AnnouncedRyzen AI Max PRO 400 (Gorgon Halo)

AMD · Ryzen AI Max PRO 400 Series

AMD Ryzen AI Max PRO 490

12 Zen 5 cores, 32-CU Radeon 8050S graphics, and 192 GB unified memory for uncompromising mobile AI workloads.

Local LLM Inference3D RenderingEngineering SimulationData ScienceVisual Computing

Cores / Threads

12/ 24

Base / Boost

3.2/ 5 GHz

PCIe Lanes

16

L2 Cache

12MB

L3 Cache

64MB

TDP

55W

Socket

FP11

Verdict

8.9/ 10

89

Quick Verdict

An exceptionally powerful mobile APU that bridges the gap between portable computing and datacenter-class memory capacity, perfectly suited for enterprise AI and heavy rendering workloads.

Best for:Local LLM Inference3D RenderingEngineering SimulationData ScienceVisual Computing

Overview

Launch

2026

Status

Announced

Generation

Ryzen AI Max PRO 400 (Gorgon Halo)

Market

Commercial AI PC / Mobile Workstation

About this CPU

The AMD Ryzen AI Max PRO 490 is a high-performance 12-core APU for mobile workstations and commercial AI PCs, featuring a 32-CU Radeon 8050S integrated GPU, an XDNA 2 NPU, and support for up to 192GB of unified LPDDR5X memory.

As the mid-tier offering in the Ryzen AI Max PRO 400 family, the 490 combines 12 Zen 5 CPU cores, a 32-CU RDNA 3.5 integrated GPU, and a dedicated XDNA 2 NPU. Its defining feature is the 256-bit memory bus supporting up to 192GB of LPDDR5X, allowing up to 160GB to be allocated as VRAM for AI or GPU tasks.

This makes it uniquely capable of running large language models and complex 3D datasets locally on a laptop or mini PC, bypassing the VRAM limitations of traditional discrete mobile GPUs. It is built specifically for enterprise users and creators who need desktop workstation capabilities in a portable form factor.

Specifications

ArchitectureZen 5 / RDNA 3.5 / XDNA 2
Manufacturing ProcessTSMC 4nm FinFET
Cores / Threads12 / 24
Base Clock3.2 GHz
Boost Clock5 GHz
L3 Cache64 MB
TDP55 W
SocketFP11
Memory TypeLPDDR5X
Memory SpeedLPDDR5X-8533
Memory ChannelsQuad-Channel (4)
Max Memory192 GB
PCIe Version / LanesPCIe 4.0 × 16
Integrated GraphicsYes
Quad-Channel16 PCIe Lanes
Target Audience
GamersStreamersContent CreatorsDevelopersWorkstation UsersOffice UsersStudents

Performance

Productivity
93Excellent

Exceptional multi-threaded performance comparable to high-end desktop CPUs, vastly accelerating compilation and rendering tasks.

Virtualization
85Very Good

Great for running multiple VMs, though ECC UDIMM support and PCIe lane counts are lower than HEDT platforms.

Gaming
72Good

While the 8050S iGPU is powerful for integrated graphics, it competes with entry-level discrete GPUs and is not intended for high-refresh 4K gaming.

Efficiency
86Very Good

Highly efficient at default TDP, but can draw significant power (up to 120W) when CPU and GPU are fully loaded simultaneously.

GamingGood
  • Radeon 8050S offers performance similar to a dedicated RTX 4050/4060 laptop GPU in rasterization.
  • Capable of smooth 1080p and 1440p gaming in most modern titles.
  • Ray tracing performance is limited compared to discrete alternatives.
  • Benefits from ultra-fast LPDDR5X memory bandwidth.
CreatorExcellent
Autodesk MayaBlender (Cycles & EEVEE)Adobe Premiere ProDaVinci ResolveSolidWorks Visualize
AI / MLExcellent
  • Capable of running large language models locally that require massive memory pools.
  • Unified memory architecture allows AI models to bypass traditional VRAM limits.
  • XDNA 2 NPU handles lightweight, persistent AI tasks efficiently.
Industry Impact
Gaming
Low
Workstations
High
Content Creation
High
Virtualization
Moderate

Architecture

TSMC 4nm FinFET

Process Node

Gorgon Halo

Codename

12C / 24T

Core Config

64 MB

L3 Cache

55 W

TDP

Architecture Overview

Built on TSMC's 4nm FinFET process, the Ryzen AI Max PRO 490 utilizes a sophisticated SoC design tailored for high bandwidth and compute density. The CPU complex features 12 full-performance Zen 5 cores, offering significant IPC improvements and 24 threads of parallel processing power optimized for heavy compute workloads. The inclusion of a 256-bit memory interface enables the massive 192GB LPDDR5X memory capacity, which is critical for feeding the 32 Compute Unit Radeon 8050S iGPU. This unified memory architecture means the CPU and GPU share a single, large pool of high-speed memory, eliminating the costly data transfers between discrete VRAM and system RAM. Furthermore, the XDNA 2 NPU provides dedicated AI acceleration, offloading background AI tasks efficiently while the CPU and GPU focus on real-time rendering or simulation logic. The FP11 socket integration ensures tight coupling between the silicon and the platform.

CPU Design

12 Zen 5 cores with SMT delivering 24 threads, designed for high-throughput workstation tasks.

Memory Subsystem

256-bit LPDDR5X controller supporting up to 192GB unified memory at 8533 MT/s.

PCIe & I/O

16 PCIe 4.0 lanes from the CPU for high-speed NVMe and peripheral expansion.

Overclocking

Locked multiplier; tuning is limited to cTDP adjustments and memory profiles.

Generation Comparison
AMD Ryzen AI Max PRO 390AMD Ryzen AI Max PRO 490
  • Increased maximum memory support from 128GB to 192GB.
  • Higher memory speed support (LPDDR5X-8533 vs LPDDR5X-8000).
  • Updated RDNA 3.5 integrated graphics (Radeon 8050S vs Radeon 890M equivalent).
  • Upgraded XDNA 2 NPU for higher AI TOPS.

Key Highlights

12 Zen 5 CPU Cores
High-performance core count designed for heavy multi-threaded rendering and compilation workloads.
32-CU Radeon 8050S Graphics
Powerful integrated GPU capable of driving multiple 4K displays and accelerating 3D rendering tasks.
192GB Unified Memory
Supports up to 192GB of LPDDR5X, with up to 160GB allocatable as VRAM for massive AI models.
XDNA 2 NPU
Dedicated AI accelerator for efficient, sustained AI inference and Copilot+ PC experiences.
AMD PRO Technologies
Enterprise-grade security, manageability, and firmware resilience for commercial deployment.
Strengths
  • Massive 192GB unified memory support for AI and 3D workloads.
  • 12 Zen 5 CPU cores deliver desktop-class multi-threaded performance.
  • Powerful 32-CU Radeon 8050S integrated graphics.
  • Dedicated XDNA 2 NPU for AI efficiency.
  • Enterprise-grade PRO manageability and security features.
Weaknesses
  • Extremely high system cost due to memory and silicon expenses.
  • Limited PCIe 4.0 lanes compared to desktop workstations.
  • Locked multiplier restricts traditional overclocking.
  • High power draw under full load requires robust cooling.
  • Niche product; overkill for standard productivity users.

History

Launch Date
2026
Status
Announced
Generation
Ryzen AI Max PRO 400 (Gorgon Halo)
Market
Commercial AI PC / Mobile Workstation
The Story

The Ryzen AI Max PRO 490 emerged from AMD's strategic evolution of the Strix Halo project, which originally sought to combine high-end CPU and GPU compute into a single mobile chip. As the demand for local AI processing exploded, the limitations of traditional 128GB memory ceilings became apparent for enterprise users running massive parametric models. </br>AMD responded by introducing the Gorgon Halo refresh, specifically updating the memory controllers to support 192GB of LPDDR5X.

The PRO 490 was positioned as the sweet spot for commercial users, offering 12 Zen 5 cores and a 32-CU integrated GPU. This provided a more cost-effective and thermally manageable solution compared to the flagship 16-core 495, while retaining the massive memory capacity. It represents a pivotal step in AMD's push to transform portable systems into legitimate AI workstations.

Improvements over Previous Generation

  • Increased maximum memory support from 128GB to 192GB.
  • Higher memory speed support (LPDDR5X-8533 vs LPDDR5X-8000).
  • Updated RDNA 3.5 integrated graphics (Radeon 8050S vs Radeon 890M equivalent).
  • Upgraded XDNA 2 NPU for higher AI TOPS.

Alternatives & Competitors

AMD Ryzen AI Max+ PRO 495
Choose the 495 if you need the maximum 16 CPU cores and 40-CU iGPU for the heaviest workloads.
Apple MacBook Pro with M4 Max
Best alternative if you are embedded in the Apple ecosystem and need high unified memory, though max memory is lower (128GB).
Desktop Workstation (Threadripper / Xeon)
Choose this if you need ECC memory, massive PCIe expansion, and don't require a portable form factor.
High-end Gaming Laptop with RTX 4080
Better choice if your primary focus is gaming and heavy 3D rendering that benefits from powerful discrete GPU rasterization.
Cloud AI Compute Instances
If you only need massive AI compute occasionally, renting cloud GPUs may be more cost-effective than buying a 192GB local machine.
Direct Competitors
Apple M4 MaxIntel Core Ultra 9 285HApple M3 UltraIntel Core i9-14900HXQualcomm Snapdragon X Elite

Should You Buy It?

Recommended for the right buyer

Enterprise users, AI researchers, and creative professionals who need to run large language models locally or render complex 3D scenes on a portable machine.

Avoid if…

  • You are looking for a standard laptop for web browsing and office tasks.
  • You prioritize battery life above all else.
  • You need extensive PCIe expansion for multiple discrete cards.
  • You are building a budget-friendly system.

Use Cases

Local AI Inference (70B+ Parameter Models)
Excellent
4K/8K Video Editing
Excellent
3D CAD/CAE Simulation
Excellent
Software Compilation
Very Good
Casual Gaming
Good

Interesting Facts

The 'Halo' codename signifies AMD's strategy to unify CPU and GPU compute with a massive memory pool, similar to Apple's Ultra chips.

It can allocate up to 160GB of its 192GB memory pool as VRAM, dwarfing the 16GB VRAM typical of high-end mobile discrete GPUs.

This chip allows laptops to run AI models with over 100 billion parameters locally without internet access.

The FP11 socket is a BGA connection, meaning the chip is soldered directly to the motherboard and not user-replaceable.

The XDNA 2 NPU is designed to handle 'Agentic AI' tasks, where the AI acts autonomously on the device.

Despite being a mobile chip, its multi-core performance rivals many desktop CPUs from the previous generation.

The Radeon 8050S integrated GPU features 32 Compute Units based on the RDNA 3.5 architecture.

AMD positions the 'Max' branding specifically for these high-memory halo products, distinguishing them from standard Ryzen AI chips.

The configurable TDP ranges from 45W to 120W, allowing OEMs to tune the chip for thin-and-lights or thick mobile workstations.

It natively supports USB4, which is electrically compatible with Thunderbolt 4.

People Also Ask

What is the AMD Ryzen AI Max PRO 490?

It is a 12-core mobile APU from AMD designed for workstations and AI PCs, featuring support for up to 192GB of unified memory and a powerful integrated GPU.

How much RAM does the Ryzen AI Max PRO 490 support?

It supports up to 192GB of LPDDR5X memory integrated into a unified pool accessible by both the CPU and GPU.

Can the Ryzen AI Max PRO 490 run large language models locally?

Yes, with up to 192GB of unified memory and 160GB allocatable as VRAM, it is specifically designed to run very large AI models locally on the device.

What graphics does the Ryzen AI Max PRO 490 have?

It features the AMD Radeon 8050S integrated graphics, which is based on the RDNA 3.5 architecture and features 32 Compute Units.

Is the Ryzen AI Max PRO 490 good for gaming?

While it has powerful integrated graphics, it is primarily designed for workstation and AI workloads rather than high-end gaming.

What socket does the Ryzen AI Max PRO 490 use?

It uses the FP11 socket, which is a BGA (Ball Grid Array) socket, meaning it is permanently soldered to the laptop or mini PC motherboard.

Does the Ryzen AI Max PRO 490 have an NPU?

Yes, it includes an XDNA 2 Neural Processing Unit (NPU) for efficient AI task offloading.

What is the difference between Ryzen AI Max PRO 490 and 495?

The 495 has 16 CPU cores and a 40-CU GPU, whereas the 490 has 12 CPU cores and a 32-CU GPU. Both support up to 192GB of unified memory.

What TDP does the Ryzen AI Max PRO 490 have?

It has a default TDP of 55W, but it is configurable by the manufacturer between 45W and 120W depending on the cooling capacity of the system.

Can you overclock the Ryzen AI Max PRO 490?

No, it has a locked multiplier and does not support traditional overclocking, though power limits can be adjusted within the cTDP range.

Frequently Asked Questions

Does the Ryzen AI Max PRO 490 support ECC memory?

No, it supports LPDDR5X which does not typically feature ECC in the traditional sense, though PRO technologies offer system resilience features.

Can I upgrade the RAM in a laptop with the Ryzen AI Max PRO 490?

No, the LPDDR5X memory is soldered directly to the motherboard and cannot be upgraded after purchase.

How many displays can the Radeon 8050S iGPU drive?

It can drive up to 4 external displays simultaneously via USB4/DisplayPort 2.1 and HDMI 2.1 connections.

What is the maximum boost clock of the CPU?

The maximum boost clock for the Zen 5 CPU cores is up to 5.0 GHz.

Is the Ryzen AI Max PRO 490 a desktop processor?

It is primarily a mobile and mini-PC processor (BGA FP11 socket), though it can be used in small form factor desktop workstations.

What PRO technologies are included?

It includes AMD PRO Manageability, AMD PRO Security (including AMD Shadow Stack and Memory Guard), and AMD PRO Resilience.

Does it support PCIe 5.0?

No, the processor provides 16 PCIe 4.0 lanes from the CPU.

How much L3 cache does the processor have?

It features 64MB of L3 cache, shared among the 12 CPU cores.

What process node is used for the Ryzen AI Max PRO 490?

It is manufactured on TSMC's 4nm FinFET process.

Can the GPU allocation be adjusted by the user?

Yes, in the BIOS or through AMD software, users can typically adjust how much of the 192GB memory pool is allocated to the GPU (up to 160GB).