Artificial intelligence startup Kneron’s new chip is set to be used in so-called AI PCs as early as this year. The San Diego-based company, backed by Qualcomm and Foxconn, unveiled its third-generation chip on Tuesday, which it says can run on affordable PCs due to its lower energy and cost requirements.
The KL830 chip is a neural processing unit (NPU) that works alongside the central processing unit (CPU) and graphical processing unit (GPU) to run AI tasks. The chip is designed to work with transformer models, which are the driving force behind AI systems that can generate content such as text, images or even music. The NPU also provides personalized GPT functionality, which means people can tailor the capabilities of the GPT to serve specific purposes, the company said.
Kneron says it’s currently “working closely” with major PC brands but declined to name which ones. However, a demo version of the KL830 is already inside an HP machine, which the company showed off at the Computex conference in Taiwan.
The KL830 provides the equivalent of 10 trillion operations per second (TOPS) GPU, combining computing power from CPU, DSP and NPU at 8-bit. PCs using Kneron’s new chip meet the International Data Corporation‘s (IDC) description of a “hardware-enabled PC,” or one with less than 40 TOPS (trillions of operations per second) performance, which typically enables specific AI features within apps to run locally.
In addition to hardware-enabled PCs, IDC has identified “next-generation AI PCs,” which include an NPU with a higher 40-60 TOPS performance and an AI-first operating system (OS). IDC also expects “advanced AI PCs,” PCs that offer greater than 60 TOPS of NPU performance, to appear in the coming years. No such products are available yet.
AI PC sales are set to grow from 50 million units in 2024 to over 167 million by 2027, accounting for over 60%, according to the IDC. Gartner, a tech research and consulting firm, predicts global shipments of AI PCs and AI phones will hit 295 million units in 2024, more than 10 times the figure of 29 million in 2023.
Kneron says it’s currently working closely with several of the biggest PC OEMs, who will be the first customers of their latest chip. The demo version at Computex is a HP machine.
Like its chip-making rivals, Kneron seeks to take advantage of the hype around generative AI, which began in late 2022 when ChatGPT captured global attention for its ability to respond to text prompts in real-time as a human would. The ensuing race to create generative AI applications has driven unprecedented demand for cutting-edge chipsets to power those programs. Case in point: In the last few days, Nvidia, AMD and Intel have unveiled next-gen AI server chips as competition in the trending sector rises.
Unlike cloud-based AI, Kneron’s chipsets power consumer electronics and cars on-device or on the “edge.” AI that runs on-device has the benefits of speed and security, whereas cloud-based AI can field more complex tasks or questions.
Several chips, including Nvidia and Qualcomm, are on the market that can run large language models (LLM) on the edge. However, Kneron says its new chip is leading the way in terms of power efficiency and costs. The KL830 has a peak power consumption of only 2 watts.
Apart from PCs, the KL830 is also available through a USB dongle that enables any device — whether a broadband router, IoT camera or classic PC — to become edge AI-enabled and perform AI tasks locally.
Watch this: Everything Announced at Intel’s Lunar Lake AI Chip Event
Editors’ note: CNET used an AI engine to help create several dozen stories, which are labeled accordingly. The note you’re reading is attached to articles that deal substantively with the topic of AI but are created entirely by our expert editors and writers. For more, see our AI policy.
+ There are no comments
Add yours