| Mobile Web

Search results for Flux

AI & Enterprise

KAIST cuts AI chip cooling energy by 90 percent

Researchers at the Korea Advanced Institute of Science and Technology (KAIST) have developed a technology that cuts the cooling energy needed to remove heat from artificial intelligence semiconductors to one-tenth of current levels. The method combines a coolant manifold and microchannels inside a chip to improve liquid cooling efficiency and reduce energy loss. KAIST said the design delivered a coefficient of performance of 106,000 using room-temperature water and can be applied without major new production-line investment.

AI & Enterprise

Prism ML unveils Bonsai Image 4B image-generation AI model that runs locally on iPhone

Prism ML has unveiled an image-generation AI model, Bonsai Image 4B, that can run locally on an iPhone. The company released two open versions, a 1-bit model using -1 and 1 and a ternary model using -1, 0 and 1, after retraining FLUX.2 Klein 4B. Memory demand falls from 7.75GB to 0.93GB and 1.21GB, respectively. A free iPhone app is available, and the model also runs on Windows 11 PCs and macOS.

Industry

TSMC dominance shows cracks, boosting Samsung foundry opportunity

The foundry market is entering a new phase as TSMC’s grip shows signs of weakening amid advanced packaging bottlenecks, opening room for rivals. Counterpoint Research said global foundry revenue rose 16 percent last year to $320 billion on steady AI chip demand. It expects TSMC’s share to slip this year, with Samsung and SMIC moving in. Samsung is pushing 2-nanometre production, while SMIC expands capacity and Intel seeks large AI customers.