PDA

View Full Version : Efficiency, but at what price.



ashiekh
09-21-2020, 07:56 PM
If I have understood correctly, modern PC power supplies achieve efficiency but at the cost of reduced life times (correct me if I am wrong).

If so, which series should I be looking at if reliability and longevity are more important to me than efficiency or voltage stability?

Jon Gerow
09-21-2020, 08:05 PM
If I have understood correctly, modern PC power supplies achieve efficiency but at the cost of reduced life times.

How/where was that conclusion made?

ashiekh
09-21-2020, 08:13 PM
How/where was that conclusion made?


You in the last few weeks; or am I imagining it... something about complexity affecting the life.

Jon Gerow
09-21-2020, 08:31 PM
You in the last few weeks; or am I imagining it... something about complexity affecting the life.

I said complexity impacts longevity. Not efficiency.

But that's true of anything. Take a VW Rabbit compared to a Cadillac with a Northstar V8, for example.

ashiekh
09-21-2020, 08:34 PM
I said complexity impacts longevity. Not efficiency.


Ah, I guess I supposed efficiency came at the cost of complexity.


Then again I have one of the early VTEC-E engines, and it just keeps on running.

Travis
09-22-2020, 07:27 PM
Temperature of key components affects longevity too. Simplicity is good, but running hot is bad. There's no simple answer.

99wjtx
09-22-2020, 08:22 PM
Temperature of key components affects longevity too. Simplicity is good, but running hot is bad. There's no simple answer.

I think I might rather have a more complex unit built with higher quality components, tolerances and manufactured with better quality control than a simpler unit of poor design built with shoddy components on a slap-dash assembly line with an inspector asleep at the wheel. But yeah - bells and whistles and "an attitude of constant improvement" have always made products previously perfected to suit a particular purpose arguably worse than a predecessor iteration of that same product.

The real problem is that there is no consensus on what constitutes "the perfect product" for just about anything. Since this is a power supply forum, some people want longevity. Some people want efficiency. Some people want low cost or value ... so designs, attributes - qualities, get prioritized differently. Frustratingly, to get the best whatever attribute *you* most want in just about any product, you have to accept a lot of other "qualities" that you may not want which may negatively impact the experience or value for you personally. Case in point - suppose I want the most efficient ATX power supply on the market. Why? I'm just weird like that, but I don't want or need a 1600 watt power supply. So, I nevertheless end up having to buy a much more powerful PSU than I need just because I feel as though I MUST have the most efficient PSU on the market. No manufacturer is going to put its BEST efficiency design into anything but it's top-tier product. So, I must accept unnecessary added complexity just to get the efficiency that I want!

Now, if power supply company A's big boss suddenly said "let's build the most efficient ATX-format power supply not only ever created but also at the very bleeding edge of applied physics and today's technology ... a real clean-sheet moonshot, a 10 year development cycle ..." but let's also put it in a cheap, flexible ABS plastic chassis, purposefully down-build the regulation and give it too few connectors. Then let's give it a ridiculous name and paint the whole thing pink. Once we're done, let's market it on a loss-leader margin to the entry level tier of the market as a revolution in power supply efficiency. That could probably be done but ... at what cost? For one thing, certainly his job.

In the ye olde days, if you wanted arguably the best engineered, best built car in the world, you bought maybe a Mercedes. You didn't give a hoot about "luxury" but you wanted the best built, longest lasting car you could get. Unfortunately, nobody builds a car like that but with cheap seats and manual AC. So, you have to endure the ridicule of your "peers" (chosen for you not by you of course) for such "egregious excess and indulgence", and you have to pay for a lot of attributes you don't want or need and which are actually a liability for you. "I mean, wow Bob! A luxury car!??! I didn't know you were so vain!" and "I saw Bob's new $120,000 car having to be towed the other day! LOL!"

So maybe the answer is ... "well, it depends." Complexity is in itself virtually never a favorable attribute from an engineering perspective unless your specific objective is complexity itself. Or, your customer is a soccer mom or bells and whistles geek.

But we all knew all of this already. OK, back to lurking for another 12 months or so.

ashiekh
09-23-2020, 09:33 AM
Perhaps part of the problem is that some people feel that half the ripple is twice as good, but one can go on halfing indefinitely. So money may get spent on a feature that is actually of little utility.

Concerning cars, I have a Honda Civic and I would argue it is efficient, reliable, long lasting, well engineered and not pricey.

mdk777
09-23-2020, 10:07 AM
Perhaps part of the problem is that some people feel that half the ripple is twice as good, but one can go on halfing indefinitely. So money may get spent on a feature that is actually of little utility.

Concerning cars, I have a Honda Civic and I would argue it is efficient, reliable, long lasting, well engineered and not pricey.

discussions of diminishing returns can go on forever.

Skipped the NVME GEN 4 because there was no difference in performance yet (see SAMSUNG 980 PRO reviews out today)

RE: Cars....Audi had a similar reputation for performance and high engineering.
I have an A4 Quattro that has required its purchase price in maintenance over the last few years (only 100,000) miles.

I wanted to get a used V8 Genesis for the same price. ( yeah, not great gas mileage and no high tech) Wife hated it....she wanted the A4.

So, yeah...I told her the V8 Genesis would drive another 100,000 miles with nothing but oil changes...but instead I learned the cost of replacing failed Audi Turbo, Starter motor, oil burning engine issues, etc. etc.

So...it is always a mixed bag.

For you History buffs....Steam engines in the 1930-1950 were incredibly efficient.
I doubt that that any diesel electric that exists TODAY can match their energy conversion efficiency.

However, the maintenance cost, the capital cost and support cost is what ended the steam age....not superior technology.

ashiekh
09-23-2020, 10:17 AM
For you History buffs....Steam engines in the 1930-1950 were incredibly efficient.
I doubt that that any diesel electric that exists TODAY can match their energy conversion efficiency.

However, the maintenance cost, the capital cost and support cost is what ended the steam age....not superior technology.


I would question that claim.


Physics Concepts and Connections 5th edition, Art Hobson P 140

Efficiencies

Steam locomotive


Best Possible 20%



Actual 10%


Diesel auto/truck/locomotive


Best Possible 48%



Actual 30%

99wjtx
09-23-2020, 12:07 PM
I would question that claim.


Physics Concepts and Connections 5th edition, Art Hobson P 140

Efficiencies

Steam locomotive


Best Possible 20%



Actual 10%


Diesel auto/truck/locomotive


Best Possible 48%



Actual 30%


At the risk of drifting the thread further off topic, might there be some confusion here because of the fuel that was used to generate the steam? Were later steam engines sometimes powered by something other than coal? Oil, diesel, etc.? If so, that might explain the discrepancy here.

ashiekh
09-23-2020, 12:29 PM
At the risk of drifting the thread further off topic, might there be some confusion here because of the fuel that was used to generate the steam? Were later steam engines sometimes powered by something other than coal? Oil, diesel, etc.? If so, that might explain the discrepancy here.



The best possible figures are based on temperatures and thermodynamics (T_in - T_ex)/T_in where temperatures are in Kelvin

I will let the moderators decide if we are off track or not.

But one other thing, some African counties at civil war would buy old steam engines as they are easier to maintain against terrorist attack (one can machine up replacement parts) and being external combustion engines they could burn almost any fuel.

I seem to recall some very large (and slow) diesel engines can approach 50% efficiency in practice, but such engines are more suited to ships and ground based applications rather than cars/trucks/locomotives

Which leads to sterling engines that are also external combustion, so might be the best of both worlds.

Things are rushed out here, so some of what I write may be in error.

99wjtx
09-23-2020, 01:08 PM
So the takeaway here would be that achieving desired end-product attributes is usually a matrix of many variables with varying weights and priorities. The question on my mind remains to what extent MUST complexity be incorporated if efficiency is the ONLY primary consideration when designing a PSU? I speak as an end-user and in no way understand the intricate technical aspects of PSU operation and design but I suspect all the components work in concert to achieve the end result, meaning a likely by-product of very high efficiency may include exceptional regulation as well, but I don't know to what extent that would be true.

What I'm driving at is, it would seem to me as though complexity is often not strictly necessary when attempting to achieve one or a few primary outcomes but rather complexity becomes a factor for various other reasons - namely, demand for certain features or other performance characteristics that might be considered of less importance than those primary ones from the perspective of a typical informed buyer.

Like, again, to use the automotive example, if the auto industry had been singularly focused on longevity ONLY, then I suspect that by the 1980's or even earlier an internal combustion engine vehicle could have been built to last upwards of a million miles with only minimal maintenance but we never got that because a) it would have cost a lot more (yes of course - stainless steel, nylon-based plastics rather than ABS, tighter tolerances in engine assembly, etc.,) but also b) because of other competing priorities like features, styling, performance/power, emission regulations, etc.

So, it's often not necessarily the limitations of science and engineering that are the problem. Rather, it's the competing and conflicting demands and priorities placed on design and construction that tend to be the issue.

The bottom line seems to be that if you want the most efficient PSU, you have to buy the manufacturer's top of the range model and get all the other things that go along with that (e.g. - complexity) whether you like it or not. That's annoying.

ashiekh
09-23-2020, 02:02 PM
Gallium Nitride or running transistors hot might be a way to efficiency without added complexity, although if I recall correctly (doubtful) there is complexity for the designer as one turns Gallium Nitride off as opposed to on as with normal transistors (depletion mode transistors are normally on and need to be turned off). LLC might be considered an added complexity for efficiency; the experts are on this forum and can correct me.

I have an old HP desktop that is 12V only and efficiency was gained by simplicity in that instance (not having to generate other voltages).

So as you correctly point out, the issue is not as simple as efficiency implies complexity.


Concerning stainless steel, I am not sure such steels have very good fatigue characteristics. My own car is over 22 years and over 220,000mi and still running well. Then there is the issue of catalytic converters which actually reduce engine efficiency.

99wjtx
09-23-2020, 02:48 PM
Well, speaking of stainless steel, there are obviously a variety of potential applications for automotive purposes, some of which would be better than others. I was referring to it mostly in the context of a solution for the frame and body panels as it pertains to the "rust belt". There should be a number of acceptable stainless formulations for that purpose and as far as I know they were never widely used because, from what I understand, in automotive mass-production stainless steel body panels would have cost quite a bit more to manufacture. It's "harder to work with" as they say.

I happen not to live where rust is a constant concern. My vehicle strategy is to throw quantity at the problem for the sake of redundancy. I have three vehicles, all of which are about 20 years old and just below or above 200,000 miles. I find that if I'm willing to make occasional repairs myself, this works out better economically AND actually better ensures I have transportation than if I only owned one late model car (which might break anyway despite being newer, and which costs quite a bit more than the other three combined including the additional repair costs associated with their ownership).

Likewise, I'm in the process of building a new primary PC, finally, after having long relied on a very reliable first generation Core I7 build circa 2010. None of my primary components from that build have ever given me any trouble - motherboard (a mid-range Gigabyte UD3 model), the RAM (relatively slow Kingston Hyper-X DDR3 modules), or the several hard drives I've added to it over the years (an Intel SATA SSD circa 2011, Hitachi 2TB spinners, etc.) or, seemingly, the Antec Earthwatts 600 PSU from that era which still seems to be working OK, and that despite the sometimes questionable power delivery I've gotten over the years from my rural electric power company.

I acquired an ax1600i PSU a while back because I like the idea of owning arguably the best PSU out there and supporting the cause of its creation in any small way I can ... anddddd ... I got a good deal on it. It will go into my new build soon. That Antec is still doing OK, although I think it may be having some minor regulation issues that are still within the acceptable limits of the ATX specification. Personally, I like the older school thinking in the earlier days of PCs where stability and longevity tended to eclipse bleeding edge performance at the sake of those attributes. So, I confidently choose the best PSU I can get (excess total wattage be damned), powering a board with superior VRM but maybe only with a mid-range processor and GPU. Gallium nitride components and all the reviewers completely geeking out over it several years ago is all I needed to be bewitched by the Corsair flagship PSU. It's likely different for the next guy.

ashiekh
09-23-2020, 02:53 PM
Well, speaking of stainless steel, there are obviously a variety of potential applications for automotive purposes, some of which would be better than others. I was referring to it mostly in the context of a solution for the frame and body panels as it pertains to the "rust belt". There should be a number of acceptable stainless formulations for that purpose and as far as I know they were never widely used because, from what I understand, in automotive mass-production stainless steel body panels would have cost quite a bit more to manufacture. It's "harder to work with" as they say.

I happen not to live where rust is a constant concern. My vehicle strategy is to throw quantity at the problem for the sake of redundancy. I have three vehicles, all of which are about 20 years old and just below or above 200,000 miles. I find that if I'm willing to make occasional repairs myself, this works out better economically AND actually better ensures I have transportation than if I only owned one late model car (which might break anyway despite being newer, and which costs quite a bit more than the other three combined including the additional repair costs associated with their ownership).

Likewise, I'm in the process of building a new primary PC, finally, after having long relied on a very reliable first generation Core I7 build circa 2010. None of my primary components from that build have ever given me any trouble - motherboard (a mid-range Gigabyte UD3 model), the RAM (relatively slow Kingston Hyper-X DDR3 modules), or the several hard drives I've added to it over the years (an Intel SATA SSD circa 2011, Hitachi 2TB spinners, etc.) or, seemingly, the Antec Earthwatts 600 PSU from that era which still seems to be working OK, and that despite the sometimes questionable power delivery I've gotten over the years from my rural electric power company.

I acquired an ax1600i PSU a while back because I like the idea of owning arguably the best PSU out there and supporting the cause of its creation in any small way I can ... anddddd ... I got a good deal on it. It will go into my new build soon. That Antec is still doing OK, although I think it may be having some minor regulation issues that are still within the acceptable limits of the ATX specification. Personally, I like the older school thinking in the earlier days of PCs where stability and longevity tended to eclipse bleeding edge performance at the sake of those attributes. So, the best PSU I can get, powering a board with superior VRM but maybe only with a mid-range processor and GPU. It's likely different for the next guy.


I had an old car where cracks were forming in the body (the body of modern cars is the structural frame)

Galling is terrible with stainless steel; use a stainless nut and bolt dry and they will seize.

What's an i7? ;-) I'm still using the Core 2 duo/quad

99wjtx
09-23-2020, 03:04 PM
Well, like most things I speak of metallurgy only as a mild consumer/enthusiast, but apparently stainless 304 (for example) is good enough from a fatigue standpoint that it would work (but there are many stainless formulations out there). Tesla chose it for that forthcoming truck and SpaceX for Starship, with pretty substantial launch and flight stresses. Musk is on record as saying they expect Starship endurance toward hundreds of uses (https://www.cnbc.com/2020/09/01/elon-musk-spacex-starship-to-fly-hundreds-of-missions-before-people.html).

If it's good enough for that application, it should be OK for an automotive body panel. :)

ashiekh
09-23-2020, 05:35 PM
The DMC DeLorean had a stainless steel body

Now way off topic for a power supply forum.

99wjtx
09-23-2020, 07:41 PM
Yes, but for what it's worth, to your point earlier about potential fatigue issues, it wasn't a unibody design apparently. However, (getting wayyyy off topic here) whoever said the automotive industry *needed* unibody construction in the first place. :) It was supposed to reduce complexity and I guess it did, unless you had to repair it!

Getting back on topic, maybe what we need are PSUs with stainless steel unibody frames. Or should that be aluminum ... One giant heatsink chassis ... No wait, copper! lol OK, I'm done.

ashiekh
09-24-2020, 10:18 AM
There is also the other side of the coin; power consumption. The future may be one of low consumption CPU and GPU chips.

99wjtx
09-24-2020, 10:54 AM
Yes, for sure. The computer industry as a whole seems relentlessly mindful of power consumption but the processor manufacturers particularly so. They seem to throttle designs or radically rework chip topologies to avoid simply increasing performance by increasing power. Portable devices have driven a lot of that, obviously, but I think we *could* have gotten a gradual increase in power consumption over the years instead of or in conjunction with die shrinkage but that really hasn't happened except where truly necessary (a few products here and there notwithstanding).
Anyway, I only later realized which forum this is. You probably started this thread to discuss PSU topology theory directly related to the relationship between complexity and efficiency so my apologies for any drift my posts encouraged. As a non-engineer, I have virtually nothing directly relevant to contribute to that conversation!

ashiekh
09-24-2020, 08:58 PM
I feel chip consumption relates directly to PC power supplies.