I need a little help choosing a new Power Supply.

[doomzday]

First I would like to say thanks to SPCR for bringing sanity back to my ears. I greatly respect and appreciate the feedback on this board when it comes to silencing the computer with quality products, and power supplies are no different.

I have always tried to strike the right balance for me when it comes to Overclocking and quiet computing.

I currently have built a i7 920 its overclocked currently to 4ghz. It has 4 harddrives, and 2 burners. I have one 5850 but plan on adding a second one.

For cooling im using one 120mm exaust fan on 5v line. The Cpu I am using a ultra 120 extreme rev C. with a Scythe 120mm Slip Stream Kaze-Jyuni PWM Fan.

The current psu is a corsair hx520 which I have been happy with, but soon as I add the second 5850 video card I am not so sure it will be able to handle it.

So whats a good PSU these days thats reasonably quiet and will be able to power my system. I am planning on sticking with the Corsair brand if they are still somewhat quiet, as there easy to obtain locally. Which wattage should I look at, the hx650,750, or 850?

I noticed corsair seems to be using two manufactures of there powersupplies, should I pick one model over the other because of that?

Thanks.

[NeilBlanchard]

Hello & welcome to SPCR,

I'd recommend buying or borrowing a Kill-A-Watt meter and plug your current system into it. This will give you the AC consumption of the system -- and then if you subtract the approximate efficiency loss of your power supply (you can reference the SPCR review for this), then you'll know how much DC power you are using -- which how PSU's are rated.

My guess is the Corsair will be fine?

[JamieG]

From what I can tell, each 5850 requires 2 x 6-pin PCIe power plugs, so you will need to look at a PSU that has 4 of these plugs.

In addition, you should definitely check out some reviews to see who is the manufacturer of the Corsair-branded PSUs you are looking at. From memory, Corsair uses CWT (=Chanel Well Technology IIRC) and Seasonic as manufacturers, with the HX520 being a Seasonic unit. The CWT units may be noisier, generally speaking.

You might want to also look at a few Seasonic-branded PSUs - the X-750 (the X-650 PSU has been reviewed by SPCR here) and the M12D 850W.

Either of these units would serve you really well for a quiet high-end gaming PC.

[audiojar]

Look around at some of the actual measured power people's systems pull, I think there is a thread about it stickied somewhere. Nobody is pulling anything near 500W, it's even a challenge to break 300W. Another video card isn't going to double your draw or anything. Your PSU is fine. You can also get adapters if you need more PCIe power plugs.

[doomzday]

Thanks guys.

This is what I love and respect about SPCR. You won't find someone here yelling I need a grossly over powered psu like many other forums.

[Parappaman]

Well, 2x HD5850 on an overclocked i7 9xx may actually be enough to overload the HX520. And since you'll anyway need to purchase additional cables to run such a setup (I can't stress enough on how those molex adapters can be dangerous), I would switch to one of the latest offerings in the 650w mark, like the Seasonic X-650 or the much cheaper (but still excellent) Corsair HX650W. They will be more than enough even if you push those cards to the maximum, and still be silent at idle.

The HX750 uses a different design which is much more efficient than the HX650, and it will be quieter at higher loads but a touch noisier at idle. It's another great option if you got the funds and want to keep your setup open to even more extreme GPU settings, and comes at around the same price as the aforementioned Seasonic.

[b_rubenstein]

Really the best thing to do is to get a Kill-A-Watt and download OCCT. Run the OCCT PS stress test. The stress test runs Linpack (CPU Burn) and a graphics test that runs the video card as hard as Furmark. The power draw is probably 25% higher than any real life application and/or game will draw. To estimate how much power a second video card would draw, I would subtract the power used in plain OCCT (same as Prime95) from the power used in PS stress. This is probably a reasonable estimate of the additional power used by a second card. Add that power to the amount measured in the PS stress test to for a figure to work with for what your system would use with two cards.

A pair of 5850's may push the HX520 (made by Seasonic) beyond its limits during typical game play. It certainly would running the OCCT PS stress test. Xbit ran a i7-920 OC's to 4.1 Ghz with a 5850 running Linpack & Fumark and it drew 483w: http://www.xbitlabs.com/articles/video/ ... html#sect0

[doomzday]

I can't say thanks enough guys. I started browsing these forums years ago after tossing in a 60mm delta black label fan. It didn't help much with overclocking, and it was the last loud fan I ever owned.

I will diffently try to pick up a meter and OCCT PS stress test the system.

I will probably pick up a new psu anyways. Im kind of jonesing for a new one, after reading these new reviews. A safe bet to say is I do NOT need a killowatt psu hahahaha.

I have had this hx520 sence they released them, due to reading the forums here. Its been a great solid power supply, and I have another system she will get tossed in.

[audiojar]

Take a look at the thorough PSU fundamentals article. On page 4 it lists some actual measured draws of some example systems. The highest end system draws 256W under full CPU and GPU load. Granted, it only has one video card, but that card is among the highest wattage and the CPU is rated at 130W (they are even over-volting it 10% too). There is no way a 2nd card is going to add enough to cause problems. Not to mention that actual games aren't going to draw as much as a full stress test.

From what I can see it seems that the evidence shows that your system will max out at well below 400W, and probably much less. If anybody has evidence of a comparable i7 SLI system drawing more I'd be happy to see it.

[b_rubenstein]

Follow the link I included above.

It makes no sense to run a power supply anywhere near its rated output. Once the load goes much above 60% it starts to heat up and the fan starts spinning up and will become audible. With dual cards that system can draw close to 500w playing a real game. Current game play can reach 70% of the power draw that the stress tests will.

Just because you can do something doesn't mean that you should.

[audiojar]

That completely depends on the efficiency curve. Many quality PSUs maintain high efficiency up to near max load. If that's the case then it wont heat up any more than a higher rated PSU with the same efficiency at that draw.

That link is lacking in a lot of specifics. For instance, they did the testing using a "special power supply" of unspecified efficiency. They don't say whether the measured draw is AC in or the actual DC. Those sorts of things matter.

[b_rubenstein]

It's DC power. They have a whole piece on their site explaining how they test power consumption of MBs and video cards.

It's very poor engineering practice to choose a piece of equipment that may wind up running at or possibly beyond its rated capacity.

A PS that is 80% efficient delivering 500w still has to dissipate 100w of heat. Do you know what the ambient temp is for that PC? Do you know the intake air temp of the PS will be in that case? Do you know what temp will be if the components in the PS? Have you calculated how much the life of those components should be de-rated based on those temps?

[boost]

I think the PSU is too weak for heavy oc and 5850 CF.
Here are some power figures for a very similar system:
http://www.xbitlabs.com/articles/video/ ... html#sect0
This is the test setup:
http://www.xbitlabs.com/articles/video/ ... html#sect0
The system draws 450 watts ingame and over 615 when fully stressed. If you add 20% to that figure for future upgrades I would recommed a 750 watt PSU.

[stromgald]

I think the PSU is too weak for heavy oc and 5850 CF.
Here are some power figures for a very similar system:
http://www.xbitlabs.com/articles/video/ ... html#sect0
This is the test setup:
http://www.xbitlabs.com/articles/video/ ... html#sect0
The system draws 450 watts ingame and over 615 when fully stressed. If you add 20% to that figure for future upgrades I would recommed a 750 watt PSU.

It's not clear in the article, but I'm 90% sure that they're measuring the system power draw at the power outlet (i.e. the PSU input). PSUs are rated at the output (i.e. the mess of cables within the case providing DC power). So, assuming an 80% efficiency, the real draw of the system in gaming would be 360W and 492W when fully loaded. So, a 520W PSU would probably work fine (albeit cutting it close when running FurMark + LinPack64).

Unless you're planning on doing alot of bench testing on your computer, I think you might be able to scrape by with the 520W unit. Corsair also has a 620W unit in the same product line if you're partial to the corsair hx line and want to upgrade. Depending on how long you hold it at those high power draw levels could cause heat buildup in the case and PSU fan ramp-up, but that depends more on the PSU's heatsink sizes and fan than the actual power electronics.

With all that being said, I think your best bet is still to buy a Kill-a-watt meter and test it. Please be sure to post your results.

The only website that I know that tries to break down power draw by component (i.e. internal draw) is SPCR, and they make it very clear what they're measuring. All other sites I've seen just use a Kil-A-Watt meter at the outlet.

[boost]

The power measurements from Xbitlabs are NOT at the power outlet.
Please read the article and the links on testing equipment before drawing a conclusion. The power figures from Xbitlabs are the most detailed I could find anywhere. Here's a link to the testing equipment for your convenience:
http://www.xbitlabs.com/articles/cooler ... html#sect0
It's true you might be able to "scrape by" with your current PSU, I recommend at least a good 750 watt power supply, if you intend to upgrade.

[stromgald]

My mistake. I went to their Testbed and Methods page of the article and they didn't mention anything about their power testing methodology.