Power-hungry Ampere could lead to PSU shortages

[Bullwinkle M]

My dualcore Sandy Bridge desktop averages 45 Watts at the wall using the ATX power supply

Switching to a Pico Power supply lets me run the same computer at 23 Watts average

Add a DC/DC converter for the 32" Full HD monitor and I can run both all day from a single battery

Try that with your RTX 3080!

No, really......I wanna see it!

 

[Evernessince]

No, just you.

No, you stated it consumed 54.3w. That was incorrect. Plain and simple. Admit it and move on.

In any case, you're again staring at a tree and losing sight of the forest. There are much lower-power CPUs than the 1600/2600/3600s. My original point was that the blanket statement "a 3080 requires at least a 750w PSU" is simply incorrect. It all depends on the components chosen.

Oh really? Your original post said this:

" There will be plenty of 3080 based systems which will be absolutely fine with a 600w or even 550w supply. "

To which I replied

" 600w should barely be fine assuming no OC and a quality power supply. 550w is not regardless of PSU. "

and then you replied again, stating that 550w would be fine:

" You're still assuming that 125w Intel CPU. Assume a 65w Ryzen, and a 550w is fine-- again, depending on mobo, ram, & drives of course. "

So if you only took issue with 750w systems why did you spend this entire thread talking about how 550w was enough? No, you made a bad point and now you are trying to back out of it.

550w is not enough for a 3080 in a majority of situations. No argument from the fringes will over-rule that.

 

[Endymio]

550w is not enough for a 3080 in a majority of situations.

Of course I never said "a majority". But there will be plenty of systems in which it will be enough. Your links certainly don't demonstrate otherwise -- take that last one, which shows a max system power draw of 133w, in a system with a 65w TDP CPU and a 165w TDP GPU. Put a 3080 in that system in place of the GTX 980, and I guarantee you it will run just fine with a 550w PSU. QED.

 

[Evernessince]

Of course I never said "a majority". But there will be plenty of systems in which it will be enough. Your links certainly don't demonstrate otherwise -- take that last one, which shows a max system power draw of 133w, in a system with a 65w TDP CPU and a 165w TDP GPU. Put a 3080 in that system in place of the GTX 980, and I guarantee you it will run just fine with a 550w PSU. QED.

That's a CPU only test.

TechSpots own numbers closely resemble the total system power consumption figure I gave earlier: https://www.techspot.com/review/2099-geforce-rtx-3080/

523w at full load

Did you not read any reviews of the 3080?

 

[Endymio]

TechSpots own numbers closely resemble the total system power consumption figure I gave earlier: https://www.techspot.com/review/2099-geforce-rtx-3080/

523w at full load

523 watts is less than 550 watts. Furthermore, that Techspot test system is with a 3950x, a CPU that draws nearly double the power of a 3600. Again, your own sources hoist you by your petard.

 

[OptimumSlinky]

Are you expecting to roll with <1000W PSU for the rest of time? As technology gets faster, requirement for more power reflects that. No matter how efficient the chips are. Give it a few more decades and we will start seeing 1500W to 2500W PSUs to power your super hungry rigs.

Not necessarily.

Back in 2010, most PSUs were 750+ but components actually got more efficient over the generation. My Ryzen 3600X uses far less energy than my i7-930 did.

Ampere is an anomaly because nVidia said screw efficiency and hot-rodded Samsung's 8nm chips to avoid losing the performance crown.

 

[Evernessince]

523 watts is less than 550 watts.

Assuming Stock settings, you have a minimum benchmark like build with only 1 drive, and no additional power consuming peripherals (like a USB stand mic). Not accounting for capacitor aging either, which is 5%. That's 27.5w the first year, meaning you'd be over your maximum power output after your PSU hits 1 years of age).

Here's capacitor aging by year:

1 Year - 5%
2 Year - 2.5%
3 year - 1%
4 Year - 0.5%
5 Year - 0.25%

Who in the world would use a PSU with only 27w of headroom? Silicon lottery alone can account for a difference that large and put you over, let alone the other factors mentioned. Buy an aftermarket card? Your screwed.

Furthermore, that Techspot test system is with a 3950x, a CPU that draws nearly double the power of a 3600. Again, your own sources hoist you by your petard.

The 3950X consumes significantly less than mid range Intel offerings:

Intel Core i7-10700K Review: Taking the Gaming Shine Off Core i9

Core i9 gaming at Core i7 pricing

www.tomshardware.com

You are trying to paint the 3950X as a power guzzler but that's simply false.

 

[Endymio]

Who in the world would use a PSU with only 27w of headroom?

Your math is bad, for two separate reasons. First, I am sure the TechSpot review you cited is measuring power by PSU socket draw, not what it is supplying to the system. A PSU's rating is what it can deliver, not what it draws. Assuming the PSU is 90% efficient (optimistic, but let's go with it), @523w draw, the PSU is supplying 52w less than that.

Second, you are still glossing over the fact that a 3950x (TDP 105w) is far from the lowest-wattage CPU out there. Assuming again our hypothetical 3600 (TDP: 65w), the headroom in that case would be 27 + 52 + 40 + 4 = 123w. That's 123
watts less than the 550w rating. Is 123 watts enough headroom? Surely.

While I admire the temerity of your posting not one, or two, but three separate links that contradict you, the fact remains that, in some systems, a 550w PSU will indeed be just fine. Also, you may want to recheck your "calculations" on your capacitor aging chart -- the math doesn't work the way you think it does.

In closing, I'm going to ask Steve Walton (or any other passing Techspot editor) to confirm that their reviews measure system power by wall draw. Otherwise I'm sure you'll spend ten or twenty posts trying to debate that point also.

 

[Mister_K]

Already have a 1500W in my PC (Intel HEDT) .

Sure with SLI and high consumption CPUs you can already reach that, but for a single GPU/general gaming CPUs this will take a while. But yeh, essentially, performance/power = higher power consumption = bigger, beefier PSU requirements.

 

[Ludak021]

yes but every fan boys knows the more powerful your PSU the more powerful your PC is. So thanks to my 1500w platinum unit, my AMD RX 470 gives me AMD 5700XT performance. It's also why I recommend a 1000w unit (seasonic of course because I have never heard of another manufacturer) as the bare minimum just to boot up a PC.

Corsair RM series has almost flat-line efficiency. It's not 1999...Aside ffrom that, I do not really care if eff. is for example 89% or 87%, or even 85%. Those are very small differences for one PC that occasionally draws a lot of power.

 

[HafizNafes89]

So, for a <700W PC, one needs 1000W PSU. Riiight. I've pushed crossfire 280x on i7 with 700W PSU without issues.

it's always better to get but more than what your pc pull from the wall because peak wattage can differ based on situation and application. Doom heavily uses the GPU but sometimes the CPU isn't utilized as hard. Expect wattage to increase as both GPU and CPU are revving up.

 

[Ludak021]

I would always leave at least 20% overhead from theoretical maximum that a PC can pull. so , for 700W pull, 850W PSU is starting to be plenty already. 1000W is overkill.

 

[GTvon]

Geez. Not sure why I went with it, but I chose an 850W PSU for my i9 build last year....just making the cut for an i9 / 3080 combo.

I always go with an 850 watt power supply every rig I build for my family and myself.
I just like to be future-proof.

 

[GTvon]

I would always leave at least 20% overhead from theoretical maximum that a PC can pull. so , for 700W pull, 850W PSU is starting to be plenty already. 1000W is overkill.

If plenty is good Overkill is better.?

 

[Ludak021]

Well, sure, depends on how much you are paying for it. If it's a little bit more, sure, but in two years, something better will come along (like corsair RM series came along and put an end to power efficiency wildly fluctuating curve) and make your investment loose any realistic re-sale value.
Depends on how much money you have really.

 

[Evernessince]

Your math is bad, for two separate reasons. First, I am sure the TechSpot review you cited is measuring power by PSU socket draw, not what it is supplying to the system. A PSU's rating is what it can deliver, not what it draws. Assuming the PSU is 90% efficient (optimistic, but let's go with it), @523w draw, the PSU is supplying 52w less than that.

Second, you are still glossing over the fact that a 3950x (TDP 105w) is far from the lowest-wattage CPU out there. Assuming again our hypothetical 3600 (TDP: 65w), the headroom in that case would be 27 + 52 + 40 + 4 = 123w. That's 123
watts less than the 550w rating. Is 123 watts enough headroom? Surely.

In closing, I'm going to ask Steve Walton (or any other passing Techspot editor) to confirm that their reviews measure system power by wall draw. Otherwise I'm sure you'll spend ten or twenty posts trying to debate that point also.

So your reply is to assume efficiency, assume that power is measured at the wall, and to cherry pick a low power CPU that likely wouldn't be paired with this GPU to begin with?

Also, you may want to recheck your "calculations" on your capacitor aging chart -- the math doesn't work the way you think it does.

So you admit it does indeed occur but don't take that into consideration?

How small diameter electrolytic capacitors in power supplies can impact reliability and cost

Recently a member of TDK-Lambda’s technical marketing team experienced first-hand just how much influence small diameter electrolytic capacitors can have on long-term power supply reliability. Unfortunately he picked February to have his central heating system upgraded and Britain’s...

www.newelectronics.co.uk

Once again, you provide an assumption with nothing to back it up. You do indeed realize that capacitor again exists yet refuse to acknowledge it because it hurts your point.

Look at the post history, when was the last time you actually tried to backup anything you've said with proof?

 

[Endymio]

So your reply is to assume efficiency, assume that power is measured at the wall, and to cherry pick a low power CPU that likely wouldn't be paired with this GPU to begin with?

"Assume" efficiency? Not sure what you mean. Are you implying that 90% efficiency near max load is an unrealistic assumption? Or are you claiming that the amount a PSU draws from the wall = the amount it supplies? Either way, you're wrong.

As for the silliness that a 3600 would never be paired with a 3080, if you search on this site for completed builds with a 3600 + 2080, you'll see quite a few. I'm sure at least a few of those individuals will upgrade to a 3080:

PCPartPicker: Completed Builds

And, of course even if true, it doesn't impact my point because -- as your own link showed -- even a 3950x CPU + 3080 runs at substantially less than 550 watts, as long as the test system doesn't load up too much ram or additional drives, of course.

How small diameter electrolytic capacitors in power supplies can impact reliability. ...You refuse to acknowledge [capacitor aging] because it hurts your point.

I never refused to "acknowledge" it; I merely pointed out that whatever math you thought you were doing in that chart was wrong. Regardless, it in no way affects my point. I mentioned a hypothetical 550w PSU -- not a hypothetical several year old PSU.

As for your latest link, not sure how you feel an article on power supplys in home boiler systems has anything to do with this argument, but hey, you've posted stranger remarks than this before.

 

[Evernessince]

As for the silliness that a 3600 would never be paired with a 3080, if you search on this site for completed builds with a 3600 + 2080, you'll see quite a few. I'm sure at least a few of those individuals will upgrade to a 3080:

PC Part picker "builds" are virtual only. Each build on that website does not correlate to anything in real life.

Even the steam survey is a better measure of builds out in the wild

Steam Hardware & Software Survey

Only 21.71% of all CPUs on the site are AMD, and of those an even smaller fraction are going to be a 3600.

I never refused to "acknowledge" it; I merely pointed out that whatever math you thought you were doing in that chart was wrong. Regardless, it in no way affects my point. I mentioned a hypothetical 550w PSU -- not a hypothetical several year old PSU.

So you are also assuming that everyone has a brand new PSU as well then? That would be an assumption against what is common, as most people don't upgrade their PSUs too often.

As for your latest link, not sure how you feel an article on power supplys in home boiler systems has anything to do with this argument, but hey, you've posted stranger remarks than this before.

As it turns out, power supplies require electrolytic capacitors for a variety of devices. Is it really that surprising that power supplies for other devices use the same kind of electrolytic capacitors? I would say you learned something but you'll probably just ignore it now that it doesn't support your pre-concieved notions.

 

[Endymio]

PC Part picker "builds" are virtual only. Each build on that website does not correlate to anything in real life.

You've outdone yourself this time. See those photos attached to each build? You think they're render mockups or something?

So you are also assuming that everyone has a brand new PSU as well then?

Everyone? I said, nor implied, anything of the sort. The argument really isn't that complex. Perhaps English is not your native language?

As it turns out, power supplies require electrolytic capacitors for a variety of devices.

Which of course has no relevance whatsoever to the discussion at hand.

 

[mailpup]

@Evernessince and @Endymio, your personal argument has gone on long enough. If you feel the need, please continue your argument via PM. Thank you.

 

[grumblguts]

All that £$£" given to AMD over the years and Over night power usage isn't a problem.
How ironic.

 

[Avro Arrow]

All that £$£" given to AMD over the years and Over night power usage isn't a problem.
How ironic.

I couldn't have said it better myself. You're 100% on-point. It really shows just how ingrained the anti-AMD bias is in the tech press. I remember the tech press just toasting the R9 Fury and Fury-X over power use but they don't dare say anything against nVidia.

The only ones who truly have the guts to tell it like it is are Jim from AdoredTV, Paul from "Not an Apple Fan", Steve Burke from Gamers Nexus and of course Steve and Tim from Hardware Unboxed (although they are often far more polite than is sometimes warranted).