Question regarding power supply heat output

[hat]

So, let's assume my current power supply is 80% efficient, and my rig draws 350w, pulling ~440w from the wall. Let's also assume this is a constant load, all day every day. Does this mean that 90w is lost as heat? Now, let's assume I replace it with a unit that is 90% efficient. This means the rig now pulls ~390w, roughly 50w less than it currently does. Would this mean I'm generating 50w less of heat? Is this any meaningful amount at all?

 

[Voluman]

Its not heat only, but yes. For me its a yes, it is meaningful.

 

[Bill_Bright]

Any energy consumed by the PSU that is lost due to inefficiency is indeed, lost in the form of heat.

However, some energy, not much, but some of the difference between what the computer (motherboard, graphics, CPU, RAM, drives, case fans, CPU fan, attached USB devices, etc.) draws and what the PSU pulls from the wall will go towards the operation of the PSU's own fan. But again, that is very little comparatively. So for all practical purposes, yeah, that ~90W and ~50W are wasted in the form of heat.

That wasted heat may not be a bad thing in the middle of winter in Ohio as it can help keep your computer room warm and toes toasty. But in the summer time, it will cause your home's AC unit to work harder, compounding the costs of those inefficiencies.

 

[hat]

Any energy consumed by the PSU that is lost due to inefficiency is indeed, lost in the form of heat.

However, some energy, not much, but some of the difference between what the computer (motherboard, graphics, CPU, RAM, drives, case fans, CPU fan, attached USB devices, etc.) draws and what the PSU pulls from the wall will go towards the operation of the PSU's own fan. But again, that is very little comparatively. So for all practical purposes, yeah, that ~90W and ~50W are wasted in the form of heat.

That wasted heat may not be a bad thing in the middle of winter in Ohio as it can help keep your computer room warm and toes toasty. But in the summer time, it will cause your home's AC unit to work harder, compounding the costs of those inefficiencies.

The rig is already drawing... quite a bit from the wall, so it dumps a fair bit of heat. If I get cold, there are ways I can make it produce more heat, of course.

I think I know what I need to know now. I'm going to need a power supply anyway, so I may as well spring for the Titanium unit. As it's going in a rig that's constantly loaded, and I have somewhat of a concern about heat output, I think it's the right choice.

 

[Bill_Bright]

I'm going to need a power supply anyway, so I may as well spring for the Titanium unit.

That's probably a waste of money! Gold is plenty. In most cases, it takes many years to make up the difference in purchase price with any savings you might achieve in energy costs. Gold is already rated for 90% efficiency at 50% load. Titanium only gets you 4 extra percentage points.

So Titanium just isn't worth it. I might get Platinum, but only if the supply was available at a "too good to pass up" discount price.

 

[Deleted member 185158]

So, let's assume my current power supply is 80% efficient, and my rig draws 350w, pulling ~440w from the wall. Let's also assume this is a constant load, all day every day. Does this mean that 90w is lost as heat? Now, let's assume I replace it with a unit that is 90% efficient. This means the rig now pulls ~390w, roughly 50w less than it currently does. Would this mean I'm generating 50w less of heat? Is this any meaningful amount at all?

2x 1070s and a 2600K??
Rig specs true on 500w??

Not platinum you need like Bill here is saying. You need a tad more headroom, that PSU will run cooler.
So a Gold at 650w might be more ideal than another 500w platinum.

 

[hat]

Specs are accurate. Maybe I'll consider a Gold unit depending on what I see when the time comes.

 

[Bill_Bright]

A common mistake is some folks believe 80 PLUS certification implies better quality power and better reliability. It does not. There is nothing in the 80 PLUS certification criteria that requires those PSUs to have better regulation, better ripple suppression, lower noise or longer lasting components.

While consumers may expect all those things because they end up paying a premium for those supplies - especially those with higher certifications, it just is not a given. There is no reason not to expect a quality Gold supply (or even a Bronze) to last just as long as a Platinum or Titanium supply.

 

[John Naylor]

1. There is no way to have a discussion on "metal or 80 + rating" w/o knowing what you pay for power and relative pourchase costs where you live. We looked at an example last week where the Platintum was cheaper than a gold, so if buying new, it would be the proverbial no brainer. On the other hand, since you already paid for the PSU in use, it would be very difficult justify the cost of replacing a PSU as you already paid for the one you have now. It is worth noting that since this 80 Plus thing came to be, we have seen many lower ratings disappear in some lines. The more enthusiasts lines... just don;t bother making Bronze and Silver anymore because to get the performance that their users expect with a particular line , the components they need to deliver that performance, a higher 80 plus rating will be a given already. So what we often saw in the past was Gold rated PSUs that were hovering around the same price as Silver Now we are seeing Platinum rated PSUs for the same, or even cheaper than their Gold models dependinmg on sales, rebates and promotions.

2. When you say it is 80% efficient... do you mean it has and 80+ rating or do you mean it has a "lower than bronze rating ?

3. Yes it is heat ... some energy is used for the PSI fan but that's so miniscule to be insignificant.

4. Your PSU is about undersized w/ twin 1070s each drawing about 193 watts in peak gaming. Using your parts list i came up with a recoommended PSU size of > 650 watts (616) watt draw) which would put you in the 750 watt category ... with no OC on the 2 GFX cards.

MSI GeForce GTX 1070 Gaming X review

We review the MSI GeForce GTX 1070 Gaming X armed with 8GB GDDR5 graphics memory. Now we all like the reference founders edition cards, but be honest with me .. everybody really waiting to see the b... Hardware Setup | Power Consumption

www.guru3d.com

Here is Guru3D's power supply recommendation:

• GeForce GTX 1070 / 1080 - On your average system the card requires you to have a 500 Watts power supply unit.
• GeForce GTX 1070 / 1080 SLI - On your average system the cards require you to have a 750 Watts power supply unit.

If you are going to overclock your GPU or processor, then we do recommend you purchase something with some more stamina. And remember, a PSU is the most efficient at 50% load. So if you use 400 Watts on average (SLI), the most energy friendly power supply would be an 800 Watt model.

5. As much as the "metal" rating, where your draw is on the curve if of significance. At past full load when gaming ... you are at the lowest efficiency point. You could pick up 2 % goin g from Silver to gold ... but you could pick up 3% going from 100% load to 50%.

6. And yes, the "metal" rating is just that a guarantee of efficiency, nothing else.

- Yes, getting a better efficiency rating usually will require better quality componentry
- Yes, getting better componentry will often deliver better performance
- Yes, getting lower heat generation will often deliver better performance
- Yes, getting a better efficiency rating will increase costs ... but will people pay for better efficiency of there is no corresponding bump in performance ?
- Yes, better efficiencies generally lead to lower noise since, all other things being equal, the fan runs less often.

As such, when ya check the reviews... significantly more often than not, and almost always within the same brand / model line, increased efficiency models will tend have better performance.

For example.... Ripple Data from Cybernetic web site

Focus Plus Gold 850 watter:

50% Load = 17.0 mV 9.9 mV 17.3 mV 6.9 mV
70% Load = 19.5 mV 10.5 mV 19.8 mV 8.4 mV
100% Load = 25.7 mV 18.4 mV 25.9 mV 16.2 mV

Focus Plus Platinum 850 watter:
50% Load 15.2 mV 8.3 mV 8.3 mV 5.4 mV
70% Load 13.4 mV 9.9 mV 9.4 mV 6.5 mV
100% Load 17.3 mV 11.8 mV 12.0 mV 7.8 mV

However, you should be lees confident about crossing model lines and even less so crossing brands. Being a math geek, when we work with users, our approach is not to tell them what to buy, but to give them the tools such that they can decide on their own what to buy.

I just reformatted it and added 2 additional examples .... around here, we use the middle one (24 cents per kwhr) so I added one for average US cost and Euro folks. Also added a bit of "commentary". Hope I copy / pasted all the formulae correctly. In case you might find it useful, spreadsheet file is attached.

 

[Bill_Bright]

That's pretty long winded and mostly a repeat of previous comments. That said, I agree with most, except,

- Yes, getting lower heat generation will often deliver better performance

There's no correlation between heat generation and performance - as long as it is not an "overheated" scenario. Heat generation is, first and foremost, directly related to the amount of power being produced. And that is simply a function of the demands put on the PSU. How much heat is then due to the efficiency of the supply.

When the computer demands more from the PSU, the PSU will produce more heat. When the demand is lower, the PSU will produce less heat. That has absolutely nothing to do with the quality of the output/performance. The absolute best performing, most efficient, El Primo Supreme 80 PLUS Optimalnium power supply WILL produce more heat whenever the demand put upon it goes up. And it will produce less when that demand goes down. Performance is not a factor.

Also, there is nothing stopping a generic, budget PSU that has a bell shaped (instead of flat) efficiency curve and a peak efficiency of just 65% from outputting tightly regulated +12.00VDC, +5.00VDC and +3.30VDC ±0.1% voltages with ripple suppression down to an amazing 10mV p-p. In other words, it can deliver outstanding output voltage "performance" and still generate lots of unwanted heat due to lousy efficiency.

Plus, if the computer is demanding 350 watts, a 500W 65% efficient PSU and a 500W 95% efficient PSU will both deliver that exact same required 350 watts. And both can do so with the exact same voltage regulation, tolerances, and ripple suppression. The only difference is the 65% efficient PSU will pull from the wall ~538W (538 x .65 = 349.7) while the 95% efficient PSU will pull ~368 (368 x .95 = 349.6).

I acknowledge you said "lower heat generation will often deliver better performance". While "lower heat generation" and "better performance" often are seen together, better performance is not because of lower heat generation. And lower heat generation does not suggest better performance.

Also, you said

4. Your PSU is about undersized w/ twin 1070s each drawing about 193 watts in peak gaming. Using your parts list i came up with a recoommended PSU size of > 650 watts (616) watt draw) which would put you in the 750 watt category ... with no OC on the 2 GFX cards.

That's fine but no where did the OP say two 1070s will be used in his new rig. In fact, he specifically said,

let's assume my current power supply is 80% efficient, and my rig draws 350w

 

[hat]

Two 1070s at 80% is 240w, I figure 110 is a generous out of my ass estimation, given the CPU is idle. Thanks for the replies, I'll be looking at titanium and gold units when the time comes. I'll see what prices look like and weigh my options.

 

[Bill_Bright]

titanium and gold

If looking at both titanium and gold, might as well look at what's sitting in the middle too; platinum.

 

[DeathtoGnomes]

I wholeheartedly disagree when someone says a higher rating is not worth the cost over a lower rating. Talking about ROI for PSUs is plain silly. If you consider the saying, "you get what you pay for", then count how many times people post on TPU in regards to PSU failures, you should 2nd guess those that are saying Gold+ is better than Platinum+. As long as the buy does the research and compares the benefits and minds the review ratings, finding a PSU within the budget should not be an issue.

I'd also like to add that electronic heat is expelled in cabling, its a small amount for sure, but that also depends on the size of the wire.

 

[Deleted member 185158]

I wholeheartedly disagree when someone says a higher rating is not worth the cost over a lower rating. Talking about ROI for PSUs is plain silly. If you consider the saying, "you get what you pay for", then count how many times people post on TPU in regards to PSU failures, you should 2nd guess those that are saying Gold+ is better than Platinum+. As long as the buy does the research and compares the benefits and minds the review ratings, finding a PSU within the budget should not be an issue.

I'd also like to add that electronic heat is expelled in cabling, its a small amount for sure, but that also depends on the size of the wire.

No heat should be in the cables. If there is, the load is too great for the wire gauge.

A hot running PSU, over 40-45c, may derate the PSU and cause unwanted ripple.

Im blind, so I purchase big PSU head room. But not wanting to lead OP in the wrong direction, should not buy too large unless intending to utilize the potential later.

 

[kapone32]

On my HX1200! I set the fan speed to a nice custom profile and I never see it go above 40 C. My principal on PSUs have always been buy the most your money can afford based on your needs. I had 2 580s in crossfire (before that 2 7950s) and my 850W PSU handled them fine. I decided to go with Threadripper and Vega 64 (crossfire) and so decided on a 1200W PSU. It is my thought process that keeping the PSU below it's limits is a good thing and all of my Corsair PSUs (one from 2005) are still going strong on systems I built for friends and family.

 

[Grog6]

Something everyone always misses on PSU efficiency is that it's a curve.

This is from Corsair, for their RM750:
$139.99msrp.

This is a Seasonic Platinum 1000W :
MSRP of $259.99

Note that both of these peak about 50% load; this is pretty normal for a mid range PSU.
But this one doesn't drop below 90%.


This is an AX1200i PSU:

This is a high efficiency PSU, about the same as the top one, but a little better..
$399.99msrp.

To get higher efficiency, you have to use more expensive construction techniques:

• Better Magnetic materials (https://www.mag-inc.com/Media/Magnetics/File-Library/Product Literature/General Information/ps-01.pdf?ext=.pdf)
• Better Rectifiers (Schottky, SiC types, Synchronous rectification)
• More complex drive circuitry (Bootstrapping, non-saturating, Synchronous)
• Better capacitors (Low-esr, Low leakage, solid poly)

There's a bunch of ways to improve efficiency by adding cost.

I've designed power supplies that had to run in vacuum; there's nowhere for heat to go, so it has to be ~97%+ efficient.
That gets Really expensive.

Takeaways:
If you want to optimize your PSU purchase for lowering the heat wasted in your room, buy one that is ~2x what you'll be drawing most of the time, as that puts it in it's optimum efficiency range.

If you want to optimize for keeping your room warm in the winter, size for 80% of load, or 20% of load.

You'll notice the more expensive power supplies are going to be Server type power supplies; if you have 100 of them on a room, you want to minimize waste heat.
I hope this helped!

 

[DeathtoGnomes]

No heat should be in the cables. If there is, the load is too great for the wire gauge.

without getting in a huge wall of text and math, I'll just say do more research. LOL

 

[Deleted member 185158]

without getting in a huge wall of text and math, I'll just say do more research. LOL

Don't need to research something that I've already experienced plenty of.

Daily rig 850w with GTX 980, no cables are warm at all.

1000w with big TEC, gets cables real hot.

Bad grounds make cables hot.

Crappy PSUs that are barely large enough for system draw run warm cables.

Properly setup and sufficient PSU, never should have warm cables.

Warm running PSU, direct indicator is heavily loaded and likely not large enough for the system.

The heat sinks in the PSU are not directly connected to the cables, therefor the cables do not dissipate PSU heat, nor should they.

But please link in any literature for my viewing pleasure. I'd gladly read it!!!

 

[John Naylor]

That's pretty long winded and mostly a repeat of previous comments.

As I have often said, I prefer not to tell people answers but educate them on how to determine them on their own... and to make sure it's not misleading. And while topics were previously covered, much of the wording was unclear or misleading, I felt it was necessary.

There's no correlation between heat generation and performance - as long as it is not an "overheated" scenario.

1. That's simply not true ... electrically, there is a correlation, resistance changes with heat ... electron flow changes with heat. It's gradual and small changes might not even be measurable with small increases ... but it's not a step curve. I have measured resistances in cables in the labortaory. I have measured quality of electric output of municipal power generators and at various points in a distribution system. Look at the Ripple Data ... as load increases, heat is iincreasing and ripple is increasing. Various factor influence this and heat is not excluded from them.

2. In my work space, acoustical noise is a performance criteria. Most folks look at Bronze (85%) and Platinum (92%) and say it's only 7% difference, In actuality the difference is almost double 8% heat vs 15% heat .... the bronze has to remove almost twice as much heat ad the platinum ... and yes, that's quite noticeable.

Also, there is nothing stopping a generic, budget PSU that has a bell shaped (instead of flat) efficiency curve and a peak efficiency of just 65% from outputting tightly regulated +12.00VDC, +5.00VDC and +3.30VDC ±0.1% voltages with ripple suppression down to an amazing 10mV p-p. In other words, it can deliver outstanding output voltage "performance" and still generate lots of unwanted heat due to lousy efficiency.

I expect then it will be very easy to provide numerous examples ?

That's fine but no where did the OP say two 1070s will be used in his new rig. In fact, he specifically said,

let's assume my current power supply is 80% efficient, and my rig draws 350w

Well, ya know what they say about assumptions .... better work on that reading comprehension before pointing figers ... to quote a phrase " no where did the OP say there was a new rig". Rest of us only saw talk about an existing rig and a new PSU.

Post # 1
"So, let's assume my current power supply is 80% efficient, and my rig draws 350w ....."

Post #4
"The rig is already drawing... quite a bit from the wall, so it dumps a fair bit of heat. .... I'm going to need a power supply."

That's why I don't like to rely on assumptions, I'm a vegetarian, don't like eating crow. Have any idea how OP came up with that assumed 350 ? Same place I and the other posters who pointed this out did. He doesn't have to say it in every post, it's in his profile.

He explained what we took for granted on how he came up with that 350 watts in Post #11 ... 240 + 110 = 350

Two 1070s at 80% is 240w, I figure 110 is a generous out of my ass estimation, given the CPU is idle.

My mindset on the original post was that I can not verify those assumptions. I think most folks would infer that the assumption presented was based upon his system specs, it is certainly the customary approach here when no mention is made of a new box or proposed list of system specs. He made no mention of such and specifically said this was about replacing a power supply in an existing rig.

However, I had some doubt as the PSU in his system specs was clearly inadequate for the componentry reflected in those system specs. A proper analysis of the type needed to address his concerns can not be performed with incomplete or impractical assumptions. Any such analysis based upon those assumptions would be grossly misleading.

-No Power costs given
-Unrealistic usage pattern assumed (full output 24 hours per day ?).
-No power supply wattage provided
-No componentry listed or mention for a new rig
-Current PSU was under powered based upon only info provided

Is it proper to size a PSU based upon the assumption that CPU is in an idle state ? s it proper to assume consistent usage 24 hurs per day ? Is it proper to size a PSU based upon a GFX card wattage of 120 when TPU says the 1070 *averages* 175 watts ? Is it proper to ignore the MoBo (30-40 watts), storage devices, RAM, USB connected items, fans and CLC cooler ? Is it proper to ignore the fact gaming level power usage only occurs when you are gaming ... and ignore the idle state ? The answer to that last one is yes .... you are drawing so little power the load / efficiency is insignificant.

Simply put there is no possible way to perform analysis to address the concerns put forth in the Ops original and subsequent posts with part of the data assumed and the rest missing. Like most of the participants here, I inferred that this was a PSU replacement for an existing sure ... or was info just missing ? I could take a guess and provide a vague "probably a waste of money " answer but, I thought it better to cover the thought process that can be used with any PSU and and any set of componentry. To ***illustrate this example***, I chose to use the set of componentry he is already familiar with ... the system detailed in his system specs.

In short, it is the "process" that is important here, not the example. If I gave you an example on how to calculate voltage drop in a conductor and gave the formula (Vd =mV/A/m x L x Ib ÷ 1000 ) using 100m as L in the example, you'd undoubtedly say "bah ... this is no good, he probably has 50m in his case" ... so what ? The "process" works regardless of the numbers. The point to be made and the reason for the "long winded explanation" is short answers like "probably not worth it" all to often can not be relied upon. It certainly wasn't in the example given as the Platinum was clearly the better buy ... in this instance. By identifying the data necessary to make this determination and how to use it, anyone, at any time, with any electric costs , with any componentry can make this determination and walk away knowing that it is correct for their individual circumstances.

 

[Deleted member 185158]

@John Naylor

I also was thinking to mention about a PSU rating in terms of the sales advertisement of "Peak" wattage vs actual sustainable usage.

How do feel about the aspect of buying a 500w PSU because the calculator says so. Do these PC PSU wattage calculators take into account it's telling you to purchase a "peak" wattage amp? edit: PSU?

 

[John Naylor]

Something everyone always misses on PSU efficiency is that it's a curve.

I think that's a cultural thing ..... the mom in Walmart shopping for a XMAs present for lil Johnny doesn't know ... but anyone who has built a PC would be hard pressed not to have picked this up.

But in my experience, anyone who know the significance of the words "PSU Efficiency" and "80 Plus" knows it's not a constant. This is very apparent looking at the 80 plus age on wiki ... and much of the discussion on PSU selection revolves around how much headroom to provide so as to get away from the efficiency drop at 100% loadings. I have never had a forum seen a forum discussion where anyone appeared to belive that it was a fixed number and I was managing hardware forums in the early 90s.

There is also a lot of misinterpretation with the word "flat" .... that is not, generally, intended to mean a flat line but a flat curve. In those images, the Y axis is greatly exaggerated (about 20 to 1) to show the differences

How do feel about the aspect of buying a 500w PSU because the calculator says so. Do these PC PSU wattage calculators take into account it's telling you to purchase a "peak" wattage amp? edit: PSU?

It depends on the source ... I have a UPS which reads out put power on an LCD, so I know what it draws. I also have a kil-o-watt meter in my toolbox

Kill A Watt Meter - Electricity Usage Monitor | P3

Most of the online calculators I find useless. I find the outervision one very reliable:

Maintenance

Another way is to simply add up the Power Consumption from TPU reviews or other surces.

CPU - Take from reviews ... include overclocking

GPU - Take from reviews.... some reviews will show overclocking settings including the Power Limiter Slider. If it says 120% on the lsider, add 20%.

Mobo (Low end) - 25-35 watts
Mobo (Hi end) - 35-45 watts
Mobo (Extreme) - 45-75 watts

RAM (PairDDR4) - 2.5 watts

SSD - 1 watt
HD - 10 - 12 watts

Optical Drive - 30 watts

1200 rpm 140mm fan - 1 - 3 watts
1800 rpm 120mm fan - 3 - 6watts

Sound Card - 8 watts
Ethernet card - 3 watts

KB and Mouse - 2 watts

USB Devices - 5 Watts

CLC - 5-10 watts

Real WC Pump - 8- 23 watts

Typical, mimimum is to go about 50 watts higher (600 ==> 650 w PSU. As Voltage Instability and Ripple increase as you get close to the 100% load point, I recommend going to 1.25 (600 goes to 750 w PSU) the calculated total for moderate overclocking and 1.50 for extreme overclocking

 

[Grog6]

True.

But I always see calculations done with a flat number, when the reality varies a lot from theory.

I think my post probably doesn't help most people here; but it might.

I also wanted to shine a light on why that Server PS is $300, and a similar one from "Cheapo MFG Co" for $50 is the same wattage.

I have yet to see a PSU for a computer with synchronous rectification, but I bet it's in a server somewhere.

Something else about Ripple voltage on the PSU: the raw voltage from the PSU is only used in a couple of spots, mostly hard drives.
In almost all circumstances, the PSU is feeding Another Switching power supply on the mobo to make the final voltage.
Very few things use 12V, 3.3V, or 5V anymore.
CPU, GPU, Ram, Southbridge and associated chips; they all have their own local power supply; that's how we can tweak them in the bios.

I was trying to come up with something that runs directly off the PSU, and only Hard Drives come to mind. And most of those have switchers on board these days.
This 10yo Seagate I have in my hand has a switcher on the 5V bus, right off the connector.

Can you think of anything?

 

[Deleted member 185158]

I was trying to come up with something that runs directly off the PSU, and only Hard Drives come to mind. And most of those have switchers on board these days.
This 10yo Seagate I have in my hand has a switcher on the 5V bus, right off the connector.

Can you think of anything?

Fans? Have a stack of Deltas all rated for an amp each.

 

[DeathtoGnomes]

Don't need to research something that I've already experienced plenty of.

This I highly doubt.


I did manage to find something assuming you can comprehend it.

Heat conduction

Heat conduction is the transmission of heat across matter. Heat transfer is always directed from a higher to a lower temperature. Denser [1] substances are usually better conductors; metals are excellent conductors. The law of heat conduction also know as Fourier's law states that the time rate...

engineering.fandom.com

Heat Loss Insulated Electric Wire Equations and Calculator

Heat Loss From an Insulated Electric Wire Equation and Calculator:

www.engineersedge.com

 

[Deleted member 185158]

This I highly doubt.


I did manage to find something assuming you can comprehend it.

Heat conduction

Heat conduction is the transmission of heat across matter. Heat transfer is always directed from a higher to a lower temperature. Denser [1] substances are usually better conductors; metals are excellent conductors. The law of heat conduction also know as Fourier's law states that the time rate...

engineering.fandom.com

Heat Loss Insulated Electric Wire Equations and Calculator

Heat Loss From an Insulated Electric Wire Equation and Calculator:

www.engineersedge.com

Yes I understand wires are a conductor. But you where implying that heat is purposely dissipated through the wires. This is not the case. The wires generally get warm from carrying the current more so than the mosfets releasing heat and a very less than 1% of that heat makes it through the PCB tracings into the wires.

The engineering physx is great and all, but you don't even know what wires are being used in your PSU. makes any calculation impossible.

So operating range of most PSUs is 40-50c. It's loaded. You're carrying tons of current across 2x 1070s which require a bunch of seperate small wires on the Original poster's 500w PSU and yea, the wires may get a little warm. From one end to the other, not just near the PSU, because if your theory is correct, the heat would dissipate before it reached the connector at the end of the wire.

So you loose current due to resistance, which is how the heat is created. I also know that current does not run INside the wire, but rather on the wire.

My wonderment on some deals where the PSU is overkill but seems to run warm.... or was it the case heat was being partially dumped through the PSU rather than out the case fans. A lot of guys like a "positive" case pressure, but this is an example of how to warm a PSU externally.