Will I see any performance increase?

Like others have said LLMs mostly use VRAM, they can use system RAM if you’re running them on CPU, but that’s ridiculously slow.

It will however increase the speed of your compile times, which is especially useful if you’re compiling something large like the Linux kernel on a regular basis.

I’m also worried about not having ECC RAM.

If you are using it purely for LLMs, if it’s going to get bit flips, it’ll happen in VRAM.

If you are compiling large things for customers, I’d recommend ECC, just in case, e.g. you don’t want a bricking firmware from a bit flip. But according to EDAC and my TIG stack, my server’s ECC RAM has never even detected an error in the past year, if I understand EDAC properly, so it’s really not important.

AFAIK fast startup only affects shutdown, clicking restart will always do a full reboot. Shift clicking shutdown will do a full shutdown like you said, but shift clicking restart will start recovery mode.

If the HOA’s router supports UPnP/NAT-PMP/PCP then you might be able to use that to get some ports forwarded.

Is it possible to send the hint from OPNsense itself?

Yes, to me it sounds like you’re already getting a big enough prefix from your ISP (all devices getting a /64), but you’ll have to request a bigger prefix from OPNsense. I believe it should give you the options to do this when you set the IPv6 mode to DHCPv6 on OPNsense, but I can’t say if your ISP router will handle it.

I have also added all Cloudflare IPs in Jellyfin’s known proxies

You should only need to add the IP of the last proxy before reaching Jellyfin, which would be Caddy.

ADCs, DACs, IO extenders

These should all work without kernel drivers. For example, here’s a user space python library for ADS1*15 ADCs, or Nuvoton MS51 IO Expanders. Unless you need very specific timing or require the kernel to know about it, you shouldn’t need a kernel driver.

Idk, with I2C if it’s not something that needs a kernel level driver, there usually isn’t a problem with interacting with it from user space, for example basically all RAM RGB controllers are I2C and OpenRGB has no problem with them. I’m pretty sure I’ve only ever used an I2C device tree overlay for an RTC.

Also I2C/SMBus is present everywhere on x86, like some graphics cards expose it through their HDMI ports, even some server motherboards have a header for it; but for GPIO I’m unaware of any motherboards that expose it, so good luck researching the chipset and tracing out the pins.

If you can’t get the VPS to work, alternatively there’s Cloudflare but last I checked streaming was a little out of their free terms. With it, you should just have to set your AAAA record and make the cloud orange, that way Cloudflare will proxy it, and IPv4 will work. There’s also Cloudflare tunnels which lets you host websites without port forwarding anything.

Google does too, although I only know of it being used for domains.google, which got killed.

I’ve got a catch-all setup to go straight to my spam folder, OP could do something similar.

They were expecting it to not be Android, but something more custom. Like I feel even just bare bones Linux would’ve been more acceptable.

Debian testing has ‘updated’ to 5.6.1+really5.4.5-1 anyway, so as long as you’ve updated within the past few days it will have been downgraded to 5.4.5.

I believe 5.6.0 was in Debian testing for almost a month too.

In theory PWAs can be configured to run offline, whether they’re doing that I don’t know.

The desktop app looks like it’s electron though.

Ampere CPUs use normal DIMMs, and don’t have integrated storage, like any other CPU. So you can have the best of both worlds (although idk about power conservation, they are efficient though).

I searched it up:

It seems the smaller tube is pressurised, the larger bulb is under a vacuum. Some of the other diagrams also show a getter in the bulb, so it’s a proper vacuum tube level vacuum.

Idk, it looks like it’s part of the same question, so solving 19+14 is a little pointless when solving 19+45.

I might be completely wrong, but I’ve heard that a key is only a few hundred dollars, and once you’ve got it you can sign whatever you want. I think ReactOS also used to offer free driver signing for open source projects.

So I guess if ReactOS can afford one, so can most anti-cheat companies.

ZLUDA originally only supported on Intel since it was designed by an Intel employee, but AMD hired him to make it work for AMD instead. So in a way Intel is somewhat important here.

A lot of external status services just send a HTTP request to a certain url, if it succeeds then it’s up, if it errors or times out then it’s down. They also usually let you check if TCP ports do the usual handshake thing if you aren’t using HTTP.

The response time can also be used to check if a site is running slower than usual too, and if you have a use for it you can usually specify the required response code for success.

Although I wouldn’t be surprised if GitHub has some per-server analytics they can also use to estimate the load, but Instatus would work as described above.

Sometimes these sorts of things are referred to as health checks, if you’re looking for search terms. For example Docker can be set up to poll a container’s web server every few minutes, and mark it as unhealthy it if it stops replying using the HEALTHCHECK instruction in the Dockerfile.