A comparison of stock vs efficient gaming

[The red spirit]

Yes, but in idle, we're talking about maybe 50 W total system power consumption. Maybe. With 80% efficiency, it's 60. That's not what's killing your wallet.

You still spend hours with computer either in idle or barely loaded. And you would still overpay for PSU in such case.

 

[AusWolf]

You still spend hours with computer either in idle or barely loaded. And you would still overpay for PSU in such case.

I somewhat agree on the overspending part. My PSU recommendation tends to be: 1. look at CPU+GPU peak power consumption (either at stock, or overclocked, but NOT undervolted). 2. Add 100 W. 3. Choose the best quality PSU that meets the number you got. Going for a higher wattage model than you need only gives you some headroom in case you want a faster CPU or GPU later.

As for hours in idle - if there's an extra 10 W consumed due to PSU inefficiency, then 100 hours in idle makes up for 1 kWh that costs around 0.28p in the UK. If you run your computer in idle 24/7, that's 744 hours in a 31-day month, which will cost you an extra £2.08 for that month, or £25 for a year. And that's when you have your PC switched on 24/7 for the whole year, mind you.

 

[The red spirit]

I somewhat agree on the overspending part. My PSU recommendation tends to be: 1. look at CPU+GPU peak power consumption (either at stock, or overclocked, but NOT undervolted). 2. Add 100 W. 3. Choose the best quality PSU that meets the number you got. Going for a higher wattage model than you need only gives you some headroom in case you want a faster CPU or GPU later.

As for hours in idle - if there's an extra 10 W consumed due to PSU inefficiency, then 100 hours in idle makes up for 1 kWh that costs around 0.28p in the UK. If you run your computer in idle 24/7, that's 744 hours in a 31-day month, which will cost you an extra £2.08 for that month, or £25 for a year. And that's when you have your PC switched on 24/7 for the whole year, mind you.

Actually there's some parasitic power use when computer is "off". Of course ErP tends to reduce it, but it doesn't eliminate that. So actually, as long as it is connected to wall physically, it costs you. Otherwise reasonable analysis of price.

 

[AusWolf]

Actually there's some parasitic power use when computer is "off". Of course ErP tends to reduce it, but it doesn't eliminate that. So actually, as long as it is connected to wall physically, it costs you. Otherwise reasonable analysis of price.

Of course, you need a miniscule amount of power to make sure your PC turns on when you press the power button. That's normal. But what has it got to do with PSU efficiency?

 

[Mussels]

You still spend hours with computer either in idle or barely loaded. And you would still overpay for PSU in such case.

overpay, sure.That said, sometimes you don't actually pay more for the extra wattage - and since PSU's do decay slowly with age, if you're keeping it for 10 years you might as well get the extra

But as far as efficiency goes, 1-2% of 60W is just... not worth worrying about. Having my second monitor plugged in at the wall 'sleeping' uses more.
It's like saying i should increase my idle wattage to get better efficiency, while ignoring the fact i'm using more power to do so.

Gah and then things like this happen
[56k dial up noise in my brain gets louder]

 

[R-T-B]

overpay, sure.That said, sometimes you don't actually pay more for the extra wattage - and since PSU's do decay slowly with age, if you're keeping it for 10 years you might as well get the extra
View attachment 252393

But as far as efficiency goes, 1-2% of 60W is just... not worth worrying about. Having my second monitor plugged in at the wall 'sleeping' uses more.
It's like saying i should increase my idle wattage to get better efficiency, while ignoring the fact i'm using more power to do so.

Gah and then things like this happen
[56k dial up noise in my brain gets louder]
View attachment 252394

Platinur. Nice. I personally prefer Titaniyum.

 

[Mussels]

Platinur. Nice. I personally prefer Titaniyum.

That's just where i cropped it lol.

They also have an issue with spelling Aviator

~80FPS 100% of the time, or 140 99% with 1% dips?
If i just capped that to 80FPS, i'd get 80FPS 100% of the time and have a really smooth game session without any fluctuations, and it'd save whatever GPU and CPU power was spent rendering the extra.
View attachment 252316

So... that 99% FPS value looks really familiar to what TPU has in the 5800x3D review
(These are borderlands 3 results, - the actual value will vary per title!)

(and 12900Ks results for the intel comparison)

There comes a point where even with a top tier GPU, the rest of your system really dictates a level where past that, you just cant maintain higher FPS perfectly. Maybe thats 85FPS for me and 106FPS for a 12900K, but the idea of limiting below that value and keeping the thermal and TGP headroom to maintain it perfectly has some appeal, especially if you have a Gsync/freesync monitor/TV

 

[AusWolf]

That's just where i cropped it lol.

They also have an issue with spelling Aviator
View attachment 252396

What's wrong there? Avaitor... if you have to vait for something, it does the job for you.

 

[Pumper]

Tried undervolting my 3060Ti to 0.925, but it only saved ~2W (199W stock vs 196~197W with uv) and did nothing for temps, performance stayed the same, even when the boost was higher and more stable with lower volts. I was testing with Superposition.

3900X is at -0.1v offset and 95W eco mode, but the eco mode did not make any difference - still get 128W at full load and 7250-7300 point in R20.

FPS capped at 72 on 75Hz monitor. Rarely see anything over 300W full system load from the wall when gaming, mostly hovering in 160-200W range.

 

[Mussels]

Tried undervolting my 3060Ti to 0.925, but it only saved ~2W (199W stock vs 196~197W with uv) and did nothing for temps, performance stayed the same, even when the boost was higher and more stable with lower volts. I was testing with Superposition.

3900X is at -0.1v offset and 95W eco mode, but the eco mode did not make any difference - still get 128W at full load and 7250-7300 point in R20.

FPS capped at 72 on 75Hz monitor. Rarely see anything over 300W full system load from the wall when gaming, mostly hovering in 160-200W range.

0.925V is still really high (relatively, it's all ampere) - my 3090 is only at 0.8v normally with my 1.6GHz summer clocks at 0.74v

 

[Taraquin]

Tried undervolting my 3060Ti to 0.925, but it only saved ~2W (199W stock vs 196~197W with uv) and did nothing for temps, performance stayed the same, even when the boost was higher and more stable with lower volts. I was testing with Superposition.

3900X is at -0.1v offset and 95W eco mode, but the eco mode did not make any difference - still get 128W at full load and 7250-7300 point in R20.

FPS capped at 72 on 75Hz monitor. Rarely see anything over 300W full system load from the wall when gaming, mostly hovering in 160-200W range.

Find highest stable frequency at 900mv or lower, then you get powersaving if that's what you want

 

[The red spirit]

Of course, you need a miniscule amount of power to make sure your PC turns on when you press the power button. That's normal. But what has it got to do with PSU efficiency?

It has nothing to do with efficiency, I was commenting about your electricity cost calculations, but ErP is still somewhat related to efficiency.

overpay, sure.That said, sometimes you don't actually pay more for the extra wattage - and since PSU's do decay slowly with age, if you're keeping it for 10 years you might as well get the extra
View attachment 252393

But as far as efficiency goes, 1-2% of 60W is just... not worth worrying about. Having my second monitor plugged in at the wall 'sleeping' uses more.
It's like saying i should increase my idle wattage to get better efficiency, while ignoring the fact i'm using more power to do so.

Gah and then things like this happen
[56k dial up noise in my brain gets louder]
View attachment 252394

That decay is really small and some manufacturers already take it into account before writing declared wattage. Your PSU comparison to me makes no sense. You only compare very very overpriced units, thus it's not weird to see stupid pricing. Those will never pay off and no buying PSU for decade isn't a bright idea either. ATX 2 standard gets updates, requirements for cables changes, demands for different rails change and etc. A decade old PSU today wouldn't have ErP, no support for low Intel C state, likely no 8 pin EPS. In short it will be quite obsolete and furthermore Intel is now working an ATX 3 standard and there is ATX12VO standard, both of them have uncertain future.

Anyway, here's what properly priced unit should cost you:
Maitinimo blokas BitFenix 650 W, 135 mm - 1a.lt
86 EUR for fully modular, 80+ Gold unit, which was reviewed here at TPU.

 

[Mussels]

Oh god.
Just watched GN's GPU power spike video

Heres a good reason to undervolt and disable the boosts: because we avoid this shit (exactly 20 minutes in)

This is JUST the GPU's PCI-E cables alone

(340) The Brewing Problem with GPU Power Design | Transients - YouTube

 

[AusWolf]

It has nothing to do with efficiency, I was commenting about your electricity cost calculations, but ErP is still somewhat related to efficiency.

Exactly, so don't pull a straw man.

Edit: As for off states and ErP, I don't think any PSU is efficient at such low power levels, but it doesn't matter because those power levels are incredibly low anyway. You're not saving anything with efficiency there.

Oh god.
Just watched GN's GPU power spike video

Heres a good reason to undervolt and disable the boosts: because we avoid this shit (exactly 20 minutes in)

This is JUST the GPU's PCI-E cables alone
View attachment 252437
(340) The Brewing Problem with GPU Power Design | Transients - YouTube

Did they also test with programs that are not famous for killing GPUs?

 

[The red spirit]

Exactly, so don't pull a straw man.

Edit: As for off states and ErP, I don't think any PSU is efficient at such low power levels, but it doesn't matter because those power levels are incredibly low anyway. You're not saving anything with efficiency there.

There is quite a bit that was done. High end power supplies perform a bit better at tiny loads. ErP partially is related too. And the biggest one is ATX12VO, which was made more or less specifically for that problem.

 

[R-T-B]

And the biggest one is ATX12VO, which was made more or less specifically for that problem.

And I still question it's logic. Moving the parts to the mobo manufacturer land is not likely to result in a net boost to efficiency over a quality standard PSU.

 

[eidairaman1]

When gpus by themselves are nearing what all other components combined draw, that is a cause of concern.

 

[AusWolf]

When gpus by themselves are nearing what all other components combined draw, that is a cause of concern.

Not nearing - they have already reached that level. A CPU can consume 200-250 W in extreme cases. The rest of the system is another 20-30 W. With SSDs and hard drives, let's call it about 300 W for the whole system. Even current gen high-end graphics cards need more than that.

Since multi-GPU is dead, I think nvidia and AMD are moving that level of performance into single high-end GPUs. There's no need for average consumers to look above x60, maybe x70 level graphics cards nowadays.

 

[The red spirit]

And I still question it's logic. Moving the parts to the mobo manufacturer land is not likely to result in a net boost to efficiency over a quality standard PSU.

Then how come we saw drastic power usage reduction in LTT video?

 

[R-T-B]

Then how come we saw drastic power usage reduction in LTT video?

What are the components being compared? Is the standard psu in the comparison a quality one? Is the ATX12VO unit being compared on equivalent hardware? Does the board in question even USE 5v and 3.3v rails (any real complex PC will need these voltages at some point)? If it does, how good is its conversion circuits? So many variables. And given it's a LTT video I really don't need to watch to guess he's failed to answer more than one of these rendering the figures useless. Someone with more time to waste than me may prove me wrong, of course.

 

[InVasMani]

I've been advocating for undervolting CPU/GPU since LGA775 personally since that's when I picked up a fanless PSU and was concerned with noise, heat, and power draw and it's a great way to help with all of those things.

 

[eidairaman1]

I've been advocating for undervolting CPU/GPU since LGA775 personally since that's when I picked up a fanless PSU and was concerned with noise, heat, and power draw and it's a great way to help with all of those things.

Even in Overclocking I use baby steps on vcore, glue logic ram, because eventually, heat destroys electronics

 

[The red spirit]

What are the components being compared? Is the standard psu in the comparison a quality one? Is the ATX12VO unit being compared on equivalent hardware? Does the board in question even USE 5v and 3.3v rails (any real complex PC will need these voltages at some point)? If it does, how good is its conversion circuits? So many variables. And given it's a LTT video I really don't need to watch to guess he's failed to answer more than one of these rendering the figures useless. Someone with more time to waste than me may prove me wrong, of course.

You are just behaving like LTT here. Being careless and ignorant. Anyway, there aren't that many videos or reviews of those units and LTT's video in this case is not the worst resource. Anyway, if you are really that anal about sources, there's OC3D review of ATX12VO wattage:

ATX12VO Tested - The Future of Power Supplies? - OC3D

ATX12VO – A Drive for Efficient Computing Last year, Intel and ASRock teamed up to create the world’s first ATX12VO motherboard, adopting a new standard which has the potential to change computing forever. At this point, it cannot be denied that global warming is a thing. Because of that, the...

www.overclock3d.net

Nearly 2x gains in low CPU loads and at least 50% gains at idle. That's quite substantial to me, considering that you aren't sacrificing any performance to get massive gains in efficiency. So it turns out that LTT was perfectly correct and you were wrong.

 

[Mussels]

Exactly, so don't pull a straw man.

Edit: As for off states and ErP, I don't think any PSU is efficient at such low power levels, but it doesn't matter because those power levels are incredibly low anyway. You're not saving anything with efficiency there.


Did they also test with programs that are not famous for killing GPUs?

They did. games.
Furmark used *less* since it doesnt surge, it's a constant sustained load


That 12VO test showing the reduced consumption at low load/idle sitatuations really is a game changer, now i dont wanna buy a new PSU til they're more readily available

 

[R-T-B]

You are just behaving like LTT here. Being careless and ignorant.

No, I just don't have time to waste on LTT, that's the honest truth.

Other sources I may consider, and as you have provided, I will look into it. Stand by.

EDIT: Ok, that certainly is interesting. My immediate question is simple: Why? Because from a electrical stand point this makes zero sense. Are the mobos conversion electronics far lower wattage than the PSUs? That might make sense.

That 12VO test showing the reduced consumption at low load/idle sitatuations really is a game changer, now i dont wanna buy a new PSU til they're more readily available

My evga SuperNova T2 is nearly 8 years old now, hard to believe. I will probably replace it with something when all this PSU standards dust settles.