Hacker News Reader

Comment by bioemerl | Hacker News Reader

I'm confused, I've built a couple of computers and I've never had issues with the power supply. The only weird issue I've had so far was with a first generation ryzen processor causing hard locks on the computer when it was left on for more than a couple of days.

Also with some bad memory causing really really weird issues all across the system.

Power supplies? And some of the cheap ones I've used have just done their job and not caused any problems.

Do you have any examples of what's going on with this?

View on HN

I just recently had a problem where writes to M.2 SSD wouldn't work due to a faulty PSU. Everything else was fine.

Took me 3 SSDs, 2 motherboards and 2 PSUs to figure out what the cause of the problem was.

It was a ThermalTake TR2 500W PSU. No graphics card (integrated graphics) so 500W should have been fine.

Power supplies are like car tires. It's not the right place to penny pinch. Always buy from reputable vendors. Thermaltake is not one of them.

I wasn't aware of that. This is my first time assembling a PC after probably 10 years, and I remember TT when they were making (good) coolers.

It's probably that they made good coolers and then decided to capitalize on the name by selling everything under the moon—i.e. reselling someone else's stuff.

Oh, yeah! The T-word again! I've had nothing but trouble with ThermalTake PSUs! One even tried to kill me. Never buying that garbage again.

The total wattage of a PSU isn't the issue. As another comment points out, it's the number of rails and amperage per rail. If the motherboard is drawing a lot of amps because of the CPU and you stick another high current device on it, you will run into issues. If you had a SATA SSD that drew the same current you wouldn't likely had any issue since it would have been on a different rail.

When possible I try to buy from companies that make a product, as opposed to just name/market/make cool stickers.

ThermalTake is just a reseller, Season is an example of a company that makes power supplies. Seasonic sells to others for relabeling, but also sells direct to consumers.

A 500W power supply doesn’t mean much when the important voltage rails see such a small amount of that wattage.

If you get a stock grey metal 500W power supply and try to run a modern graphics card you’ll have a far worse time than some Corsair 400W PSU.

I haven’t build a PC in a decade so maybe this is old news.

I think a lot of people still don't know this stuff and it's quite reasonable to expect people not to know this stuff.

Not everyone is an electronic engineer, and if the overall wattage doesn't give the information people need, the sales/marketing should be forced to provide the necessary details right there on the box in "plain language"; RAM speeds are still a pet-peeve of mine.

That said, I've been building PC's for ~25 years and I'm yet to have a PSU fail on me. Anecdotal I know, and with some skew as I've always tended to build near or at the top end.

That said, the Ryzen hard lockups were a real disappointment for me, after spending north of £3500 on my desktop in the OG Ryzen era with an R7 1800X, I was left with a machine that frequently hard-locked when compiling code on all cores and AGESA is such a mess that if it boots at all, it takes several minutes at times to make it past POST.

I really hope the next gen build I make, whatever it is is more stable because I'm not planning to go back to Intel any time soon.

What caused the hard lockups in your situation?

It's believed to be a fault with the Ryzen chips, though don't believe AMD ever officially addressed it (although they were RMAing them repeatedly for hte issue). I sent it back a couple of times for replacements but the lottery wasn't in my favour, and although the one I ended up with is better, it's not perfect;

https://www.phoronix.com/review/new-ryzen-fixed

I was an early adopter, and I accept that but it still left a bitter taste none the less, and ultimately it just meant that my pretty expensive computer barely got used because I hated dealing with it (still have it but barely boot it these days).

It's more like: If you buy a very old PSU it won't devote enough to 12v.

The stock grey metal models are going to be out of date, but they're not that out of date. At this point I'd expect them to have a reasonable rail balance for a modern computer.

I can't speak to what the other poster was talking about but NVidia 3000 series caused a huge stir because it would jump in power consumption much faster than previous generations and many power supplies couldn't provide the wattage fast enough and would brown out the system. It was especially bad for low and mid-tier PSUs that were nearing their rated wattage limit at peak power.

edit: I misrembered some aspects of this, The 3000 series was actually spiking much higher than rated TDP and tripping overcurrent protection versus just browning out. NVidia did recommend 850w PSUs for that first release of cards for that reason but iirc some 850w still had issues.

I put heavy blame on Nvidia for that one. There is no reason their cards should be pulling these huge transient spikes. By putting these cards out with this flaw, then pointing at the power supply manufacturers, they are causing a problem and blaming someone else for it.

The power supply is just doing its job, when you far exceed its rating for even a short amount of time it's going to try to shut down to protect the computer because it thinks something inside of your computer is burning up.

you're right, I forgot about the fact the spikes were far in excess of the TDP of the card, NVidia should have taken a lot more responsibility rather than let PSU mfgs take a lot of the heat.

It (3090) actually tripped out my quite expensive quite high-end seaSonic prime titanium 850 watt power supply. But it was perfectly fine on my old EVGA bronze 850 watt

My team ran a desktop fleet for a few years. Power supplies were a top 3 failure component, and tended to go in waves when an OEM got screwed by counterfeit capacitors or other components.

My guess was that bad power was a contributor to many other issues, but the nature of the SLO was such that more complex issues resulted in a device swap. Also, any kind of dirty environment drives higher AFRs.

Two builds is nothing. Since everything computers do is powered by electricity, a good PSU is paramount and literally the only thing I never go cheap with. Consider yourself lucky because a bad one can easily wreck your entire system.

Cheap cases used to come with crap power supplies even 15 years ago. It's great that's no longer the norm. Everyone wised up a bit on this one

Not really two. I have used 6 power supplies in total.

Maybe I've just been lucky or I've just been buying the quality stuff, two of the ones I'm running now are 1,200 Watts, the other two are in computers that don't use much power because they are basically idle servers all the time. They are also semi reasonable quality.

But I have been running a computer on a 450 w power supply that was very much having its limit pushed.

I also did run on a bottom of the barrel super cheap 500 w supply that I bought in the middle of 2020 when all of the supply crisis was happening.

That is why I was curious. My experiences didn't line up with what the OP was talking about, so I was interested to see examples of what was going on.

If you're not using a SeaSonic, you're still using a cheap power supply.

I think PSUs are a very common factor in system instability. If you're suffering from lock ups or reboots and you aren't overclocking, I would look to the PSU first. That being said, you can go a little crazy if you pay too much attention to detailed PSU benchmarks. there's a threshold where a PSU is good enough. I mainly look at the efficiency curve when shopping for a PSU. A more efficient supply won't get as hot for a given load, and so its components should last longer, and should have more headroom to degrade before they eventually do fall out of spec.

I reused a 10 year old "gold" 850W Corsair PSU in my zen3 build. It had some noticable inductor whine when the system was idle and I moved the mouse. After a year I got a 6800xt GPU, and it kept trucking along for another few months. I did some GPU overclocking and stress testing one Saturday, and my PC wouldn't start the next day. It had a good run for PSUs I suppose; it outlasted its 7 year warranty. I bought another 850W Corsair PSU to replace it, with "titanium" efficiency rating and a 10 (or was it 12?) year warranty.

99.99% of the time there is no problem. But if there are electrical gremlins present then there is a good chance the PSU may be to blame.

I see issues all the time with external power cubes. Devices goes wonky, look up the spec, find it's DC 12v @ 2 amps or whatever and replace the power supply. About 90% of the time the device is happy again.

Reply on news.ycombinator.com