Month: January 2026
-

Microsoft Maia 200 Beats Blackwell’s Efficiency
Microsoft is deploying its second-generation AI accelerator. The new Maia 200 NPU targets inference, with raw performance similar to that of the Nvidia Blackwell but requiring much less power. Maia only employs Ethernet, compared with other accelerators that employ a different technology for local (scale-up) interconnect. Microsoft has withheld details; its statements indicate that Maia…
-

A Prodigy Turns 10: Tachyum Unveils a Universal Processor
The tachyon is a faster-than-light particle, which cannot exist without violating known physics laws. Tachyum is a company promising to perform as well as a CPU, GPU, or NPU, which also shouldn’t be possible. Also threatening its existence, the company sued Cadence in 2022 after finding blocks (IP) licensed from the EDA giant were deficient.…
-

Byrne-Wheeler Report Discusses Nvidia Context-Memory Device
Joe and Bob discuss Nvidia’s CES disclosures and funding news for AI startups like Etched and Upscale AI, as well as Marvell’s acquisition of XConn. Additionally, they explore OpenAI’s partnership with Cerebras, SiFive gaining access to NVLink, and Global Foundries’ acquisition of Arc from Synopsys.
-

Cadence HiFi IQ Supports Both Immersive Audio and Lightweight LLMs
The Cadence Tensilica HiFi IQ is a high-performance, sixth-generation audio DSP designed to handle classical signal processing and modern AI workloads. Built on the Xtensa LX8 architecture, it offers a leap in AI throughput and energy efficiency, suiting it to chips for high-end automotive infotainment systems.
-

Byrne-Wheeler Report Discusses Nvidia-Groq Deal
Mike Demler joins Bob and Joe to discuss AI-chip juggernaut Nvidia rapturing key Groq staff and licensing the struggling startup’s innovative NPU technology.

