Understanding Professional Review PSU Ratings

[flong]

Hello:

Recently on a thread I got into a discussion on PSU requirements and several power consumption graphs were posted from professional reviews. One of the graph statisticss was for power consumption when running crysis. It showed 777W on the graph.

One comment was made that "at the socket" the actual power draw was 640W which does not make sense to me but I am not an electrical engineer.

In general, I would like to understand when I see a power usage graph from a professional review, what is the actual demand on the PSU if we use the information from the graph. In this example, if 777W was what was listed for the I-7 920 system what wattage would the PSU have to put out to make that wattage and still deal with the inefficiencies of the system?

I hope that my question is clear enough. I would like to better understand the professional reviews and what the actual power requirements for a PSU are. For example, if the PSU is 90% efficient, doesn't that mean it has to put out 111.11W to supply 100W to the system? So to supply 777W for Crysis (SLI 580 system) wouldn't the PSU have to put out 777/.9 = 864 W?

Thanks

 

[ubercake]

You are correct in your assumption about efficiency and power requirements. The power from the socket/outlet is always going to be higher than what's required by your components. The efficiency of the power supply will determine the overhead. The review, as you're described, doesn't make sense.

Based on my own experience, I have an i7-960 and a 580 SLI setup. My system never requires more than 800W from the wall and I have a Corsair 950tx that is 8x% efficient. That's with the 580s running in the high 90% utilization range during BFBC2.

 

[flong]

You are correct in your assumption about efficiency and power requirements. The power from the socket/outlet is always going to be higher than what's required by your components. The efficiency of the power supply will determine the overhead. The review, as you're described, doesn't make sense.

Based on my own experience, I have an i7-960 and a 580 SLI setup. My system never requires more than 800W from the wall and I have a Corsair 950tx that is 8x% efficient. That's with the 580s running in the high 90% utilization range during BFBC2.

So if you have 800W from the wall and lets say your PSU is 85% efficient; then what is the PSU actually supplying? Would that calculation be 800W x .85 = 680W? So though the outlet is supplying 800W, actually your system is using 680W and the PSU is supplying 680W, is that correct?

 

[Leaps-from-Shadows]

Those two numbers don't make any sense to me either. Maybe the numbers were reversed by whoever commented? That would make much more sense -- if 777W was drawn from the socket and only 640W was actually required by the hardware, then the power supply in question was roughly 82.37% efficient in those conditions. That sounds about right for a good power supply today.

I always thought that power supply efficiency referred to how much of the AC power from the socket was converted into DC power for the system versus how much was wasted as heat. The AC draw from the socket will always be higher than the DC power supplied to the hardware.

Take a look at the five-test graph in this recent review of a Sentey 850W power supply. In Test 1, the Sentey was supplying 175.5W (the Total line) and it was drawing 198.2W from the socket (the AC Power line). It's efficiency in that test was 88.5% (the Efficiency line).

If the hardware requires 777W, then the power supply will deliver 777W. The efficiency will be in how much is drawn from the socket to supply that 777W to the hardware. If the power supply is 90% efficient, it will draw 863.33W from the socket.

Edit: Scooped! Oh well...

 

[ubercake]

So if you have 800W from the wall and lets say your PSU is 85% efficient; then what is the PSU actually supplying? Would that calculation be 800W x .85 = 680W? So though the outlet is supplying 800W, actually your system is using 680W and the PSU is supplying 680W, is that correct?

You got it. I question the validity of the review you read. Where did you see it?

 

[flong]

OK guys, I was stupid in how I posted this. Here is the link to the graph I was talking about. It is a review of the power consumption of a GTX 580 on a base I-7 920 machine which the review lists.

If you look at the third graph down it shows the "Furmark" load rating as 777W for a 580 SLI system. My question is does Anandtech mean that at the socket they measured 777W going to the PSU or do they mean that the socket power to the PSU is 777/.9 (the assumed efficiency of the PSU) = 864W?

When Anantech lists these wattages are they the actual wattages that the card(s) are using?

Here is the link: http://www.anandtech.com/show/4008/nvidias-geforce-gtx-580/17

 

[ubercake]

OK guys, I was stupid in how I posted this. Here is the link to the graph I was talking about. It is a review of the power consumption of a GTX 580 on a base I-7 920 machine which the review lists.

If you look at the third graph down it shows the "Furmark" load rating as 777W for a 580 SLI system. My question is does Anandtech mean that at the socket they measured 777W going to the PSU or do they mean that the socket power to the PSU is 777/.9 (the assumed efficiency of the PSU) = 864W?

When Anantech lists these wattages are they the actual wattages that the card(s) are using?

Here is the link: http://www.anandtech.com/show/4008/nvidias-geforce-gtx-580/17

I would say the 777W here is the total draw from the wall. This is very consistent with mine. I was just telling someone the other day the highest draw I've ever seen from my PC (in signature with 2 580s) using an appliance load tester is in the 770Ws. This is with both cards in the high 90%s load as measured in GPUz with maxed out AA, AA transparency turned on and every other detail at max in BFBC2. Additionally, all BFBC2 uses all 8 threads of the i7-960. Furmark pushes that GPU limit and holds it there.

They don't really explain their methodology, but if we assume the PSU is 90% efficient at that level of usage:

777W(from the wall during Furmark) * .9 = 699.3W actually used by the system components

More realistically, with the worst case scenario and a 80-plus power supply, you'll be operating at 80% efficiency:

777W(from the wall during Furmark) * .8 = 621W actually used by the system components

The thing is, even on 'The Test' page in the review, they make no mention of the PSU? I didn't read the entire article, but to me, if you're going to post something like power consumption, you need to let people know what kind of efficiency you're working with up front. How can you compare this to your setup if you don't know the efficiency of power supply they're using? I don't understand how an article gets published like this. Maybe I missed something?

 

[shanky887614]

in simple terms, a power supply is a transformer

when ever a transformer increases or decreases voltage (from 110/230 down)
there is energy that is lost as heat

the eficency rattings just mean that it is going to be cheaper to run really in the long run and you shouldnt really concern yourself with them

trust me im doing an electrical course and

if your confused about efficeny, have a look at this

http://en.wikipedia.org/wiki/Power_factor

i know its generally ussed for lighting but this will give you a basic understanding because the principles are the same

if in doubt buy a more powerfull power supply than you need, i generally aim to get at least %50 over the max draw of my system

so if my system will draw 500w ill buy at least a 750

(you can buy exactly what you need but id rather spend extra money than spen ages working out power rattings)

 

[flong]

I would say the 777W here is the total draw from the wall. This is very consistent with mine. I was just telling someone the other day the highest draw I've ever seen from my PC (in signature with 2 580s) using an appliance load tester is in the 770Ws. This is with both cards in the high 90%s load as measured in GPUz with maxed out AA, AA transparency turned on and every other detail at max in BFBC2. Additionally, all BFBC2 uses all 8 threads of the i7-960. Furmark pushes that GPU limit and holds it there.

They don't really explain their methodology, but if we assume the PSU is 90% efficient at that level of usage:

777W(from the wall during Furmark) * .9 = 699.3W actually used by the system components

More realistically, with the worst case scenario and a 80-plus power supply, you'll be operating at 80% efficiency:

777W(from the wall during Furmark) * .8 = 621W actually used by the system components

The thing is, even on 'The Test' page in the review, they make no mention of the PSU? I didn't read the entire article, but to me, if you're going to post something like power consumption, you need to let people know what kind of efficiency you're working with up front. How can you compare this to your setup if you don't know the efficiency of power supply they're using? I don't understand how an article gets published like this. Maybe I missed something?

Thanks, I am glad that a number of things are not even clear to you about how Anandtech set up their review - it makes me feel less like an idiot ha, ha. What you say makes a lot of sense, especially since you own a similar system.

It became a big discussion detail on another thread when I recommended the Corsair AX 1200 for a system like yours in lieu of the Corsair HX850. I felt that the AX 1200 would be more efficient, cooler and quieter than the HX 850 which starts to ramp up the sound and lose efficiency at 538W.

Which of these two units do you think would be a better choice for your system?

 

[ubercake]

I always choose the safer option. I don't really need a 950W supply, though I have one. I guess I feel more at ease with having some headroom as opposed to pushing any limits.

I'd go with the 1200 given the two options you offer, but as we see in all cases, the 850W would be just fine.

If it was an either-or situation by which I could either choose better video cards or the higher wattage power supply, I'd choose the better video cards and go with the 850W.

 

[flong]

I always choose the safer option. I don't really need a 950W supply, though I have one. I guess I feel more at ease with having some headroom as opposed to pushing any limits.

I'd go with the 1200 given the two options you offer, but as we see in all cases, the 850W would be just fine.

If it was an either-or situation by which I could either choose better video cards or the higher wattage power supply, I'd choose the better video cards and go with the 850W.

Thanks, I appreciate your input. The HX 850 efficiency drops to 84% around the 650W mark and the fan starts to ramp up which produces some noise. It is a great unit, I own it, but for such a high demand card as the GTX 580, I felt that the AX 1200 would be better. I know that 850 would power the system, it just won't do it as coolly, as quietly and as efficiently as the AX 1200.

The remarkable thing is, if your minimum non-gaming demand is above 225W, which is not very much, the AX 1200 is the better choice if you go by efficiency. This is because at 225 W the 1200 is at 87.5% efficiency. From there it quickly goes up to 90% + efficiency and it stays there up to past its 1200W rating.

This is remarkable performance if you think about it because the big 1200W unit is MORE efficient at low power than most 750W and 850W PSUs which is usually not the case. All PSUs begin to drop dramatically in efficiency near the 200 W mark.

Couple this with a cost around $250.00, a 7-year warranty, it is fully modular and the AX 1200 is a terrific value. Even though I run only one GPU, I think that I may get it for my next build.

 

[ubercake]

Thanks, I appreciate your input. The HX 850 efficiency drops to 84% around the 650W mark and the fan starts to ramp up which produces some noise. It is a great unit, I own it, but for such a high demand card as the GTX 580, I felt that the AX 1200 would be better. I know that 850 would power the system, it just won't do it as coolly, as quietly and as efficiently as the AX 1200.

The remarkable thing is, if your minimum non-gaming demand is above 225W, which is not very much, the AX 1200 is the better choice if you go by efficiency. This is because at 225 W the 1200 is at 87.5% efficiency. From there it quickly goes up to 90% + efficiency and it stays there up to past its 1200W rating.

This is remarkable performance if you think about it because the big 1200W unit is MORE efficient at low power than most 750W and 850W PSUs which is usually not the case. All PSUs begin to drop dramatically in efficiency near the 200 W mark.

Couple this with a cost around $250.00, a 7-year warranty, it is fully modular and the AX 1200 is a terrific value. Even though I run only one GPU, I think that I may get it for my next build.

Sounds good. At any power requirement, that 1200 seems as if it will use less power. Seriously, all too often people don't even consider the warranty as any part of the value. I've had to deal with Corsair in the past with a bad stick of RAM and they had a new one to me within two days of reporting it. That 7 years makes the 1200 a great deal. You'll be able to use that with at least a couple of systems.

 

[flong]

http://www.overclockersclub.com/reviews/asus_hd_6950_review/16.htm

OK this is what drives me crazy about professional reviews - read this review of the Asus 6950 - but look closely at the power consumption ratings of the GTX 580 that are also shown. They are almost TWICE as much as what we have been discussing for a single 580 card. What the heck is going on. Does anyone really understand these reviews???????