Posts by Dave81

GitHub ponders kill switch for pull requests to stop AI slop

Bottlenecks

The vast majority of the time spent on developing production quality large system software is spent on Q&A and debugging, not typing out code. So if AI speeds up the typing part but increases bugs or reduces understanding of new code, making code harder to maintain, it will ultimately slow a project down. You cannot introduce code that humans haven’t reviewed and fully understand. That’s why AI for now is a great assistant for looking things up or learning how to do esoteric things, but by no means a replacement for high quality human software developers. AI is mostly useful for writing code in these contexts when using it reduces the amount of bugs introduced per feature, and it’s from that perspective that it should be put to use.

Musk’s DOGE probed by top watchdog after poking around Uncle Sam's systems

Re: Bloated

Register readers are too lefty to appreciate this

Million GPU clusters, gigawatts of power – the scale of AI defies logic

Nuclear fusion

This is why Musk needs to get behind nuclear fusion rather than claim solar can cover our future needs. AI will have unlimited energy demands. What we really need is an infinite supply of safe, clean energy.

Sorry if I don’t speak in the sarcastic tone that’s expected in these comments

Just how deep is Nvidia's CUDA moat really?

I don’t get it

Why don’t these vendors just build a CUDA compiler or PTX assembler for their own GPUs? Is there some sort of IP issue blocking their freedom to do this?

Intel execs discuss the possibility of spinning off foundry

Not clear

Whenever I hear about Intel spinning off its foundry business, it sounds like they’re thinking of becoming fabless like AMD. But I don’t that’s what they mean. I think it means that the part of their business targeting external foundry customers will be spun off. The article needs to clarify this.

Who had Pat Gelsinger retires from Intel on their bingo card?

Mixed bag

He did well helping Intel execute in their process roadmap, but his strategy to turn Intel into a foundry was doomed from the start. The clear winning strategy is their old biz model where they earn exceptional margins by designing and also manufacturing leading products. It fell apart when they lost their process lead. They need to get that process lead back so they can get back to winning. The foundry thing never made sense. That’s lower margin, requires more command and control management and employees willing to take a lashing to stay competitive. I don’t think western employees are built for that.

The problem for Intel is that CPUs aren’t the biggest chip market anymore, so they can’t generate the cash needed to have the leading process tech if it’s just being used to dominate CPUs. They’ll need to win in GPUs to be number one like the good old days. The best strategy for them is to make sure they take the process lead and then use that fleeting opportunity to make the best GPU for AI on the market. They pull that off and they can become a trillion dollar market cap company. Otherwise any other outcome will result in them eventually falling behind tsmc again, and they’ll eventually become irrelevant

AMD grabs a quarter of x86 market with desktop gains, but server growth slows

Absolute numbers?

Would be great to know the absolute numbers here in terms of units sold and dollar value

Ex-Intel board members make an ill-conceived case for spinning off Foundry

Don’t be a foundry you fool

Intel, if you can hear me, please listen carefully. Your original business model of making and selling your own chips produces the best margins. All you need to do is manufacture your products at a node or two ahead of TSMC, do that for AI GPUs, and you’ll own that market and become a multi trillion dollar market cap company. You’ll generate way more cash than TSMC and be able to use that to maintain your lead against them. You’ll always be at a disadvantage offering foundry services since your customers are also your competitors. Stop wasting your resources chasing this foolish low ROI pipe dream. A foundry business with its lower margins also requires military like command and control to be competitive, which is the kind of corporate/work culture that meshes well with employees in Asia, not the West. You’re just never gonna be great at this so stop wasting your time, and lead with the business model that once allowed you to be the most profitable chip manufacturer on earth. If you instead chase government handouts to fund a business model that doesn’t make sense, supposedly for national altruism, you’ll eventually end up dead or insignificant, and America will be far worse off for that. And btw, now that GPUs are the biggest chip market, leading in CPUs alone won’t provide the cash to maintain you’re manufacturing lead long term. So make sure you get your priorities straight and harness 18A to win in GPUs. This will be your last chance

IBM Cloud to 'uplift' prices by up to 29 percent

Compute prices should fall

Moore’s law has made each generation of tech less expensive than the previous for the same unit of work/functionality. The cloud ought to follow the same trend. If not then there’s a serious flaw in that model.

Biden urged to completely cripple AI chips to China

Black market

They’ll just end up smuggling them in at a higher price.