Use a surface cleaner first. You wanna get the majority of the crud broken up and off first, wiping towards the middle then go get the whole thing with a glass cleaner using a microfiber cloth if you don’t have a squeegee.

Probably unplug three of those fans and block off the case openings to get some sort of laminar flow.

Switch to a more efficient power supply. Yours is platinum but it may be worth it to look into a list of low standby power units.

Maybe take out half your ram. It probably doesn’t matter though, I’m thinking about the need for lpddr to get past the 1tb barrier on old servers.

The b850 boards will have efficiency settings in the bios that will make the biggest difference.

Before you change anything though I would recommend first putting a watt meter inline and measuring.

Theres a bunch of stuff you can manually do in software but its fractions of a percent and makes your experience very nonstandard.

The biggest thing is probably gonna be turning off the computer when you’re not using it. It’s not like you’re gonna be spinning down a bunch of hdds…

A datacenter on land that relies on water for cooling is using fresh water. Aside from the insane and undesirable use of fossil water or groundwater, fresh water has a much higher ratio of surface area exposed to air and volume than the ocean does.

That ratio is important because you can only evaporate off the surface, so the same volume of water would evaporate faster in a frying pan than it would in a saucepan given the same conditions.

Freshwater is also smaller than the ocean by many orders of magnitude. That’s important because a smaller body of water will heat up more than a larger one given the same conditions.

Freshwater is also generally speaking moving towards the ocean somehow in a complex process called the water cycle. It might flow down a hill into the ocean, it might drip through aquifers to the ocean and it might evaporate and fall as rain either in the ocean or somewhere else where it takes some other path towards sea level.

That last part is important because in America rivers like the ones in Colorado and California have been reduced in volume so much due to of a bunch of manmade events and earthworks that they now lose a higher portion of their volume to evaporation as opposed to flow. That’s a big deal because the Colorado for example flows south but water vapor that comes off of it is blown east. So now much more of the rivers water is going into the desert as opposed to reaching the ocean. The water cycle has been disrupted.

Freshwater is also comparatively super rare and necessary for life on land.

So if you were water cooling a datacenter it would be better to use the huge ocean with much lower surface area to volume ratio and much more volume than to use the rare freshwater that will get heated up much more by the same energy, evaporate faster, we already know can have its water cycle disrupted and all life on land relies on.

Now a person might ask “what about the sea life, doesn’t it matter?” Of course sea life matters, but the way that sea life handles an increase in temperature that’s localized to one specific area is to just go away from there, like it does when undersea magma vents pop up. When the freshwater gets too hot it all evaporates away, there’s no water and the land animals die.

A person might say: “well don’t animals die when the oceans get hot too?” And they’re right! An average increase of .5 degrees c in sea surface temperatures would lead to massive coral bleaching. But enough energy to raise the sea surface temperatures .5c is phenomenally huge. Like getting more sun huge. Because that’s what’s raising it, getting more sun. Datacenters produce so little heat energy in comparison to that level of power that it’s not a concern.

The number one culprit of undesirable cornering behavior is incorrect tire pressure.

Why did you delete the post with system specs, @[email protected]?

That information would really help people to give you advice and your pc isn’t as unique as you think it is…

Here’s how you could have figured it out yourself:

Boot something else, read the logs.

Use a usb to boot some other operating system (or the same one) and access the logs of your system which will show where it’s failing and how.

Another person put that you could add nofail to fstab, you could also do this by moving the devices or services to nonessential branches of systemd.

Stop reading this thread and buy this thing or something like it.

There are at least three things in between the wall and what the os tries to do before you start fiddling around in the settings. Did the thing you changed take effect? Did it stay in effect? Is the cpu actually doing what you ask it? Can you even trust what the cpu is reporting back to you? The motherboard?

Don’t just start fucking around with stuff before you put a watt meter in line. Everything else is just guesswork.

I’m almost 100% I have a 390 card running with nvidia drivers on Debian stable. Will have to double check when I get to that device.

My intent was to provide an explanation for why someone might consider your reply about Chinese manufactured ram being a supply chain risk to be inaccurate.

To be 100% clear as a bell: ram is not a supply chain vector people need to be worried about.

Ram would be a really hard component to supply chain attack. It doesn’t store anything when powered off, so you’d need another chip on the board that can store your attack and that’d stick out like a sore thumb.

It also requires incredibly low latency, so low that trace lengths need to be optimized in order to deliver data accurately. So stream manipulation is out the window.

You’re left with searching through the contents looking for something juicy and that requires some kind of extra sore thumb chip that can’t go fast because it doesn’t have a heatsink.

Plus it’s been standard practice to harden the memory of libraries and programs and even operating systems to avoid stuff like the old Intel hyper threading attacks for at least fifteen years now, so there’s a reduced attack surface.

No one’s supply chain attacking your ram.

Hang on, I’m gonna add a suspicious new component onto a part that is incredibly expensive and heavily scrutinized specifically for speed and latency that will bit bash the I/o.

You add a piece of code (to ram, which famously does not hold information while unpowered).

Which scans for a specific very big prime number (finding large primes quickly would completely invalidate the world’s cryptography and therefore banking, that’s why people are afraid of the quantum boogeyman).

You look for any process and inject into stdlibc any backdoor of your choice (just any process, doesn’t need elevated permissions, assuming they use libc, assuming the backdoor hasn’t been patched out from the other end, defeated by any of the dozens of software integrity checks that have become standard).

No one is doing what you’re describing. If you would like to verify that for yourself there’s a ton of well documented back and forth on various platforms that indicates a complex and long running effort.

the core of Wayland development, including those directing volunteer work have historically famously been people in the employ of some company directly tied to linux in some way.

There are many, many people contributing to get screensavers working on Wayland and it’s been an uphill battle for at least one decade I’m personally extensively aware of (but according to the discussions it goes all the way back) because it doesn’t matter if you write something that works, the person who maintains whatever moving target of a protocol or api or framework will just drop the part you need after arguing with you for a month that you shouldn’t be doing what you’re doing.

Even if you cleverly hook into a some part that’s too deeply embedded in the Wayland monolith and too widely used by everything else to simply grind off, there’s no standard complement of Wayland components to design for and every major distribution of Wayland bits has enough tiny differences that by the time you’ve convinced, cajoled and wooed the first fiefdom into letting its users do something they’ve expected for forty years you gotta turn your attention to the next one who, rather than seeing the fragile peace you and your former enemy have struck over battle scarred handshakes and leading with diplomacy has barred the gates, manned the towers and released crocodiles into the moat. “You may have deployed the snakelike perfidy needed to get those guys to accept screensavers, but we will never accept your terms!”

Look at screenshots for a phenomenal example of how this goes down. If you have a program that takes a screenshot you don’t want to go fishing around for some indication of what particular method of waylandly accessing a frame buffer is installed and set up on the system or include a new dependency for your at one time simple seeming program, yet those are your options and when the ding dongs at gnome broke it they condescended to the users that they shouldn’t be taking screen shots!

And now I bet you’ll be in here saying “well, they’re right, users shouldn’t be able to take screen shots it’s too dangerous!”

lol “The corporate overlords of Wayland don’t owe you feature parity, snowflake and that’s a good thing” is a hell of a take.

To be happy. The canvas can be more than a blank slate waiting to be transformed into beloved good screen or hated bad screen. Back on x11 the screensaver can also be the locker, so if you need to know it’s not a piece of crap you can pick a locker that works good (like xscreensaver).

People have been complaining that xscreensaver doesn’t work in Wayland for a decade that I’m aware of, the answer has always been “you don’t need that”. Its not that people don’t want it, its that Wayland devs have actively resisted it.

Welcome to using Wayland, where no one cares what you want and it’s out of scope and actually you’re out of line for wanting it, get with the times!

You can graft xscreensaver in but the extent to which it will work is wholly dependent on your environment. Wayland provides the protocols, not implementations, so you may be running something that surfaces hooks for making xscreensaver work and you may not.

Lest this just be about showing pretty pictures, there have been numerous times when some desktop environment has decided to staple the pretty pictures part of xscreensaver on to its own locker (most famously to get matching window decorations when the computer asks for a password) and ended up with very funny bugs (most famously the “smash keyboard to bypass lock”).

If you’re doing campaigns or the “serious” multiplayer I think vcmi is still a non starter.

It was a couple of years ago at least. Of course hota has dropped two new towns since then so that may have changed…

You’d have to pull out a lot of stuff to make a system that wouldn’t run on x86 64.

I think what you’ve done is cool and I was not trying to denigrate or minimize your work.

My intent was to help someone who seems like they don’t have 25 years of newsgroup discussion about what constitutes a distribution and why under their belt understand why it was perfectly reasonable and defensible to get all “woah there buddy” about it.

Well, first off it’s all completely based around this one persons hardware and needs, using personal keys instead of those in the care of an organization.

There’s nothing wrong with making your own cool Linux is stitched together from the pieces you need.

It’s just something short of a distribution.

The op isn’t even doing the “distribution” component, their isos are just torrents hosted by the internet archive.

Which isn’t an insult, it’s a laudable achievement to put together an os, it just might fall short of a distribution.

Think about it like this: if you swapped the engine and drivetrain of a Silverado into an old jeep and replaced the body panels with those of a bronco carefully bent and shaped to fit the new geometry did you make a new model of car? No, of course not. It’s cool, and I want to see and drive it, but you didn’t make The Homer, you made a custom car.

If you started a business modifying other people’s jeeps with ls engine and blazer body swaps then do you have a new model of car? The many shops that do this in real life would like you to think so, but their creations remain legally registered as jeeps and no one except the dorkiest of owners refer to them as Homers.