Feed aggregator

Intel Corp. unveils new details about its next-generation client processor for AI PCs, the Core Ultra series 3, code-named Panther Lake, which is expected to begin shipping later this year. The company also gave a peek into its Xeon6+ server processor, code-named Clearwater Forest, expected to launch in the first half of 2026.

Core Ultra series 3 client processor (Source: Intel Corp.)

Panther Lake is the company’s first product built on the advanced Intel 18A semiconductor process, the first 2-nanometer class node manufactured in the United States. It delivers up to 15% better performance per watt and 30% improved chip density compared to Intel 35 thanks to two key advances—RibbonFET and PowerVia.

The RibbonFET transistor architecture, Intel’s first in over a decade, delivers greater scaling and more efficient switching for better performance and energy efficiency. The PowerVia backside power delivery system improves power flow and signal delivery.

Also contributing to its greater flexibility and scalability is Foveros, Intel’s advanced packaging and 3D chip stacking technology for integrating multiple chiplets into advanced SoCs.

Panther Lake

The Core Ultra series 3 processors offer scalable AI PC performance, targeting a range of consumer and commercial AI PCs, gaming devices, and edge solutions. Intel said the multi-chiplet architecture offers flexibility across form factors, segments, and price points.

The Panther lake processors offer Lunar Lake-level power efficiency and Arrow Lake-class performance, according to Intel. They offer up to 16 CPU cores, up to 96-GB LPDDR5, and up to 180 TOPS across the platform. They also feature new P- and E-cores, along with a new GPU and next-generation IPU 7.5 and NPU 5, delivering higher-performance and greater efficiency over previous generations.

Key features include up to 16 new performance-cores (P-cores) and efficient-cores (E-cores) delivering more than 50% faster CPU performance versus the previous generation; 30% lower power consumption versus Lunar Lake; and a new Intel Xe3 Arc GPU with up to 12 Xe cores delivering more than 50% faster graphics performance versus the previous generation, along with up to 12 ray tracking units and up to 16-MB L2 cache.

Panther Lake also features the next-gen NPU 5 with up to 50 trillion of operations per second (TOPS), offering >40% TOPS/area versus Lunar Lake and 3.8× TOPS versus Arrow Lake-H.

The IPU 7.5 offers AI-based noise reduction and local tone mapping. It delivers 16-MP stills and 120 frames per second slow motion and supports up to three concurrent cameras. It also offers a 1.5-W reduction in power with hardware staggered HDR compared to Lunar Lake.

Other features include enhanced power management, up to 12 lanes PCIe 5, integrated Thunderbolt 4, integrated Intel Wi-Fi 7 (R2) and dual Intel Bluetooth Core 6, and LPCAMM support.

Panther Lake will also extend to edge applications including robotics, Intel said. A new Intel Robotics AI software suite and reference board is available with AI capabilities to develop robots using Panther Lake for both controls and AI/perception. The suite includes vision libraries, real-time control frameworks, AI inference engines, orchestration-ready modules, and hardware-aware tuning

Panther Lake will begin ramping high-volume production this year, with the first SKU scheduled to ship before the end of the year. General market availability will start in January 2026.

Recommended Intel’s confidence shows as it readies new processors on 18A

Clearwater Forest

Intel also provided a sneak peek into the Xeon 6+, its first 18A-based server processor. It is also touted as the company’s most efficient server processor. Both Panther Lake and Clearwater Forest, built on Intel 18A, are being manufactured at Intel’s new Fab 52, which is Intel’s fifth high-volume fab at its Ocotillo campus in Chandler, Arizona.

Xeon 6+ server processor (Source: Intel Corp.)

Clearwater Forest is Intel’s next-generation E-core processor, featuring up to 288 E-cores, and a 17% increase in instructions per cycle (IPC) over the previous generation. Expected to offer significant improvements in density, throughput, and power efficiency, Intel plans to launch Xeon 6+ in the first half of 2026. This server processor series targets hyperscale data centers, cloud providers, and telcos.

The post Intel releases more details about Panther Lake AI processor appeared first on EDN.

Broadcom Inc. launches the Tomahawk 6 – Davisson (TH6-Davisson), the company’s third-generation co-packaged optics (CPO) Ethernet switch, delivering the bandwidth, efficiency, and reliability for next-generation AI networks. The TH6-Davisson provides advances in power efficiency and traffic stability for higher optical interconnect performance required to scale-up and scale-out AI clusters.          

The trend toward CPOs in data centers is to increase bandwidth and lower energy consumption. With the TH6-Davisson, Broadcom claims the industry’s first 102.4 Tbits/s of optically enabled switching capacity, doubling the bandwidth of any CPO switch available today. This sets a new benchmark for data-center performance, Broadcom said.

(Source: Broadcom)

Designed for power efficiency, the TH6-Davisson heterogeneously integrates TSMC Compact Universal Photonic Engine (TSMC COUPE) technology-based optical engines with advanced substrate-level multi-chip packaging. This is reported to dramatically reduce the need for signal conditioning and minimize trace loss and reflections, resulting in a 70% reduction in optical interconnect power consumption. This is more than 3.5× lower than traditional pluggable optics, delivering a significant improvement in energy efficiency for hyperscale and AI data centers, Broadcom said.

In addition to power efficiency, the TH6-Davisson Ethernet switch addresses link stability, which has become a critical bottleneck as AI training jobs scale, the company added, with even minor interruptions causing losses in XPU and GPU utilization.

The TH6-Davisson solves this challenge by directly integrating optical engines onto a common package with the Ethernet switch. The integration eliminates many of the sources of manufacturing and test variability inherent in pluggable transceivers, resulting in significantly improved link flap performance and higher cluster reliability, according to Broadcom.

In addition, operating at 200 Gbits/s per channel, TH6-Davisson doubles the line rate and overall bandwidth of Broadcom’s second-generation TH5-Bailly CPO solution. It seamlessly interconnects with DR-based transceivers as well as NPO and CPO optical interconnects running at 200 Gbits/s per channel, enabling connectivity with advanced NICs, XPUs, and fabric switches.

The TH6-Davisson BCM78919 supports a scale-up cluster size of 512 XPUs and up to 100,000+ XPUs in two-tier networks at 200 Gbits/s per link. Other features include 16 × 6.4 Tbits/s Davisson DR optical engines and field-replaceable ELSFP laser modules.

Broadcom is now developing its fourth-generation CPO solution. The new platform will double per-channel bandwidth to 400 Gbits/s and deliver higher levels of energy efficiency.

The TH6-Davisson BCM78919 is IEEE 802.3 compliant and interoperable with existing 400G and 800G standards. Broadcom is currently sampling the Ethernet switch to its early access customers and partners.

The post Broadcom debuts 102.4-Tbits/s CPO Ethernet switch appeared first on EDN.

Omnivision says its new image sensor captures sharp HDR video in any light, combining speed, clarity, and efficiency for premium phones and action cameras.