September 28, 2008

September 28, 2008 by yorksranter 0 comments

stupidity tax

I’m relatively laying off US presidential politics this year, unlike in 2004 (and aren’t you glad?), but this sticks out: John McCain gambling serious money, and apparently considering casino execs his “friends”. Long before the point that he gambles with them and apparently wins, which is bad enough, you’ve got to wonder, haven’t you? They’re…

September 28, 2008 by yorksranter 0 comments

Piracy Drivel Watch

Just to note that one issue of the Sun this week managed to rehash the much debunked pirates will knifecrime our asylums!!! story, yet again. It’s still drivel, of course; after all, here are those excitingly amoral Frenchmen, about to wire their pirates up to the ship’s power bus….or not. The latest lot are heading…

September 28, 2008 by yorksranter 0 comments

Mr 10% meets the 27%ers up town

OK, someone’s left an armed UAV in the changing rooms. Is it you, Harrowell? No? Speak up? America? Turkey? Italy? Well, it would be interesting to ask somebody how many Predators the RAF possesses at the moment, compared to a few weeks ago. Relatedly, this is wrong: Pakistan, Afghanistan, nuclear proliferation and trans-national Islamic terrorism…

September 28, 2008 by yorksranter 0 comments

Big miscarriage of justice of our times, phase 2

This is sick, but perversely reassuring. With the great miscarriages of justice of the 70s, the first phase was that the judge, the cops, the Home Office, and all right-thinking people agreed, and nobody took seriously that the victims might be innocent. This lasted a long, long time; but then we passed into the second…

September 28, 2008 by yorksranter 0 comments

The shock of the old

I’ve been blogging longer than the Rude Pundit.

China's two ways of building AI GPU ecosystems to compete with Nvidia
Even if "partially compatible" with CUDA, there will be a greater "performance loss," resulting in the product price/performance ratio lagging behind Nvidia, and at the same time, because CUDA contains many of the proprietary features of Nvidia's high-end AI GPU hardware, which is difficult to be reflected in the other Chinese AI chips.
Arm capitalizes on ASIC opportunities as demand for customized chips grows
the demand for ASICs will gradually increase. Based on their current operating scales, it will be difficult for existing specialized ASIC manufacturers to fully digest such huge demand. As long as Arm finds the right partners, establishes the platform first, and then adds support from the previously accumulated mobile device ecosystems, it can quickly build […]
NeuReality Launches Developer Portal for NR1 AI Inference Platform - High-Performance Computing News Analysis | insideHPC
Interesting that NR is talking about "high-volume, high-variety" workloads
OpenAI dominates B2B AI market with over 600K ChatGPT Enterprise users
Brad Lightcap, COO of OpenAI, says the number of registered paid users for ChatGPT Enterprise has surged from 150,000 in January 2024 to over 600,000 in April.
Samsung Foundry plans to provide 'one-stop-service' to compete with TSMC
Samsung's Taylor plant, which is expected to start production in the second half of 2024, will have its front-end manufacturing process completed in the US. However, semi-finished products still must be shipped back to South Korea for back-end processing. Once the US advanced packaging plant is in operation, Samsung can complete all wafer production in […]
Baidu launches ERNIE 4.0 Tool and AI toolkits, amid speculation of cooperation with Apple
Ernie 4.0's training throughput is 5.1 times more efficient, with queries per second increasing by 105 times, while inference cost has been reduced to 1% of the original...ERNIE Bot has amassed more than 200 million users since its launch on March 16, 2023, and the Ernie API currently handles more than 200 million queries per […]
UAE's G42 chooses US over China for Nvidia chips and US$1.5 billion investment from Microsoft
Peng Xiao, CEO of G42, revealed that the collaboration plans include G42 utilizing Microsoft's Azure cloud computing platform as the cornerstone for its AI product development and deployment. Through this partnership, G42 aims to establish partnerships with more businesses globally in the AI and cloud computing sectors. Smith stated that Microsoft will utilize G42's data […]
Rapidus outlook dimmed by TSMC, Samsung 2nm fab projects in US
Rapidus, which plans to build its wafer fab in Hokkaido, northern Japan, expects to obtain orders from US chip vendors or government-related customers under some sort of understanding between the US and Japanese government, the sources said. The new fab projects announced by TSMC and Samsung mean that 2nm manufacturing processes will soon be available […]
Can Samsung's 2nm process reduce the gap with TSMC?
Samsung is preparing to mass-produce the second generation of its 3nm process in the second half of 2024 and the 2nm process in 2025. South Korean media speculate that the two fabs in Taylor, scheduled to commence operations in 2026, will become pivotal for Samsung's development of the 2nm process.
Mario Draghi: Radical Change—Is What Is Needed - Groupe d'études géopolitiques
The approach we took to competitiveness in Europe after the sovereign debt crisis seemed to prove his point. We pursued a deliberate strategy of trying to lower wage costs relative to each other – and, combine this with a procyclical fiscal policy, the net effect was only to weaken our own domestic demand and undermine […]