© ROOT-NATION.com - Use of content is permitted with a backlink.
Large language models (LLMs) require substantial amounts of memory. To increase capacity, users of AI applications often rely on multi-processor systems. However, this approach has its limits. Sometimes the GPU offloads computations to the CPU, which brings CPU cache and system memory (DRAM) performance into focus. As a result, overall performance in AI workloads depends on the entire GPU–CPU–DRAM chain, highlighting opportunities for optimization.
ASUS R\&D labs demonstrated this on the latest AMD Ryzen processors. Recently, AMD introduced the Ryzen 9 9950X3D and Ryzen 9 9900X3D, featuring 3D V-Cache technology, which elevates the capabilities of the series. During extensive testing, ASUS identified a set of settings that significantly improve performance in AI workloads. These optimizations were implemented in the BIOS feature called AI Cache Boost, which can deliver up to a 12.75% performance increase in LLM tasks.
Read also: ChatGPT-5 from OpenAI: What the New AI Model Can Do
TABLE OF CONTENTS:
Activating AI Cache Boost
The feature is available on ASUS AMD 800-series motherboards paired with AMD Ryzen 9000 Granite Ridge processors. It is enabled through the BIOS under the Extreme Tweaker section. Among the optimizations is an increase of the Infinity Fabric clock (FCLK) to 2100 MHz, which boosts bandwidth between cores, cache, and memory modules – an improvement particularly beneficial for handling large LLM workloads.
For users who combine AI workloads with gaming, AI Cache Boost has virtually no impact on gameplay. For added stability, the FCLK frequency can also be manually reduced.
Performance with AI Cache Boost
Test configuration: ROG Crosshair X870E Hero motherboard, NVIDIA GeForce RTX 5090, and 2×16 GB DDR5 RAM. The Ryzen 9 9950X3D, 9900X3D, and 9800X3D were tested. In Geekbench AI (ONNX DirectML), performance gains ranged from 4% to 8% depending on the model. Using an RTX 4090, the highest increase – up to 7.57% – was seen with the Ryzen 7 9800X3D. In the UL Procyon AI Computer Vision Benchmark, the same processor achieved a 12.75% boost.
Read also: VRAM Memory: What It Is and Do You Really Need More Than 8 GB?
Turbo Game Mode
ASUS also tested combining AI Cache Boost with Turbo Game Mode, which optimizes the system in two ways:
- It disables the second CCD (Core Complex Die) on processors that have two CCDs, reducing latency in applications that don’t use all cores.
It turns off SMT (Simultaneous Multithreading). In some scenarios, disabling SMT improves efficiency in workloads that don’t require many threads.
These steps reduce latency in workloads that don’t require maximum thread counts. Turbo Game Mode should be enabled selectively, as it lowers overall multithreaded performance. When combined with AI Cache Boost, the Ryzen 9 9950X3D showed performance gains of up to 19.85% compared to default settings.
Conclusions
The new AMD Ryzen 9 9950X3D and 9900X3D processors significantly expand options for performance enthusiasts. With 32 and 24 threads respectively, these chips combine strong multithreaded performance with the benefits of AMD’s 3D V-Cache technology.
ASUS motherboards with AMD 800-series chipsets allow these processors to reach their full potential through specialized BIOS optimizations. AI Cache Boost, in particular, is an effective tool for accelerating AI workloads, including large language models.
To make the most of the new AMD processors, it’s recommended to consider the full range of ASUS AM5 motherboards – from high-end X870E and X870 models to more budget-friendly B850 and B840 options.
Read also:
- Motorola Moto G86 Review: Smartphone That Outclasses the Competition
- JSAUX accessories for portable consoles review, part 2: Backpack, docking station with M.2 SSD and adapters