Well, I wouldn't pick solely based on that list, because it's nowhere near granular enough to tell which PSU is better, but it can at least tell you which PSU not to get, which is nice to narrow things down.

The tiers are not ideal in my opinion because they roughly map as follows:
F: Fire hazard.
E: Trash
D: Still trash, also pretty much no units there so why bother differentiating?
C: Works, but worse performing half I'd never buy, while those going beyond the minimum requirements are actually good budget PSUs.
B: Barely more stringent performance requirements than C, but excludes all older/simpler designs on principle.
A: Indirect minimum efficiency (and price) requirement, with still very lax performance requirements.

The end result is that e.g. an FSP Hydro G Pro (80+ Gold, ~150€ for 1000W) that I'd never buy because it's so thoroughly mediocre to underwhelming at that price and the EVGA T2 / Super Flower Titanium (80+ Titanium, ~250€ for 1000W), which I'd absolutely recommend if it makes sense in proportion to the budget, end up in the exact same tier.

#3931
That all sounds very reasonable.
2.5G is pretty much standard on AM5 mobos, so you shouldn't have any trouble getting that.
Same with 2x SATA and an M.2 NVMe slot.
I'm not sure what you consider a decent amount of USB ports at the back, shouldn't be a problem either though.
Front IO depends on the case. If the case doesn't have a USB-C port at the front, there's nothing the mobo can do. Just pay attention whether it's USB 3.1/3.2 or 3.0. 3.x uses the 20-pin Key-A connector, 3.0 uses a very different, incompatible 19-pin. Most mobos only have one or the other, with the newer ones usually having Key-A, like the Meshify 2 needs.
Enough PCIe lanes are a given, though if you want PCIe 5.0 you'll need a B650E or X670E mobo. There are no GPUs with PCIe 5.0 yet though, so for now it won't actually do anything.

Do not buy PSUs by brand. Two reasons:
1. The high-end platforms of one manufacturer generally being good does not mean all of them are good, and their budget platforms might always suck. They're just not going to send out review samples for those.
2. Brand != manufacturer. EVGA neither builds nor designs PSUs, they just slap a new label on something they bought from the usual suspects in China or Taiwan. Seasonic is an actual manufacturer but for some reason has started outsourcing the production of their low-end units and generally started using much worse designs for them. E.g. the S12III is worse and more expensive than the much older S12II.

So always check reviews for PSUs. And not the low effort "it's a PSU and it works", you'll want one of those where it's 20 pages of measurements that you don't understand. Then you simply skip to the conclusion where they'll tell you what that actually means, whether the PSU is actually worth the price, great, got some minor flaws that shouldn't affect you, or is actually missing important stuff like overcurrent protection, acting up under high load etc.

If this only happened to a single prem player who just happened to be a scout main and playing sniper at the time, then I need you guys to come up with some conspiracy theories.

Not even budget french (belgian) or russian either.
We've got such a good mystery and yet so little drama. Surely we can do better than that.

Depends on overclocking.
Stock it's around 250W for the 3070 and 125W for the 9700K, no problem.
OC on the 9700K would push it to around 180W, if you don't do anything crazy with the 3070, that's still fine.
If it's all watercooling with the 9700K somewhere around 250W and the 3070 far beyond 300W, then I'd get worried.

I must commend your PSU choice by the way, I'm using the 520W version in my NAS/server.
It's a budget PSU, but budget as in affordable 80+ Bronze, not as in "might catch fire". It will actually deliver 620W without issues.
Well, around 20% more than 620W but at some point overcurrent protection would shut it down, because it actually got that. Because it's a good PSU.

Rule of thumb, unless you're way up there with big servers and RAM in the Terabytes, or Intel is playing stupid market segmentation games again (not applicable here), you're basically limited by DIMM size.
If the mobo supports only DDR4 UDIMMs that go up to 32GB and you got 2 slots, you got 64GB max, and with 4 slots 128GB max and so on.

Now if the mobo is old enough that only 16GB sticks were available at launch, then the listed capacity in contemporary sources will be lower, but there's usually nothing stopping you from putting in 2x32GB anyway.

Your board supports 64GB of RAM, as it should.
https://www.msi.com/Motherboard/B450M-GAMING-PLUS/Specification
Obviously doesn't help if you really need a full 128GB, but it very much does support 64GB.

#3917
MSI B450M what? I mean buying a mobo with only two RAM slots is almost always a mistake because those suck, all the corners have been cut, but even those should allow for 64GB.

I'd argue that being able to actually run a program instead of not being able to run it at all is a 1/0 = infinity% increase in performance. Mathematically that should just about justify 300 USD.

I don't think DDR5 mobos being 100 USD more more expensive is the problem? Just get a 7600X instead of a 5800X3D and it works out to roughly the same price and performance.

Is the job so terrible that it's not worth investing 300$ to be able to work from home? You'd spend much more than that on commuting.

Wanting a new monitor and GPU (+PSU) is completely unrelated, that's a you problem, don't blame your job for it.

I have outsourced that task to kindred, but just to make sure...

Don't.
Unless you're running TF2 on max settings at 4K or something equally strange, a new GPU will do nothing.

It's not.
Yes, do overclock the CPU.
That CPU is already the best you can get without replacing the mobo (and possibly RAM), so unless you've got money to throw at the problem, I'd keep it.

Yes.
You also need to read a guide on overclocking a Kaby Lake CPU, I think.
https://www.overclock.net/threads/kaby-lake-overclocking-guide-with-statistics.1621347/

I don't think you'll need anything special for the mobo. Just make sure it's either a 700 chipset, or can be updated without a CPU installed or you might run into trouble since the 600 need a BIOS update to support the 13000 CPUs.
B760 would be the cheapest of the new chipsets and completely fine for your purposes.
B660 is fine too, H610 I'd avoid.
Definitely go for one that supports DDR5.

For RAM I'm not entirely up to date, but I think at this point DDR5-5600 or sometimes even 6000 can be reasonably priced enough to get that instead of sticking with the minimum 4800.
2x 8GB should be enough, but I won't stop you if you get 2x 16GB. Though RAM is only going to get cheaper, so if you've got a mobo with 4 slots, I'd stick with 2x 8GB for now.

Yeah, nothing wrong with wanting an upgrade, as long as you're doing it for the right reasons.

Ok, so a 390 should be somewhere around a 1650 to 1650 Super or 5500 to 5500 XT. If that's still good enough keep it, if not, unless you need a huge upgrade, a used GPU will probably be cheaper.

I think you could and should reuse everything apart from CPU, RAM, and mobo. If there's nothing wrong with your HDD/SSD, PSU, cooler, and case, then I would keep them unless you're trying to sell the entire old build.
Maybe add a new NVMe SSD, but that's it.

GPU encoding would mean basically no CPU load, and that obviously makes a huge difference for a CPU-bound emulator.
https://www.youtube.com/watch?v=CLqpVImLPGE&t=305s
https://www.youtube.com/watch?v=H0pCpNT4b-Q&t=620s
Basically, thanks to NVENC (the nVidia hardware encoder), CPU encoding (x264 in particular) for streaming stopped being worth it years ago. By now AMD (AMF, formerly VCE) has caught up and even surpassed nVidia in some cases, but it's pretty even. And then there's Intel Quick Sync. It's the one thing the Intel GPUs do really, really well, because they do it better than anyone else. It beats x264 medium preset. You can not afford a CPU that would even run a 1080p 60 fps stream via x264 without choking the emulator completely.
So I do highly recommend going this route.

Yeah, if CXBX still doesn't like Ryzen, and you're fine with GPU encoding and don't want to overclock, then there's no reason not to get an Intel CPU.

Again, the i3-13100 would be a fine upgrade. Should be roughly 80% faster single threaded and you do get 8 threads instead of just 4 as with the 4690K. Still 4 actual cores though. 3.4 GHz base clock, 4.5 boost. 190 CAD.
Then there's a couple of i5s. All with 6 big cores (12 threads), but the i5-13400 only gets 4 small cores on top, the others get 8. Not going to matter unless you go for something highly multithreaded (18+ threads), but worth keeping in mind anyway.
Anyway, so the 13400, 13500, 13600, and 13600K get 2.5/2.5/2.7/3.5 base clock (1.8/1.8/2.0 for the small cores), 4.6/4.8/5.0/5.1 boost.
With the 13400 at 310-320 CAD and the 13500 at 330 CAD, I'm going to declare the 13400 to be pretty pointless.
The 13600 is for some reason not available at all and the 13600K costs 410 CAD, which really isn't worth it if you're not overclocking.

The lower base clocks are to fit 6+ cores into 65W TDP, while at boost clocks they're allowed to draw >150W. Most mobos have very generous settings in that regard by default, and if not it's easily fixed, so I'd mostly pay attention to the boost clocks.
So the 13500/13600 should be 6-11% faster even single threaded than the 13100 when properly cooled, but paying 75-100% (if you can find a 13600) is obviously only worth it if you're actually getting something out of the extra cores.
For CS:GO, TF2, and Dota, I don't they'd matter, but they can be nice for newer games and e.g. Counter Strike 2 performance might be of interest to you, since you're getting it for free.

I could've made this a lot shorter, but I want you to be able to make an informed decision instead of just slapping down "get an i3-13100 or i5-13600 and use Quick Sync" and leaving it at that.

I usually only check tftv every couple of months, but somehow always right after someone asks a hardware question. Someone asked one one day before you did, so I thought I didn't need to check the pc build thread, since I thought that was why I had the urge to check.

Anyway, a couple of preliminary question:
What's your current build?
Do you actually want to upgrade or is it just because you can't figure out what's wrong?
Do you want to reuse any hardware?
Are you fine with GPU encoding? That should make things a lot easier for the CPU and by now it's comparable and sometimes better than CPU encoding.
Is that information about CXBX doing better on intel recent? Because almost everything that made AMD CPUs suck really badly for certain things only apply to the pre 2017 architectures, and the last thing that I think would be relevant only applies to CPUs with more than 4 cores from before 2020, and more than 8 cores since. I'd definitely check just in case.
Are you going to overclock?

If you're going for as cheap as possible, I'd actually look into a non-overclockable Intel CPU with an iGPU.
Ironically, despite Intel GPUs being considerable worse for anything else, they're really, really good at encoding, consistently beating x264 for streaming (unless you've got an absurd CPU) last I checked.
So if your GPU is still good enough for your purposes, or you can get a cheap used one that is, and you reuse everything but the CPU, RAM, and mobo, you could probably get an i3-13100 and get a significant upgrade for less than half your budget. Next step up would be an i5-13600, if you want/need 6 big cores, but I don't think you do.

Eh, I've seen it.
If you've got two main targets in the über you're not going to be switching between them every 10 milliseconds, you'll try to switch as little as possible, both because a solo über is obviously much better, and because breaking the beam that often would add up.

So if you haven't flashed your secondary target in a while because of that, you need to flash it again before searching for tertiary ones or you might not get that last flash in.
But if you have flashed both primary and secondary and turn away to look for more right as the über runs out, you're not really paying attention to the position that might get you killed after the über.

Basically, don't get greedy. A quick flick to the side to über a second or third target at the end of the über is easily doable and should be done, but spinning in a circle to try and flash everyone can and will go wrong fairly often.

How you picture the end of über flashes in your head: Free über for the whole team.
How it actually goes: You turn to flash the flankers and get surprised by how fast the über drains when it's on 5 people simultaneously. You don't manage to turn back in time, drop both of your main über targets, push's fucked.

Seriously, at the end of your über you flash the second scout or whoever else is coming in for post über cleanup and is within reach, and that's it. You take that, and you're happy with it.
If you can manage two extra flashes, probably because it was a solo über, then great, but don't get greedy.

Wasn't all this already known?
I don't even know where I ready anymore, because this was over a decade ago. If I had to guess it was either Sigma or PYYYOUR who wrote about it.

Über still flickers for exactly 1 second after the charge runs out or the beam connection is broken, and every player with those flickers on them increases über drain rate by half the normal one.

Hence keeping exactly 3 players constantly flashed drains at twice (1 + 2x 0.5) the normal rate and gives you 4 seconds of über, which an extra second of flickering at the end. So the last one runs out after 5 seconds, down from the usual 9, the others slightly sooner, depending on how fast your last round of flashes was. Those last flashes are indeed almost free.

And yes, if you flash a player just once, then never touch them with the beam again, you lose 0.5 seconds of über.

It is known.
But I approve of the testing. You had not good source, set up tests, and your results are, as far as I know, absolutely correct.