Nvidia 50xx

[ddrueding]

Kind of on topic. I just downloaded the current quality software and driver for my 4070.
I'm wondering what kind of settings you use in the Nvidia software to maximize visual acuity?
I'm also wondering why all of a sudden I have 3D acuity and sharpness I was never close to before????

The latest nVidia driver claims to fix a number of bugs related to their vibrance modes and HDR under some circumstances. That may have influenced your experience. Normally display drivers should not significantly alter the experience of watching media.

 

[Santilli]

Thanks. The new Studio driver is installed.

 

[Mercutio]

Microsoft, one of the most well financed and wealthy companies to ever exist on Earth, started charging for Copilot Code tokens at their actual monetary value yesterday instead of heavily subsidizing them on June 1, 2026. Today, many Copilot Code users have discovered that their day-one use of tokens was at times more than half the tokens allowed by their subscription tier.

Microsoft can't make AI work from a financial standpoint. Cost of Inference does not go down. Newer models require more hardware that needs more power and more water to operate, and even the datacenters theoretically under construction now for Blackwell architectures will not be adequate for next-generation Rubin hardware, nor is it expected that support could be retrofitted. This was the case for Blackwell as well. Patterns in markets also mean that there's not much incentive to fire up older hardware. It takes up too much space and costs too much to operate compared to the newer systems, so the second the next-gen thing is released, all spending will move to that.

Oracle is on the hook for $50 billion annually for datacenter construction in the name of AI. That's the GDP of Libya, and it's doing that by taking on debt. Oracle isn't exactly a growing concern. Most of its income comes from rapacious license agreements from the same customers.

Current estimates say that 50% of purported datacenter construction, at least in the US, has either never started or has been paused shortly after starting. There are almost no planned datacenter construction sites in the US with no ongoing opposition from the communities where they're planned.

Ride share company Uber spent its annual AI token line-item budget in under four months.

Does this sound like a healthy industry to anyone? Anyone?

 

[jtr1962]

Does this sound like a healthy industry to anyone? Anyone?

No it doesn't. I think the tech companies thought they could make a profit off AI because they would have legions of customers more than willing to pay them enough to make a profit, even given the exorbitant costs you mentioned. It turns out for most people AI is a solution in search of a problem. Companies have even discovered that it costs more to replace an employee with AI than to just keep them on the payroll.

I think AI will eventually be viable, but not until we have chips which are at least two orders of magnitude more efficient than what exists now. An AI data center shouldn't require the power of a small city.

 

[sedrosken]

The bubble is ripe for popping. My question is, when it does, will we see companies crawling back to the consumer market, or will they simply keep servicing the datacenters with the goal of keeping compute in the cloud where we'd have to rent it?

I don't hold my breath on cheap used hardware flooding the market because it already takes forms not easily usable by an end consumer. This isn't like the crypto craze in that way.

 

[Mercutio]

"Crawling" might be a strong word. But what else will they do if rackmounted compute hardware isn't consuming every bit of capital expenditure in tech? Without OpenAI and Anthropic driving demand, why vastly overspecialize in high bandwidth memory?

The bubble popping will be UGLY. Not just because some of the major players in tech will have at the very least, very bloody noses, but because institutional investors will see mega-billions of dollars in wealth disappear. Tech stocks related to AI have been the only real area of growth in the last several years, so if you have a Vanguard or Fidelity investment account, a defined-benefit pension or just fancy yourself an independent investor, that money is going to go poof the same time people with real money realize they've been fleeced. Normal people aren't going to be able to retire. People are going to lose their homes over this.

No, nobody can run an HGX200 at home. nVidia's PCIe Compute cards don't have display outputs on them, although I've read there are ways around that. Nobody is going to put HBM GDDR7 on workstation-class PC hardware. Maybe there will be wins for availability of high speed network devices or lower cost colocation or cloud hosting services, But I think memory manufacturers are probably going to collude to keep prices high. They've done that before and it would make sense for them to keep prices high as long as they can. What do we do with compute nodes? Don't know, don't care. They'll probably get turned into real time face recognition systems that governments use, or computer vision systems to identify more surfaces to cover with ads while operating self-driving delivery trucks or something.

My hope is that Nvidia and the AI firms lose 90% of their market cap and one or more of the bubble giants like Meta or Oracle winds up in an Enron death spiral. These companies aren't too big to fail and moreover, we need to see them humbled. They already act with utter impugnity with regard to consumer or customer rights versus their obligations. If we don't, we're just going to wind up with something worse than William Gibson or Neal Stephenson's tech dystopias.

 

[jtr1962]

As an aside I was going to invest in a mutual fund which holds the stocks of companies providing electricity generation for AI data centers. The returns over the last two years were something like 50% annually. Now I think I'll pass. Only a matter of time before this bubble pops. And it IS going to be really ugly when this happens. About the only thing I see saving it is if we come up with chips which are much more efficient than what exists now. There are some ideas to make this happen:

How Sparse Computing Can Make AI 70x More Energy Efficient

🚀 AI’s Hidden Superpower: How “Zeros” Could Redefine the Future of Intelligence Artificial Intelligence is growing at an incredible pace. Every few months, we hear about larger, more powerful models—models with billions, even trillions of parameters.

www.linkedin.com

Hopefully this is implemented at scale before the bubble pops. Right now I have over $30K in NVidia as part of one of my IRAs.

 

[Mercutio]

I don't think we're getting vastly more efficient anything unless there's some staggering breakthrough in materials science. Germanium-based wafers or the like. We're already deep into diminishing returns; 5nm to 4nm didn't move the efficiency needle much for CPUs or GPUs, and I don't think we'll see a huge jump going to 3nm products, either.

I do think power generation and green energy are good places to put money. Much as the current administration hate wind and solar power, those things are only going to get cheaper and easier, and the should be relatively stable even if most of the market is about to have some kind of coronary incident. Boring and stable won't have enormous returns, but you should at least get a positive ROI on that stuff.