The current Labour government here often seems to be criticised for being ideological and not pragmatic.

Those are strange accusations to level at the government. Really, pragmatism is a type of ideology, being more the means justify the ends, where as ideological is generally that the ends and means are inseparable.

The current labour government doesn't appear to have any coherent ends. To me that's neither ideological nor pragmatic. They just have means which they kind of flap about at random.

Also the investment culture here is garbage compared to the US.

You don't need any custom sensor hardware to read the disk as the original drives have signal lines that provide all that's needed. They usually have some automatic gain control, but it usually gives the right result.

You still need to connect the drive to a cheap device like the Greaseweazle (USB), which reads the raw flux reversals from the surface (plus timing information) as they're picked up by the drive head. All data decoding is done in software on the host system, which can handle ANY encoding type, and be smarter about error recovery.

The disks themselves sometimes need a bit of manual cleaning to be in a good position to read. Not everything can be recovered but older disks are generally more reliable than cheaper newer media.

I'm reading a lot of "it's simple, just get a...". If you read the article, it says they're "associated with an early Mac computer". That almost certainly means these a GCR formatted disks, and need a drive that can do variable speed rotation.

It's not impossible obviously, but it's likely the best way to do this is with a vintage Mac itself. Which then implies hooking up a mass storage device of some kind to that Mac so that it can be transferred to something more modern. So not super rare and impossible, but definitely fiddly.

I mean, if he's just talking about generating 3d assets and the like, then maybe?

Finally we can get legs!

Plus also vaccines aren't 100%. In the OP's mind, he was "told" vaccines are protection and anything less than 100% is a LIE.

But look, we're 5 years on, he's clearly got strong opinions and hasn't taken the time to learn. He ain't listening.

Woke people in shambles because they cannot name things like they wish.

So the woke are for free speech and you (anti-woke) are against it. Thanks for finally admitting that!

Sorry, carnivore here who still make his own ice cream a couple times a year.

You make your icecream out of meat? Or are you actually an omnivore?

Which in a world where nitrates are claimed as evil

Here's an advil to help with your hurt butt: O

Nitrates aren't "claimed as evil". Nitrates combined with amines (i.e. from protein), and especially at high heat form nitrosamines, which are carcinogenic. Fried cured meat (e.g. bacon) will create quite a lot of nitrosamines. Eating a bunch of raw veggies especially without much protein won't. None of these thing is up for debate, certainly from you[*].

Using foolish loaded language does not change the facts or the underlying science.

They are not "claimed as evil", they are known in certain cases to be carcinogenic. Fucking deal with it, bro. Eat bacon or don't, I don't care. If you want to raise your risk of cancer slightly to have delicious fried bacon, you can and you can, like me, own that choice.

But don't kick up dust and edge towards science denialism simply because you don't like what it has found. Reality won't change when you cast shade. Saying shit like "claimed as evil" won't actually change the cancer risk.

You do understand that, right? Your words and feelings don't affect reality.

[*] Just because sdience can always change from new evidence doesn't mean you personally have some. You don't.

I mean a beyond meat burger sure, but an old fashioned spicy bean burger? Those are not only pretty good, bit also just normal food. I'm also kind of annoyed that the good veggie burgers have gone out of fashion. I like meat burgers and iv like veggie burgers, but I've gone off the beyond meat types.

This aggressively stupid legislation would stop spicy bean burgers being called as such, despite decades of precedent.

Yeah if you want to run large transformer models a lot, you'll want something a bit beefier than a 2080Ti. With that said, there's a lot more out there than large transformer models.

AMD is not going to prioritize older hardware when all the money is on newer cards.

Their priority has ALWAYS been to prioritize the cards where the money is. The strategy hasn't worked for them, an has lead to a market where they're a minor player and everyone is pissed off with NVidia.

The problem is they're/were expecting people to either jump in at the expensive end or (now) run a canned LLM. There's so much more to ML than LLMs, and they have always (or until very recently if reports are to be believed) have made it obnoxiously hard to get started using their hardware.

Nvidia basically works for every one so they have a very deep base supporting each other, finding and fixing bugs, pushing into the odd corners etc etc, because NVidia have worked for that. AMD haven't and while their stuff is now better supported, it's still kind of obnoxious and if reports are true somewhat buggier.

AMD's searching for the money, they should be searching for the userbase

You can run a local instance very well on a 7900XTX, 9070XT, or 395+ and do very well. Many people already do, without particular difficulty.

The 395+ is intriguing, having lower specs but vastly more RAM than any other affordable GPU. Thing is if AMD weren't quit so useless, anyone vaguely intrigued could grab a random APU laptop and kick the tyres a bit to see what that performance profile is like. But AMD's not supporting the seed of an idea stage.

Your London mayor is a Muslim.

So? I'm a Jew. Also not a racist piece of shit, so I don't mind having someone who's brown and from a different religious background. He's been doing a pretty decent job, good enough that he'd get my vote again.

You are being invaded by the boatload, your own people are starting to rebel against you, and yet you push harder -- arrest anyone who dares voice a differing opinion

Ah yes, alternative facts land. Right wing domestic abusing racists attack the police: 25 arrests. Grannies holding cardboard signs quietly that support Palestine and attack the governments abhorrent use of the terrorism act: 2000 arrests.

You see people flying the Union Jack as racists.

So?

You see your own people in a very negative light.

Only the racist ones.

Next time read the supported hardware list.

As opposed to somehow scoring literally any GPU made after about 2018. You are not bolstering your point here.

This reads like an ad for NV based on tropes from a decade ago.

10 years or 10 days? If AMD have finally fixed their shit (and I do hope so, but they haven't with their supposed AI chips that's for sure) it's recent, much more than 10 years.

Trawling through reddit, it seems like there was a spate of "OMG pytorch/rocm actually works now" from about a year ago.

And from 2 years ago AMD officially starts supporting a few RDNA cards:

https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fpytorch.org%2Fblog%2Famd-e...

10 years? bollocks mate innit

1. Buy any (yes, any) Radeon 9xxx or 7xxx card and install it

That's already several steps behind where NVidia are. About the oldest card which will work well with Pytorch is something from the 2080 architecture. I didn't upgrade it for the latest project. Still works well. But even so, the last few times I checked, it was CDNA official and "lol good luck but maybe" for RDNA. I'm glad they've changed that for the better, though it's not enormously obvious.

3. If you use Linux, follow those simple instructions here: https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Frocm.docs.amd.com%2Fproj...

That's a lot more to do than "pip install pytorch". Maybe after an "apt install nvidia-driver-550".

After that wodge of stuff, you then need to do this lot:

https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Frocm.docs.amd.com%2Fproj...

So here's the thing, I can do all of that easily. A lot of people getting started in ML aren't experienced Linux sysadmins. They want to follow a simple tutorial and do "pip install pytorch". Given AMD's reputation as the second choice very few people won't then be put off with that lot up there?

Nvidia on the other hand ... Well, lets hope you won't need the external driver from Nvidia, or else ... prepare for some real pain.

I've got 3 machines with GPUs and use cloud machines. They all "just work". This pain you speak of doesn't appear to exist for me.

Amd's biggest problem is getting people to use their software so they can get a foot in the door and take over from cuda.

This is indeed AMD's biggest problem.

However the main problem with AMD is... AMD.

AMD desparately want to replace CUDA so people can buy an AMD GPU instead of an NVidia one and run their existing code on AMD GPUs. What they really care about is the expensive, high margin datacentre or pro GPUs.

This is deeply stupid for a number of reasons.

First let's accept the obvious truth that currently ML is the lion's share of the market. Sure there's a long tail of other stuff out there, but really to be available for a substantial fraction

Here's how you do the common thing on an NVidia GPU.

1. Get a GPU. Really doesn't matter how. But a 1650 second hand from CEX. Use the 40whatever in your laptop. Hire an H200 on a cloud machine. Get a Jetson. Literally anything in between. Doesn't matter.
2. pip install torch
3. go to town

Here's how you do it on AMD:
1. Get a GPU.
2. pip ins... wait docker what?
3. Find it doesn't work.
4. Check compatibility lists.
5. Find only GPUs way out of your price range for a quick experiment will work. Oh also fuck you if you have a laptop with an AMD GPU or even better an AI chip? Want to do AI on an AI chip? Fuck you.
6. Find maybe your chip might work because it's the same microarchitecture stupid code name
7. Fuck with docker some more because you hate yourself
8. Discover that your chip was supported, but suffered from the turbo speed deprecation cycle
9. Try an older version.
10. Fuck with docker some more
11. Cry.
12. Yeet your AMD gpu into the sun
13. Get an NVidia GPU

All that shit makes perfect sense from AMD's perspective. Why bother targeting gaming GPUs? That's not where the big money is. And since you're only doing pro ones, there's no harm in pro level difficulty, and limited hardware support. Why not just sell nothing but heaps of high margin MIbullshit to big people. They've got the resources to make the most of them.

What they can't seem to grasp is NVidia provide the low effort on ramp to their shit. You can go from a notion to an idea to a thought to a hack to a slightly bigger hack, to a project, to a business, to a large business easily. Or alternatively people flow out of universities into companies taking new found ML education with them which is all on NVidia because that's accessible to them and they use it because it's no risk to use what you know.

AMD are purposely missing the big base which feeds the upper levels. Plus millions of small projects with millions of hackers will hammer on their ecosystem and reveal all the annoying bugs and tough cases which will help it also not be a bit of a buggy mess (pytorch I love you but you suck).

It's crazy: AMD are demanding people jump through inane hoops to use their stuff from a position of being completely behind in the market.

Hmm - that site mixes operating systems with SSL usage on the same graph. But the other thing is - it's stats about public facing internet accessible sites. The majority of Red Hat clients are RHEL are internal or data centre, non-public. I'm struggling to get a link that works, but the 2025 estimate is around around 43% market share, and I'm honestly surprised it's that low.