Have you ever experienced a power supply failure?

[Blue4130]

[EOL] PSU Tier List rev. 14.8

PSU Tier List 4.0 rev. 14.8 (END OF LIFE) Last Update: 27-07-2021 Legend : Gray - EoL/obsolete and/or otherwise not recommended for purchase. Green - small form-factor (gold and blue colors are disregarded due to scarcity of SFX PSUs) Gold - best units in the tier (includes requirements for blue ...

linustechtips.com

They

I personally don't put a whole lot of trust in "the list". They don't do any testing and it's more of an aggregate similar to websites like *Best *** of 2021!!"

Also, reviews test single units not multiples. Not every psu, even same model, will not test the same. It's up to you as the consumer to use as many metrics as possible and make an informed decision.

 

[Sempron Guy]

hmm here's a few I can pull from memory

1. generic psu I bought around 2008 to power my Sempron 3000+ and Geforce 7600GST. Fan on the psu didn't work at first boot. This was the last generic psu I ever bought.
2. OCZ ModXstream 600w. One day it decided to die out. No early signs that it's gonna go. Had it for around less than 2 years if I remember correctly. It was powering an overclocked Pentium E5200 + HD4850 which I later upgraded to an unlocked Athlon X2 -> Phenom II X4 quad w/ unlocked L3 + GTX 460
3. Corsair VX550. Got a lemon unit.
4. Deepcool DE 500w from my sister's rig that I bought. Choose it cause my sister's rig literally can be powered by a potato (apu set-up) It died in peace after 2 years.

 

[GreiverBlade]

iirc, none of my "unpainted unsleeved" one ever failed ... (heck i still keep some for retrorig sake, i even have a LC-Power 350W ghetto modded with a blue led 120mm fan fo laugh and for P4/C2D rigs purpose, still holding strong )

the only event where i suspected a PSU failure, well that would be the recent one (and still i can't be 100% sure it was not the mobo or the CPU/RAM) was with my previous Seasonic M12II 750

and i got a sh!ttons of "firehazard" PSU, Thermaltake smart series, chieftec (hahaha .... my current Proton 1000W is a Chieftec although the platform used which Sirtec is the OEM, seems healthy ) LC-Power, Corsair CX (tho i might have been lucky and got one with the "good" platform instead of the "bad" one, probably a DSAII/III ) yet none of them failed

by no mean that recent failure (if it is one) make me think Seasonic is bad ... she did hold 5yrs+

 

[The red spirit]

And i got a sh!ttons of "firehazard" PSU, Thermaltake smart series, chieftec (hahaha .... my current Proton 1000W is a Chieftec although the platform used which Sirtec is the OEM, seems healthy ) LC-Power, Corsair CX (tho i might have been lucky and got one with the "good" platform instead of the "bad" one, probably a DSAII/III ) yet none of them failed

I wouldn't consider them being bombs. Thermaltake or Chieftec is still more trustable than a complete no name brands. LC-Power is definitely not something I would trust. You can look at something that is truly sus:

Power Supply ATX AK-B1-500 500W

Akyga® AK-B1-500 ATX power supply 500W is one of the basic power supplies in our offer dedicated to power desktop computers

www.akyga.com

It retails at measly 20 Euros. Only has half power on 12 volt rail, no 80 plus rating and etc... I would be really cautious with it.

 

[freeagent]

I have an injured TT SPM850.. it was a decent PSU. I overloaded it for too long making it not a a good PSU. Doing the exact same thing as my TX-850.

 

[eidairaman1]

It is Nike's opinion and what he/she thinks is a best practice based on personal experience. It sounds similar to the fear mongering that exists in other forums, but we should be respectful and provide a valid rebuttal to these advises.

I am always up for the best PSU available for a given system, but not everything has to be top of the line when you're just going to open a doc, some emails, watch videos or movies, and play some very casual games.

Its why bronze exists. Good clean power is a must for any system no compromise.

 

[Shrek]

Good clean power is a must for any system no compromise.

Devil's advocate... which component uses the power supply without first regulating it? the CPU has a VRM.

Not saying clean power is not important, but there is always ripple so we must compromise (1% ripple max on the 12 and 5 volt lines according to the ATX specs at 10 Hz - 20 MHz)

Which brings in an interesting trick; capacitors fail to do their job at high frequency because they have a small inductance, but two identical capacitors in parallel double the capacitance but half the inductance, so a nice technique is to use two smaller capacitors rather than one larger (the smaller capacitors already have lower inductance even before the halving).
I recall the first time I saw a circuit diagram with a small capacitor in parallel with a large one and it took some time before I realized that the function of the small capacitor was to cover high frequencies where the large one could not.

 

[Valantar]

Devil's advocate... which component uses the power supply without first regulating it? the CPU has a VRM.

Not saying clean power is not important, but there is always ripple so we must compromise (1% ripple max on the 12 and 5 volt lines according to the ATX specs at 10 Hz - 20 MHz)

The CPU VRM expects reasonably clean power already. If it didn't, baseline requirements for on-board filtering would need to be much higher, driving up motherboard prices. It's always easier to clean power at the source - that way you need one beefy filtering stage rather than several. This is both less wasteful and just plain smarter. Of course high end gear will give you several whether you need them or not, but that's besides the point.

Too much ripple into your CPU or GPU VRM will cause it to operate less efficiently and optimally, at best leading to more heat output (and thus more waste + worse VRM performance), at worst leading to outright failures over time.

Its why bronze exists. Good clean power is a must for any system no compromise.

Yep. Efficiency is great, but clean power is a necessity. Of course a balance is ideal - otherwise we'd all be running gigantic linear regulators.

 

[eidairaman1]

The CPU VRM expects reasonably clean power already. If it didn't, baseline requirements for on-board filtering would need to be much higher, driving up motherboard prices. It's always easier to clean power at the source - that way you need one beefy filtering stage rather than several. This is both less wasteful and just plain smarter. Of course high end gear will give you several whether you need them or not, but that's besides the point.

Too much ripple into your CPU or GPU VRM will cause it to operate less efficiently and optimally, at best leading to more heat output (and thus more waste + worse VRM performance), at worst leading to outright failures over time.


Yep. Efficiency is great, but clean power is a necessity. Of course a balance is ideal - otherwise we'd all be running gigantic linear regulators.

Yup the psu is the first stage of protection, compromising on that can fry the rest of the system. Considering there are many lands where the powergrid is trash or just getting old.

 

[Shrek]

It's always easier to clean power at the source - that way you need one beefy filtering stage rather than several. This is both less wasteful and just plain smarter.

I beg to differ; one must also clean near the source since devices like RAM and CPU are throwing noise into the system after the power supply.

The CPU can throw in noise at the GHz range while the power supply stops at the MHz range.

 

[Valantar]

I beg to differ; one must also clean near the source since devices like RAM and CPU are throwing noise into the system after the power supply.

The CPU can throw in noise at the GHz range while the power supply stops at the MHz range.

Sure, but I said one beefy filtering stage. AC/external noise sources is where you're the most likely to get significant electrical noise after all. If your internal components are creating voltage ripple in the high double digits, there's something seriously wrong with them. So while you clearly do need some peripheral filtering as well, doing all/most of the filtering at each end is bound to be inefficient (at least in terms of materials) and will put the onus on each component to be good at filtering rather than passable. That to me sounds like a recipe for higher prices with worse quality overall.

 

[Shrek]

Absolutely, I am all for good regulation at the supply (never said otherwise); I was only questioning the 'no compromise' approach.

Good, not brilliant; given that there are noise sources on the mother board. You are right, demanding brilliant supply regulation will just give rise to high prices with no benefit, i.e. go down to a noise level maybe a bit less than that already on the mother board, but it is wasteful to do much better. To use your words "This is both less wasteful and just plain smarter."

I did begin with the words "devil's advocate..." exactly to avoid an argument.

 

[freeagent]

If I keep pushing I might

 

[eidairaman1]

If I keep pushing I might
View attachment 216445

What you talking about willis?

 

[Bill_Bright]

It's always easier to clean power at the source - that way you need one beefy filtering stage rather than several. This is both less wasteful and just plain smarter.

I beg to differ

I'm with Andy on this. Certainly a "beefy" filtering stage at the source is extremely beneficial, but I totally disagree that only one is needed - at least not with sensitive, high-density, low DC voltage devices like microprocessors, memory modules and the such. If one was enough, why do even budget motherboards have several filtering stages and the higher end motherboards have multiple stages of filtration?

If your internal components are creating voltage ripple

Wait! What? How are a computer's internal components - which run on DC - creating ripple? Ripple, by definition, is "the inherent ac component of the output voltage caused by the internal switching of the power supply."

Ripple is a product of AC voltage. It cannot be created by devices running on DC. It can, however, affect the stability of DC devices.

One "beefy" filtering stage is plenty good for things like fan motors. But for very sensitive devices like CPUs, several stages, each improving the voltage a little bit more is the only way to go.

 

[witkazy]

Inter tech died on me around 775 socket era and that is about it .Pleasently surprised with Phantex in my last build , all plugs fit in sockets like some magick lube was pre applied

 

[ArdWar]

Wait! What? How are a computer's internal components - which run on DC - creating ripple? Ripple, by definition, is "the inherent ac component of the output voltage caused by the internal switching of the power supply."

To be fair, logic devices consume power discontinuously. For dynamic logic mostly at the clock cycle frequency. Of course in engineering term it isn't usually referred as a ripple, but noise is noise no matter if it came from source or sink. The cycle, being at MHz range, not only affect the device's operation but also might affect nearby devices due to EMI. Being at MHz range also means filtering is somewhat easier with ceramic capacitors both at the CPU substrate for example, and close coupled capacitors at the socket backside. This particular filtering is also a case where one single central filter at the source won't work, since the lengthy conductors' resistance and inductance make the filtering ineffective.**

One "beefy" filtering stage is plenty good for things like fan motors. But for very sensitive devices like CPUs, several stages, each improving the voltage a little bit more is the only way to go.

One beefy filtering stages work if your load is purely linear (like you said, motors mainly), not changing and/or homogenous. It won't work as well once you add nonlinear loads like non-PFC power supplies. For example, if you use CFL or LED lighting I doubt the bulb's PSU use PFC at all (newer LED bulb might use some fancy PFC trick tho).

** Edit: The same concept also apply for mains power. If your appliances make a lot of high frequency noise (for example through hard switching, rectification, or old school dimming circuit), one central filtering won't be as effective as local filtering.

 

[freeagent]

What you talking about willis?

Its only a 750w unit that's running a pretty barebones setup.. unless my math is off and by off I mean backwards. That was F@H on the CPU and GPU

 

[The red spirit]

Its only a 750w unit that's running a pretty barebones setup.. unless my math is off and by off I mean backwards. That was F@H on the CPU and GPU

That explains the confusion. 610 watts with 750 watt PSU is totally fine.

 

[Darmok N Jalad]

Nothing catastrophic here. 2 total, both were Antecs. One was a SFX that just stopped working after maybe 4 years—the board it was powering still works to this day. The other was DOA, so that probably doesn’t even count.

 

[Bill_Bright]

The other was DOA

Shaka, when the walls fell.

 

[GreiverBlade]

I wouldn't consider them being bombs. Thermaltake or Chieftec is still more trustable than a complete no name brands. LC-Power is definitely not something I would trust. You can look at something that is truly sus:

Power Supply ATX AK-B1-500 500W

Akyga® AK-B1-500 ATX power supply 500W is one of the basic power supplies in our offer dedicated to power desktop computers

www.akyga.com

It retails at measly 20 Euros. Only has half power on 12 volt rail, no 80 plus rating and etc... I would be really cautious with it.

yeah true, although Thermaltake Smart series had some bad reputation at the time i got it and Chieftec well ... i quite like the price/perf ratio of the Proton, although the OEM is trustable, a bit like the CX series from Corsair had "bad" units outside the CWT DSAII/III platform OEM

as for LC-Power, in europe it's not much a no name ... at least their Metatron series were quite good (which is not the 350W i have )
example of a review conclusion of the LC1000 V2.3 Legion X2 of the Metatron series
"For currently just under € 140, the customer receives a PC power supply that largely fulfills the promised properties and can provide more than just the specified 1000 W. This is a welcome development and LC Power should follow this path with the other models as well. With a targeted search, the enthusiast will surely find alternatives, albeit at slightly higher prices. But the differences are not great."

 

[freeagent]

That explains the confusion. 610 watts with 750 watt PSU is totally fine.

I think I am still going to get another unit with a bit more beef.

 

[The red spirit]

as for LC-Power, in europe it's not much a no name ... at least their Metatron series were quite good (which is not the 350W i have )
example of a review conclusion of the LC1000 V2.3 Legion X2 of the Metatron series
"For currently just under € 140, the customer receives a PC power supply that largely fulfills the promised properties and can provide more than just the specified 1000 W. This is a welcome development and LC Power should follow this path with the other models as well. With a targeted search, the enthusiast will surely find alternatives, albeit at slightly higher prices. But the differences are not great."

I'm from Europe too (Lithuania) and despite seeing various LC powers in online shops I haven't really seen anyone having it in their PC. They are very underrepresented in reviews and etc and I haven't heard anything about them. They also have tendency to make power supplies that are really uncomfortably cheap, look like some rebranded units from last decade. They look too sketchy to even try.

 

[silentbogo]

I'm not gonna count those that exploded in my workshop... Only the ones that died in regular household use.
1) Anyone remembers Dynex? For non-americans: it's an equivalent of FSP Qdion...
Back then I was overseas and needed any PC ASAP, so I went and bought a cheapest prebuilt at Bestbuy. 2 months later a PSU exploded. Did an RMA and they've replaced it with another Dynex, and it also exploded with sparkly fireworks and lots of magic smoke. Gave up on RMA and bought a good-ole Antec (back then it was still good).
2) Coolmax CUG-950B , which is a shitty PSU per-se, but it served me well for over 12 years. I bought it for my very first HEDT rig(Nehalem and a pair of GTX275's in SLI!!!). This one died quite recently, and quietly. Wanted to throw it into my cheap-ass HP Z620 ghetto-rig, but it simply refused to power up. It's just a failed standby rail, so I might fix it some time in the future(along with missing modular cables).
3) My trusty Seasonic SSR-450RT. Same thing as coolmax - standby rail started acting up, especially after power loss or hitting the back switch. Eventually it stopped working all together.
BTW, I've managed to fix it about a week ago, all while playing a Russian Roulette with infamous new Gigabyte PSUs.... It was so cheap that I simply could not resist (I mean it was even cheaper than no-name and bottom-of-the-barrel garbage). I guess news have caught-up with retailers and they are dumping them at loss(or at cost, depending on how you look at it ).
4) Do laptop power bricks count?