MSI PRO MAX EDGE AI+ launches with Ryzen AI Max+ 395 to run large language models locally

MSI PRO MAX EDGE AI+ Runs Large AI Models Locally

MSI has launched the PRO MAX EDGE AI+, a compact desktop system designed for local artificial intelligence inference. The machine combines an AMD Ryzen AI Max processor with a large unified-memory pool, allowing it to accommodate language models that would exceed the dedicated graphics memory available on most consumer GPUs. Configurations extend to the AMD Ryzen AI Max+ 395, which combines CPU cores, RDNA 3.5 integrated graphics and an XDNA 2 neural processing unit. The graphics section provides up to 40 compute units, while the NPU is rated for as much as 50 TOPS. MSI lists up to 126 TOPS of combined AI performance for the complete platform. That combined figure represents the theoretical output of multiple processing engines rather than the performance of one component. The results obtained in real applications will depend on whether the software can use the CPU, Radeon GPU and NPU effectively. Model format, inference engine, numerical precision and memory bandwidth will also have a substantial effect on performance. The PRO MAX EDGE AI+ can be configured with up to 128 GB of LPDDR5X-8000 unified memory. As much as 96 GB can be assigned as variable graphics memory, giving the integrated GPU access to a substantially larger memory allocation than an ordinary consumer graphics card. Because the CPU and GPU share the same physical pool, data does not always need to be duplicated between separate system and graphics memory. MSI says the 128 GB configuration can run large language models with up to 120 billion parameters. This maximum will generally require some form of quantization because the operating system, inference software and context cache also consume memory. Model architectures with identical parameter counts can furthermore have different practical requirements. The system is housed in a four-liter enclosure and uses MSI Frozr AI Pro Cooling together with the company's Glacier Armor thermal design. The cooling arrangement is intended to control temperatures during sustained inference sessions. Independent testing will be necessary to determine its long-duration performance, operating temperatures and acoustic output. Multiple PRO MAX EDGE AI+ desktops can also be linked in a distributed cluster. MSI claims that a clustered configuration can run models containing up to 670 billion parameters by combining the memory and processing resources of several systems. The achievable performance will depend on the number of connected nodes, networking hardware and support within the selected inference framework. Running models locally can keep prompts, documents and other sensitive information away from external cloud services. This makes the system potentially useful for private document analysis, local assistants, software development and retrieval-augmented generation. However, local processing does not automatically provide complete security; organizations must still implement suitable storage encryption, access controls and network policies. MSI has not disclosed pricing or a detailed regional availability schedule. Those details will be important when comparing the system with conventional GPU workstations, multi-GPU servers and rented cloud-computing services. Specification MSI PRO MAX EDGE AI+ Processor platform AMD Ryzen AI Max 300 Series Maximum processor AMD Ryzen AI Max+ 395 Integrated graphics AMD RDNA 3.5, up to 40 compute units Neural processor AMD XDNA 2 NPU NPU performance Up to 50 TOPS Combined AI performance Up to 126 TOPS System memory Up to 128 GB LPDDR5X-8000 unified memory Variable graphics memory Up to 96 GB Single-system model support Up to 120 billion parameters, subject to model format and quantization Clustered model support Up to 670 billion parameters across multiple systems Chassis volume Approximately 4 liters Cooling MSI Frozr AI Pro Cooling and Glacier Armor Pricing Not announced Source: MSI

MSI Launches PRO MAX EDGE AI+, Featuring Up To Ryzen AI Max+ 395 For Delivering 126 AI TOPS Of Power

MSI has one more AI mini PC for enthusiasts that can deliver high-performance computing and carry out AI workloads locally. MSI Debuts PRO MAX EDGE AI+ Mini PC, Supporting Multi-System Clustering for Handling LLMs With Up To 670 Billion Parameters Some vendors still deploy Ryzen AI 300 chips in their mini PCs due to their powerful computing capabilities and strong AI performance, making them perfect for small AI systems. MSI has already released a couple of AI mini PCs based on the latest AMD chips, and has now rolled out another one, featuring the Strix Halo chips. MSI says the PRO MAX EDGE AI+ is a powerful AI mini PC that offers exceptional AI performance and also keeps your data safe through enterprise-grade privacy and security. The mini PC brings the power of AMD Ryzen AI Max+ 300 chips and offers up to 16-core/32-thread chips like the Ryzen AI Max+ 395 processor. This chip is the flagship SKU in the lineup, bringing Radeon 8060S integrated graphics, which features 40 Compute Units that make gaming and carrying out intensive graphical workloads effortless. The processor has built-in XDNA 2 NPU, which offers 50 TOPS of AI power, while the combined AI power of the Ryzen AI Max+ 395 peaks at 126 TOPS, enabling the system to perform intensive local AI workloads. The MAX EDGE AI+ offers up to 128 GB LPDDR5-8000 unified memory, sufficient for running Large Language Models with up to 120 billion parameters. However, users can further boost AI performance through multi-system clustering, where multiple PRO MAX EDGE AI+ devices can handle LLMs with up to 670 billion parameters for even faster processing. To ensure the system can run cooler, the PRO MAX EDGE AI+ uses Frozr AI Pro cooling and MSI Glacier Armor inside a compact 4-liter chassis that keeps overheating at bay at sustained workloads. News Source: MSI Follow Wccftech on Google to get more of our news coverage in your feeds.