weebull

It would be, yes. x86 had very few registers, so anything you could do to free them up was vital. Arm 32bit has 32 general purpose registers I think, and RISC V certainly does. In fact there's no difference between 32 and 64 bit in that respect. If anything, 64-bit frame pointers make it marginally worse.

Anything running a full database server is not micro.

Firat it was people pushing plows, then horses pulled them, and now machines do the work a hundred people used to do.

This technology is no different.

It places the minimum bar at known place, and that known place is anti-consumer. So now all companies know that they don't have to any more than have a certified repair center program.

What's that you sa? The requirements on certification are impossibly high. Oh noes.

Right, so latency is less of an issue (because control flow is predictable) and performance becomes about bandwidth.

8087 - vector edition.

The hardware is still on the chip.

> 1. The hardware isn't nVidia's primary moat. CUDA is.

People really don't appreciate this. So many times I see people comment that AMD should implement CUDA and not try to get software to move standard.

AMD have a massive task ahead because they NEED to get CUDA unseated just to play on a level playing ground, let alone win.

In dollars, for sure.

In units sold?

> I wonder, is the dual-use of GPUs a significant engineering bottleneck for Nvidia et al?

Not much. GPUs became array processing engines in the early 2000s and the amount of residual fixed function graphics hardware is tiny now.

The biggest thing is that a lot of network evaluation is being done at lower precision (BF16/ FP8 / ternary) now, but my understanding is that training the networks still requires FP32. As these companies are really targeting the training market the high precision modes needed for graphics can't be thrown away. If a company was to target the end user, then you could optimise for lower precision, but end users find dedicated neural net acceleration a luxury they can get for free by buying a GPU.

They can apply. They'd be refused on that and many more grounds.

NATO isn't a pay-to-play organisation. NATO takes no money from countries. NATO has no forces of it's own to pay for.

There is a non-binding convention that member should spend 2% of their GDP on *their* armed forces. That way when other nations call for help they can respond. It's a guideline figure though and some countries do it and some don't.

Honestly, I look at the readme and I think it's just a minimal effort layer on top of ComfyUI. It's main point seems to be "easier installation" but it's not ComfyUI that's difficult to install. It's the GPU libraries. CPU based ComfyUI is a `git clone` and a `pip install -r requirements`. Shame it's so slow on the CPU, but that's diffusion models for you.

Red were kinda the Tesla of the film camera world. New name, new tech, new price point, fashionable branding Vs existing long established players.

Thing is ARRI pivoted and covered the new tech pretty well, and had the existing business links into the rental market allowed them to continue unflustered. Red got the layman hype because they seemed to make high quality available to more people at an achievable (but still high) price point. The industry didn't really care though. They rent cameras, not buy them and they were already affording the old stuff.

"Made in rust." is dying off?

Sounds like it's far more about "big" data analysis, and recognising that elyptic curves encryption has a statistically apparent signature. AI/ML was just the analysis that exposed it.

> So our lived experience of government regulation is, in large part, positive.

I think I'd nuance it a bit. We see it as necessary. Necessary because otherwise powerful companies take advantage I think most can point to bad places in life for government regulations, but when the opposition is more powerful than most governments you need to fight with the most powerful thing you have.

If labour relations are a problem in American then there's no way they are expanding into Europe.

This is the crux of it in my opinion. 90% of my time using pandas seems to go on data formatting and cleaning to get info into the perfect form for pandas to do a relatively simple set of operations on. It gives we computational speed in exchange for a ton of fiddly tweaking.

The biggest thing for me is that data frames are collections of series, and not collections of rows. That just seems wrong on such a fundamental level to me.

> We've came to the happy state of ecosystem where tools are available to everyone and not just those who pay Microsoft for access.

I've done similar with an 8GB 7600, and I've run SD1.5, SDXL and Stable Cascade on it with ComfyUI. To get beat speeds I need to not run the desktop. That way I get all 8GB for inferredce, and control it from another computer. I think the 16GB 7600XT could be a nice entry level card for ML applications.

This is something I've yet to see a software testing framework do. Compare the results of two different implementations separated in time. i.e. two different revisions out of source control, using one as the golden reference for the other.

No it won't. GPUs are good at ml partly because of the huge memory bandwidth. 1000s of bits wide. You won't find connectors that have that many terminals and maintain signal quality. Even putting a second bank soldered on the same signals can be enough to mess things up.

Don't know about 3.0, but Cascade has different level of safety between the full model and the light model. Full model is far more prudish, but both completely fail with some prompts.

I think the AMD 8600XT is a mod in this direction, otherwise there was little point in releasing it.

GPUs need a decent virtual memory system though. The current "it runs or it crashes" situation isn't good enough.

I'd like to understand where this difference originates from. It can be very confusing working with Germans (for example).

...but is it Tinie Tempah proof?

Now you see I thought it was correcting a potential "misspelling" and stopping you from generating scat images. I had to search to find out you were talking about iconography appropriated by fascism.

My bad.

I find your perspective interesting, because it somewhat matches my own on the diffusion based image models.

Would you say that the output has changed, or that the output was originally novel to you and therefore useful, but now that same output style is no longer novel?

I've certainly noticed that the output I get falls into a pattern a lot of the time and that rush of being presented with something new has gone.

Is this the one that just recently allowed women to take part?