Modders have found a method to unlock previously restricted artificial intelligence processing performance on Nvidia GeForce RTX 20, 30, and 40-series graphics cards. This modification removes a performance cap placed by Nvidia on specific artificial intelligence calculations, enabling supported graphics cards to finish operations up to twice as fast.
According to VideoCardz, the RTX 4090 reaches double its previous peak performance in specific artificial intelligence tests, while the RTX 3090 Ti demonstrates improvements of up to 80 percent in certain calculations. These enhancements primarily apply to artificial intelligence tasks rather than video games.
How Modders Unlocked the Performance
Nvidia GeForce RTX graphics cards contain specialized hardware called Tensor Cores, which assist in processing artificial intelligence workloads. However, Nvidia restricted the speed of certain calculations on consumer graphics cards. Specifically, calculations using FP32 accumulation operate at half the speed of similar operations using FP16 accumulation.
These restrictions do not apply in the same manner to some professional Nvidia graphics cards. Modders bypassed these limits by altering low-level graphics processing unit settings prior to the start of the Nvidia graphics driver.
This modification avoids replacing the graphics card hardware or changing its firmware permanently. The adjustments are temporary and clear away when the graphics card loses power.
RTX 4090 and RTX 3090 Ti See Major Improvements
Early testing reveals significant enhancements in specific artificial intelligence calculations. The peak processing performance of the RTX 4090 increased from approximately 165.6 trillion operations per second to 331.6 trillion operations per second in a specific Tensor Core test.
Meanwhile, the RTX 3090 Ti recorded an improvement of approximately 80 percent in BF16 calculations used for artificial intelligence processing. These numbers represent specific artificial intelligence computing tests rather than general graphics card performance. Other supported graphics cards also showed enhancements, though results vary depending on the model and workload.
What This Means for AI and Gaming
The unlocked performance could benefit individuals who use Nvidia graphics cards to run artificial intelligence applications on their computers. Tasks involving complex mathematical calculations, including artificial intelligence image generation and processing prompts for local chatbots, can become faster.
Yet, performance enhancements remain generally smaller in real-world applications compared to benchmark results. For instance, some RTX 3090 tests demonstrated improvements of roughly 14 percent to 36 percent in artificial intelligence prompt processing. Chatbot text generation speed stays largely unchanged because that process depends more on memory bandwidth than raw processing power. The modification does not provide a similar performance boost in video games.
Not All Nvidia Graphics Cards Are Supported
The modifications currently target selected models from Nvidia RTX 20, 30, and 40-series graphics cards. Most available tools are designed for Linux, though experimental Windows versions are also under development. Support varies across different models, and the modifications are not confirmed to work on every graphics card in these generations. Nvidia newer RTX 50-series cards are not currently covered by confirmed working methods.
These tools remain experimental and carry risks, including system crashes, startup failures, and possible hardware problems. The discovery indicates that older Nvidia graphics cards can deliver higher artificial intelligence processing performance without new hardware, though actual benefits depend on software and tasks.
