Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Ryzen AI MAX+ 395 local AI systems use unified memory, so the Radeon 8060S GPU and the CPU draw from the same physical LPDDR5x pool. You do not add a separate VRAM module; instead, you set how much of system memory the platform makes GPU-accessible. On Windows, the normal route is AMD Software: Adrenalin Edition → Performance → Tuning → AMD Variable Graphics Memory, followed by a reboot. If your computer does not show that control, use the vendor’s BIOS/UEFI setting, commonly called UMA Frame Buffer Size or iGPU Memory Size. AMD documents up to 96 GB for graphics workloads on a 128 GB system, but the largest value is not automatically best: leave memory for Windows, drivers, context, and other applications.
This guide explains what the setting changes, how to configure it, how much to reserve for different local-AI workloads, and why Linux-specific values such as 120 GB GTT memory should not be treated as universal consumer settings.
As an Amazon Associate I earn from qualifying purchases.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →What does GPU memory mean on a Ryzen AI MAX+ 395?
The Ryzen AI MAX+ 395 is a shared-memory platform. AMD lists 256-bit LPDDR5x memory, up to 128 GB of total system memory, and Radeon 8060S integrated graphics in the processor specifications. The GPU-accessible amount is therefore a portion of that same physical pool, not a separate bank of dedicated VRAM. You can confirm the platform’s headline specifications in AMD’s Ryzen AI MAX+ 395 product page.
When a local model is loaded, the runtime needs memory for weights, the key-value (KV) cache that grows with context, temporary tensors, and the operating system. A larger GPU-accessible slice lets more layers remain on the GPU. A smaller slice leaves more immediately available to the CPU and desktop. If the model does not fit in the GPU-accessible portion, a runtime can use mixed CPU/GPU loading, but that generally changes the performance profile compared with full GPU offload.
Popular on Amazon
- 【90 Degree Full Function USB C Adapter】Supports the latest USB4 standard for up to 40Gbps of bandwidth, 8K video support, and up to 240W Fast charging; Support USB PD 3.1 Extended Power Range (EPR) including 140W, 180W and 240W. Your productivity is easily doubled. *Warm Tip: For the best fit, please make sure the USB-C port is fully exposed; thick or bulky cases may affect compatibility.
- 【Reliable Quality】Made from high-quality, scratch-resistant aluminum alloy, delicate metal frame. get excellent wear resistance and heat dissipation. Non-Slip Design for easier insertion and removal.
- 【Small and portable】Ultra thin size, ultra compact design, a petite physique that matches the minimalist design of your modern laptops. Coming with a slim figure that won't block the port next to the one it's plugged into.
- 【Compatibility】Widely work with most usb c, thunderbolt 4/3 devices. Compatibilit for ROG Ally, Steam Deck, Switch, MacBook Pro, Quest 3 VR headset, Surface Book 2, iPhone 18/17/16/15, Samsung Galaxy S23/S22/S21, Note 20/10, Google Pixel, ChromeBook, etc.
- 【Excellence Customer Support】If you experience any issues or cable failure, simply reach out to our dedicated customer service team. We provide a prompt response within 18 hours, providing you with the necessary technical support to resolve any concerns.
AMD describes the principle in its ROCm material: with a higher GPU memory allocation, and depending on system-vendor and BIOS support, more layers can be offloaded to the GPU. That is an allocation choice, not a promise that every model will run faster or that every application will use the entire reserved amount.
How much memory can the Ryzen AI MAX+ 395 allocate to graphics?
On a 128 GB system, AMD has published a maximum of up to 96 GB dedicated to graphics workloads. This is a platform ceiling documented by AMD, not a recommendation to select 96 GB on every machine. Your firmware may offer fewer choices, and a lower-capacity configuration cannot expose 96 GB.
AMD’s AI setup guidance also recommends leaving roughly 20% of system RAM for the operating system. On a 128 GB computer, 20% is about 25.6 GB. A 96 GB graphics allocation leaves approximately 32 GB for the OS and applications, which is why that value can be practical on a high-memory system. The right choice still depends on your model, context length, desktop workload, and whether other services are running.
| System memory | Published or illustrative GPU-accessible range | Memory left for OS and applications | How to interpret it |
|---|---|---|---|
| 128 GB | Up to 96 GB in AMD’s published platform guidance | About 32 GB at a 96 GB setting | A high-memory local-AI configuration; verify that your firmware exposes the value |
| 128 GB | 64 GB in an AMD ROCm benchmark environment | About 64 GB before runtime overhead | An example configuration, not a universal optimum |
| Other capacities | System-vendor and firmware dependent | Varies with the selected value and running software | Use the choices your own Adrenalin panel or BIOS provides |
The 64 GB figure comes from an AMD ROCm article describing a 128 GB unified-memory system running Ubuntu 24.04 LTS, ROCm 7.2.1, and Ollama 0.20.x. It demonstrates one workable environment; it does not establish a recommended value for every model or operating system.
How do I change GPU-accessible memory in Windows?
Use AMD Software: Adrenalin Edition first. The control is called AMD Variable Graphics Memory and appears only when the system firmware and OEM software expose it.
- Save work and close memory-heavy applications. Changing the allocation requires a restart, and unsaved work will be lost if you reboot without closing it.
- Open AMD Software: Adrenalin Edition. Use the installed AMD Software entry from the Windows Start menu.
- Open Performance, then Tuning. Look for the AMD Variable Graphics Memory section or control.
- Select an allocation. Choose a value that fits the model and context you intend to run while preserving headroom for Windows and other applications. Do not assume that the highest displayed number is the correct choice.
- Apply the change and reboot. The new GPU-accessible amount is normally applied after the restart.
- Verify after Windows starts. Reopen Adrenalin and your AI runtime, and check that the reported graphics-memory or unified-memory limit reflects the new setting. Different runtimes label shared memory differently, so compare the value with the setting you selected rather than expecting a discrete VRAM module.
AMD notes that some OEM systems hide AMD Variable Graphics Memory. If the menu is absent, that is a platform-support difference rather than an indication that the processor lacks unified memory. Continue with the BIOS/UEFI method below and consult the computer maker’s current documentation for the exact menu.
Free tools Windows power users keep installed
One-click scans. No signup required.
Also popular on Amazon
- 5-in-1 Connectivity: Equipped with a 4K HDMI port, a 5 Gbps USB-C data port, two 5 Gbps USB-A ports, and a USB C 100W PD-IN port. Note: The USB C 100W PD-IN port supports only charging and does not support data transfer devices such as headphones or speakers.
- Powerful Pass-Through Charging: Supports up to 85W pass-through charging so you can power up your laptop while you use the hub. Note: Pass-through charging requires a charger (not included). Note: To achieve full power for iPad, we recommend using a 45W wall charger.
- Transfer Files in Seconds: Move files to and from your laptop at speeds of up to 5 Gbps via the USB-C and USB-A data ports. Note: The USB C 5Gbps Data port does not support video output.
- HD Display: Connect to the HDMI port to stream or mirror content to an external monitor in resolutions of up to 4K@30Hz. Note: The USB-C ports do not support video output.
- What You Get: Anker 332 USB-C Hub (5-in-1), welcome guide, our worry-free 18-month warranty, and friendly customer service.
What if AMD Variable Graphics Memory is missing?
Use the system firmware as a fallback. Restart the computer and enter UEFI/BIOS using the key shown during startup; the key and screen layout vary by manufacturer. Search menus under names such as UMA Frame Buffer Size, Integrated Graphics, iGPU Memory Size, or Dedicated GPU Memory. AMD’s own guidance uses these labels as examples, not as a universal menu map.
- Enter UEFI/BIOS and locate the integrated-graphics or UMA memory setting.
- Record the original value before changing anything. A photo of the page makes recovery easier.
- Choose a supported value that leaves adequate system memory. Firmware options can be expressed in gigabytes, megabytes, or a short list of presets.
- Save changes and reboot.
- Check Windows and the AI runtime. If the machine fails to boot correctly or graphics become unavailable, return to UEFI/BIOS and restore the recorded value.
Do not expect every Ryzen AI MAX+ 395 computer to offer the same maximum, increments, or labels. OEM firmware can impose a lower ceiling than AMD’s 96 GB platform figure, and a BIOS update can change available choices. Follow the documentation for your exact model before applying a setting.
How should I choose an allocation for local models?
Start with the model’s actual memory requirement, then reserve operating-system headroom. There is no single allocation that AMD or the available documentation identifies as optimal for every model, quantization, context length, or runtime.
1. Estimate the workload, not just the model-file size
Model weights are only one part of the requirement. Longer context increases the KV cache, and generation can need temporary workspace. A model that appears to fit when loaded may still fail when you increase context or run multiple requests. Leave additional margin instead of setting the allocation exactly equal to the file size.
2. Keep roughly 20% available to the OS
Use AMD’s roughly 20% rule as a starting point. Windows, drivers, the runtime, browser tabs, file caches, and background services all consume memory outside the GPU allocation. If you routinely run a desktop, a development environment, and a local model together, a conservative allocation can be more usable than the maximum.
3. Decide whether full GPU offload is necessary
Smaller workloads may fit entirely in the GPU-accessible portion. Larger workloads can use a mix of GPU and CPU memory when the runtime supports it. More GPU-resident layers can reduce the amount of work performed by the CPU, but mixed loading may still be the practical way to run a model that exceeds the available slice.
Also popular on Amazon
- Sleek 7-in-1 USB-C Hub: Features an HDMI port, two USB-A 3.0 ports, and a USB-C data port, each providing 5Gbps transfer speeds. It also includes a USB-C PD input port for charging up to 100W and dual SD and TF card slots, all in a compact design.
- Flawless 4K@60Hz Video with HDMI: Delivers exceptional clarity and smoothness with its 4K@60Hz HDMI port, making it ideal for high-definition presentations and entertainment. (Note: Only the HDMI port supports video projection; the USB-C port is for data transfer only.)
- Double Up on Efficiency: The two USB-A 3.0 ports and a USB-C port support a fast 5Gbps data rate, significantly boosting your transfer speeds and improving productivity.
- Fast and Reliable 85W Charging: Offers high-capacity, speedy charging for laptops up to 85W, so you spend less time tethered to an outlet and more time being productive.
- What You Get: Anker USB-C Hub (7-in-1), welcome guide, 18-month warranty, and our friendly customer service.
4. Increase in measured increments, not all at once
If your current model is constrained, raise the allocation one firmware preset at a time, reboot, and confirm that the desktop remains responsive and the runtime still initializes. This makes it easier to identify a setting that leaves too little memory for the OS.
5. Treat context length as a separate decision
Reducing context can lower KV-cache pressure without changing firmware. If a model loads but crashes when a long prompt is supplied, test a shorter context or fewer concurrent requests before assuming that the BIOS allocation itself is wrong.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →What is the difference between Windows allocation and Linux UMA or GTT memory?
Windows’ Variable Graphics Memory and BIOS UMA settings select the GPU-accessible portion exposed by the platform. Linux can use the same unified-memory architecture, but some advanced configurations also adjust kernel and driver bookkeeping. Those paths are not interchangeable.
AMD’s ROCm documentation describes a 128 GB Ryzen AI Max+ environment with 64 GB GPU-accessible memory on Ubuntu 24.04 LTS, ROCm 7.2.1, and Ollama 0.20.x. In that kind of UMA setup, a smaller model may be fully GPU-offloaded while a larger one uses both CPU and GPU memory. The software versions and memory value are examples from the article, not guaranteed requirements or performance results for your installation.
A separate 2026 AMD technical article about a specific Framework Desktop Ryzen AI MAX+ 395 128 GB system reports a BIOS iGPU Memory Size of 512 MB and Linux parameters ttm.pages_limit=30720000 amdgpu.gttsize=120000, exposing 120000M of GTT memory in that cluster configuration. The same article states that the system’s BIOS dedicated-VRAM limit is 96 GB. The 120 GB figure is a Linux TTM/GTT configuration for that particular setup, not a new universal BIOS maximum or a consumer recipe.
Before changing Linux kernel parameters, check the current kernel, amdgpu driver, distribution, bootloader, and system documentation. An invalid value can prevent the graphical session from starting or reduce stability. Keep a known-good boot entry so you can remove the parameters if the system does not start normally.
Recommended Free Tools
Also popular on Amazon
- Portable and powerful USB-C HUB: BENFEI USB Type-C HUB, with super-soft and knot-free silicone woven design cable, meets most mobile office needs. Compact, lightweight, stylish, and powerful portable USB C Hub equipped with 1 x HDMI port, 1 x 100W charging, and 3 x USB ports. 18-month warranty, 24-hour response, to ensure you feel at ease when using our product.
- Design centered on comfort and reliability: Thanks to BENFEI's end-to-end in-house cable production capability, in-house PCBA and assembly capability, using the industry's most advanced silicone woven design and process, 20cm cable in length, no knots, super-soft, the HUB is easy to use in all scenarios: laptop, tablet, stand etc. Super-soft, 25000+ life cycles, to meet your daily carrying and office needs.
- 100W Charging: Support up to 90W USB C pass-through charging via Type-C port to keep your laptop powered. 10W is reserved for other interface operations. No data and video function on the Type-C port.
- 4K HDMI Display: The HDMI port supports media display at resolutions up to 4K 30Hz, keeping every incredible moment detailed and ultra vivid. Please note that the C port of the Host device needs to support video output.
- Transfer Files in Seconds: Transfer files and from your laptop at speeds up to 10 Gbps with USB A 3.2 port. Extra 2 USB A 2.0 ports are perfectly for your keyboards and mouse.
How can I tell whether a model fits?
Use the runtime’s load message and system monitors together. A successful model load does not prove that the full workload is GPU-resident; the runtime may have placed some layers in CPU memory. Look for separate indicators for GPU layers, unified memory, or CPU/GPU placement when your application provides them.
- Fits comfortably: the model loads, the intended context is usable, and the desktop retains substantial free memory.
- Borderline: loading succeeds but longer prompts, multiple requests, or other applications trigger allocation failures or heavy system pressure.
- Too large for the selected slice: the runtime refuses to load, reports an out-of-memory error, or falls back to extensive CPU loading.
Change one variable at a time: allocation, context length, quantization, or concurrent requests. That makes failures easier to diagnose and avoids mistaking a software limit for a firmware limit.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What are the common failure modes?
The Variable Graphics Memory control is not present
Use the OEM BIOS/UEFI route. Check for UMA Frame Buffer Size, Integrated Graphics, iGPU Memory Size, or a similarly named option. If no option exists, the vendor may have fixed the allocation or restricted changes through firmware.
The selected value is not available
Firmware exposes only supported presets. A 128 GB system can still have a lower vendor-defined maximum, and smaller systems cannot use the 96 GB figure from AMD’s 128 GB guidance. Install only firmware updates intended for your exact model and recheck the vendor’s release notes.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Windows becomes sluggish after increasing the allocation
Reduce the setting and reboot. The graphics slice is shared physical memory; reserving more for GPU workloads leaves less for Windows and applications. Close background programs and keep the allocation near the point where the target model fits with headroom.
The model loads but generation fails at longer context
Lower the context length, reduce simultaneous requests, or choose a smaller/less memory-intensive model configuration. The KV cache and temporary buffers can exceed the margin that was available during initial loading.
Also popular on Amazon
- 【Read before You Buy】 Note: This adapter is USB 2.0 version and cannot transmit HDM, VGA, or other video signals. Nevertheless, they are fully capable of charging devices,duplicating files, and can be seamlessly integrated with USB-C earphones, wireless network adapters, and SD card adapters.
- 【Easy to Use and Safe】Thanks to modern manufacturing techniques, we've made these adapters sufficiently compact so that they can remain inserted in your device without noticeable protrusion or inconvenience. This means you don't have to constantly plug and unplug them, reducing the likelihood of wear and misalignment, and making them more convenient to carry around.
- 【Reliable Quality】Engineered for enhanced durability, durable aluminum alloy shell material,get excellent wear resistance and heat dissipation,the male USB C and female USB connectors & the male USB and female USB C connectors are designed to be plugged and unplugged up to 20000 times. Provides data transfer, device charging, memory space expansion, and additional computer peripherals.
- 【Think more for you】Do you need to connect a USB-C-enabled device to an odd device that only has USB-A ports? Then these USB-A male to USB-C female adapters are made for you.Simply insert the adapter into the USB-A port on the device, and you can instantly connect your USB-C device to the device that originally didn't have the correct connection port.
- 【Trusted xiwxi Quality and Service】As a responsible brand seller, you will receive xiwxi's 12-month Quality assurance and lifetime friendly customer service, you're not just buying a product,you're investing in peace of mind with all-round quality and exceptional customer service.
Linux reports less usable memory than expected
Distinguish BIOS-dedicated memory from driver-managed GTT memory. A Linux article’s 120000M GTT value does not mean every system can expose 120 GB, and it does not replace the firmware’s own limits. Verify driver and kernel messages before changing parameters.
Should I always set a 128 GB system to 96 GB?
No. Ninety-six gigabytes is AMD’s published upper figure for graphics workloads on a 128 GB system, not a universal best setting. Select it only when the model and context need that capacity and the remaining memory is sufficient for the operating system and applications. For a desktop that must remain responsive, a lower value such as the 64 GB example documented by AMD may be a better starting point, provided it fits the workload.
What does a higher allocation actually improve?
A higher allocation can let a runtime offload more model layers to the GPU. AMD ROCm describes this as a way to obtain additional performance, subject to system-vendor and BIOS support. It does not increase the physical memory installed in the computer, and it cannot guarantee a particular token rate, latency, or model size. CPU speed, GPU architecture, runtime support, quantization, context, and concurrent work still matter.
Safe configuration checklist
- Confirm the computer has a Ryzen AI MAX+ 395 and note its installed memory capacity.
- Check whether AMD Software exposes Performance → Tuning → AMD Variable Graphics Memory.
- If it does not, read the OEM’s BIOS/UEFI instructions for UMA or integrated-graphics memory.
- Write down the original firmware value before changing it.
- Reserve roughly 20% of system RAM for Windows or Linux and other applications.
- Choose enough GPU-accessible memory for weights, context, and runtime overhead rather than using the maximum automatically.
- Apply the change and reboot.
- Verify the value in the operating system and in the AI runtime.
- Test the intended context length and concurrency, not only the initial model load.
- Keep a recovery path: the previous BIOS value and a boot option without experimental Linux parameters.
Key takeaways
The Ryzen AI MAX+ 395 does not have user-replaceable dedicated VRAM for local AI. It uses unified LPDDR5x memory, with a configurable GPU-accessible portion controlled by Adrenalin when supported or by an OEM BIOS/UEFI setting when it is not. AMD documents up to 96 GB for graphics workloads on a 128 GB system, while its guidance recommends leaving about 20% of RAM for the operating system. Configure the smallest allocation that gives your model and context enough room, then increase it only when the workload requires more GPU-resident layers. Linux TTM/GTT values such as 120 GB belong to specialized configurations and should not be copied blindly.
Frequently Asked Questions
Can I add VRAM to a Ryzen AI MAX+ 395 laptop or desktop?
No. The Radeon 8060S uses the processor’s shared LPDDR5x memory pool, so there is no separate VRAM module to install. You can only change the GPU-accessible portion exposed by the firmware and operating system.
Does changing UMA memory erase my files or reinstall Windows?
Changing the graphics-memory allocation normally changes a firmware setting and requires a reboot; it does not reformat storage. Save work first, and record the old value so you can restore it if the system behaves unexpectedly.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows 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 reinstallWhy does my AI program show less memory than the BIOS setting?
The runtime may reserve memory for drivers, the operating system, workspaces, or the KV cache, and it may report only currently available memory. Shared-memory accounting also differs between Windows and Linux, so the application’s figure will not always equal the firmware preset.
Can a BIOS update change the available GPU-memory choices?
Yes. Allocation controls and their limits are vendor- and firmware-dependent. After an update, recheck the OEM documentation and confirm the selected value rather than assuming the previous presets remain unchanged.

