PSU Discussion - Enough vs Future Proof

[Tristerin]

I see this quite commonly on this forum where people state that a minimum wattage is good enough.  Which is correct, Im not here to argue that, but to discuss why we say its enough, versus trying to ensure there is Headroom on the Watts. 

 

Quite commonly as well I see us suggesting processors, and GPUs with this same thought process - you want headroom for the peaks.  You don't want to be pegged at 90-100% usage.  But that same logic doesn't seem to apply to PSU (and of course it doesn't - its performance isn't the same as a CPU/GPU) - however we also in the same breath are always saying - get the next best thing.

So why not suggest a PSU that isn't going to only be useful today?  The trends that I decided to take a look at (just today only) and from my gut (20 years PC gaming) (not to mention all the Laws and Principles you can read on this stuff!) says that to "Future proof" or not be "Wasteful" one should purchase a PSU that will be usable today, tomorrow, and 3-5 years from now. 

 

List of NVIDIA GPU's - TDP tracking started at GeForce 8 series - you can see a trend:

https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units

 

Now when I see this I clearly see that they threw more power at the GPUs but over the last 3 series we have even keeled out at 250w - but isn't the problem with the RTX series that the 250w is being used already...without even firing up the Ray Tracing and Tensor cores (which will need power to utilize)?

 

So the trend suggest to me that whatever replaces the next RTX card will be a 350-400w card.  That would be my guess based on the trends at a glance.

 

List of AMD GPU's - TDP tracking started at x1300 series - you can see a trend:

https://en.wikipedia.org/wiki/List_of_AMD_graphics_processing_units

 

The last series of cards for AMD have finally evened out the wattage usage (RX 4 and 5) with sensible power solutions, but you can VR (just using my testimony as my reasoning here) on the last 3 series of AMD GPU's (R9 RX4 RX5 higher end models .... all with the higher watts in the family)

 

So just taking an hour to look at these things, I have to wonder why there is so much discussion against Overhead within a PSU purchase.

 

DISCUSS!  Tell me why Im stupid!  Tell me what angle I am missing!

[Crunchy Dragon]

I just about always suggest a PSU that will hold up with future upgrades.

 

This is the reason I recommend a 500-550W unit in even lower end builds.

[Stefan Payne]

22 minutes ago, Tristerin said:

I see this quite commonly on this forum where people state that a minimum wattage is good enough.  Which is correct, Im not here to argue that, but to discuss why we say its enough, versus trying to ensure there is Headroom on the Watts. 

 

How do you know what the future will bring?

So you waste 50.-100€ on a PSU for "future proofing" and then a year or two after you bought it there is a change that makes the PSU basically useless, what then?

And that is something you miss, that there are changes that make older PSU a danger with modern components because the way stuff works.

 

Just look at the power saving mechanisms!
The older GPUs before ~2012 had very simple power saving mechanisms, basically a 2D Mode and a 3D Mode. And the clockrate stays the same within the mode. So in 3D, you have the 3D Clockrates...

 

Today we have "intelligent" power saving mechanisms, look for race to idle. With modern PSU you have dozens of changes per second. That's not something most older PSU are up to.

 

And look at 10 years before 10 years ago. 1998. What kind of PSU had we then? 
The "named brands" only did 350W, maybe 400W at most, if you're lucky. But that really started in the 2000s...

And look at the power distribution. 8A on +12V for a 300W PSU. Today we have up to 25A. Because the power switched from 5V towards 12V - wich also made the PSU higher wattage while using cheaper components because the Amperage is what makes it expensive. And back in the day we had up to 45A on +5V - wich is around 225W - something that's possible with only 19A on +12V. so with the same components you have more than double the wattage...

 

And there are also other things to consider. Nobody really needs or wants the 3,3V Rail. 5V is only used because the ATX Connector only has 2 pins with 12V!

 

And ATX is around since 1996!

And was made with the thought that you could use the 3,3V Rail from the PSU for the CPU VCore (later I/O) and Memory voltage...

 

 

So the buttom line is:
No, moar Watt is NOT more future proof, its just a waste.

What is future proof? nobody knows!

Only time will tell and we do not know what the future will bring because it is in the future. 
But there will be dramatical changes that will make the PSU you have obsolete and not usable with upcoming hardware. A new Power Spec that gets rid of most voltages and only leaves the +12V and maybe +5V and the +5V Standby rail is highly likely...

 

And there will be changes, there always were...

[Tristerin]

25 minutes ago, Stefan Payne said:

How do you know what the future will bring?

So you waste 50.-100€ on a PSU for "future proofing" and then a year or two after you bought it there is a change that makes the PSU basically useless, what then?

And that is something you miss, that there are changes that make older PSU a danger with modern components because the way stuff works.

 

Just look at the power saving mechanisms!
The older GPUs before ~2012 had very simple power saving mechanisms, basically a 2D Mode and a 3D Mode. And the clockrate stays the same within the mode. So in 3D, you have the 3D Clockrates...

 

Today we have "intelligent" power saving mechanisms, look for race to idle. With modern PSU you have dozens of changes per second. That's not something most older PSU are up to.

 

And look at 10 years before 10 years ago. 1998. What kind of PSU had we then? 
The "named brands" only did 350W, maybe 400W at most, if you're lucky. But that really started in the 2000s...

And look at the power distribution. 8A on +12V for a 300W PSU. Today we have up to 25A. Because the power switched from 5V towards 12V - wich also made the PSU higher wattage while using cheaper components because the Amperage is what makes it expensive. And back in the day we had up to 45A on +5V - wich is around 225W - something that's possible with only 19A on +12V. so with the same components you have more than double the wattage...

 

And there are also other things to consider. Nobody really needs or wants the 3,3V Rail. 5V is only used because the ATX Connector only has 2 pins with 12V!

 

And ATX is around since 1996!

And was made with the thought that you could use the 3,3V Rail from the PSU for the CPU VCore (later I/O) and Memory voltage...

 

 

So the buttom line is:
No, moar Watt is NOT more future proof, its just a waste.

What is future proof? nobody knows!

Only time will tell and we do not know what the future will bring because it is in the future. 
But there will be dramatical changes that will make the PSU you have obsolete and not usable with upcoming hardware. A new Power Spec that gets rid of most voltages and only leaves the +12V and maybe +5V and the +5V Standby rail is highly likely...

 

And there will be changes, there always were...

 

I don't know what the future will bring - I can only look at trends - did you not look at the trends over the past since they have tracked TDP?  I agree with CPU's - I cant see a trend so far until you hit the massive core counts/thread counts (Threadripper) but that seems to be the direction for more cores and threads...so isn't that the TREND?

 

Why are we ignoring trends?  Isnt that how we should gauge the future?  Its obviously not an equation that has a factual answer - but it does have trends that can genuinely point us in the right direction until there is a Technological Breakthrough - that you cant TREND nor PREDICT.

 

Example - NASDAQ - no one ignores the trends.

 

EDIT - we cannot ignore that the RTX 2080 is at its max power draw already without engaging Tensor or Ray Tracing Cores...so then what happens (other than downclocks on the 20XX series) until they add the 21XX series with higher TPD - at least that's going to be my educated guess from the information available to me on the interwebs.

[SkyHound0202]

If your PSU must supply enough wattage, your system might not work in the first place. Leave a few headroom if you desire for future expansion. But:

Higher Wattage ≠ Future Proof.

You should also take into account other things like efficiency, personal preference and expansion options matters:

-Are you going to run it 24/7 or just a few hours a day? If the answer is 24/7 then you should consider a 80 Plus Platinum/Titanium over Gold.

-Do you like silence? Then you probably should go fanless or a PSU with a good fan (preferable dual ball bearing ones) or '0 RPM mode'. Or perhaps you want RGB or smart monitoring via software?

-Are you using multiple GPUs, then is your desired PSU going to provide enough 12V rails? Or you need a bunch of drives that requires more on 5V.

 

You mention that the Power Consumption might be consistent for a while now, but did you realized that the Power Efficiency (GFlops/watts) has increased dramatically. To put that into perspective, GTX 1060 3GB is on par with GTX 970 on the computational level, yet the 1060 consumes 20% less power.

With improvement on manufacture nodes and new architectures, I can say that power requirement will more or less stay the same, but the performance you get will be much more eventually.

[Stefan Payne]

26 minutes ago, Tristerin said:

I don't know what the future will bring

Exactly and thus every talk about "future proof" is fortune telling at best and bullshit at worst.

So why bother at all??
Because a Seasonic X-560 probably won't work well with modern systems either...

 

26 minutes ago, Tristerin said:

- I can only look at trends - did you not look at the trends over the past since they have tracked TDP? 

That is useless when the specification is reaching its limits and there are problems with ATX popping up left and right and better boards have additional connectors (PCIe usually) to circumvent that. 

 

So what is more probable:
a) that a 550W will not be enough for a decent mid range PC in the future

b) ATX will be scrapped and we will see something new/fresh with less pins and voltages (wich will make PSU a bit cheaper as well!)

 

I bet on b) because I see the limits and Problems of ATX, the burned ATX connectors that happen frequently...

 

What you also forget is that higher power consumption dramatically increases the cost of a system...

So if you do not think about the Chip and Memory of a graphics card, the difference between a 150W and a 250W Graphics card is enormous! Just look at the Heatsinks, that have to be more complicated, use either more heatpipes or a Vapor Chamber. More and higher quality voltage regulators. More capacitors. All that increases the cost.

Just look at a Radeon HD7850 or 7870 and compare that with a 7970(GHz).

 

26 minutes ago, Tristerin said:

Why are we ignoring trends?  Isnt that how we should gauge the future? 

Why are you ignoring the trends??
Why are you ignoring the facts about ATX?

The trends are that the big system integrators already scrapped ATX for some of their business machines and use single voltage Power Supplys. Some use a 5V Standby rail, some use a ~12V Standby Rail. And it is probable that this will switch to more and more systems. And move towards the higher end as well.

Maybe they will add a 5V for drives in higher end systems but there is no need for a 5V Rail to the Motherboard. It is only needed for Harddiscs.

There is no need for the 3,3V Rail anywhere. 
There is no need for the -12V (wich is why it will be removed next).

 

There is only a need for +12V and a Standby rail. The signs are already on the horizon.

And there are many people posting stuff about burned ATX Connectors because the +12V Connectors are overloaded and burned.

 

The facts are that ATX is a piece of shit for modern systems. It was made 22 years ago for what they thought made sense then! But it doesn't today.

 

And there were people talking about the Intel Design Guide version 1.4 wich require different supervisior ICs, maybe other changes as well...

 

And that will be here in 2 years. So your "moar watts = Future Proof" is already bullshit because it will not be. No PSU you can buy today will (probably)...

 

26 minutes ago, Tristerin said:

Example - NASDAQ - no one ignores the trends.

We ain't talking about imaginary, virtual things without any knowledge about that stuff.

We are talking about technical stuff. That are two different things.

 

And what you see is a fragmentation of the Market. Not all systems use ATX compatible PSU. Some Systems use +12V Only PSU. Some even totally break compatibility with 12V Standby Rail. 

 

So why are you ignoring those systems??

 

26 minutes ago, Tristerin said:

EDIT - we cannot ignore that the RTX 2080

Highest end, totally overpriced card that is not interesting for the masses.

The cheapest 2080 is listed for 750€...

So should we argue with max. OC Power Consumption of a 2990WX??

Or should we argue with more mainstream components that are 300€ or less??

26 minutes ago, Tristerin said:

is at its max power draw already without engaging Tensor or Ray Tracing Cores...so then what happens (other than downclocks on the 20XX series) until they add the 21XX series with higher TPD - at least that's going to be my educated guess from the information available to me on the interwebs.

Your problem is that you focus only on power consumtion while ignoring other stuff that is happening right now. Like the burned ATX Connectors.

Like the Business Machines from Dell + HP that use a 12V Only PSU. Like the mentioned Intel Design Guide 1.4...

 

 

And I don't find it probable that the ATX connector will increase another 2-4 pins to add the really needed 12V Pins. I see that its more probable that it will be getten rid of and we will see something like we see on DELL and HP Boards with an 8-10pin connector with only 12V and the "Communication" stuff (power Good, PSU On)...

 

[Tristerin]

3 minutes ago, Stefan Payne said:

snip

 

I cant disagree that future technology should bring us more efficiency - your guesses in that snip as are...trendy as mine - however I agree with you that the way of the ATX is archaic at best.  Ive oft thought that the next technological breakthrough will not be PCs as we know them today.  Thus the discussion but you cant just dismiss the actual trends.  So given that what you are talking about is trends dying, new trends starting (since they are new they cant be graphed and used to make logical guesses like having a trend that has been happening for 20 years though.  And to ignore that is...well...not a discussion).

 

Im pretty sure my original post was just skimmed through.

 

I mention the NASDAQ, which is a HUGE deal to the world in its success and failures, predicted solely by trends - yet this is imaginary and cant be discussed?  Maybe to someone who hasn't played, and doesn't understand why the trends are important to gage your next big move of monies. 

 

But to me, the trends seem fairly...predictable.  Did you actually look at the links and actually look at the TDP of GPU's over the last 10 years or no?  I haven't graphed them yet myself but they seem fairly...noticeable. 

 

 

[Stefan Payne]

18 minutes ago, SkyHound0202 said:

-Do you like silence? Then you probably should go fanless or a PSU with a good fan (preferable dual ball bearing ones) or '0 RPM mode'. Or perhaps you want RGB or smart monitoring via software?

I disagree with most of this.

The 2BB Fans are rather loud to annoying. So if you want it quiet, you need a high quality advanced sleeve bearing like an FDB or the friction less magnetic bearings.

 

And in the end, it is more important what the PSU was designed for and if "quiet operation" was in the spec sheed of the Contractee and the PSU is modified to be quiet(er) and has a modified fan controller...

 

And the 0 RPM mode can sometimes cause more harm than good. A CWT made SFX unit is proof of that (ie Enermax Revolution SFX and other similar things).

[Tristerin]

Due to what I see as obvious trends...the next RX and RTX series will be 100w or more or power requirement.  Just from the trends I would easily place a $50 bet on that.

[Princess Luna]

If I'm buying a super high end video card that will use more power than the PSU I already own, then I'll buy a new PSU? I mean seriously, you gonna pay over 1000 dollars on a GPU, you can't pay another 100 for a good PSU to go along it?

 

If you even need to, the TX650M I got for my original i7 8700 + 1080 Ti is still sufficient for the much more power hungry setup with an i9 and 2080 Ti.

 

I use around 550w~560w on full load with what is pretty much the most power consuming mainstream system you can put together right now.

[Tristerin]

15 minutes ago, Princess Cadence said:

If I'm buying a super high end video card that will use more power than the PSU I already own, then I'll buy a new PSU? I mean seriously, you gonna pay over 1000 dollars on a GPU, you can't pay another 100 for a good PSU to go along it?

 

If you even need to, the TX650M I got for my original i7 8700 + 1080 Ti is still sufficient for the much more power hungry setup with an i9 and 2080 Ti.

 

I use around 550w~560w on full load with what is pretty much the most power consuming mainstream system you can put together right now.

I understand your Use case, and similarly the same reason I have the HX850i (warranty, tier 1, future proof using trends as my proof) - to ensure I wont find myself next year gasping for Watts without having to pay again for a unit that delivers power should something fail, a new must have comes out etc.

 

Im trying to change my angle of thinking to understand why the average user would ever purchase a 550w PSU.  This discussion was prompted by being told it was "wasteful" to go over that (basically) whereas I see the waste in purchasing two different PSUs in a couple years timeframe (environmentally, and financially) 

 

Maybe Im thinking to much about the GPU itself, and less about CPU power (which at a glance trends to be more efficient year to year until you get to HIGH core count) improvements etc.  But I cant see that getting MUCH better than it is because this forum is for the Average and Power users typically.  So we game, we edit, we produce...which will, if we want to do it efficiently (time = money and/or time = spent gaming we want it to be enjoyable) require us to upgrade...every 2-3 generations at least.

 

And as the trends show...next generation of GPU's (if there is any credence to the trend, and if you don't believe there is that's fine but Im a data miner in my RL job, and statisticians, etc exist for a reason) will be power hungry.  Im still going to bet 100w over current gen - the next RX refresh will be the first to prove this theory/trend or disprove it.  But how do you think AMD will get more out of the Polaris beyond smaller size...well...more watts (my opinion)

[Velcade]

I like to play in this region...

[Tristerin]

Just now, Velcade said:

I like to play in this region...

Same - my PSU fan doesn't even turn on.  Or at least not often or long enough that over 1 year time the dust filter is spotless (checked 2 weeks ago)

[tonymc]

7 hours ago, Stefan Payne said:

So what is more probable:
a) that a 550W will not be enough for a decent mid range PC in the future

b) ATX will be scrapped and we will see something new/fresh with less pins and voltages (wich will make PSU a bit cheaper as well!)

 

I bet on b) because I see the limits and Problems of ATX, the burned ATX connectors that happen frequently...

There is no need for the 3,3V Rail anywhere. 
There is no need for the -12V (wich is why it will be removed next).

 

There is only a need for +12V and a Standby rail. The signs are already on the horizon.

 

 

We are talking about technical stuff. That are two different things.

 

And what you see is a fragmentation of the Market. Not all systems use ATX compatible PSU. Some Systems use +12V Only PSU. Some even totally break compatibility with 12V Standby Rail. 

 

So why are you ignoring those systems??

 

Highest end, totally overpriced card that is not interesting for the masses.

 

And I don't find it probable that the ATX connector will increase another 2-4 pins to add the really needed 12V Pins. I see that its more probable that it will be getten rid of and we will see something like we see on DELL and HP Boards with an 8-10pin connector with only 12V and the "Communication" stuff (power Good, PSU On)...

 

It seems that his views are very accurate because they have a lot of logic, but my question is the following:
Where did you get that information from? What reference do you have to affirm this?

It is also true that the manufacturers of complete systems such as Dell and HP have always done things their own way, totally incompatible with the standards in terms of design of their systems.

[tonymc]

8 hours ago, Tristerin said:

I see this quite commonly on this forum where people state that a minimum wattage is good enough.  Which is correct, Im not here to argue that, but to discuss why we say its enough, versus trying to ensure there is Headroom on the Watts. 

 

Quite commonly as well I see us suggesting processors, and GPUs with this same thought process - you want headroom for the peaks.  You don't want to be pegged at 90-100% usage.  But that same logic doesn't seem to apply to PSU (and of course it doesn't - its performance isn't the same as a CPU/GPU) - however we also in the same breath are always saying - get the next best thing.

So why not suggest a PSU that isn't going to only be useful today?  The trends that I decided to take a look at (just today only) and from my gut (20 years PC gaming) (not to mention all the Laws and Principles you can read on this stuff!) says that to "Future proof" or not be "Wasteful" one should purchase a PSU that will be usable today, tomorrow, and 3-5 years from now. 

 

 

 

 

Friend, my English is not very good, but according to what I understand of all that I must say the following.

a-You should not worry that your source today, is not what you need for the future, because in the end you should not have a PSU for more than 5 or 6 years no matter how good it is. The problem is that there are components that are degraded over time (for example the capacitors), therefore the operation or better said parameters that begin to change. Also the PSU is a critical component because it is a common factor for the whole system. My recommendation, do not wait for it to fail completely, change it working.

b-The increase in consumption I do not think is so aggressive, considering that the manufacturing processes will be smaller .. We are going for 10 or 7 nm.

c-In any case, the most aggressive systems for energy consumption are not those of the majority, but of the minority, those that generate TOP TOP configurations.

regards

[tonymc]

8 hours ago, SkyHound0202 said:

You mention that the Power Consumption might be consistent for a while now, but did you realized that the Power Efficiency (GFlops/watts) has increased dramatically. To put that into perspective, GTX 1060 3GB is on par with GTX 970 on the computational level, yet the 1060 consumes 20% less power.

With improvement on manufacture nodes and new architectures, I can say that power requirement will more or less stay the same, but the performance you get will be much more eventually.

Yes,I believe that this will continue to be so in the future.

[17030644]

7 hours ago, Tristerin said:

I have to wonder why there is so much discussion against Overhead within a PSU purchase.

 

The problem is NOT having some overhead.

 

Having some room to upgrade in the future is fine because the usual trend is to get better parts which in some cases requires more wattage. And overclocking stuff rather than running it at stock is definitely going to increase power requirements.

 

The problem is when people get to the point where they get a 750W+ PSU for an APU build because "it's more future proof"

 

So "future-proofing" is often used in such a stupid way because people lack reasoning for it to be future proof they just say so probably because they heard it somewhere.

 

There is usually little to no price difference between from 450 to 650W units at least in the high-end but once they go 750W or above they're either spending significantly more money or giving up quality which is even worse. For what? 750W+ is a niche thing, you need to SLI/Xfire or run beast CPUs to take advantage of it.

 

It's easier to overestimate than underestimate.

 

Again, it's hard to go too low, there are not many good PSUs under 450W and even if there were some, many don't have enough connectors for what you need so you usually overestimate from the get-go. Why go higher when you already have plenty of room?

 

If they want more future proofing then they should a better quality PSU or buy better CPU/GPU as those age way faster.

 

Just think about it, if you need 750W or more you wouldn't want to risk such a rig with an entry-level PSU, even with a mid-range one. Efficiency would probably be important as well. As for the other one, if you can't afford that kind of hardware right now, what makes you think you soon will?

 

Even if you did, in fact, go with a massive upgrade to an HEDT build, don't you think you could also afford a PSU replacement in that case? For now focus on HAVING A GOOD PC now, not on having lots of W that do nothing for you.

 

 

[17030644]

that's it

[17030644]

6 hours ago, SkyHound0202 said:

-Do you like silence? Then you probably should go fanless or a PSU with a good fan (preferable dual ball bearing ones) or '0 RPM mode'.

 

dual ball bearing is not meant for silence AFAIK

 

and 0 RPM mode is more of a gimmick than anything else.

 

Some PSUs claim fanless mode and it only works for the first 50W of load and then the fan goes wild and straight up to 50dBA

 

So you should look at the fan behavior across different loads and temps

[17030644]

5 hours ago, Tristerin said:

whereas I see the waste in purchasing two different PSUs in a couple years timeframe (environmentally, and financially) 

 

I see the waste in going overkill

 

watts you don't need, having to cut down on your other hardware, being less efficient in some cases.

 

5 hours ago, Tristerin said:

Same - my PSU fan doesn't even turn on.  Or at least not often or long enough that over 1 year time the dust filter is spotless (checked 2 weeks ago)

That's just your particular situation, not everyone's. That doesn't apply to all 850W psus.

5 hours ago, Velcade said:

I like to play in this region...

This is another good example, you're taking a particular PSU, on the other hand, the 850G3 runs its fan at maximum speed even under light loads. It's definitely not quieter than a 450W bitfenix whisper M despite being higher wattage, and it's definitely not quieter than a fanless PSU.

 

 

So those kinds of assumptions just don't make any sense and you don't need to be genius in order to understand it.

HIGHER WATTAGE ISN'T QUIETER, NOR MORE EFFICIENT, NOR SAFER, NOR MORE FUTURE-PROOFING NOR BETTER AND IT WON'T LAST LONGER. MORE WATTAGE IS JUST MORE WATTAGE, PRETTY STRAIGHTFORWARD isn't it

If you want your PSU to last long, then get a quality PSU

If you want to be safe, then buy a safe PSU

If you want a more efficient PSU, buy a more efficient PSU

If you want your PSU not to make lots of noise, just buy a PSU with a proper fan control or go fanless.

 

It's that simple

 

[Stefan Payne]

6 hours ago, Tristerin said:

I cant disagree that future technology should bring us more efficiency

I'm not talking about efficiency par se.

But getting rid of 3,3V and reduce the 5V rail to 10A or less would greatly increase the Efficiency.

Last time I've seen a measurement about that was about 2-3% or so...

 

6 hours ago, Tristerin said:

I mention the NASDAQ, which is a HUGE deal to the world in its success and failures, predicted solely by trends - yet this is imaginary and cant be discussed?  Maybe to someone who hasn't played, and doesn't understand why the trends are important to gage your next big move of monies. 

No, its just two completely different things.

One is technical and based in reality.

The other is more or less just speculation.

 

If someone in some position has gas, the NASDAQ falls. 

6 hours ago, Tristerin said:

But to me, the trends seem fairly...predictable.

To me it is not.

And its always the same, TDP increases with the last generation of a process node and decreases significantly.

And with the increase in cost, I wouldn't expect more than 175W for under 300€.

A couple of years ago you could have gotten the 300W Monsters for that...

 

6 hours ago, Tristerin said:

Did you actually look at the links and actually look at the TDP of GPU's over the last 10 years or no? 

I don't need to, I know them, have most of that in my head.

And it also depends on when you start.

When you start in the beginning, you'll see a dramatical increase in power consumption - amd then it abruptly stops with the first DX10 generation and doesn't move anywhere. May be +/- 50W here and there at the top end but that's about it...

But that's with ignoring the price...

 

If we look at the sub 300€ level things look different. While we had around 200-250€ in the 28nm area, it drops to way under 200W with 14nm and the 250W TDP Cards got rediculously expensive...

 

6 hours ago, Tristerin said:

I haven't graphed them yet myself but they seem fairly...noticeable. 

only the last generation.

But that's to be expected.

Will drop with 7nm again.

[Stefan Payne]

2 hours ago, tonymc said:

It seems that his views are very accurate because they have a lot of logic, but my question is the following:
Where did you get that information from? What reference do you have to affirm this?

It is also true that the manufacturers of complete systems such as Dell and HP have always done things their own way, totally incompatible with the standards in terms of design of their systems.

Experience.
And the fact that the AT Standard only lasted roughly 10 years while ATX is already 22 years old. Though extended twice with CPU Connectors and PCIe Connectors. That wasn't in the original spec.

 

The thing that Dell and HP are doing don't look that proprietary from afar, though I don't have many systems from both manufacturers to research it further.

But the fact is that all big PC OEMs are using +12V Only PSU for their Business Class systems.

It might even be that there is some kind of unofficial standard between those guys...

[Tristerin]

14 hours ago, Stefan Payne said:

Experience.
And the fact that the AT Standard only lasted roughly 10 years while ATX is already 22 years old. Though extended twice with CPU Connectors and PCIe Connectors. That wasn't in the original spec.

 

The thing that Dell and HP are doing don't look that proprietary from afar, though I don't have many systems from both manufacturers to research it further.

But the fact is that all big PC OEMs are using +12V Only PSU for their Business Class systems.

It might even be that there is some kind of unofficial standard between those guys...

Great points - so the trend would say that ATX should have been gone already, given the delta between the two.  From my understanding AT was around purely for brute force power to the mainboard?

 

Since it is not, and each generation of mainboard being produced for the consumer/prosumer is still for ATX with no end in site of that being the case (PC is meant to be universally compatible for hardware/software to work as much as possible together...another discussion just lightly touching on it so not looking to be tarred and pitchforked for not going in deeper on that)

 

So to me it seems you are more inclined (given the information you have and your testimony) to think that to Future-Proof a PSU in any degree is not worthwhile because the ATX standard should be/is going away.

 

Is that accurate?

 

17 hours ago, tonymc said:

Friend, my English is not very good, but according to what I understand of all that I must say the following.

a-You should not worry that your source today, is not what you need for the future, because in the end you should not have a PSU for more than 5 or 6 years no matter how good it is. The problem is that there are components that are degraded over time (for example the capacitors), therefore the operation or better said parameters that begin to change. Also the PSU is a critical component because it is a common factor for the whole system. My recommendation, do not wait for it to fail completely, change it working.

b-The increase in consumption I do not think is so aggressive, considering that the manufacturing processes will be smaller .. We are going for 10 or 7 nm.

c-In any case, the most aggressive systems for energy consumption are not those of the majority, but of the minority, those that generate TOP TOP configurations.

regards

This makes a TON of sense to me.  I love logic.  But hasn't the capacitors, components, and technology overall seen some great leaps and bounds (Warranties are going 10 years for good PSUs now) for efficiency/longevity - and the same can be said about our Electrical Grid in most first and second world Countries no?  So the PSU (correct me if I am wrong) isn't getting "attacked" with Power Surges like in the 90's.

 

The other thing I have to question is - how many PSU's have you seen go out?  Testimonial wise - I have seen 2 PSU's go poof.  Both times...nothing happened to any portion of my system (and both times the PSUs were old generic, I didn't know what I was doing, PSU's) sans the PSU - so even the technology back then (early 2000s) didn't give me any sort of feeling that a PSU can take my system with it...it busted a fuze on one and just shot the craps on another.

 

So when I hear - Components go bad and if the PSU goes it *may* take the system with it.  Is there any data that can give us more information about PSU's and taking out components with them?  If we use the *it may take components with it* analogy at all times with no data (for me I only have my experience first hand to talk about PSU's going out) than to me that's the Techie that typically just regurgitates what he has been told.  (admittedly that's most of my tech experience, but then I apply it in real situations to verify it)

 

To try to understand how critical of a component it is compared to any other component I would have to say it is the same criticality of a Motherboard, CPU, and RAM.  Without which none of this works (hierarchy of needs for a PC I guess?) - but I never see anyone say to replace a motherboard (CPU/RAM) because its old (using the logic of why to replace a PSU - capacitors etc) - but for upgrades.

 

Thanks for all the information sharing folks, Im here to be learnt a thing-or-two!

[tonymc]

4 hours ago, Tristerin said:

The other thing I have to question is - how many PSU's have you seen go out?  Testimonial wise - I have seen 2 PSU's go poof.  Both times...nothing happened to any portion of my system (and both times the PSUs were old generic, I didn't know what I was doing, PSU's) sans the PSU - so even the technology back then (early 2000s) didn't give me any sort of feeling that a PSU can take my system with it...it busted a fuze on one and just shot the craps on another.

 

So when I hear - Components go bad and if the PSU goes it *may* take the system with it.  Is there any data that can give us more information about PSU's and taking out components with them?  If we use the *it may take components with it* analogy at all times with no data (for me I only have my experience first hand to talk about PSU's going out) than to me that's the Techie that typically just regurgitates what he has been told.  (admittedly that's most of my tech experience, but then I apply it in real situations to verify it)

 

The decision is yours. The doctor says, "do not smoke that harms your health" and, you can say, I will continue because "nothing has happened to me until now" and in truth nothing could ever happen but in honor of the truth, what you have done is "decrease your luck percentage", to put it in some way.
The problem is not when the PSU is turned off due to problems, but after a long time of use, it is no longer the same in the stability parameters, although at first glance everything is fine.

However, you must also take into account that you do not have the PC on all the time, it is not the same 24/7 than 6 or 8 hours a day.
Personally I have PC from the Pentium2 Slot 1, and a PC component has never died in my hands, except a PSU that exploded in my 478. I do not know if it has to do this, but I have always been strict with the time of use of the PSU, maybe it's just luck, or maybe I've helped decrease my percentage of failures.

The degradation of components, I have not invented, is to devote some years to electronic repair and has allowed me to see some things with my own eyes.

 

[Tristerin]

3 minutes ago, tonymc said:

The decision is yours. The doctor says, "do not smoke that harms your health" and, you can say, I will continue because "nothing has happened to me until now" and in truth nothing could ever happen but in honor of the truth, what you have done is "decrease your luck percentage", to put it in some way.
The problem is not when the PSU is turned off due to problems, but after a long time of use, it is no longer the same in the stability parameters, although at first glance everything is fine.

However, you must also take into account that you do not have the PC on all the time, it is not the same 24/7 than 6 or 8 hours a day.
Personally I have PC from the Pentium2 Slot 1, and a PC component has never died in my hands, except a PSU that exploded in my 478. I do not know if it has to do this, but I have always been strict with the time of use of the PSU, maybe it's just luck, or maybe I've helped decrease my percentage of failures.

The degradation of components, I have not invented, is to devote some years to electronic repair and has allowed me to see some things with my own eyes.

 

So I will focus on just the bold words:

Obviously the decision is mine. 

 

Agreed that the components do degrade over time

 

What percentage of failure?

 

We've gone off topic but I believe there is a lot of great information, to fill the delta of WHY little headroom on a PSU is commonly suggested.

 

1.) Degrading components can cause a failure, why wait for that failure (and to see what that failure may cause)

2.) Purchasing a high wattage PSU you should be replacing (see #1) is a waste - because #1

3.) Future-proofing isn't a thing (we all actually know that as Prosumers - you can only mitigate the future but the future will be coming either way!)

4.) If you can afford the high wattage next generation items, you can also afford to replace the PSU at that time should your wattage be insufficient

 

Those are all very sound reasons - any I missed or didn't extrapolate on enough?