What would your dream Power Supply Unit have?

[Scott2010au]

What would your dream Power Supply Unit have?

The 'dream' PSU would have the following features (v1.3):

■ Be both ATX12V and EPS12V compliant
■ Silent running, or silent up to a specified loading
■ Have a 5 year warranty, ideally a 7 year warranty
■ Modular connectivity
■ Support for 'three' CPU sockets (in cable line, in case one fails), each requiring one 4/8 connection
■ Support for at least two video cards, each requiring 'up to' two PCIe 6/8 connections
■ Be at least 71% efficient at 40 watts load (anything less is unrealistic)
■ Be at least 80% efficient at 225 watts load and above (up to max spec), I would consider lowering this to 85 watts depending on build
■ Have ripple and noise that exceeds the ATX / EPS specifications (all of them, including 'lifted' retrospective future standards) by at least 20%, I mean 'exceeds' in the good way!
■ Has received non-fake 80 PLUS Certification (at least standard, or bronze)
■ Has received non-fake ATI Crossfire-X Certification
■ Has received non-fake NVIDIA SLI Certification

I will amend the above when replies are added.

 

[MRFS]

In addition to all of the above goodies,
it would have a set of internal sensors
that detect key voltages, temps, wattages
and fan speed(s).

This data would be fed back to the motherboard
via a standard USB 2.0 cable that plugs into the
rear panel of the PSU.

Of course, there would need to be a GUI program
that displays this data, much like the ASUS PC Probe program.

If/when measurements reach variable threshholds,
the program could issue audible warnings
and/or record key events in a database.

There are a large number of uses of such data
e.g. tracking start-up wattages as HDDs spin up,
helping to optimize PSU capacity selection, and so on.


MRFS

 

[Scott2010au]

Be RoHS compliant.

http://www.rohs.gov.uk/

 

[Scott2010au]

70 percent sufficient? Have you ever heard of 80 percent efficiency?

Yes, I have, and 80 PLUS does not cover loads that low.

Having a PSU that is 80% efficient from 1 watt loading, to 100% loading, would be completely unviable to produce in a given form factor.

It would also cost a fortune!

FYI: Some PSU's become significantly more efficient at temperatures well above 23 Celsius.

There are a few, starting to reach consumer channels, that will reach peak efficiency at over 40 Celsius and only at high loads (ie: above 225 watts).

You could say that I'm trying to gauge the typical consumer thought process, and break it down into different segments / market personalities.


That said, yes ~ 70% efficiency does 'sound' low, just not in the context specified.

IMHO: A sane system would limit GPU power consumption to 340 watts, ideally over at least two dedicated/isolated rails limited to 20 Amps each. With two video cards that is 170 watts per card, far more than the stock 65 to 75 watts a PCIe x16 slot provides. If using four connections (6 pin or 8 pin PCIe power) that averages out to about 85 watts each.

That is a HUGE amount of video grunt for a consumer, even a heavy gamer.

I hear a few people around these forums have built systems with three, or four, video cards, and have system loads of OVER 500 WATTS !

A system that used 1200 watts (externally, not internally) would use half the juice a typical home circuit can provide (240 Vmax, 10 Amps, 50 Hz), which typically has three power outlets. So realistically there should be a 800 watt ceiling on devices when one thinks about it rationally.

Of course, I am asking for community input

■ What does a 'consumer' look for in a PSU?
■ How does it differ between different people?

 

[hunter315]

The nVidia and ATI certifications are pretty meaningless, there are some low end 500W unit that are SLI certified, what are you going to SLI on them? A couple of 9600GT? I would much rather a unit get a good review from Jonnyguru or Hardware secrets than be SLI and Crossfire Certified.

Also, only being 70% efficient even that low is pretty bad, the corsair AX1200 can pull off 84% efficiency at only 122W of load, and will hold about 90% efficiency for normal load range(20% to 100%), you are setting the bar real low with only needing them to be 80+ basic for the full range, even cheaper units can handle it these days with DC-DC design.


Personally, my dream PSU can be summed up as an AX1200, you would never need to worry about upgrading your PSU ever again!
http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story&reid=189

 

[Scott2010au]

Please point me to a link with charts, tables, graphs, etc that has a PSU that is over 80% efficient at 40 watts loading, with a ceiling wattage of at least 500 watts - ideally 650 watts or more.

This I gotta see!

 

[hunter315]

The Seasonic X560 will do 82% at 62W, its hard to find anyone who checked 40W because many reviews will only include a 10% mark which is at least 50W for a unit of 500W. I imagine it would do pretty close to 80% at 40W, im going to search through some of the other ones and see if i can find one that they tested even lower.
http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story2&reid=202

The X400 pulled off 81% at 37.6W, its a bit weaker than what you were looking for but it handles its full range with 80+ Gold efficiency,
http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story2&reid=200

Oh, and it doesnt even have a fan.

 

[geofelt]

A label that says:
Seasonic
PC P&C
Corsair
XFX
Antec

Perhaps some others.

 

[Scott2010au]

That's amazing research hunter315.
- Cheers!

I was going to suggest people look for 200 watt units that are 80 PLUS Certified, that way the 20% loading would cover 40 watts. (As 200 watts * 0.2 = 40 watts). As a 'work around' so to speak.

Imagine a LAN/'net café with 30 to 50 of these beasts running near full load, the sheer amount of power used & heat generated would be astonishing... at least by corporate office standards!


As for the the labels most of those are companies that do sell re-branded SeaSonic units, at least in their upper mid-range products. (Their cheaper units will be manufactured by someone else).

PC Power & Cooling I am not so sure about any-more though, they used to be highly recommended, even in server builds. Maybe they had a supplier with a bad batch of capacitors, like Dell, recently?

 

[Scott2010au]

What about a switch to select between single, dual, and multi rail over-current protection?

I think the Silverstone Zeus line has a feature like that...

 

[hunter315]

PC Power & Cooling was bought by OCZ recently so their designs have change some and they arent top of the line anymore.

For something like an internet cafe the X400 would work quite well, it would keep noise down and the gold level efficiency keeps heat to a minimum. The efficiency at lower levels isnt nearly as important as the efficiency at higher power levels, a unit thats 70% efficient at 40W consumes 57W, if it was 80% it would consume 50W, its going to take a long time for those extra 7W to show a difference, but at 200W if its only 80% it will pull 250, but if its 85% it only pulls 235 saving you 15W, the mid ranged efficiency has a far bigger effect on the overall savings, but still, only being able to do 70% at 10% load level would be pretty pathetic for a modern unit.

A switch to combine the rails really isnt too necessary as rarely will you be running so close to max capacity that you need to carefully balance your rails like that, normally they have already taken care of the rail balancing, if you need to combine it into a single rail you really need to move up to a bigger unit.

 

[Scott2010au]

Best answer selected by Scott2010au.

 

[cjl]

A system that used 1200 watts (externally, not internally) would use half the juice a typical home circuit can provide (240 Vmax, 10 Amps, 50 Hz), which typically has three power outlets. So realistically there should be a 800 watt ceiling on devices when one thinks about it rationally.

First, in the US (where a lot of us are), the standard is 120VRMS, 60Hz, 15A, which gives a maximum power of 1800 watts. However, that doesn't mean that if there are 3 plugs per circuit, you should limit each device to 1/3 of that. The vast majority of devices pull a tiny fraction of the available power from the circuit, so on a given circuit, you might have (for example) your 1200 watt computer (which, I might point out, only draws 100-400 watts at idle, depending on components), a cell phone charger (a couple watts), and a lamp (30-150 watts). Even though your computer pulls far more than 1/3 of the line's capacity, the total capacity is never exceeded. It would be very rare for people to have enough high draw devices on the same circuit to actually cause a problem.

Also, the highest power draw in most people's homes (out of things that don't have a dedicated circuit) is probably the microwave. A fairly basic unit can draw a kilowatt, and a good one here in the US can draw 1775 watts (I've measured mine at 1775 watts). This is pretty much 100% of the capacity of one circuit, and yet you very rarely hear of people blowing circuit breakers because they wanted to heat their cocoa. Basically, I wouldn't worry about it too much - you'd have to try pretty hard to exceed a circuit's capacity with your computer (especially if you have 240V).


(Oh, and a minor nitpick - 240Vmax isn't used anywhere. It's 240VRMS, which gives 339Vmax, and 679Vpp)

 

[cjl]

I don't need to show you links for this. Most psu will have a efficiency rating way over 80 percent at 40 watt load.
Proof? De rating of transistors. At that load depends on the room temperature aswell their efficiency will be high.
Now increase the load and watch the efficiency drop as higher temps come into play

Nope. Most power supplies drop precipitously in efficiency at low load. Just look at the JonnyGuru tests - he tests all PSUs at power levels from standby all the way up to 100%. Here's an example:
http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story2&reid=212

At standby, the efficiency is around 70%, at 10% load, the efficiency is around 81%, but from 20% up to 100%, the efficiency is over 86%, peaking up around 90% at a bit under 50% load. That's pretty typical behavior for modern power supplies - low efficiency at low load, with good efficiency from 20% or so on up, peaking around 50% load.

 

[cjl]

Please point me to a link with charts, tables, graphs, etc that has a PSU that is over 80% efficient at 40 watts loading, with a ceiling wattage of at least 500 watts - ideally 650 watts or more.

This I gotta see!

Well, it wasn't tested down to 40W, but here's an 800 watter that's still at over 85% at 80W. I'd imagine it's probably close to 80% at 40W, but it's hard to say for sure.

http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story3&reid=205

 

[cjl]

Of course you can design a power supply to be efficient at 40W. However, you specifically said "Most psu will have a efficiency rating way over 80 percent at 40 watt load". Most PSUs are designed to take a load of at least 300W, and your statement was wrong.

(Oh, and it isn't that hard to make a computer that only pulls 40W idle. A Lynnfield based system with a low end or an integrated GPU will do that quite easily)

 

[cjl]

lol its not wrong. rule of the thumb. Estimated power draw from hardware x 2 equals maximum efficiency.
Again if your going to plan a low load of 40w from a psu then 80-100w will be efficient. Efficiency don't get influenced by power draw a lot it gets influenced by temperatures.

This statement is correct. It's also not what you said before.

You said that "Most psu will have a efficiency rating way over 80 percent at 40 watt load". That's a factually wrong statement. Stop trying to weasel out of it.

 

[stingstang]

My dream PSU would have a hot chick despenser. Anyone top that?

 

[shovenose]

ahaa I'm not weasel out of anything.
Let me prove you wrong again
The power supply used on the PC uses an even better approach: it is a closed loop system. The circuit that controls the switching transistor gets feedback from thepower supply outputs, increasing or decreasing the duty cycle of the voltage applied to the transformer according to the PC consumption (this approach is called PWM, Pulse Width Modulation). So the power supply readjusts itself depending on the consumption of the device connected toit. When your PC isn’t consuming a lot of power, the power supply readjusts itself to deliver less current, making the transformerand all other components to dissipate less power

Ill suggest you go look up about switching power supplies. Pc don't use linear power supplies

my dream psu would make this guy not try to argue about false statemtns...

lets face it youre wrong.

 

[cjl]

ahaa I'm not weasel out of anything.
Let me prove you wrong again
The power supply used on the PC uses an even better approach: it is a closed loop system. The circuit that controls the switching transistor gets feedback from thepower supply outputs, increasing or decreasing the duty cycle of the voltage applied to the transformer according to the PC consumption (this approach is called PWM, Pulse Width Modulation). So the power supply readjusts itself depending on the consumption of the device connected toit. When your PC isn’t consuming a lot of power, the power supply readjusts itself to deliver less current, making the transformerand all other components to dissipate less power

Ill suggest you go look up about switching power supplies. Pc don't use linear power supplies

Yes, a SMPS can output less than full output reliably. However, at 10% or lower load, the efficiency drops. A lot. The total power dissipated drops, yes, but so does the efficiency. Pretty much all computer power supplies capable of 500+ watts at 85% or more efficiency will be down below 80% by the time you get down to 40 watts (Look at www.jonnyguru.com - you might learn something).

And you're still trying to weasel out of it.

 

[Scott2010au]

Yeah, but in an Internet / LAN café, or corporate enterprise environment with a 'stable' 10 Amps per circuit you wouldn't really want to be hammering it too much.

In those particular scenario's it is more predictable that each power outlet is going to have a load between X and Y. (Versus home use).

Of course, it depends on what people here consider 'normal' home usage scenario's too.


His info isn't all bad, he just isn't humble in his opinions.
In a way I respect that as it helps push progress forwards.


Of course in six years time a typical office computer may be using 40 to 85 watts internally, and slightly more (+25%) at the outlet. Today's modern power supply units should be able to last six years.

Considering that 80GB to 160GB is ample for most Government computers, in a few years an SSD that size will cost peanuts compared to mass produced HDD's (which will probably transition to 2.5" or maybe even 1.8" by that time in 'low cost' machines).

However the performance of the above 'office computer' will be remarkable by today's standards. (I'm rather forward thinking).


If he wants to post he can post, I don't have to read it all but I will certainly try to read at least some of it.

 

[Scott2010au]

Oh, How'd you convert between V_RMS and Vmax ?

 

[cjl]

Oh, How'd you convert between V_RMS and Vmax ?

Well, anytime the voltage follows a sine wave profile, Vmax will always be sqrt(2) times Vrms. The standard voltage specified for home circuits (and AC circuits in general) is RMS.

 

[Scott2010au]

Don't you mean Vrms will always be sqrt(2) times Vmax?
If I'm wrong you'll need to explain it to me sometime, or link to a diagram or graph.

 

[cjl]

Nope. Vmax will always be greater than or equal to Vrms (assuming no DC offset). Sqrt(2) is larger than 1 (specifically, it's 1.414), so Vrms couldn't possibly be sqrt(2) times Vmax.

(Not to sound patronizing or anything, but you do know what rms is, right?)