Please satisfy my curiosity - PSU question

[IVIYTH0S]

Oh look this thread again

[Gr4vitas]

Lol at the champs replying to me.

[jammin]

OK, interesting thread.

 

Firstly I should make a note before I continue with my opinion on the subject. That is that I happen to be in the position of using an 850W PSU in a build that probably draws around the 300W mark that was mentioned in the original post (and I suspect it would be less, though I haven't measured it). This is entirely because of the amount I was paying for the PSU (if it had been full price, I wouldn't have got it). My upgrade plans are also uncertain, so it is good to have the flexibility.

 

Right, with that out of the way, let's see what we have here.

 

Is it a good idea to buy a power supply that is rated obviously way over what your system requires or is likely to require in future?

 

I think we could approach this from a few angles.

 

Cost

Assuming you are aiming for equal quality, more capacity is usually going to cost you more. From this point of view it makes little sense to spend more, when you are going to see no benefit for your money. Some of that may be negated by component de-rating (where an over-specced PSU will help) and the possibility of transferring it to a more power hungry build somewhere down the line. Those are relatively weak reasons though (particularly if you are the type of person who usually builds with one GPU in mind).

 

Cost is the strongest argument against in my opinion

 

 

Efficiency

Selecting a PSU with an efficiency curve well matched to your build is a pretty sensible thing to do, and it is true that figures do tend to drop off at low PSU loads. Recent high quality units are getting better at this as well though, with some already over 80% at a load of 10%.

 

Another thing to note is that at lower loads, % efficiency also has less impact on power being drawn from the wall.

For example:

 

Take two Gold rated PSUs - PSU 1 is a 560W model and PSU 2 is a 1200W model

 

At a load of 120W DC

PSU 1 is 87% efficient - 138W (rounded up) drawn from the socket

PSU 2 is 84% efficient - 143W (rounded up) drawn from the socket

 

So the difference is 5W

 

At a load of 300W

PSU 1 is 91% efficient - 330W (rounded up) drawn from the socket

PSU 2 is 88% efficient - 341W (rounded up) drawn from the socket

 

So the difference is 11W

 

So at both loads the difference isn't that much, and in terms of wattage it is likely to actually be smaller at lower loads because we are dealing with smaller numbers.

Note that the numbers above are roughly based on review measurements, rather than pulled from thin air (PSU 1 and PSU 2 if you were interested).

 

I'd say that arguing it from an efficiency point of view is fine, but the difference isn't night and day (though there is one), and I think that cost is certainly a much stronger argument.

 

 

Quality

 

One could argue that units with higher wattage ratings are also generally of higher quality.

I'd say that there is a certain amount of truth in this, but that there are modern high quality PSU platforms available in a good range of ratings these days, so it should be easy to find a quality unit that matches your power requirements.

 

 

Conclusion

 

The biggest argument against buying a PSU that is considerably over-rated for your build is that it is going to cost you more money for potentially little to no real benefit. It is money better spent elsewhere in my opinion.

Edited March 16, 2011 by jammin

[Gr4vitas]

OK, interesting thread.

 

Firstly I should make a note before I continue with my opinion on the subject. That is that I happen to be in the position of using an 850W PSU in a build that probably draws around the 300W mark that was mentioned in the original post (and I suspect it would be less, though I haven't measured it). This is entirely because of the amount I was paying for the PSU (if it had been full price, I wouldn't have got it). My upgrade plans are also uncertain, so it is good to have the flexibility.

 

Right, with that out of the way, let's see what we have here.

 

Is it a good idea to buy a power supply that is rated obviously way over what your system requires or is likely to require in future?

 

I think we could approach this from a few angles.

 

Cost

Assuming you are aiming for equal quality, more capacity is usually going to cost you more. From this point of view it makes little sense to spend more, when you are going to see no benefit for your money. Some of that may be negated by component de-rating (where an over-specced PSU will help) and the possibility of transferring it to a more power hungry build somewhere down the line. Those are relatively weak reasons though (particularly if you are the type of person who usually builds with one GPU in mind).

 

Cost is the strongest argument against in my opinion

 

 

Efficiency

Selecting a PSU with an efficiency curve well matched to your build is a pretty sensible thing to do, and it is true that figures do tend to drop off at low PSU loads. Recent high quality units are getting better at this as well though, with some already over 80% at a load of 10%.

 

Another thing to note is that at lower loads, % efficiency also has less impact on power being drawn from the wall.

For example:

 

Take two Gold rated PSUs - PSU 1 is a 560W model and PSU 2 is a 1200W model

 

At a load of 120W DC

PSU 1 is 87% efficient - 138W (rounded up) drawn from the socket

PSU 2 is 84% efficient - 143W (rounded up) drawn from the socket

 

So the difference is 5W

 

At a load of 300W

PSU 1 is 91% efficient - 330W (rounded up) drawn from the socket

PSU 2 is 88% efficient - 341W (rounded up) drawn from the socket

 

So the difference is 11W

 

So at both loads the difference isn't that much, and in terms of wattage it is likely to actually be smaller at lower loads because we are dealing with smaller numbers.

Note that the numbers above are roughly based on review measurements, rather than pulled from thin air (PSU 1 and PSU 2 if you were interested).

 

I'd say that arguing it from an efficiency point of view is fine, but the difference isn't night and day (though there is one), and I think that cost is certainly a much stronger argument.

 

 

Quality

 

One could argue that units with higher wattage ratings are also generally of higher quality.

I'd say that there is a certain amount of truth in this, but that there are modern high quality PSU platforms available in a good range of ratings these days, so it should be easy to find a quality unit that matches your power requirements.

 

 

Conclusion

 

The biggest argument against buying a PSU that is considerably over-rated for your build is that it is going to cost you more money for potentially little to no real benefit. It is money better spent elsewhere in my opinion.

 

Thanks Jammin, I was too lazy

[Waco]

Conclusion

 

The biggest argument against buying a PSU that is considerably over-rated for your build is that it is going to cost you more money for potentially little to no real benefit. It is money better spent elsewhere in my opinion.

Thanks for your post!

 

I'm surprised those two units have such a high efficiency at low loads - perhaps 80 Plus Gold PSUs are better in that area than I expected (or those are two exceptional PSUs).

[SpeedCrazy]

I'm surprised those two units have such a high efficiency at low loads - perhaps 80 Plus Gold PSUs are better in that area than I expected (or those are two exceptional PSUs).

+1 And i settled for a bronze certified. though i probably pull about 50-60%.

[jammin]

Those do happen to be two very good units, so it would be interesting to do the same comparison with others. The differences could well be bigger. It was mainly and example to show that once you use a reasonably low load the actual difference in watts will tend to get smaller. Plus it obviously matters where your system is going to idle (the figure might be below 100W for example).

 

Let's try another, slightly more skewed example.

 

In this case we'll take a 1000W Enermax Galaxy and a 400W Corsair CX400. Note that the Enermax was released about twice as long ago as the Corsair.

 

At a load of 90W DC

 

The Enermax is about 65% efficient (ouch) - 138W from the socket

The Corsair is about 81% efficient - 111W from the socket

 

27W is certainly a much less negligible difference.

 

I did cherry pick somewhat to find a bigger difference there.

Edited March 16, 2011 by jammin

[Waco]

I did cherry pick somewhat to find a bigger difference there.

Can't fault you for that.

[DanTheGamer11]

I heard that a PSU operates at max efficiency at half the load?

[Boinker]

I heard that a PSU operates at max efficiency at half the load?

 

Not neccesarily but that may be true for some. They have an efficency curve that drops out normally near 75% Just like everything has. The more power you produce the more that is lost is not the case as I have seen in the past. It all depends on the psu model and make. Quality is going to make the largest difference in the long run and will be a determining factor for the remainder of your system.

 

Why more wattage. I know alot of people have claimed I do not need a psu as large as mine. I could in theory run my system with a decent 850w but I would rather not. The reason for that is to not only keep the power supply at the upper end of its effieciency curve and also to keep it away from 85 or better % load at all times. The reason for this is due to the heat it produces. Power supplies are really no different then say a processor, ram, HDD, or a motherboard even. The cooler you keep them the longer they will last. Capacitor aging will happen as it always does but will be kept to a min when you have less temperature involved.

 

Where i stand: If you know exactly where your total power consumption stands I always recommend getting a psu rated for 200w more then you need. The reason above explains my madness and It just another IMHO. And this all depends on the Quality of the manufacture.

 

BTW I have two questions just for fun and to stir thought.

 

Has anyone attempted to use the PSU tester in OCCT to see if it actually does make a decent load? I figure once I get a Kilowatt plug ill test my complete system to see what it does.

 

And do you actually know or have documentation to prove that your PSU is rated at its Continuous or Peak wattage?

[Waco]

Has anyone attempted to use the PSU tester in OCCT to see if it actually does make a decent load? I figure once I get a Kilowatt plug ill test my complete system to see what it does.

I didn't have any way to test the power used by my computer / PSU at the time - but when I ran the test for an hour I actually heard the fan in my Silencer 750 spin up past idle for the first time ever.

[paulktreg]

Just found this and I have to agree with jammin.

 

Just one or two things I'd like to add....

 

If a power supply manufacturer wants 80 Plus certification for their latest power supply it will be part of the design brief that it hits maximum efficiency at 50% DC loading whilst still maintaining the required levels at 20% and 100% DC loading. The 80 Plus specification is also pretty specific about how this load is spread across the 3V3, 5V0 and 12V rails but it appears to be pretty representative of a modern set up with the emphasis placed on the 12V rail loading.

 

Fit a large power supply by all means, it's not going to hinder PC performance in any way, it will run cool and quiet but the running costs will be marginally higher.

 

Take a rig that pulls 125W idle and 275W with GPU and CPU fully loaded from the power supply. Take two power supplies, both 80 Plus Bronze certified (not uncommon these days for an high end power supply), one 450W and the other 850W. This will mean efficiency levels of 82%, 85% and 82% at 20%, 50% and 100% DC loading. How much will the PC cost to run at idle and at maximum load for 8 hours a day 24/7?

 

I'll take the cost of electricity in the USA at $0.12/kWh, after a little research this seems to be a good figure although some pay more and some pay less.

 

The 450W power supply.

 

125W = 28% DC load which I'll approximate at 83% efficiency = 151W from the wall = 0.151kWh = 0.151 x 0.12 = 0.018$/hr = 0.144$/day = $52.56/year

275W = 61% DC load which I'll approximate at 84% efficiency = 327W from the wall = 0.327kWh = 0.327 x 0.12 = 0.039$/hr = 0.312$/day = 113.88/year

 

The 850W power supply.

 

125W = 15% DC load which I'll approximate at 81% efficiency = 154W from the wall = 0.154kWh = 0.154 x 0.12 = 0.185$/hr = 0.148$/day = $54.02/year

275W = 32% DC load which I'll approximate at 83% efficiency = 331W from the wall = 0.331kWh = 0.331 x 0.12 = 0.040$/hr = 0.320$/day = $116.80/year

 

The difference at the end of the day is marginal but when you can buy a perfectly adequate 450W for $55.00 and a 850W one for $120.00 which also costs more too run, which way would you go?