80 Plus Certification power graph

[Shrek]

I look at the power graph from 80 Plus Certification and my mind wonders about the knee in the graph...
but there is actually none, the first part of the graph covers 30% and the second part the other 50%

I find this unprofessional in the extreme; what manner of Engineer made this?

 

[_roman_]

That graph just covers more the bad efficiency.

I write this quite often. I want to see the efficiency in 5 Watt steps from 0 Watt to e.g. 150 Watt for a PSU. In that area the efficency is worse.


Your graph just shows the wasted Wattage but not the very bad efficiency.

I bought the 750W one: You may pay attention to the efficiency numbers. 20 Watts are possible for a Mainboard / CPU / DRAM / NVME

Verlustleistung - Wasted power in my words
red line - effizienz - efficiency at 230V AC 50Hz = central europe for example

It's german: https://www.tweakpc.de/hardware/tests/netzteile/enermax_revolution_df/s03.php


I do not care for a good efficieny at 5 % of 1050 Watt PSU. That is far too much Wattage already. The real game is for 5 Watts in 5 Watt steps to 105 Watts on the output.

10 Watts on the output and 10 Watts Wasted is awful. Your Diagram in post #1 hides taht fact very nicely. I prefer the Percentage line. Something around 50% efficiency is very, very bad.

 

[Shrek]

Yes, but your graph has a linear scale, the one I showed does not.

But I agree, one likes to see the low loading behavior.

 

[Bill_Bright]

Sorry Shrek but your post makes little sense because there is no context to it. Where did that chart come from? What specific PSU did that engineer use? What is its efficiency curve? What exactly is the chart supposed to illustrate? We don't know because we have no clue of the context of the article the chart was used in.

As far as I know, there is no PSU anywhere that has a perfectly linear power curve so that graph may represent a specific PSU perfectly.

What I do know for a fact is that "most" PSUs have a bell shaped efficiency curve with a peak efficiency obtained somewhere between 50 and 70% load. "Most" tend to be less efficient below their peak than they are above that peak. So that curve may be accurate, depending on where that supply's peak lies.

Because computer loads are constantly varying, computer PSUs (ideally) have a fairly (NEVER perfectly) flat efficiency curve from 20% (10% for top certs) up to 100%. But without knowing more about the specific PSU used in that chart, we have no context to go on.

Now before someone jumps in with some anecdotal exceptions, I intentionally said several time, "most". That leaves room for exceptions - but exceptions don't make the rule.

Yes, but your graph has a linear scale, the one I showed does not.

Huh? Do you see a perfectly straight line? I don't. I see a huge curve.

 

[Shrek]

Yes, I see a perfectly straight line; here is the output plot on a linear scale

 

[Bill_Bright]

Huh? What line are you looking at in _roman_'s graph?

The red one is not straight.

The bar tops are not straight across either.

Why did you include your own graph again in your reply? That makes no sense. And there is no straight line there either!

And you say the first part only covers 30%, that is wrong. Just because it starts at 20%, that does not mean it is not included in the calculations. Regardless, you even note the "knee" indicating it is not a straight line.

But ONCE AGAIN, you have failed to tell us where your graph came from so we have no context to base anything on.

 

[Shrek]

No, I am looking at the graph I posted.

Here is the source

https://www.clearesult.com/80plus/sites/80plus/files/manufacturer-certificate/xfx-technology-xps1200wxts2-3801.pdf

 

[adilazimdegilx]

80 PLUS standard does not cover below 20% load (apparently except for the Titanium but that is added later) and break points for standartizations are 20%, 50% and 100%. Expected ratings changes at those points so it is normal for PSUs to have a bend at 50% in graphs because it's not a middle point. Though that's just not it. Percentage difference between 20-50% and 50-100% also changes. So that should also effect the graph.
It would be aesthetic more satisfying if they'd choosen 60% instead of 50% but it still wouldnt be a perfect straight line because of the slight change in efficiencies after middle point.

80 PLUS certification specifications and ratings | CLEAResult

80 PLUS is a certification program for internal power supply units (PSUs). Learn more about the 80 PLUS ratings, power supplies and FAQs.

www.clearesult.com

 

[Shrek]

• Original graph

• Fixed output graph

 

[_roman_]

Your pdf file https://www.clearesult.com/80plus/s...tificate/xfx-technology-xps1200wxts2-3801.pdf

Most stupid thing I ever saw

a straigt line from 20% to 50%. A straight line from 50% to 100%.

They just assume the efficiency is identical. That is plain wrong.

Better make X to see the datapoint. a dot is harder to see. Connecting those few points with a line is just plain wrong.

It's also technically wrong to draw a graph like that which starts with 20% without the special symbol.

 

[adilazimdegilx]

Technically there is still a bend. Because before and after 50%, efficiency ratings change slightly. It wouldnt be straight line even if they picked the middle point and didnt introduce a artificial bend.

 

[Shrek]

I am only questioning their presentation.

 

[Bill_Bright]

But there is no bend, that is just an artifact of the presentation.

Says who?

And I disagree. Once again, no PSU has a perfectly straight output across a variety of loads.

If you look at the requirements for 80 PLUS certifications, not a single one from Basic 80 PLUS (aka White), all the way up to and including the new 80 Plus Ruby have a perfectly straight curve. They all vary a few percentage points between the different points.

For example, 115V Titanium must be at least 90% efficient at 10% load, 92% at 20% load, 94% at 50%, and 90% at 100% load. That's a spread up to 4% - NOT perfectly straight.

The graph, while perhaps a bit crude and very generalized, is correct for illustration purposes.

 

[Shrek]

I'm so tired of arguing.

 

[Bill_Bright]

Let's agree they could have used a better graph, ok?

 

[Shrek]

That was my ONLY point, but it means I don't trust the 80-PLUS people.

I recall a talk by a SMPS engineer who said that if there were any jumps in the efficiency curve, that indicated an issue with the design; this thought was falsely triggered by the 80-PLUS document.

 

[Bill_Bright]

They did say it was provided by a 3rd party.

I think they are still trustworthy. The problem is people's perception that better efficiency means a better, more reliable supply. It does not. It is only a measure of efficiency at 3 or 4 points. Important, nevertheless.

A better overall rating to base decisions on would be from Cybenetics. Though it too is not perfect or conclusive.

 

[lilhasselhoffer]

So...there are two things that you are...I think...trying to communicate.
1) The graph's bottom scale is not continuous and not linear.
2) The data has an inflection point where the thing suddenly changes...or "has a knee."


To the first point, the graph is labelled. It is not lying. It probably doesn't need to show indistinguishable difference from input to output for the first 20%.

To the second point, the red bit is your loss. It's the difference between input and output...or how much you lose that isn't capable of doing anything for you. While this is conjecture, I'd suggest that it's part of the design of the PSU. Note that there are, visually four line segments (forming a "knee"). The efficiency would be (output wattage)/(input wattage)...which looks to be a pretty much increasing function...and if you're looking at a modern switching mode power supply this is probably a function of where internal heating and resistances in the switching change. This is also why power supplies are rated at xx% efficiency under yy% of the rated load...because as they bet near maximum ratings they lose efficiency.

 

[Shrek]

I believe the data has no knee (or very slight)

I question the presentation, not the data.

 

[lilhasselhoffer]

I believe the data has no knee (or very slight)

I question the presentation, not the data.

Open Excel.
Create Column A labelled "Input Wattage"
Create Column B labelled "Efficiency"
Create Column C labelled "Output Wattage"
Create Column D labelled "% Total"
Fill A in starting at 100 and in 100 increments going to 1400.
Fill B in so that 0.98 is from 100-700, and 0.96 is from 800-1400.
Calculate C by A*B
Calculate D by C2/$C$15

Graph A and C, with D being the X axis.

Looks a heck of a lot like your elbow, once you scale to 20%-100% and even more once you play around with those B efficiencies. Note that this represents a change in efficiency of 2% when you go from 700 to 800 watts of input power, representing a changing in the switching mode supply's efficiency. It's almost like the "elbow" is just a change in efficiency which is visualized by a discontinuity, which itself is a function of design and construction. I wouldn't have called it an elbow, but it is a physical feature that makes sense with switching mode supplies and having a rated efficiency against their rated load. As such, you shouldn't feel bad or tired, just that you didn't understand why this was the way it was and now do. So, please don't stress. All is well.

 

[Bill_Bright]

I believe the data has no knee (or very slight)

Based on what?

I provided a link to the criteria where there typically is a 4% variance between 20 and 100% load. Titanium even shows a 6% spread between 10% and 50% load.

Are you suggesting those are so slight that they constitute no knee?

 

[Shrek]

Now that I think about it, there should be absolutely no knee at all, as the output should be a straight line by definition.

• 20% should be 240W
• 50% should be 600W
• 100% should be 1200W

It's the presentation that troubles me.

 

[Bill_Bright]

Now that I think about it, there should be absolutely no knee at all, as the output should be a straight line by definition.

Come on, Shrek, that's silly.

Who's definition are you using? Yours? Clearly not reality's. That's not meant to criticize. Just an observation.

You are assuming the ratios remain identical from 20% all the way up to 100%. In what reality? Show us the physics that says 75% "will be" 900W.

Does your car consume the same amount of fuel ("proportionally") when accelerating from 0 to 30 as it does from 30 to 60 as it does from 60 to 90?

Yes? How? Other variables kick in.
No? Why not?

Why would electronics be different? Are there zero variables in electronics? Absolutely. Heat a major one.

 

[Shrek]

• 20% of 1200W is BY DEFINITION 240W
• 50% of 1200W is BY DEFINITION 600W
• 100% of 1200W is BY DEFINITION 1200W

this is exactly what is in the original output plot.

 

[Bill_Bright]

Okay. As you said, early, I'm tired of arguing - especially after admitting they could have used a better graph.

I'm outta here. Have a good day.