October 25, 2008

October 25, 2008 by yorksranter 1 comment

DKE

I was reading the various Boris Johnson blogs earlier this week when something struck me. It was the combination of the Boriswatch story about the basis of the whole obsession with Routemaster buses falling apart, and the news that Boris Johnson’s crazyarse idea about building an airport in the Thames Estuary was being examined “in-house”…

October 25, 2008 by yorksranter 1 comment

Homemade particle physics

It all started with this Making Light thread; Bill Higgins was musing about improvising a radiation detector, which led him to this deeply cool device – scroll down for the cunning methodology that goes with it. But James MacDonald beat him, I think, with this cloud chamber implementation: Field-expedient cloud chamber:Needed: shot glass. High-proof vodka…

First-time buyers face toughest test for 70 years, lenders say - BBC News
the most common living arrangement for an adult aged between 18 and 34 in 1997 was being in a couple with children, but now it was living with your parents.
Super Micro plunges ahead of third-quarter earnings later this month
Super Micro Computer, one of the vendors that build Nvidia-based servers, saw shares drop 23% Friday after the company broke from its recent pattern of providing preliminary results, and announced it would report earnings later this month. In January, Super Micro increased its sales and earnings guidance 11 days before announcing second-quarter financials. The company […]
Samsung doubles AP developers, introduces Google's Gemini Nano 2 into Galaxy S25
Before the Galaxy S24, earlier flagship Galaxy models did not come with a Neural Processing Unit (NPU). However, Samsung aims to provide AI services through software updates by enabling the central processing unit (CPU) and graphics processing unit (GPU) to perform functions similar to an NPU // this is wrong; some of the Exy ones […]
With robots seen as the next big thing, Taiwanese manufacturers target highly-priced key component
The reducer is one of the most crucial components of robots, as it increases torque by reducing rotational speed, thus improving robots' precision on repetitive actions. It's also one of the most expensive components, accounting for over one-third of a robot's total cost. Analysts pointed out that traditional six-axis joint robots typically pair two harmonic […]
TSMC's earnings call points to AI HPC as sole driver
IC design companies note that their end customers have begun downgrading their annual sales forecasts for low-end and mid-range smartphones and PCs. With weak demand in the low- and mid-range, they can only hope that they will see some growth by hitching on the AI bandwagon. More concerning is TSMC's gloomy forecast for automotive chips
Strong AI demand drives TSMC US site development
The first wafer fab has made excellent progress. It started engineering wafer manufacturing with 4nm process technology in April and expects to start mass production in the first half of 2025. The second fab's plans have been updated to include 2nm process production to meet increasing AI demand, in addition to the previously disclosed 3nm […]
Samsung's mobile AP market share falls behind Apple and MediaTek
In the global mobile AP chip market in the fourth quarter of 2023, MediaTek secured the top spot with a market share of 36% when based on shipment volume. MediaTek was followed by Qualcomm (23%), Apple (20%), UNISOC (13%), and Samsung (5%). Analysis believes that as the demand for system-on-chip (SoC) solutions for 5G and […]
TSMC lowers 2024 semiconductor, foundry outlook
The company anticipates stable growth in the second quarter and full year, but surprised people by revising downward its full-year outlook for the semiconductor (excluding memory) industry...revenue from AI server processors will more than double in 2024 and account for a low-teens percentage of TSMC's 2024 revenue. In the next five years, AI server processors […]
Beat Nvidia with RISC-V? Samsung expands its AI chip unit in Silicon Valley
Getting into RISC-V and grinding an inference ASIC. I wonder if the ASIC is using their NPU from mobile, as there was a major security issue with that which caused them to block third-party access to the API
AI giants wrestle with data drought amid LLM race
He speculates future models like GPT-5 may require even more data, ranging from 60 to 100 trillion tokens. Villalobos indicates that despite utilizing all available high-quality textual and visual data, future AI models may still face a gap of 10 to 20 trillion tokens in data // as Naveen Rao says, data efficiency is *the* […]