Guide: How to find a good PSU

citay

Pro
SERGEANT
Joined
Oct 12, 2016
Messages
9,840
The four main points are:

1) Look for an 80 PLUS Gold specification (or better, if you wish). The price increase over 80 PLUS Bronze isn't big anymore, and 80 PLUS Gold is quite a lot more efficient.
Not only will this save money on the electricity bill, it will also run cooler (less power going to waste), require less airflow (which makes less noise), and last longer.



2) Look for a long warranty period (ideally 5 years or more), implying the manufacturer's confidence in the product.

3) Check the PSU tier list, which gives a rough indication of many PSU models' quality. Ideal is Tier A or B.

4) Optional, but recommended: Read a proper review of the PSU model you may want to buy.
For example, when you build a silent system, you also want a silent PSU. But there are some good PSU models (80 PLUS Gold and long warranty) that might still make a bit more noise than you want. So by reading a review, you can make sure that there are no dealbreakers. However, don't rely on amateur reviews where they are using just a PC with no test equipment. PSU reviews have to be done by people with knowledge and using the proper equipment, like these for example.

I usually recommend the following PSU models because i can vouch for them:
Best mid-range: be quiet! Pure Power 11 (CM/FM) or Seasonic Core GM <- both Tier B
Best upper mid-range: be quiet! Straight Power 11 (Platinum) or Seasonic Focus GX/PX <- both Tier A
Best high-end: be quiet! Dark Power 12 (Pro) or Seasonic TX <- both Tier A

Of course there are also good models from other brands. Each brand sells a variety of models with sometimes vastly different quality, so always look at specific PSU models.
For example, EVGA sell a lot of PSU models, from the very bad (EVGA W1) to the halfway decent or mediocre (EVGA BQ, GD, GQ) to the very good (EVGA G6).

Why don't i recommend a "best low-end / cheap" model? Well, those all have downsides. As i'll explain later, there aren't really any good deals with PSUs, instead, you pretty much get what you pay for. So i recommend spending a bit more and getting a PSU model that can last way longer, all the while running more quietly, running more efficiently, providing cleaner power to the components, and so on. Even for an office PC, i would get a quality PSU, but the rated wattage can obviously be quite low then.

For the best - or should i say "most acceptable" - cheap PSU, you should look at what's available for your budget, and first filter the models with help of the PSU Tier List.
Disregard anything worse than Tier C as well as anything from the "what to avoid" list further below in this post, then try to find the best model within your price range.
Prefer PSUs from well-known PSU brands, and see if you can find a review that will list a specific model's pros and cons.
Examples of acceptable cheap PSU models: be quiet! System Power 9/U9/B9 (only those with 400W or more), Seasonic B12 BC, Corsair CV with 650/750W, EVGA GQ.

The minimum amount of money you have to spend on a PSU for it not to be complete trash starts around 40-45 EUR/USD. For that - depending on your country of course - you may be able to get one of the cheap models i mentioned above, or models of similar quality. In some other countries, PSUs may be more expensive, so the minimum amount you have to spend might start at 50 or 60 EUR/USD. And this is just for a PSU that will not stop working after a short time or cause harm to your components.

Again, there is just no way to save a lot of money on PSUs. With the CPU or RAM, you can get a slower model and the result is not severe, just the performance will be lower. But when you go too cheap on a PSU, you basically put all your other hardware at risk, among many other other negative effects like bad efficiency, "dirty" output voltages that go out of spec under load, things like that. Out of all the components to go too cheap on, the PSU is probably the worst one, it will always come back to haunt you later.


What to avoid completely:
- Noname PSU brands in general. Just stick to well-known PSU brands like Seasonic, be quiet!, Corsair etc., and find a good model from them (not all of their models are good).
- PSU models you can't find on the PSU tier list, or that are worse than Tier C. But even Tier C isn't ideal and has to be avoided for a nicer system, go for Tier A or B then.
- PSUs without 80 PLUS certification, or with just 80 PLUS White (=below Bronze), or with 80 PLUS Bronze "EU" (230V only). The full 80 PLUS Bronze certification is the minimum for a halfway decent PSU. Some no-name brands even use fake 80+ certificates like "82+". Just stick to a model that's high enough on the PSU Tier List.
- PSUs with a price that seems too cheap for their specifications (from Chinese shops for example). There are no "great deals" with PSUs. If it's very cheap, it's also bad quality.
- PSU brands that are only sold in a specific region, like NOX, Acbel, Artis, Pichau, Redragon..., because they are often bad quality and can't compete on the worldwide market.
- Cheap cases that come with a PSU inside. The low price guarantees that such a PSU is of terrible quality.

And finally, avoid PSU brands that have completely ruined their reputation, like Aerocool, Aresgame and Gamemax. They are known for "review samples manipulation" and "post-review design downgrades". This means that they send out a nice PSU (a so-called "golden sample") for the purpose of being reviewed, but the PSU that they eventually sell to the public will use cheaper and worse components, or even leave out components, and is not of the quality that was shown in the review.

And they might submit a golden sample PSU for the 80PLUS testing (efficiency certificate), it achieves the certificate, but then they sell a worse PSU with the exact same model name to the public, and keep the sticker with the 80PLUS certificate on it (which isn't valid for that new model anymore). The PSU they sell to the public ends up being borderline dangerous for your hardware in the end, and can easily fail.

PSU authority Jon Gerow aka "JonnyGuru" (now head of PSU R&D at Corsair) had this to say about an AeroCool PSU, for example:

Screenshot 2022-05-04 at 17-14-48 AeroCool Integrator 600 Watt PSU Review - KitGuru.png


When a manufacturer does this, like Aerocool, Aresgame and Gamemax have done, they are not to be trusted whatsoever, because they have no integrity as a company.


How many Watts?
The required total PSU wattage depends mostly on the CPU and the graphics card, as they are by far the most power-hungry components. It's good to read some CPU/GPU reviews to get a rough idea how much power those components will need under load, add it all up, and then get a suitable PSU with a bit of headroom (because you never want the PSU to work at the limit all the time).

Examples:
B550 motherboard + Ryzen 5600X = ~150W, Radeon 6600 or GeForce 3050: ~150W, and to have a bit of headroom, you should look at a quality 450W PSU or higher.
Z690 motherboard + Core i7-12700K = >250W, GeForce RTX 3080 or Radeon 6900 XT: >300W, and you also need more headroom, so get a quality 750W PSU or higher.

The emphasis here is on "quality". A good PSU model with less wattage is better than a bad one with more wattage.
A higher rated wattage doesn't hurt, but the PSU quality is the most important factor. If the quality is not there, an inflated wattage rating won't make it any better.

Now, a more detailed explanation what differentiates a good PSU from a bad one.

First off, you might ask yourself, why is there even a guide on how to find a good PSU? After all, the PC seems to work fine even with a "bad" PSU.
Only when the PC doesn't want to turn on anymore, it could be a reason to replace the PSU, right?

Well, it's more complicated. The main thing with PSUs is, it is not a simple case of "on=works" and "off=broken". It's a complex piece of electronics which transforms the 230V or 115V mains power (depending on your country) into 12V, 5V and 3.3V for the PC to use. There can be a number of things wrong with the quality of the output voltages which are not immediately obvious.

In cheap/bad PSUs, there will be some cost-saving going on, which can make the PSU one of the first things (if not the first) to slowly cause problems in a PC. While cheap doesn't always equate to bad, there's a strong correlation with PSUs, because almost every component is important in some way. And components with a decent quality cost a certain amount of money.

So how come that some PSUs cost so much less than others at the same rated wattage? What are the cost-saving tricks with PSUs and how do they impact the quality?

There are three common ways for the PSU brand to save money. They are often used in combination:
1) Using a simpler (more outdated and less efficient) internal design which requires less components.
2) Using cheaper, lower-quality components.
3) Leaving out certain components.

The first potential problem with a cheap PSU lies in its internal design. The cheaper a PSU model has to be, the more outdated of an internal design they tend to use. Often, they tend to settle on a so-called "group-regulated" design for the cheap models, which has too much of an emphasis on the minor rails (5V/3.3V). With almost all power being pulled over the 12V rail nowadays, the other voltages will drop considerably under load with such a design. That group-regulated design is cheaper, because it requires less components than a more advanced design. It was common a decade or more ago, but nowadays it's just not suitable anymore. It can't handle cross-load with the majority of power draw from 12V, which is exactly what happens with modern hardware. Modern CPUs and graphics cards can also cause considerable so-called "transient loads", short (but high) peaks/bursts of power draw which PSUs with a group-regulated design are ill-equipped to handle. But even with a slightly more advanced design, there are ways to use less components, for example using a half-bridge vs. a full-bridge design and so on. The cheaper and more outdated the internal PSU design, the less efficient it will be (wasting more power into heat instead of being able to supply the power to your hardware).

Apart from the internal design, of course the PSU makers can also cheap out the components in general. Here we come to another very common problem of cheap PSUs: The capacitor selection.
Capacitors are components that can buffer electricity and filter/smooth the output voltages. They are important to reduce the "voltage ripple" and provide stable, clean voltages to the components attached to the PSU. There are distinct differences in quality and longevity between certain makers of capacitors (or caps for short), and even within the product range of each capacitor brand.

For one, the operating lifetime of a capacitor depends a lot on the ambient temperature, especially with electrolytic capacitors which have a fluid inside them. Caps have different maximum temperature ratings, either 85°C or 105°C, obviously the latter tends to last longer inside a warm place like a PSU. But the different brands of capacitors and even their individual product ranges make more of a difference. The caps all have a different reputation, depending on how well they held up in the past (especially in PSUs). There are some (usually Chinese/Taiwanese) brands with a not-so-good reputation like CapXon for example, and then there are some - usually Japanese - brands with a good reputation like Panasonic or Nippon Chemicon. Historically, the latter have survived well past the warranty of the PSU, while the cheaper brands depend a lot on higher airflow within the PSU, and tend to perhaps still fail soon after the warranty ends. Finally, in some places, so-called solid caps can be used, but they are more expensive and mostly used in high-end PSUs.

In my experience, PSUs are consistently among the first component to fail in a PC, in large part owed to capacitors not being able to do their job properly anymore.
Sometimes one is greeted by this upon inspecting the failing PSU: Bulged or leaking capacitors.



Other times it might not be so obvious, but the caps might still not filter the voltages very well anymore. The problem with less filtered output voltages is that all the connected components are put under more stress and might also degrade earlier than expected. This is an often underestimated factor when using a cheap PSU: The "dirtier" the output voltages are, the more the attached hardware has to "pick up the slack", and now their filtering capacitors and circuitry are under more stress and aging faster!

It's also important where the capacitors are placed inside the PSU. In the picture above, they are directly next to a MOSFET (switching/amplifying transistor), a component that can get very hot and is actually screwed to a heatsink. So caps that are in a hot, cramped corner of the PSU are also a recipe for early failure, especially with budget caps rated for 85°C.

Of course, there can be a lot of other negatives in a cheap PSU than just the capacitors, for example a low overall efficiency. But when a cheap PSU starts to fail, it's often because of the capacitors, and the failure is often gradual. Meaning, it won't go up in a puff of smoke, it will cause instabilities at first, or even things like file corruption.

The PSU makers tend to adjust their warranty period accordingly, meaning they often won't give more than two or three years of warranty for their cheaper PSU models. In other words, that was the timeframe they deemed comfortable for the quality of capacitors they used and their placement (how much they heat up, since that will shorten their lifespan). When capacitors age, they can't properly do their part in supplying clean power to the components anymore, and can't handle transient (sudden) loads or sustained full load very well. One cheap way for the PSU makers to prevent premature capacitor aging is to use an aggressive fan curve, which will force a lot of air over the components to remove the heat more quickly. Of course that has the downside of being rather noisy.

The final cost-saving trick in PSUs is simple: Leaving out components altogether. This can be a theme throughout the whole PSU: Less input filtering components (which makes it more susceptible to interference), less components that would improve efficiency (which makes it waste more power into heat), less components that handle power (which makes it weaker under load), and even leaving out some protection circuits (which makes it dangerous to itself and your other hardware).

So, building a cheap, but reasonably good and quiet PSU is not easy at all. With five or more years of warranty on a reasonably quiet PSU, the capacitors can't be the cheapest quality, otherwise the manufacturer wouldn't be comfortable with that warranty period. That's why the warranty is one of the good indicators of PSU quality. Basically, a long enough warranty weeds out a lot of the PSUs that will fail too early and thus would be a waste of money anyway. It's better to spend a bit more and have a better, more efficient, longer-lasting PSU that will continue to output clean voltages for years.

The issue with assessing the PSU's output quality, or to detect PSU aging: To properly test it, it would need to run on a programmable power source and power meter (for example a CHROMA test station, to create exactly specified loads). Then the PSU's outputs would have to be observed under load with an oscilloscope. Here are photos of such a setup:

https://www.hardwareluxx.de/images/stories/galleries/reviews/nt-roundup-chroma-april12/chroma01.jpg
https://www.hardwareluxx.de/images/stories/galleries/reviews/nt-roundup-chroma-april12/chroma02.jpg
https://www.hardwareluxx.de/images/stories/galleries/reviews/nt-roundup-chroma-april12/chroma03.jpg
https://www.hardwareluxx.de/images/stories/galleries/reviews/nt-roundup-chroma-april12/chroma04.jpg

A good review lab will have such equipment available for PSU testing. Again, i they just use a normal PC to review a PSU, it's not a proper PSU review.

The PSU is the foundation of the entire system. The PSU has to match the other hardware, not just in wattage, but in quality as well. A budget PSU will not be outputting very clean voltages under load, especially not after a couple years of use. If you're building anything other than a cheap office PC and using a bit nicer and/or more powerful components, then those components would be subjected to unnecessary stress with a cheap PSU, not to mention the questionable long-term reliability of that setup. So the PSU, the foundation, it has to be solid for whatever PC is built on top of that.

My other guides:
RAM explained: Why two modules are better than four / single- vs. dual-rank / stability testing
Guide: How to set up a fan curve in the BIOS
 
Last edited:

rojmiller

New member
Joined
Nov 7, 2019
Messages
18
I had a situation where I am reasonably sure that my motherboard at the time actually killed several power supplies. This was back in the AMD Bulldozer era, when I had an FX-8320 processor on a Gigabyte GA-970A-D3 Motherboard. My first power supply was an Antec HCG-620M, which lasted 7 months. They replaced it, and the replacement lasted almost 1 year. This time it was replaced with an Antec EDG650. That one lasted 3 years. They replaced it with a HCG-620M.

At that point I decided that I was done with Antec. I kept the 620M as a spare, and purchased a Thermaltake Toughpower Grand RGB. That PSU lasted 1 year.

I became very suspicious of my motherboard at that point, because my wife had an identical computer, with zero power supply failures. It was time to upgrade anyway, so I upgraded both systems to AM4 motherboards and cpus, mine with the replacement Thermaltake power supply and hers with a new one. That was 3.5 years ago, and I have had zero PS failures since then.

So I have no proof that my motherboard was a power supply killer, but that string of failures on the same motherboard is unlikely to be a coincidence!
 

citay

Pro
SERGEANT
Joined
Oct 12, 2016
Messages
9,840
Usually when a PSU dies, it's because of an internal fault, which can be triggered by the power draw of the attached hardware, but is not primarily caused by it. It can also be killed by bursts and surges coming in from the mains, for example when lightning hits nearby and sends a surge of several kV through the grid, or when the starting of an industrial machine nearby causes several bursts in the grid. And when your power outlets are not properly wired, this can also be bad.

A PSU has protections on the mains side (input filtering), as well as a bunch of protection circuits on the output side (OVP, UVP, OPP, OTP, OCP, SCP...) which prevent any kind of short etc. in your attached hardware from causing a catastrophic failure inside the PSU.

So for a motherboard/system to kill not one, but several PSUs in a row, it must have had some kind of manufacturing defect which caused a lot of strain on the PSU, almost something like a short, but not quite, and not enough to draw your attention in daily use. It would certainly be very unusual, i have not read about something like that before, to be honest.

So my first thought is: What were the symptoms of the "dead" PSUs? Probably that the PC wouldn't turn on anymore? Then there still could've been a defect on the board, for example that the capacitors of the CPU VRM weren't working right, so the CPU depended on a very low-ripple 12V output from the PSU to be transformed into the 1.xx V that the CPU gets via the CPU VRM. Once the PSU seemed to have died, the voltage ripple had just increased above that unusually low threshold that the defective board required. This theory could've been confirmed or disproven if you tested the supposedly dead PSUs in another PC, in your wife's one for example (maybe you did?).

That's just a random theory of mine, i haven't heard of such a case before. What i read about all the time though: Too old and/or too cheap PSUs causing all kinds of problems, like crashes, reboots, shutdowns, file corruption, unusual behaviour like only booting after a reset, and a lot of other things. There are easily several cases each month where the PSU was not up to par and caused trouble.

So, while i find your experience interesting and unusual, and while it probably warrants some extensive googling into the reliability of those Gigabyte boards from 10+years ago, it is just that, interesting and unusual. What's far more common is the other way around: A bad-quality and/or old PSU being primarily responsible for problems, that was the main reason to write this thread.
 

rojmiller

New member
Joined
Nov 7, 2019
Messages
18
Usually when a PSU dies, it's because of an internal fault, which can be triggered by the power draw of the attached hardware, but is not primarily caused by it. It can also be killed by bursts and surges coming in from the mains, for example when lightning hits nearby and sends a surge of several kV through the grid, or when the starting of an industrial machine nearby causes several bursts in the grid. And when your power outlets are not properly wired, this can also be bad.

A PSU has protections on the mains side (input filtering), as well as a bunch of protection circuits on the output side (OVP, UVP, OPP, OTP, OCP, SCP...) which prevent any kind of short etc. in your attached hardware from causing a catastrophic failure inside the PSU.

So for a motherboard/system to kill not one, but several PSUs in a row, it must have had some kind of manufacturing defect which caused a lot of strain on the PSU, almost something like a short, but not quite, and not enough to draw your attention in daily use. It would certainly be very unusual, i have not read about something like that before, to be honest.

So my first thought is: What were the symptoms of the "dead" PSUs? Probably that the PC wouldn't turn on anymore? Then there still could've been a defect on the board, for example that the capacitors of the CPU VRM weren't working right, so the CPU depended on a very low-ripple 12V output from the PSU to be transformed into the 1.xx V that the CPU gets via the CPU VRM. Once the PSU seemed to have died, the voltage ripple had just increased above that unusually low threshold that the defective board required. This theory could've been confirmed or disproven if you tested the supposedly dead PSUs in another PC, in your wife's one for example (maybe you did?).

That's just a random theory i came up with, i haven't heard of such a case before. What i read about all the time though: Too old and/or too cheap PSUs causing all kinds of problems, like crashes, reboots, shutdowns, file corruption, unusual behaviour like only booting after a reset, and a lot of other things. There are easily several cases each month where the PSU was not up to par and caused trouble.

So, while i find your experience interesting and unusual, and while it probably warrants some extensive googling into the reliability of those Gigabyte boards from 10+years ago, it is just that, interesting and unusual. What's far more common is the other way around: A bad-quality and/or old PSU being primarily responsible for problems, that was the main reason to write this thread.
My Antec power supplies all died by sudden death - one day the computer just wouldn't turn on. I never tested them in my wife's computer, I just quickly replaced them with a spare until the warranty supply came in. The Thermaltake that died just started rebooting the computer every few minutes. Again my only test was to replace it with a spare, which worked normally.

Probably not related, but I did have to get that replacement Thermaltake replaced, even though it was never used on the old board. I have always run my computers on UPSs (Cyberpower 1000 series), and I found that when testing the UPS my computer would immediately shut down. In this case I did a cross-test with my wife's UPS powering my computer and mine her's, and my computer shut down again when tested, while hers did not (so it was not my UPS). I assumed that my power supply had a defective hold-up time on power loss, and Thermaltake replaced it.

I realize my situation was unusual, which is why I posted. I never had any problems with the failed power supplies until they shut down completely - no random reboots or other weird technical glitches.

I am curious if this has ever happened to anyone else, although it is obviously difficult to diagnose - you have to go through several power supplies over many years with a single motherboard, but the techy types that could figure out what was happening are usually the types that upgrade often. (Especially if they are running a Bulldozer CPU!)
 

citay

Pro
SERGEANT
Joined
Oct 12, 2016
Messages
9,840
The UPS is another good clue. The Cyberpower 1000 series seems to be line-interactive, that means, it's a middle ground between a standby UPS and an online UPS. While a standby UPS is the slowest to switch to battery power (still comfortable below 16 ms though), and an online UPS has no switching time, the line-interactive UPS should switch in less than 4 ms. And we know that the ATX norm for PSUs requires 16 ms hold-up time under full load! Of course, an aging PSU can easily fall below those 16 ms, when the primary capacitor(s) can't buffer enough electricity anymore. But your PC shutting down within what i assume to be a mere 4 ms window before your battery power takes over, this is more than suspicious.

We will probably never fully get behind the cause of your problems. And i can tell you that i read about all sorts of problems on all sorts of forums, i hear about it in my line of work, and replaced a fair share of failing PSUs for customers myself. Once, i even had a brand new PSU fail on me the first time we turned it on at the customer's office, with a puff of white smoke coming out the rear, and the customer standing nearby. But i have to say, something like your problem, that's probably the first time hearing that. This is what makes me think that something else contributed to the faults, and the cause of the faults was never properly diagnosed to the extent that these weird events demanded.
 
Last edited:
Top