Printable View
What's with people getting 800, 900, or even over 1000 watt PSU's? There's no way in hell, even with tri-SLI is anyone pulling that kind of power. I believe most of these calculators online are using the specs on the side of the boxes to determine the wattages needed. If a video card says it needs a system with a 400 watt PSU, that doesn't mean you need 400 watts for every video card!
I wouldn't be so sure. The Nvidia GTX 480 pulls aprox 165 watts at idle and 376Watts under load, so there is just over 1100 watts with just the cards in Tri-SLI. Then if you add a CPU that can pull up to 125W, and don't forget the RAM. Rule of thumb for RAM is about 15W per GB at full load, so if you have say 8GB there is 120W. So without even adding HDD's, motherboard chipsets, sound cards, or fans, your sitting at just under 1350W. Of course thats when everything is running at 100% load.
It will still last longer with a lesser load. The Corsair 1200W PSU that I sold to a guy in Brazil (He wanted it and sent me the link to it with a promo code for newegg) reaches maximum efficiency at 600W. He's also doing water cooling, using multiple hard drives and SSDs, overclocking an i7 950, and running multiple GPUs (everything purchased from me), so I can see why he wanted it.
The power supply also comes with a 12 year warranty. Corsair probably knows that most people won't be going very close to its peak output, or at least that's what I would guess is their reasoning for such a long warranty period.
My current system is running on a Cooler Master Silent Pro Gold 800w. And I do notice a dimming of my fans LEDS when stressing things.
140w AMD Phenom II X4 CPU - 140w
2 Gigabyte Radion 5870s - 358w + 188w (for second card)
8GB Ram - 120w
5 HDD - 100w
7 Fans - 35w
1 BD drive - 25w
Motherboard - 100w
Front LCD - 20w
Water Cooling Pump - 50w
CFFL and LEDS - 10w
That's 1146w at load. After doing the math I am very surprised that my 800w Cooler Master is keeping up. I guess they rate PSUs like they do audio amps. IE: 800w RMS and 1200 peak. I have a 1200w Thermaltake Toughpower that I know can handle 1500w easily.
The more important thing to look at today is how many amps your PSU can handle per rail. If your PC requires 60 amps on the 12v rail and your PSU can only give it 45amps then things will not work right. You run the risk of overheating the PSU and making things let out their magic smoke.
I know Sean Kalinich a writer for tweak town and editor of thebrightsideofnews.com put 4 of the Nvidia 480s in SLI and IIRC it took a separate 1200w PSU just for them. So these high wattage PSUs are needed. Maybe the average user will never need more than say 500w but power users with high end multiple GPU systems will.
Another thing I forgot to mention is the additional load USB devices put on a system. I have 2 externals hooked up 1 being USB 3.0, some nice surround sound USB headphones, a small USB fan, my iPod Touch, and my cell phone all charging at once. I bet that adds another 50-75w to the mix.
It's basically just multiple GPUs that eat all that juice. Go check tomshardware, they do power consumption on every rig they test, and have had various fermi SLI setups recently including a quad SLI rig that went well beyond the specifications of its 1000W PSU (but, impressivley, the PSU did not fail).
Oneslowz28, you very much overestimate the consumption of some of your components. Bare in mind that you'd have a hard job stressing all those components fully at the same time, and that your PSU will almost certainly have a cut out within 200W of its rated maximum.
I've personally tested some PSUs to 600W+ over their "max." It has a lot to do with how the PSU is built.
Take a look at the power limiter removal for GTX580? The GPU can be seen drawing as much as 500+ watts ,and there's a warning it can ruin card and board. Those numbers aren't as overinflated as they sound. These are tests done with high-end test equipment, not kicking back and saying "well, that looks like about..."
CJ and I are both PSU reviewers, and my review equipment stack probably has a similar net worth to your car. It includes dual oscilloscopes and wattage meters. A lot more goes into a PSU review than people think...
I've pulled over a rated maximum through a Mushkin PSU...by about 100watts... It lasted for as long as I used it like that...
Well I may not have some of the hard core gear Kayin does, but I do have the privilege of being in contact with engineers on some of these PSUs. I know for a fact that the 1200w I have can run a 1600w load. I was shown testing data from a custom 3d rendering workstation that ran 6 Nvidia workstation cards and dual water cooled xenons. Would I recommend over working a PSU? Never. But you do what you have to do sometimes. I will be switching my gaming case over to the 1200w just to be safe in a few weeks.
As a rule of thumb, always get a psu that is 20% overpowered at minimum. As long as it is of good quality IE: brand name manufacturer with good reviews and at least $70 you will be fine.
Myself along with several other TBCS staff members will be attending the 2011 Consumer Electronics Show in Las Vegas this coming January. I have a few meetings set up with several of the industry leaders, so if I get to speak with a design engineer in person I will bring up the topic of over powering PSUs and see what they have to say.
This. Almost all PSUs have an efficiency curve that peaks at 50%. The more efficient it is running, the less heat produced and the longer it will last. The longer it lasts, the less money the end user spends on PSUs. I have a 1000W Ultra X3 that runs at about 30-50% load depending on what I'm doing...it has a lifetime warranty that I suspect I will never have to use.:)Quote:
Originally Posted by nevermind1534
This is also true. The only time a theoretical maximum power draw could ever be reached would be if a user was running synthetic benchmarks on every piece of hardware at the same time. There is not a single real-world scenario that would result in everything sucking down 100% power at the same time.Quote:
Originally Posted by Blibbax
I entirely agree with Kayin. All PSUs are not created equal, the wattages tend to serve more of a marketing purpose than a technical one with most brands. I've seen "350W" PSUs sustain 500W output without any problems at all, not even a shorter life expectation (based on the observation that they're not heating up), while I've seen "600W" PSUs fail to run "400W" P4 systems loaded with one AGP card and not much else. I've never bought anything much larger than 600W simply because it then becomes more cost-effective to run two PSUs.
The biggest determinant of PSU output is going to be the oomphf of the main transformer (if it has one), robustness of the rectifier stage, and the sophistication of the switching circuitry. The regulation circuitry might have some "intelligence" which better keeps voltages in spec, but tends to have less impact on overall performance than the other stuff. Of course these are the parts that tend to cost the most, so they are the first ones to get downgraded to less capable parts by the manufacturer.
I find that only buying PSUs capable of Active PFC (more a combination of engineering approaches than a particular technology) ensures that a certain higher-quality threshold is present throughout; this is the "hardest" (most costly) qualification when engineering a PSU because it is largely defined by the interaction of everything else. Much the same reason you'll never encounter an utterly cheap DMM if RMS capability is integrated.
With Active PFC you can be confident that the PSU will work well under load, even if maybe not as good as advertised. Without Active PFC you're just drawing a wild card.
Another good indicator of overall PSU performance seems to be how well regulated the 3.3V output is. Tight PSUs tend to be tight all across the board, while looser (I've seen 'em as bad as ±0.2V) indicates poor engineering (and awful voltage regulation) all across the board.