n=25? Seriously?

This is barely passable as an early hypothesis test before you perform an actual study.

You also get UMR from AMD https://gitlab.freedesktop.org/tomstdenis/umr

There is also a bunch of other tools provided: https://gpuopen.com/radeon-gpu-detective/ https://gpuopen.com/news/introducing-radeon-developer-tool-s...

They have little experience producing code that is high enough quality it would be accepted into Linux kernel. They have even less experience maintaining it for an extended period of time.

Also LFP prices dropped enough that shipping cost from China became a significant part of the price. This will be even more of a factor should the less energy dense sodium batteries ever reach the promised $30/kWh.

- you underperform while onboarding and rapidly learning to cover whatever skill gaps you discover

- you slowly go from doing OK to being overqualified, but it doesn't get reflected in compensation or title enough and your employer has little incentive to help you reach the next stage, because you are currently overperforming compared to your pay

- you are so much past your paygrade you quit and go to a company that will pay you what you will be worth after onboarding and learning to cover skill gaps

With everything that companies provided to earn long term employee loyalty being sacrificed to stock price employee growth through multiple roles in the same company has become more of an exception.

Source: most of the companies I worked or consulted for in the past 20 years.

I would love to say that this was an exception during almost 20 years of my professional career, but it wasn't. It was certainly the worst, but also much closer to average experience than it should have been.

In the world of VC powered growth race to bigger and bigger chunk of market seems to be the only thing that matters. You don't optimize your software, you throw money at the problem and get more VMs from your cloud provider. You don't work on fault tolerance, you add a retry on FE. You don't carefully plan and implement security, you create a bug bounty.

It sucks and I hate it.

There isn't much difference between MIT and GPL unless you are selling a product that runs locally or on premisses and with the latter some companies try to work around GPL by renting servers with software on it - either as physical boxes or something provided on cloud provider marketplace.

Look at what you actually have installed on your computer - odds are that unless your job requires something like CAD, photo/video editing or other highly specialized software you have nothing made by large enterprise with exception of OS and Slack/Teams/Zoom.

Historically those cards had narrow memory bus and about a quarter or less video memory of high end (not even halo) cards from the same generation.

That narrow memory bus puts their max memory bandwidth at a comparable level to desktop DDR5 with 2 DIMMs. At the same time quarter of high end is just 4GB VRAM which is not enough for low details for many games and prevents upscaling/frame gen from working.

From manufacturing standpoint low end GPUs aren't great either - memory controllers, video output and a bunch of other non-compute components don't scale with process node.

At the same time unified memory and bypassing PCIE benefits igpus greatly. You don't have to build an entire card, power delivery, cooler - you just slightly beef up existing ones.

tl;dr; sub-200 dollas GPUs are dead and will be replaced by APUs. I won't be surprised if they will start nibbling at lower mid-range market too in the near future.

- 12VHPWR connectors are compatible with 18ga or at best 16ga cables. For 90C rated single core copper wires I've seen max allowed amperages of at most 14A for 18ga and 18A for 16ga. Less in most sources.. Near the connectors those wires are so close they can't be considered single core for purpose of heat dissipation..