weebull

Indeed. At least now it's only 3x TSMCs annual revenue.

Jim Kellar is also out to ride the AI bubble, but at least he has a company with product you can buy and a bunch of ASIC engineers with experience (Tenstorrent). $1T in his pocket would produce something, I'm sure. I just don't think you want to king-make to that degree.

The best thing of this is that it completely chips the legs out of Sam Altman in the eyes of government. (hopefully)

It's also becoming very apparent that the studios can't support the level of investment their having to make to develop AAA either. A bit like the movie studios, dropping billions on titles that can only make money back if they manage to be the #1 hit of the year.

I'm amazed we're still on discs as physical media and not usb keys. Speed (), size, and capacity would all be greatly improved. It would be a bit like the cartridges of old.

I guess it must be a cost thing.

USB3.0 can theoretically do a 25GB blue ray in 50 seconds, 3.1 is half that, and 3.2 half again.

So it's $140 dollars for 2 in the comparison. Right?

As a general philosophy, don't ban it, regulate the use of it. What if trace amount of mercury are needed for the next world changing technology? By precluding reasonable access you're ensuring that will be invented somewhere else.

..and yes, I realise "bans" are colloquisms for heavy regulation I tend to think of a ban as lazy regulation that only specifies the few small exceptions as opposed to detailed regulation that defines the areas where care needs to be exercised.

I'd recommend trying stuff that exhausts the VRAM. That seems to be where thinks get flakey for me (RX 7600 - 8GB), especially if running a desktop too.

Indeed, but I think people forget that the reason AMD have a license in the first place was because Intel's customers in the early days required a second source for it's processors.

Who owns the Cyrix x86 license these days?

Installing PyTorch with the PyTorch website instructions for AMD was pretty painless for me on Linux. I know everybodies experience is different, but install wasn't the issue for me.

For me the issue on AMD was stability in situations when VRAM was getting tight.

You think they're going to say "Hey, compute became competitive but nothing else changed performance therefore... PRICE HIKE!"? They don't have the reputation to burn in this domain for that IMHO.

Granted you could see a supply/demand related increase from retailers if demand spiked, but that's the retailers capitalising.

> because honestly Nvidia's current generation 24GB is the sweet spot price to performance

You're aware the 3090 is not the current generation? You can see why I would think you were talking about the 4090?

...because they feel that at 20+10 it achieves a superior output than at 50 steps for SDXL. They also benchmark it against 1 step for SDXL-Turbo.

The code is MIT. The model has a non-commercial license. They are separate pieces of work under different licenses. Stability AI have said that the non-commercial license is because this is a technology preview (like SDXL 0.9 was).

WTF!

On my 5900X, so 12 cores, I was able to get SDXL to around 10-15 minutes. I did do a few things to get to that.

1. I used an AMD Zen optimised BLAS library. In particular the AMDBLIS one, although it wasn't that different to the Intel MKL one.

2. I preload the jemalloc library to get better aligned memory allocations.

3. I manually set the number of threads to 12.

This is the start of my ComfyUI CPU invocation script.

    export OMP_NUM_THREADS=12
    export LD_PRELOAD=/opt/aocl/4.1.0/aocc/lib_LP64/libblis-mt.so:$LD_PRELOAD
    export LD_PRELOAD=/usr/lib/libjemalloc.so:$LD_PRELOAD
    export MALLOC_CONF="oversize_threshold:1,background_thread:true,metadata_thp:auto,dirty_decay_ms: 60000,muzzy_decay_ms:60000"

Honestly, 12 threads wasn't much better than 8, and more than 12 was detrimental. I was memory bandwidth limited I think, not compute.

Shouldn't be a reason you couldn't do a ton of Layer C work on different images, and then swap in Layer B.

> because honestly Nvidia's current generation 24GB is the sweet spot price to performance

How is the halo product of a range the "sweet spot"?

I think nVidia are extremely exposed on this front. The RX 7900XTX is also 24GB and under half the price (In UK at least - £800 vs £1,700 for the 4090). It's difficult to get a performance comparison on compute tasks, but I think it's around 70-80% of the 4090 given what I can find. Even a 3090, if you can find one, is £1,500.

The software isn't as stable on AMD hardware, but it does work. I'm running a RX7600 - 8GB myself, and happily doing SDXL. The main problem is that exhausting VRAM causes instability. Exceed it by a lot, and everything is handled fine, but if it's marginal... problems ensue.

The AMD engineers are actively making the experience better, and it may not be long before it's a practical alternative. If/When that happens nVidia will need to slash their prices to sell anything in this sphere, which I can't really see themselves doing.

> The light that burns twice as bright burns half as long - and you have burned so very, very brightly, Roy.

Define "viable"?

> Personally I want Nvidia to break the x86-64 monopoly

The one supplied by two companies?

Not income of their neighbourhood, but the cost of living in their city/state. It needs to be a big enough region that a good job gives social mobility, but introducing California wages into Ohio would cause massive inflation there, driving normal people into poverty. At the same time nobody could live in LA on Ohio wages.

$7T is more than all of TSMCs revenue ever. Last year was $69B. That was their second biggest year, and they haven't existed for 100 years yet.

I think your recollection of the Evga exit is wrong.

They already had their 4000 series boards designed. They were all set. Prototype boards have made their way into the public sphere now. Then, whilst the launch was prepping Evga [cited](https://youtube.com/watch?v=cV9QES-FUAM) that they were fed-up of working with them, and we're basically being undercut by their own supplier.

Inglewood and Santa Monica are close enough to require the same level of income in order to afford a certain standard of living. Sure the standards of living may not overlap much, but they are on the same scale.

That's not true of California and Ohio in general.

It's not just that the journalists are technically illiterate. It's also that the people making these claims are trying to deceive. You needs lakes of water to have enough mass to make gravity batteries practical. Those are called pumped-hydro. Anything talking about dropping heavy things is probably about 6 orders or magnitude too small and a scam to get government funding.

In which case, this one is a joke. It stores less than $3,000 worth of electricity (15¢ per kWh).

...and that's the problem. You can easily get distracted by the look of the words on the page. Is the heading big enough, centered, got enough white space around it. By the time you've faffed about with that, I'm onto my second or third paragraph of content.

People use things like LaTeX exactly because of this problem. Word processing software brought the problem of inconsistent formatting and layout within the grasp of everyone and boy! ...did they grab it with both hands. Systems based on plain text allow the author to concentrate on content only, without the need to format.

Personally I find LaTeX misses the mark. Too much markup is needed and it detracts. I'm a fan of asciidoc though. I just wish the templating was a little better though.

Pretty sure whistleblower protections would trump NDAs, but not in a public forum.

> In security red generally means encryption/protection is active.

That's daft. You want to highlight (red) the insecure state, and reassure (green)in the secure state.

Maybe because a gun should never be treated as totally safe.

Reprogram your brain for "Red means danger, green means safe" and you'll do a lot better.

Red at a traffic light means there's a danger so you stop. Green means it's safe to proceed.

Red on a gun safety switch means danger because it's live.

The red wire is live and dangerous. The green wire is earth and safe. (at least true in Europe).