Yazılar

Broadcom unveils Thor Ultra networking chip to challenge Nvidia in AI data centers

/in /tarafından

Broadcom has launched its new Thor Ultra networking chip, designed to help companies build massive artificial intelligence computing systems by linking together hundreds of thousands of processors — escalating its rivalry with Nvidia in the race to dominate AI infrastructure.

Unveiled on Tuesday, the Thor Ultra chip enables data center operators to connect far more AI processors than before, making it easier to train and deploy large models like OpenAI’s ChatGPT. The launch follows Broadcom’s announcement on Monday of a major deal to deliver 10 gigawatts of custom chips for OpenAI starting in 2026, further challenging Nvidia’s dominance in AI accelerators and networking technologies.

“The network plays an extremely important role in building these large clusters,” said Ram Velaga, Broadcom’s senior vice president. “So I’m not surprised that anybody in the GPU business wants to participate in networking.”

AI has become a $60 billion to $90 billion market opportunity for Broadcom by 2027, according to CEO Hock Tan, split between networking chips and custom data center processors built for companies such as Google and OpenAI. In 2024, Broadcom reported $12.2 billion in AI revenue, and in September it disclosed a $10 billion unnamed customer for its AI chips.

The Thor Ultra doubles the bandwidth of its predecessor and acts as a vital link between AI systems and the rest of the data center, improving data transfer speeds and scalability. Engineers developed it alongside Broadcom’s Tomahawk networking switches, refining every detail from power consumption to thermal management.

While Broadcom does not sell servers directly, it provides reference designs for partners to build upon. “For every dollar we invest in our silicon, our ecosystem partners invest six to ten times more,” Velaga said, emphasizing the company’s design-first strategy in the AI infrastructure market.

Google Pixel 10 Series to Feature Tensor G5 Chip with Major Upgrades, Including Custom Camera ISP

/in /tarafından

Google is set to unveil its highly anticipated Pixel 10 series later this year, powered by the next-generation Tensor G5 chip. In a significant shift, Google is expected to move away from Samsung as the primary manufacturer of its smartphone processors and turn to Taiwan-based TSMC for production. Ahead of the official launch, several details about the Tensor G5 chip’s new components have emerged, shedding light on the advancements and changes that will define Google’s upcoming flagship devices.

The Tensor G5 will build upon the foundation laid by the previous Tensor G4 chip, with Google continuing to incorporate its proprietary components for enhanced performance. According to sources close to the company, key features of the Tensor G5 will include the integration of Google’s always-on compute (AoC) audio processor, the Google Emerald Hill memory coprocessor, the Google GXP digital signal processor (DSP), and the Google EdgeTPU, all of which have been carried over from the G4. These components are expected to offer improvements in processing efficiency, audio handling, and machine learning capabilities.

A notable change with the Tensor G5 is the switch in GPU suppliers. Unlike its predecessors, which used GPUs designed by Samsung, the G5 will feature a GPU from Imagination Technologies, specifically the DXT model. This move is expected to improve graphics performance and power efficiency, contributing to a more seamless user experience. Additionally, the Tensor G5 will drop support for some older video codecs, such as Google’s BigWave (AV1) and Samsung’s MFC video codecs, in favor of a new video IP from Chips&Media, the WAVE677DV. This update could enhance video playback and streaming capabilities on the Pixel 10 series.

Perhaps the most significant change is in the camera department. Instead of relying on Samsung’s customized ISP (Image Signal Processor), Google is reportedly developing its own custom ISP for the Tensor G5, which is expected to offer enhanced image processing capabilities. This shift reflects Google’s ongoing investment in camera technology and its desire to have more control over the image quality produced by its smartphones. Additionally, the Tensor G5 will feature alternative components for interfaces and controllers, including UFS, USB3, and PWM controllers from third-party suppliers such as Synopsys, SmartDV, and Faraday Technologies, replacing Samsung’s previous offerings. These changes highlight Google’s commitment to improving the overall performance and customization of its hardware, setting the stage for a more refined and powerful Pixel 10 series.