AMD has, as far as I understand, been outcompeting Nvidia on value for a good while.

And yet their market share doesn’t increase. Because Nvidia holds a stranglehold on both the software (CUDA, RTX support, frame gen, etc) AND sheer brand recognition- gamers always complain about Nvidia and want better AMD cards, but continue to line up in droves to buy Nvidia cards no matter what.

Nvidia knows they can do whatever the fuck they want right now.

AMD is a puppydog that follows Nvidia around on the open market with 10% or less market share. If Nvidia constricts supply and causes a massive price jump and shortages, AMD will just follow the pricing curve and we will still get no GPU’s.

AMD is also 100% reliant on TSMC and VRAM suppliers, the same exact supply pressures causing Nvidia to turn off the consumer tap will come for AMD too.

Option two is not correct, option one is correct. This announcement is specifically for consumer gaming GPU’s only, it does not affect institutional datacenter customers.

This is Nvidia saying “thanks small fry, you were useful, but we’re leaving you behind now. Fight for the scraps.” Complete cartel behavior.

Also some of us are stuck with necessary stupid software that has no Linux releases and doesn’t run in proton/wine yet.

A Minecraft server is the classic.

Don’t discount just putting together a basic webpage that can be accessed at home too- something he could put together in a basic HTML editor (drag and drop) and put his favorite things on or whatever he may be focusing on (cars, animals, space, you name it).

I hate to say it, but prebuilts are the absolute best bang for your buck that you can get now. At least for another month or two until the shortages start impacting OEM inventories.

Right now you can get decent high end ish machines like this: https://www.costco.com/p/-/ibuypower-element-gaming-pc-desktop-amd-ryzen-7-9800x3d-nvidia-geforce-rtx-5070-12gb-windows-11-home-32gb-ram-2tb-ssd/4000384603?langId=-1 and like 35% of the total machine cost washes out to JUST the memory right now. You’ll need to replace the PSU and likely the water cooler for the long term, but the bones are relatively good.

1gb ram is crap. Hardware capabilities aside it’s just not enough to run anything usable for real hosting. Get an old office machine for 50-100 with 8gb or more of memory and it will do infinitely better.

They don’t really have to collaborate though. They’re proving right now that they can price out consumers by just buying all the hardware capacity up and letting the market take care of the little guys. Hardware manufacturers like Micron are obliging.

What I’m becoming worried about now is all these corporations now realizing that they can simply supply price the average consumer out of owning electronics or any kind of compute. And locking them into renting or leasing access to data center compute and keeping the power of information further consolidated in corporate interests.

You won’t ever see it in a commercial / OTS slicer implementation for a while, fucking Stratasys still has a patent on it and they love lawyers. This guy finally just said fuck it and wrote it himself.

Its more about it being a non-open standard that was unilaterally implemented by Google, and then just expected to immediately become the default because daddy Google said so.

cartel that has previously done cartel things continues to do more cartel things

more at 11

Not often, if you’re running non-fiber reinforced filaments. even standard brass nozzles can get 1000 hours on them before you start to get loss of print quality. And hardened nozzles have practically infinite life in normal plastic.

Better off getting multiple sizes though. A default 0.4mm nozzle is a nice middle ground. 0.2mm gains a ton of detail for small parts. 0.6 and 0.8mm nozzles can drasticall ly reduce print times for large parts and creates better overhang/bridge surfaces.

Google probably realizes literally nobody likes WEBP and avoids using it

AV1

The real key is buried in the middle, where they say hardware decode capabilities are going to be restricted to models with discrete GPUs… Meaning they can make a $500 upsell mandatory for the most basic of capabilities.

Oh thank god. This solves my problem of no good integrated cam hardware on the market that isn’t cloudified or a huge security hole.

That was part of it for sure, but even on the machines we reserved specifically for experienced users and were personally maintained by myself just had myriad issues. Every first layer had to be watched like a hawk. And they were bedslingers so knocking over prints was a semi common occurence too.

That means the peak is here. The insider of all insiders sold? Dump everything!

I do, and it’s one of the reasons I bought a Bambu (pre-lockdown shit of course). Despite not having many hours, the prusa mk3’s my college had in their makerspace were CONSTANTLY having issues and were a lot harder to get successful prints off of. I got an x1c when i got real money after graduating and it’s given me 1200 hours of zero maintenance except grease and a couple belt tension resets.