
Testing whether an iGPU leaves more room on a discrete GPU
If your question is whether an LLM fits better on your discrete card, compare that card’s live memory use or free budget before and after changing only the display routing. Keep the inference settings and workload stage the same, then note model-load success, stability, and performance if measured. An iGPU’s shared-memory ceiling cannot by itself show that the discrete GPU gained VRAM.
If the display is already routed through the iGPU, switching to it may not change anything. If it is routed through the discrete GPU, moving it may affect that card’s display allocations, but no gain should be assumed without measurements. Model weights, context/KV cache growth, and other applications can also change memory use while inference runs.
Rank #2
- Powered by the NVIDIA Blackwell architecture and DLSS 4
- Powered by GeForce RTX 5070 Ti
- Integrated with 16GB GDDR7 256bit memory interface
- PCIe 5.0
- WINDFORCE cooling system
Intel Shared GPU Memory Override: limited availability
Intel documents a Shared GPU Memory Override for Core Ultra Series 2 and later, with Intel Graphics Software 25.26.1602.2 and driver 32.0.101.6974 or later. Intel specifies at least 10 GB of system memory, a default setting of 57%, and a restart for changes to take effect. Intel says, “Memory not currently in use by the GPU remains available to the system,” and warns that performance may be affected. This adjustable limit is specific to the stated supported hardware and software; do not assume it applies to other Intel iGPUs, AMD graphics, Apple silicon, or Linux. Details are in Intel’s Shared GPU Memory Override requirements.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What a useful result looks like
A useful test reports the before-and-after readings by adapter and workload phase, rather than comparing a shared-memory maximum with a live-use value. Include the inference backend, model and quantization, context length, offload settings, display routing, and background processes so the comparison can be interpreted. If only the shared-memory ceiling changes while live GPU use and system RAM readings do not, that is not evidence that memory was freed for inference.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesQuick Recap
Best Value
- Powered by the NVIDIA Blackwell architecture and DLSS 4
- Powered by GeForce RTX 5060
- Integrated with 8GB GDDR7 128bit memory interface
- PCIe 5.0
- WINDFORCE cooling system
Rank #4
- AI Performance: 767 AI TOPS
- OC mode: 2632 MHz (OC mode)/ 2602 MHz (Default mode)
- Powered by the NVIDIA Blackwell architecture and DLSS 4
- Axial-tech fan design features a smaller fan hub that facilitates longer blades and a barrier ring that increases downward air pressure
- A 2.5-slot design maximizes compatibility and cooling efficiency for superior performance in small chassis
Rank #3
- Powered by the NVIDIA Blackwell architecture and DLSS 4. System Requirements: Minimum 850W PSU with 16-pin 12V-2x6 (12VHPWR) connector required. Verify before purchasing.
- Military-grade components deliver rock-solid power and longer lifespan for ultimate durability. Compatibility: 348mm (13.7") length, 3.6 slots, 4.3 lbs. Confirm case clearance and slot spacing. GPU bracket included.
- Protective PCB coating helps protect against short circuits caused by moisture, dust, or debris
- 3.6-slot design with massive fin array optimized for airflow from three Axial-tech fans
- Phase-change GPU thermal pad helps ensure optimal thermal performance and longevity, outlasting traditional thermal paste for graphics cards under heavy loads
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
