Rural island off the coast of a european country:

10g fiber for $65/mo (I don’t even think they cared, I asked for more and I think they made up a number).

House literally down the street from google in silicon valley:

Comcrap $100 for shit cable, I’m paying $250 for actual upload speed.

This country is ruled by the corrupt.

Google is pushing av1 because of patents, but 266 is just plain better tech, even if it’s harder to encode.

This same shit happened with 265 and vp9, and before that, and before that with vorbis/opus/aac.

They’ll come back because it’s a standard, and has higher quality.

Maybe this is the one time somehow av1 wins out on patents, but I’m encoding av1 and I’m really not impressed, it’s literally just dressed up hevc, maybe a 10% improvement max.

I’ve seen vvc and it’s really flexible, it shifts gears on a dime between high motion and deep detail, which is basically what your brain sees most, while av1 is actually kind of worse than hevc at that to me, it’s sluggish at the shifts, even if it is better overall.

Yeah, using the a750 the same.

Can’t wait for next Gen arc with vvc (x266) support.

Intel has excellent transcode, even in their igpus.

I use an arc750 specifically for transcode, av1 runs at ludicrous speeds, but don’t do an Nvidia, they kind of suck because they dont support vaapi, only nvenc/nvdec and vdpau.

Yeah, the ryzens are great too.

Full amd will treat you well, I’m running dual xeons and a Radeon pro with an arc 770 just for av1 encode right now.

Next round going full epyc.

Amd knocks Nvidia into a hat on Linux, the drivers are just too incredible.

With the exception of AI, where Nvidia is just plain the gold standard.

Intel is fine, it has exceptional video encoding and works.

Core i3 is fine, celeron can route, but you don’t have as much headroom, or room for firewall rules, etc. Recommend Intel x520 or mellanox cx3 or newer, though the cx2 is perfectly fine.

The bs about bsd being slower is maybe 15 years old at best?

Bsd is a monster for routing.

Run 25gbe routing, still can get by on your 4 core, but I throw some serious xeons at it anyway.

Meh, I like my udm.

I use some of the features, but mostly it just works, and it’s debian under the hood so I just ssh in and unfuck whatever needs unfucking.

It’s vastly closer to a hand-built setup than anything else, and you can spend less time worrying about security.

Ollama, llama3.2, deepcode and a bunch of others.

Using a GPU but man they’re picky, they mostly want Nvidia gpus.

Do NOT be afraid to run on the cpu. It’s slow, but for 1 user it’s actually mostly fine.

I mean why not, that worked out perfectly fine for bulldozer…

And there it is, the actual reason the owning class are jizzing themselves over AI:

Finally they never have to worry about their underlings ever again, now they truly are everything and as self-reliant and self-made as they always told everyone from private school.

No, this is clearly evil and racist and true nazism, and we must reverse it even though it doesn’t hurt anybody but the west, and therefore Russia actually wants it, but not really?

I’m sorry, the talking points are confusing here, can I have my lines again?

So, trying not to dox myself, I worked with the architect twice.

Knights Ferry was derived directly from Larrabee (GPU), P54Cs with pre-AVX-512, .

KNC was a die shrink with more cores. Both of these were PCIe accelerators only.

KNL had full Airmont Atom cores with smt4, basically meaningful cores with proper AVX-512. Also you could boot them with linux, or as a PCIe accelerator.

KNM jadded ML instructions, basically 8/16bit float and faster SIMD.

They cancelled KNH.

I interviewed some of the actual Larrabee guys, they were wild, there was a lot of talk about dynamic translation, they were trying to do really complex things, but when people talk like that it makes me think they were floundering on the software and just looking for tech magic solutions to fundamental problems.

Intel always dies when the software gets more complex than really simple drivers, it’s their achilles heel.

KNL also had the whole MCDRAM on package for basically HBM bandwidth, but that didn’t actually work very well in practice, again due to software issues (you have to pick where you allocate, and using it as an l4 cache was not always effective).

Oof, yeah, they actually had another they didn’t release, based off pentium cores with avx512, basically knights landing with software support for graphics.

They were canceling projects like it was going out of style, which is sad, that would have been amazing for Ai.

True, yet Nvidia was a nobody that arrived out of nowhere with the Riva graphics cards, and beat everybody else thoroughly. ATi, S3, 3Dfx, Matrox etc.

Actually, they didn’t.

This was their first: https://en.wikipedia.org/wiki/NV1

Complete failure, overpriced, undercapable, was one of the worst cards on the market at the time, and used quadratics instead of triangles.

NV2 was supposed to power the dreamcast, and kept the quads, but was cancelled.

But the third one stayed up! https://youtu.be/w82CqjaDKmA?t=23

I mean fine, but first gen, they can fix the features and yields over time.

First gen chips are rarely blockbusters, my first gen chips were happy to make it through bringup and customer eval.

Worse because software is so much of their stack, they had huge headroom to grow.

Arcs are OK, and the competition is good. Their video encode performance is absolutely unworldly though, just incredible.

Mostly, they help bring the igpu graphics stack and performance up to full, and keep games targeting them well. They’re needed for that alone if nothing else.

Run them in a lot of places, love them.

They’re good at being configured and forgotten about.

My main rack, that’s more complicated, I have proper gear, but mikrotik is great for everything else.

Sorry, public universities get into the act too:

https://www.buffalo.edu/tcie/grow-your-business/collaborate/research-partners.html

Oak ridge is a partnership with UTK, and one of the few ways to get literate people into Tennessee. Sandia is the same, mostly funded by private corporations or the military/DOE.

Berkeley is too, we got mips and a ton of other stuff out of that.

You’re upset they don’t profit more, then demand a bigger share from licensing!

They don’t demand more because the administrators are corrupt as all fuck, and take small rewards in exchange for selling the larger ones, but even UT has a ton of money coming in from profit sharing, and most decent research universities do.

I’d love to see us cut tuition by making universities more reliant on research and/or alumni funding, they’d end up focusing on things that lead to successful student outcomes.

First of all, they don’t, that’s how publicly funded R&D works, the IP is shared, or licensed, but the product is owned.

Have you never read about actual patent licensing for places like Stanford and MIT? How do you think those work? The professors license the patents on behalf of the schools and either get a share, or in exchange for grants.

This model works, and has worked pretty well, it’s why we rocked all research for decades.