The newest chips from Nvidia are made to run AI at home as competition from AMD and Intel approaches
Last year, Nvidia became the focal point of the artificial intelligence boom due to the fact that its costly server graphics processors—including the H100—were necessary for developing and implementing generative AI systems like OpenAI’s ChatGPT. Nvidia is currently showcasing its prowess in consumer GPUs for “local” AI, which may operate on a PC or laptop at home or in the workplace.
The RTX 4060 Super, RTX 4070 Ti Super, and RTX 4080 Super are three new graphics cards that Nvidia unveiled on Monday. Their prices range from $599 to $999. Additional “tensor cores” on these cards are intended for use with generative AI software. Additionally, Nvidia will supply graphics cards for laptops made by Acer, Dell, and Lenovo.
Nvidia’s market capitalization has risen to over $1 trillion due to a boom in sales of its corporate GPUs, which are expensive and sometimes come in systems with eight GPUs operating in tandem.
Nvidia’s main product line has always been PC GPUs meant for gaming, but the company claims this year’s graphics cards are better designed to run AI models without sending data to the cloud.
The business claims that although their primary purpose will be gaming, the new consumer-level graphics chips are also capable of handling AI applications. Nvidia claims that the RTX 4080 Super, for instance, can produce AI video 150% quicker than the previous generation model. Nvidia said that other software updates it just revealed will speed up large language model processing by five times.
“With 100 million RTX GPUs shipped, they provide a massive installed base for powerful PCs for AI applications,” said Justin Walker, senior director of product management at Nvidia, to reporters during a press briefing.
Over the course of the next year, Nvidia anticipates that new AI applications will arise to capitalize on the enhanced processing power. Later this year, Microsoft is anticipated to release Windows 12, a new version of the operating system that will further leverage AI chips.
According to Walker, the new chip can be used to eliminate backdrops from video chats or to create graphics using Adobe Photoshop’s Firefly generator. Additionally, Nvidia is developing tools that would enable game developers to incorporate generative AI into their products. One such tool would be the ability to create dialogue from non-player characters.
Nvidia will compete with Qualcomm, AMD, and Intel in local AI, despite being the firm most identified with large server GPUs, as demonstrated by its hardware releases this week. All three have revealed plans to release new CPUs with machine learning-specific components that will power so-called “AI PCs.”
Nvidia’s decision coincides with the tech sector’s efforts to determine the most effective approach to generative AI deployment. This technology is very expensive to run on cloud services and requires a significant amount of processing power.
One technological option is the “AI PC,” often referred to as “edge compute” at times, which is being pushed by competitors of Nvidia and Microsoft. Devices with more potent AI chips within will be able to run so-called large language models or picture generators in place of utilizing powerful supercomputers over the internet, but with some drawbacks and trade-offs.
Applications that can leverage a local AI model for urgent tasks and a cloud model for complex queries are suggested by Nvidia.
“While RTX tensor cores in your PC are going to be running more latency-sensitive AI applications, Nvidia GPUs in the cloud can be running really big large language models and using all that processing power to power very large AI models,” explained Nvidia’s Walker.
According to the company, the new graphics cards will comply with export regulations and can be transported to China, providing Chinese companies and academics who are unable to obtain Nvidia’s most potent server GPUs with an alternative.