I overbuy power supplies because I think, but have not tested, that it will convince them to run the fan at lower speed. This strategy has resulted in quiet, for sure, but I'm not sure that it gets you reliability.
This theory is sound. High-end power supplies tend to not turn on the fan until reaching 30-50% of their rated maximum output. Fanless power supplies are usually internally using the same components as regular high-end models with 80-100% higher output ratings, but with larger heatsinks. A high-quality oversized power supply with a fan will be silent most of the time, but keep itself cool when necessary.
( I have a semi-passive 750W power supply that has spun up its fan exactly once, when I accidentally shorted two of its power rails together and melted some wires while burning out the PCIe slot I was measuring the power draw through. The motherboard and wires had to be replaced, but the CPU, PSU and PCIe card were all fine.)
Spending days trying to chase down an intermittent reboot just to find that your 750W PSU is actually shitting the bed around ~350W is a terribly unsatisfying journey.
These days I don't know if that's entirely true, so I would at least look at reviews to make sure.
Anecdotally, Corsair has always been a top-notch company to me. Even when their products go bad, their support is awesome. I had a set of memory sticks appear to go bad, but I couldn't tell if it was my northbridge or the RAM itself. I was a cash-strapped kid, so I didn't have another machine to test with. Corsair sent out a set of known-good RAM sticks for me to test with, at completely no charge to me. When the test revealed that it was indeed the RAM that was going bad, they offered to let me keep the test sticks they sent me until my replacement came in the mail.
The OEM for most of their PSUs is CWT, but they are "Corsair custom" designs instead of off-the-shelf. Seasonic is only used for relatively few of Corsair's models.
Only one out of ~ 10 I bought ever failed on me, and that was after 2 years of no usage.
Jonnyguru's last Seasonic review: https://www.jonnyguru.com/blog/2018/07/03/seasonic-prime-ult...
Literally nothing in the bad & mediocre summary sections, and with nothing but recommendations for the entire series:
> Our look at the PRIME Ultra Titanium series has now at last come to a conclusion. We can now definitively say there’s not a bad performer in the bunch, and while all do have some very minor drawbacks I can’t see any reason not to recommend any of them.
Test the RAM in a known stable machine. If it passes an overnight memtest, then replace the flaky machine's PSU.
I avoid XMP, however, and typically manually specify the voltage that is listed in the RAM's spec sheet and run at the speed that the CPU manufacturer recommends. (So the RAM might be sold as 3200 but Intel says 2400 is the max, so I run it at 2400.)
[1] https://www.anandtech.com/show/14525/amd-zen-2-microarchitec...
You need 3600MT to match that while keeping 1:1 sync, which means XMP unfortunately.
Gaming hardware has larger economies of scale which leads to more people running the hardware and submitting RMAs. Gamers are also more likely to overvolt their hardware, leading to most consumer components being overspec'ed for heat and power.
As such, the second revision of a gaming board imo is more refined than a workstation/server board that almost never gets a new revision.
On a more practical level, consumer CPUs usually have higher clockrates and newer architectures and gaming NVidia GPUs are so much cheaper than the enterprise product lines that you can buy multiple. My workflow is faster with the cheap stuff, sometimes twice as fast.
I think you're also overestimating how big the gaming desktop components market is, and underestimating how fragmented it is. For motherboards, the economies of scale apply to the OEM custom motherboards used by Dell, HP, etc., which have far more volume than any one consumer retail motherboard model. You're probably also vastly overestimating how many gaming-oriented motherboards make it past revision 1.0. I just checked Gigabyte's top gaming product line, and out of 75 models listed, all but 8 were revision 1.0. Browsing through a similar number of their most recent mainstream consumer motherboards didn't turn up any that went beyond rev 1.0.
Some gamer-oriented motherboards are over-specced for power delivery and cooling. It's definitely common to find VRM heatsinks that are oversized gaudy sculptures, but that doesn't help product reliability. There are also tons of examples of gamer-oriented boards that advertise a large number of VRM phases, but have obvious shortcomings in their power delivery design once the heatsinks are removed and an EE takes a close look at things. And then there's the motherboard at issue here. When gamer boards are overbuilt, it's not thorough, it's usually just the most visible aspects that get beefed up (and the price).
RAM is a really obvious area where paying a bit extra for workstation-class ECC DIMMs gets you a real reliability advantage, but paying extra for high-speed consumer modules with RGB LEDs and big heatsinks does not.
RGB LEDs are still thankfully pretty uncommon on SSDs, but the handful of examples show that the impact varies from compromised performance to catastrophic instability. Most gamer-oriented SSDs with no LEDs but with flashy heatsinks also fail to outperform or outlive the best pedestrian-looking models.
Enterprise and NAS hard drives are definitely diverging from mainstream consumer hard drives in more ways than simple firmware feature gating. Some of the technologies they're using to drive density beyond consumer HDD needs have a questionable impact on reliability, but filling them with helium does help reliability, and you don't find that on a WD Black or Seagate Barracuda.
"Using AllFrame™ technology, WD Purple™ drives improve video capturing and help to reduce errors, pixelation, and video interruptions that can occur in a video recorder system."
"WD Purple 8TB, 10TB, 12TB & 14TB capacities feature AllFrame AI technology that enables not only recording up to 64 cameras, but also supports an additional 32 streams for Deep Learning analytics within the system."
"WD Purple drives with AllFrame AI technology feature a workload rating up to 360TB/year to support the Deep Learning analytics that is featured in AI capable NVRs."
https://documents.westerndigital.com/content/dam/doc-library...
I debugged this problem over a number of years... I replaced the RAM, I replaced the motherboard, I eventually replaced the CPU... still happened no matter what I did. Even if I underclocked the machine and kept the voltages the same, instant crash. It was maddening.
Exasperated, I eventually busted out an oscilloscope and looked at the waveform on the 12V supply to the CPU. When starting the AVX benchmark, there was a huge brownout. That basically explained everything; my power supply essentially turned off when the CPU started drawing a ton of power. I replaced the power supply and got lucky -- it handled it fine and I could run the benchmark. I even got some overclock out of it.
After this whole experience, I've never looked at computers the same way again. You can buy high-spec brand-name components, and it's all just a crapshoot. Maybe your computer won't crash in the middle of an important task. Maybe it will. There isn't much you can do but cross your fingers.