AI accelerator
PulseAugur coverage of AI accelerator — every cluster mentioning AI accelerator across labs, papers, and developer communities, ranked by signal.
4 天有情绪数据
-
Insta360 Mic Pro features e-ink display, AI noise reduction
Insta360 has released the Mic Pro, a wireless microphone featuring a unique e-ink display that can show custom images or logos. This mic boasts a three-microphone array with AI-powered noise reduction and offers three d…
-
Insta360 launches wireless mic with E-Ink display and AI noise cancellation
Insta360 has released its new Mic Pro wireless microphone system, designed for content creators and filmmakers. The system features two industry-first technologies: a customizable E-Ink display for identifying transmitt…
-
Google eyes direct TSMC partnership for advanced Tensor and AI chips
Google is reportedly seeking a direct relationship with TSMC, the world's leading semiconductor manufacturer, to gain priority access to advanced production processes. This strategic shift aims to bypass intermediaries …
-
Ascend-RaBitQ system accelerates billion-scale vector search with NPU-CPU architecture
Researchers have developed Ascend-RaBitQ, a novel system designed to accelerate billion-scale vector similarity search by leveraging heterogeneous NPU-CPU architectures. This approach decouples coarse ranking on NPUs wi…
-
Google unveils Broadfly TPU with novel network topology
Google has introduced a new inference-focused TPU with a novel network topology called "Broadfly" during their recent Google Cloud Next conference. This design allows for scaling up to 1,152 TPUs within a single pod. Th…
-
Google DeepMind's AlphaEvolve AI optimizes TPUs, boosts commercial AI training and simulations
Google DeepMind has announced AlphaEvolve, an AI-powered coding agent that has been integrated into its infrastructure to optimize hardware and software. The system has already improved the efficiency of Google's next-g…
-
New NPU-aware denoising model achieves high fidelity on mobile devices
Researchers have developed a novel approach for real image denoising specifically optimized for mobile Neural Processing Units (NPUs). This method uses a lightweight student network trained via knowledge distillation fr…
-
Google's Gemma 4 models achieve 3x speed boost with speculative decoding
Google has released Multi-Token Prediction (MTP) drafters for its Gemma 4 open models, which can increase inference speed by up to three times. This advancement utilizes a speculative decoding architecture, allowing a l…
-
GPT-5.5 Super App, Nebius NVIDIA Cloud, and Google TPU Sales Highlight AI Advancements
A new claim suggests that GPT-5.5, combined with Codex, can function as a "super app" with seven distinct capabilities. These features reportedly include app building, debugging, web browsing, and image generation, posi…
-
Next-gen chips promise data centers greater efficiency and AI power
Next-generation chip designs, including those optimized for AI, energy efficiency, and heat tolerance, have the potential to significantly alter data center infrastructure. Innovations in packaging, memory, and offload …
-
AI-accelerated CFD simulations adapted for IPU platform show performance gains
Researchers have adapted AI-accelerated computational fluid dynamics (CFD) simulations to run on Graphcore's Intelligence Processing Units (IPUs). The study focused on training machine learning models to predict simulat…
-
Researchers benchmark object detection models for edge devices
Researchers have benchmarked several deep learning object detection models, including YOLOv8, EfficientDet Lite, and SSD variants, on various edge computing devices like Raspberry Pi and Jetson Orin Nano. The study eval…
-
Study finds switchless networks more cost-effective for MoE LLM serving
A new paper analyzes network topologies for Mixture-of-Experts (MoE) Large Language Model (LLM) serving, finding that lower-cost, switchless networks can be more cost-effective than expensive scale-up infrastructures. T…
-
Google to sell its TPUs to some customers, who also fancy big-G GPUs
Alphabet announced a significant increase in its 2026 capital expenditure guidance, raising it to $180-$190 billion, driven by unprecedented demand for AI computing resources. The company's CFO highlighted strong growth…
-
Tenstorrent launches Galaxy Blackhole AI servers with 32 accelerators
Tenstorrent has announced the general availability of its Galaxy Blackhole AI compute platform, featuring 32 Blackhole accelerators in a 6U chassis for $110,000. The system offers 23 petaFLOPS of FP8 performance and can…
-
Google Cloud's AI compute market share rises with surging TPU demand
Google Cloud's market share is projected to increase significantly by 2026, driven by a massive surge in demand for Tensor Processing Units (TPUs). The company is expected to control a quarter of the global AI computing…
-
AHASD architecture boosts LLM speculative decoding on mobile devices
Researchers have developed AHASD, a novel asynchronous heterogeneous architecture designed to optimize large language model (LLM) inference on mobile devices. This architecture employs task-level decoupling for parallel…
-
Google unveils TPU V8 with two chips for training and inference at massive scale
Google has unveiled its eighth-generation Tensor Processing Units (TPUs), marking a significant shift by introducing two distinct chip designs for the first time. These new TPUs are engineered for specific, crucial task…
-
Tessera offers secure, near-line-rate weight streaming for edge AI accelerators
Researchers have developed Tessera, a new architecture designed to securely stream model weights to edge accelerators in Unified Memory Architecture (UMA) systems. This approach addresses the challenge of protecting pro…
-
New research explores LLM security, efficiency, and training optimization
Researchers are developing novel methods to enhance the efficiency and security of Large Language Models (LLMs). One approach, "Widening the Gap," exploits outlier injection to compromise LLM quantization, demonstrating…