PSU tier list 2.0

[kwa-e]

I think there should be ten tiers.

Really, really, really good

Really, really good

Really good,

Good

Kind of sort of somewhat above average

Just about aver.........nah, just yanking your chain.

What would the last few lowest tiers be?

New Garbage?

Old Garbage?

Fireworks?

BOOM!?

More like,

Ok

Bad

Really bad

and finally

"WTF?!?!?!"

How about an australian inspired tiering system, consisting of:

Bloody great!
Bloody decent!
Bloody disastah!

 

[superninja12]

but when is the broom going through the list ?

 

[Brandon G]

I see the 400W XFX XT sells for $30. Good option for budget? Has there been a review on any of these yet besides what Orion checked? Seems like you get what you pay for.

I just installed the XFX XT 500w in a Xeon E3 1231v3 / GTX 950 FTW rendering and light gaming machine. I would like to test it, but I have never tested a PSU before.

What programs and tests should I use. I assume I would need additional equipment to test it too?

 

[turkey3_scratch]

You can never use programs to test PSUs - only oscilliscopes (hundreds of bucks) and multimeters (can be cheap, but cheaper ones have greater error tolerance and can't work with as much amperage). Or for real professional stuff sunmoons are used.

 

[Onus]

A lot of very expensive equipment is required to properly test PSUs, including a device to apply calibrated loads, an oscilloscope, and other measuring equipment. If I had to guess, $10K worth of equipment would be a minimum (assuming you built some of the loads yourself), and closer to $40K would be more typical for a fully automated test setup. Aris?

 

[turkey3_scratch]

A lot of very expensive equipment is required to properly test PSUs, including a device to apply calibrated loads, an oscilloscope, and other measuring equipment. If I had to guess, $10K worth of equipment would be a minimum (assuming you built some of the loads yourself), and closer to $40K would be more typical for a fully automated test setup. Aris?

On Jonnyguru they always talk about how they can't spend more than a few hundred bucks on stuff, so I guess it can vary. Tomshardware probably does use as high-end equipment as possible.

 

[orionfotl]

I see the 400W XFX XT sells for $30. Good option for budget? Has there been a review on any of these yet besides what Orion checked? Seems like you get what you pay for.

The performance is still unknown, but it's a fact that it's filled with the CS brand capacitors. I can only recall seeing them in trash units or chinese nonames - I think there's a reason they were never seen in anything by the major brands. I'm not judging without reviews, though.

It also doesn't really have a complete transient filter - the platform has only one common mode choke, and the recommended amount is two.

XFX does still offer 5 year warranty on them, so maybe they will be able to hold up so long after all.


It also really bugs me that it's impossible to know the +12V rail rating, since they have conflicting values on all three models.

 

[turkey3_scratch]

@orion: What about heatsinks? If the capacitors stay cool they'll last longer.

 

[tiny voices]

The performance is still unknown, but it's a fact that it's filled with the CS brand capacitors. I can only recall seeing them in trash units or chinese nonames - I think there's a reason they were never seen in anything by the major brands. I'm not judging without reviews, though.

I saw this too. Definitely not something I'm going to recommend even if it performs decently (I doubt it will perform better than decently at best).

 

[orionfotl]

They should stay away from hot places. The +5VSB CS capacitor here sits very close to the secondary heatsink, and that circuit is active all the time.

Secondary filtering capacitors are further away from the heatsink, so they have to worry about the temps of the filtering chokes, fan RPM and the amount of air it moves. The series is 80+ Bronze and the 500W/600W versions have large heatsinks so I believe it will be fine, but still, the build quality seems to be lower than things like EVGA 500W - more like the EVGA 400W, which also doesn't have every needed transient filtering component and uses JEC and Asia'X (but also performs worse and doesn't achieve 80+ Standard efficiency). It's the Rui Sheng Yuan OEM, after all.

 

[Darkbreeze]

Hmm. Rui Sheng Yuan, my kung fu is better than your kung fu. Let's fight!!! Sorry, that was obligatory.

 

[toyftw]

A lot of very expensive equipment is required to properly test PSUs, including a device to apply calibrated loads, an oscilloscope, and other measuring equipment. If I had to guess, $10K worth of equipment would be a minimum (assuming you built some of the loads yourself), and closer to $40K would be more typical for a fully automated test setup. Aris?

Hi - $10k would probably suffice for a "very professional" PSU testing setup. Hardware Secrets which I feel
is as good as and better than most, equipment total (excluding supplies like solder, screwdrivers, etc)
in 2009 was $4736. So, in today's $$ (and they would need an upgrade on the load tester which back then
was limited 33a on the +12v rail) 10k sounds about right.

 

[ko888]

You can't beat the Chroma Power Supply Automatic Test system Model 8000 for complete qualification testing. Fully equipped it would cost you well over $70,000 US.

 

[Karadjgne]

Made by Acme I suppose?

 

[Deleted member 217926]

Let me just check the change jar.....

 

[Darkbreeze]

Made by Acme I suppose?

 

[Turb0Yoda]

...
no words

 

[Someone Somewhere]

If you wanted to be ghetto, you could get 90% of the way there with a bunch of relays and dummy loads (e.g. bulbs or power resistors) - but the loading won't be quite as granular.

You'd still need to spend $1k+ on a good AC power meter, multimeter (several), oscilloscope, accurate shunts etc.

 

[pianoplayer88key]

My brief thoughts on number of tiers (I was reminded by seeing the 10 tiers post above) is basically keep it relatively simple...

TL;DR: 3 tiers: Great for all situations ; Don't buy but keep if it came with pre-built (used only in stock config) ; Never turn on and replace IMMEDIATELY.

* - Great for anything, provided it's appropriate wattage & amperage for the hardware.
For example no using an FX-9590 and an R9 295X2s with a SeaSonic SS-400FL2. An appropriate PSU in this tier could handle Linus's 7 gamers 1 CPU system with watercooled R9 390X2s, and another appropriate one could easily handle a home/office system with an Intel T SKU CPU, iGPU, 2.5" SATA SSD, and be safe to run 25-33% over its rated wattage for twice as long as the warranty period, even in an environment over 50°C.

* - Don't buy this, but if your pre-built came with it, it's okay to use for a while:
if you don't upgrade the CPU or GPU or add more than 1 or 2 fans or hard drives, and don't stress it. (No running Cinebench, Prime95 (even 26.6 in 22nm Intel CPUs), FireStrike, no rendering, no games that get less than 60fps at 1080p at ultra on hardware that scores 300 or less in FireStrike, etc.) Be sure to replace it with a better unit as soon as the warranty expires.

* - DANGER! Do *NOT* even *TOUCH* the power cord to anything in physical contact with the PSU when you open the packaging the computer came in!
Call the local department of forestry fire squad *IMMEDIATELY* to have them dispose of the PSU, then go buy a good quality unit (see the top tier for suggestions), install it, *THEN* turn on your new computer for the first time.

 

[Someone Somewhere]

and be safe to run 25-33% over its rated wattage for twice as long as the warranty period, even in an environment over 50°C

I don't think this is a good test, unless you consider it shutting down cleanly to be a pass. Plus it's hard to actually test.

You're also penalising manufacturers for providing a PSU otherwise equal but with a longer warranty.

 

[Vic 40]

There is nothing wrong with multiple +12v rails. Delta has been using 4x rails for years including their 1300w Antec HCP masterpiece.

I understand,but you're comparing a 1300watt to a 450watt,there's a difference.Two of the four +12V rails are almost enough to use the entire Amps available with that 450watt psu.It's my opinion,you don't have to agree.

 

[turkey3_scratch]

Multirail is great when the OEM balances out the overcurrent protection properly. If link one overcurrent protection to all the PCIe cables combined, that's not going to be a good thing having all the PCIe cables on the same rail if you have 4 rails. I've seen some bad rail distributions, but I've seen probably more good than bad.

 

[toyftw]

Let me just check the change jar.....

Must be one big *&%$# & Jar!!!

 

[pianoplayer88key]

and be safe to run 25-33% over its rated wattage for twice as long as the warranty period, even in an environment over 50°C

I don't think this is a good test, unless you consider it shutting down cleanly to be a pass. Plus it's hard to actually test.

You're also penalising manufacturers for providing a PSU otherwise equal but with a longer warranty.

Of course, shutting down cleanly would be a requirement. If I was scoring JonnyGuru reviews, it would be an automatic "0" on performance if the PSU explodes or otherwise dies during testing, no matter how good everything else is. I'd add a half-point back if it wait until it's over its rated capacity to do so. In the case of testing an old PSU, I might give it another half point, and/or strip the build quality for failed caps, or something, idk exactly.

Also I don't intend to penalize PSU manufacturers, but ... I'm not sure I exactly follow what you mean? You mean two equal PSUs with different length warranties (like the same platform used by different manufacturers), the longer warranty would be penalized more?

And speaking of long warranties and using PSUs well past the warranty period, I'd like to see a PSU still be good enough for a then-new high-end gaming system when it's as old as the one below is now. Of course that assumes it was good for a gaming system when the PSU was new, and doesn't take into account new connector interfaces over time. By "good enough", I mean it would be as likely to die or kill your hardware as a Tier 1 unit on this thread's list currently is.

Speaking of older computers, a question for those who have, shall we say, "been around a while".

What PSUs, back in primarily the 1980s and into the 1990s, were as good quality, reliable, stable, for computers of their day, as current Tier 1 PSUs are for modern PCs? Of course I know that 80+ wasn't around then, and I don't know if active PFC was even around, say, when the original IBM PC, Commodore 64, or, dang I can't think of the name of one from the late 1970s that I think ran CP/M or something.

Or has it only been relatively recently, like maybe since we started seeing mainstream multicore CPUs, or even as far back as when flagship GPUs started needing double-slot coolers, that the more reputable manufacturers have buckled down on improving PSU quality?


Also speaking of multirail vs single ... my Corsair AX760, I believe, is single rail. (It's a Tier 1, so I'm not really worried about it killing anything or vice versa.)

How do the protections, or not, or capacity, or whatever, work on multi-rail vs single-rail PSUs? For example, if some rogue SSD or LED or other low-drain device powered by 12V (substitute something else with similar power draw if SSDs and LEDs aren't 12V based, I don't remember) went haywire, could a PSU protection kick in?

And could the same PSU, assuming it had enough capacity, safely run two PowerColor Devil 13 Radeon R9 390X IIs, overclocked so its single-GPU firestrike score beats four Pascal Titans in DX12 Multi-GPU? (Also slightly OT for the thread, but I'm also curious how long before we have fast enough mobile CPUs so they could render the Cinebench benchmark at, preferably 144fps at 4K, but maybe just 60fps at 1080p would be a step in the right direction - on the single-thread test? (And no fancy stuff, for the test, allowed like spreading a single thread across multiple cores and improving performance that way, although in the real world I definitely WOULD like to see that.)

(Of course that assumes you have adequate cooling on the cards - in a case like this I'm guessing you'd still be pegging at 100°C+ on liquid helium, making it only practical for a benchmark or test, not everyday use. Maybe, to heat things up more, I should have picked an older less efficient GPU, like an 8 or 9 series GeForce or 200 series, or equivalent ATI card.)

 

[turkey3_scratch]

When being tested, I think a power supply should always stressed over its rated amperage on the rails to see if overcurrent protection or overpower protection kicks in properly. If a power supply is drawn past its rated wattage and has really out-of-spec ripple and other stuff, or if it blows, that's a bad thing, because they should be implementing the OCP or OPP trip point at a lower value. I kind of get frustrated that a lot of power supply reviewers make no mention of testing the protections. Many draw 110% of its rated wattage, but very few, such as Gabriel Torres, continue going until the baby burns up or shuts of. Some power supplies like the X-Series 560W that Torres tested actually was capable of delivering 730W of power with voltages and ripple all well within spec. Protections kicked in at 730W. Protection testing should be more common in power supply tests. Having protections means nothing, having working protections at safe trip points means a lot.

That's the one benefit of the Dell/HP power supplies that are used in those desktops. Most of them I see have properly working protection circuitry. Not all, but a lot. It's the no-name brands that custom builders buy that don't have safe trip points and catch on fire. I think PSU reviewers need to be less scared about ruining their power supplies and more focused on, after everything else is finished, testing every protection.

Then again, and as end user, you should be responsible for not drawing more than your PSU is rated for, but your PSU should also be responsible for taking proper action if such an occurrence arises. It's like a dual-fold deal.

One of the things that frustrated me was this Bargain Basement PSU Roundup at Jonnyguru back from 2007 http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story&reid=71

The bad thing is that most of these PSUs were an old design made for heavy 3.3V and 5V load, and they ignored that and pulled a ton of amperage from the 12V rail. The good news is that by doing so they actually tested some protection circuitry!