There, look, I condensed it all into a neat title for you to digest. Oh, so you want details? I guess I can share those, too. It's a super long, complicated and somewhat emotional story, so let me try and condense that again, into something of an executive summary. But, who am I kidding, it'll probably be 87 paragraphs long anyway.
My 6900XT is dead. Likely memory hardware damage, caused by me taking the cooler off 87 times (a nice variation on my usual go-to high but believable number for hyperbolic exaggeration of doing something many times, that number is usually 27), trying to fix the 35-40C+ Edge/Hotspot delta, of course, I couldn't fix it - and there are no compatible water blocks or after-market coolers for this board.
So let me first start by saying, DO NOT buy the MSI Gaming Z Trio RX 6900 XT. The cooler is absolutely garbage.
That's not even half the problem, but it's nice to know that all six heatpipes are making good contact with the GPU, right? Oh wait, that's 25% of the cooler fin-array (the outsides) that aren't getting good thermal transfer with the core. Woops!
This picture also highlights the poor contact on one side of the GPU, this is likely the main driver of the high hotspot (think 100+ at stock), and variation on boards in the factory likely makes this of case-by-case severity. I simply wasn't happy with the fact that the board had literally zero OC headroom because it was hitting HS limits out of the box, at 300W. Max power limit is an express ticket to 115C and throttling slower than stock. The only way to run the card with acceptable delta is with -10% power limit at 275W, I mean, AMD's Navi 21 GPU is very efficient at 275W, it's still fast - but that's not the point. The point is the cooler is poorly designed. Might as well not be an XTXH binned die with this crap on top and no way to use alternative cooling.
Anyway, the card died. Memory artifacting, started with corruption in Fallout 76 in the menus; flickering, audio cutting out and then a full system restart, with the GPU not being detected when it boots. Remoting into that would reveal no GPU in the system; restarting again would recognise the card but Resizable BAR (Smart Access Memory) would have switched itself off.
Second, running 3DMark benchmark would cause artifacts after the benchmark ends, on the UI, then the system would literally lock up and stop, then restart. So I pulled the damn thing out and tried it in another PC: same thing. Windows booted, locked up, black box artifact then hard-lock. It's dead.
In a way, I'm glad it's dead, it might be a £1200 paper weight now, but that saga is over. And I have learned a valuable lesson: Listen to MY OWN ADVICE, when buying a Radeon, never, ever, EVER, give MSI a single penny of your money ever again until I die, and beyond if bloody need be.
Lillnex, for what it's worth, you were right. I should have listened, but I didn't, and here I am. Luckily, I still have the TITAN Xp, so It Is What It Is.
That brings me to the next part of this (already longer than I wanted, shocker) post. I'm quitting building custom/high end PCs as a hobby. This isn't a new sentiment from me, it's a long-time coming withdrawal from a love/hate relationship that has caused more stress and financial loss for me than anything else in my life combined. This was more or less the final straw. In my defence, if MSI wasn't so shit, I'd have a 6900XT right now and be happy. Or if I wasn't so obsessively impulsive, I would have listened to advice and read reviews before hand and bought another model and been happy.
That last part is a lot of the reason why I'm quitting building high end PCs. There's always that 'urge' to 'upgrade' or fiddle, change, or mess with the parts. And for me, while I enjoy it sometimes, it causes more stress and anxiety, and in this case, financial loss, than I feel is worth it. With that in mind, going forward, I'm moving into a new hobby of macro electronics - soldering, circuit boards, switches and the like. Potentially might involve Raspberry Pis and programmable functionality. All in all, much more hands-on and a lot cheaper for the scale I intend to work with.
I already know I have a passion for building cool electrical bits out of junk, so I'm not even sad to leave PC building behind. For the record, I'm still passionate about processors, technology, etc, all of that isn't changing. I'll still write stuff about the latest CPUs, GPUs, etc and even Babbles and all that. I'm just not buying the latest, or most expensive components with an unhealthy obsession like I used to.
Oh, and I'm selling all my expensive components, on top of that, to pay a friend back. I threw together something to use for PC gaming at my desktop, and that machine consists of a Ryzen 7 5700G and a TITAN Xp. Unique combo according to 3Dmark, but at least I'm still having fun with, doing things like running the memory at 4266 1:1 with the fabric, because that's something only Cezanne and Renoir on AM4 can do.
Closing thoughts, with the GPU market literally fucked into oblivion, it's not like I could get a bargain and enjoy the old days again, where you could get the latest flagship for 500 quid. Those days are long gone, and now, so is my 'need' to spend that money. Sure, I'll still be excited about upcoming hardware, I'm a techie after all. I might even be excited about buying said hardware, but only if it hits my absolutely anal price/perf ratio that I swore by before, with such champs as the Radeon RX 570, GTX 1660, etc. Though, unfortunately, it looks like such bargains are going the way of the Dodo.