Nvidia hosted the GeForce RTX AI PC Tour in Bangalore to showcase various new technologies with the potential to create immersive experiences in gaming, content creation and academics. The GeForce AI PCs leverage Tensor Cores which are dedicated AI accelerators to deliver a performance that ranges between 200 and 1300 trillion operations per second (TOPs) which is expected to beat its competitors.

Nvidia’s Tensor Core combined with the RTX 4090 laptop GPU beats the the performance of the Apple MacBook Pro which is equipped with the M3 Max chip by twice the AI performance.

John Gillooly, Technical Product Marketing Manager, Asia Pacific South at NVIDIA, demoed Nvidia ACE with a game called Mecha Break which used an on-device small language model to bring life to a gaming character and give it improved conversational skills.

The character called Yun Ni is able to identify the gamer using the computer’s camera which is powered by OpenAI’s ChatGPT-4o and understand references to other ideas in the real world while also being able to continue the conversation.

An open-source application called ComfyUI was built with Stable Diffusion where the user can take a selfie and is able to see different superhero representations of themselves within a few seconds.

Another demo app called ChatRTX helps you connect a large language model to your own data and then uses retriveal-augmented generation (RAG) to fetch a relevant answer once you type in a query onto the customised chatbot. Because it runs locally, the data also stays secure.

Separately, a report by The Verge stated that the yet-to-be-released GeForce RTX 5090 graphics card will have 32GB of GDDR7 memory. The RTX 5090 and RTX 5080 are reportedly expected to use more compute. While the RTX 5090 is said to draw 150 watts more than the 450 watts that the RTX 4090 used. The RTX 5090 will also reportedly include 32GB of GDDR7 memory on a 512-bit bus, with 21,760 CUDA cores.

Nvidia hasn’t yet confirmed when these cards will be released.