• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

How many watts do you really need?

M

M4573R

Gawd
Joined
Jul 30, 2006
Messages
522
I'm planning on building a whole new machine, and I'm curious how much power I really need. Say I get an almost top of the line intel processor, almost top of the line nvidia card, a couple dvd drives, and 1-2 harddrives, sound card, etc.; How many watts do I really need? What if I went SLI?

Thanks
 
so many variables. i am not one to suggest anything to you, but let me just say that you cant say "maybe a couple of these, a couple of those, perhaps sli..." etc. you need to know what parts exactly you want before you can figure out how much power you "need." but even then, you will limit your upgradability (yes thats a word:p), and you will limit your PSUs life span. it will also make you PSU work harder, which means hotter and louder. asone guy said recently, "there is no such thing as too much power." for what you seem to want (i7?) it sounds like anything around 750-1000W should be good. make sure you buy a quality maker, thats much more important than lights and modularity etc that just tack tons of money onto your total. make sure its 80+ certified as well. to find out how much wattage you need minimum, you can use the psu calculator in the FAQ sticky once you know whats going in your computer.
 
I kind of threw in the last things assuming the processor, video card, and harddrives take the most power.

As far as the manufacturer, there's a lot of good reviews on many power supplies, but no matter what PSU I pick, there's at least 30 reviews on newegg saying it dies very quickly or was DOA. When I built my current PC, my silverstone 750 exploded two weeks after use and fried just about everything in my case. I've veeery cautious of PSUs now.

Edit: The calculator gave me 544 watts. I'd assume that's total power suckage and I'd need 750-1000 like ekuest said.
 
yeah sounds good. i had a cheap TT 430W psu powering my rig with a big OC and tons of peripherals for 3 years. i recently upgraded to a 650W antec earthwatts. i cant stress enough how much i loved that TT. i put it through everything, it came with sleeved cables, had all the connectors i needed, just wonderful. but most people would never even consider it. i started getting random shutdowns with the EA650 soon after installing it. im sticking with it though cuz i like the extra overhead and it seems to have sorted itself out. anyway what im trying to say is that psus are the biggest gamble in your system. every brand has failures here and there. so find one that is recommended by [H] (FAQ sticky) or by a good number of knowledgeable [H] users here in the psu forum, and go with one that has a good refund/rma policy. dont trust newegg review as most people who complain on there are idiots and probably caused the psu failure themselves.
 
I'm thinking the corsair 1000HX. Pretty pricey for my taste, but it seems to be one of the best.
 
M4573R said:
I'm thinking the corsair 1000HX. Pretty pricey for my taste, but it seems to be one of the best.
Click to expand...

For only a single graphics card, it is more than you need. More than you need IS NEVER a bad thing, but you just may end up spending more than you need to spend.

For SLi/Crossfire GTX 285 or HD4890 quality 750W-900W is the comfort range, 800W/850W is probably about a sweet spot.

For SLi/Crossfire...if you get HD4870X2, or GTX295 then quality 850W-1000W would be right around your comfort zone for something that doesn't go absolutely insane with TECs and overclocking.
 
Last edited:
i have a friend that runs an i7 setup water cooled with double rads 5 HDD's in raid and 3 8800 GTX's in sli and with a kil-a-wat meter he measured only 600W at load. he runs everything on a pc power and cooling 750 i really believe that as long as you buy a name brand PSU that 750-1000 would be all that anyone would ever need.
 
M4573R said:
I kind of threw in the last things assuming the processor, video card, and harddrives take the most power.

As far as the manufacturer, there's a lot of good reviews on many power supplies, but no matter what PSU I pick, there's at least 30 reviews on newegg saying it dies very quickly or was DOA. When I built my current PC, my silverstone 750 exploded two weeks after use and fried just about everything in my case. I've veeery cautious of PSUs now.

Edit: The calculator gave me 544 watts. I'd assume that's total power suckage and I'd need 750-1000 like ekuest said.
Click to expand...

calculators can give really inflated numbers. and as MrF said you should only use it as an estimate. if you list the components you have or are going to use that would help us get a better idea of what you need.
 
dunno if I can helf but online calculators sucks

system
e7200 3,75ghz 1,3V
2x1gb ram ocz stock
gb ga-g31m-s2l
asus 9600gt TOP
1x 3,5" samsung F1 drive
1x 2,5" toshiba drive
enermax liberty 400W

systems was eating on idle 110W and on load 165W (without gpu 70/100W)

the same on q6600 65nm 3,0ghz 1,2V 211W

system with e7200 nd gpu works on matx/sfx 270W psu good
 
Power supplies tend to peak in efficiency around 50% load, so shoot for double what you think your system will draw at max load, maybe tack an few extra watts on there for upgrades. For average systems with a top end card these days, I'd say a quality 500 watt should be good. For SLI or OC'd systems bump it up to 650 or higher.
 
One thing to separate is what you need and what you ought to get. Computers don't often need nearly as much as people think. Very rare to find a computer that can actually draw 1000 watts, you need an amazing amount of parts to do that.

However what you should get is more than what you need. In general, I'd say aim for around double. Why? Four major reasons:

1) Plenty of headroom if you are wrong. If you calculate you need 500 watts and are off by 20%, well no problem if you bought a 1000watt PSU. However if you bought a 550watt PSU, well you are now over loading it.

2) PSUs often aren't designed to be operated at their max wattage long term. Good ones, like the Corsair HX1000 are and you can have it pump out 1000 watts all day long. However for others the max rating is more like an ideal max, and shouldn't be sustained. So good to have additional headroom for that reason.

3) Most PSUs operate at their ideal efficiency at about half their rated load. As you increase load past that, efficiency drops. Now it isn't a huge amount, a few percent, but why not size the PSU to operate as efficiently as it can? No reason to under size it and have lower efficiency.

4) Noise. You discover that as the PSU gets a higher load, its fan has to work much harder. In the case of the Corsair supplies their fan sits at a minimum speed until right around 50% load then starts to rise rapidly.


So it is an area that overspecing in is a good idea. Don't go crazy and get a "more is better" attitude and buy something unreasonably huge, but get more than you need.

Also something else to think about is if you get a high wattage modular supply, you get load of connectors which is nice. I've got an HX1000 and there's just tons of connectors, which means no matter what kind of things I put in my system, I've got cables for them without splitters/converters/etc. For example on some PSUs you might find that they only have 2 PCIe cables. They've got sufficient amps on the 12v rail to handle 2 GPUs, but only 2 built in PCIe connectors. So if you want to do SLI, you are using molex to PCIe converters. Not a problem on the HX1000, it's got 6 PCIe connectors total.

I do recommend modular cables in that case though, because otherwise you'll have tons of unused cables cluttering up your case. Like in my case I only have a single GPU, so I only need 2 PCIe connections. I'm happy that I can leave the other 4 off and save room. I do, however, have a bunch of harddrives, fans, and a few odd things like an X-Fi elite Pro that need connections so I can easily knock on cables for all that and not need to use splitters, which keeps the clutter down.
 
meh. im not big on modular. perhaps if i had a 1000W like that corsair and was running tri sli etc. but for a simple build i find that about 15 mins and a little thoughtfulness can make a non-modular psu look perfectly good, and not have a $40-$80 premium. on my sff build i actually plan on cutting some wires off or else sticking a couple inside the psu(outside of where the air flows). who needs 10 molex connectors? :p
 
Just go to a couple of sites like Tom's Hardware. Occasionally they do articles about testing a single component but throw out information like energy consumption.

You should always keep in mind if they are talking about consistent load or peak/spike load.
 
i use an OCZ EliteXstream 800W psu for my PC, i originally bought an OCZ GameXstream 700w psu and kept having troubles with my 3870 X-Fire setup (it would crash to desktop then reboot automatically) after replacing my PSU 4 times, I just asked to be upgraded to a better PSU, and OCZ gave me their top of the line PSU for free :)
 
M4573R said:
I'm planning on building a whole new machine, and I'm curious how much power I really need. Say I get an almost top of the line intel processor, almost top of the line nvidia card, a couple dvd drives, and 1-2 harddrives, sound card, etc.; How many watts do I really need? What if I went SLI?

Thanks
Click to expand...

A ~500 Watt PSU is enough for most high-end systems. 750 Watt if you use SLI/XFire.

Mainstream systems are generally fine with about 300-400 Watt :) According to the UPS of the main company server (3800+ X2 EE, 8500 GT, 4 GB), it idles at around 60 Watt, barely touching 100 Watt load. Admittedly the GPU isn't used much :)
 
I'm running a coolermaster real power pro 1000w I bought for $160 (after $40 rebate) at frys last october. Still going strong with no issues even after upgrading from a wolfdale and SLI geforce 260s. As for what it's supply power to now, see sig. One thing though, if you go for anything over 650, prepare for one heavy ass case! :D
 
Last edited:
If I got a whole new system here's what I planned out:

i7 920
3x2gigs ddr3 1600
3 WD 640GB caviar blacks
gtx285 ssc
1 dvd drive & 1 dvd burner
My left over x-fi extreme gamer
case with 6 120mm fans
 
MrF said:
You can get rough estimates using online calculators.
This is an example.
And I cannot emphasize enough that this should be looked at only as an estimate.
http://www.thermaltake.outervision.com/Power
Click to expand...

we don't take kindly to TT numbers round here:
http://www.thermaltakeusa.com/Product.aspx?S=1332&ID=1863
16dBA @ 60CFM? My ass.

Anyways, I've got my fare share of volt meters and surge protectors and I can tell you I've never seen a unit draw more than 550 VA. That said, I've never had more than 2 GPUs in any one system.

I firmly believe that a solid 750W unit will last as long as 12V and 5V are the standards. the Corsair 750W units fling themselves of the shelves, and for good reason, its a good unit at $100 after $20 MIR.
 
M4573R said:
If I got a whole new system here's what I planned out:

i7 920
3x2gigs ddr3 1600
3 WD 640GB caviar blacks
gtx285 ssc
1 dvd drive & 1 dvd burner
My left over x-fi extreme gamer
case with 6 120mm fans
Click to expand...
A good 600W+ unit would be plenty for that setup. If you want to go SLI in the future, I would suggest at least 750W.
 
For what it's worth, I run a ton of stuff on a fairly budget power supply. I only paid $60 or so for my Earthwatts 650 and between my Overclocked quad, 10 hard drives including scsi, crossfire, etc I put plenty of load on it.
 
Far and away the best thing to do is to get a Kill-A-Watt. put what every PS you have in a new system and take some power measurements. It's a good idea to make a measurement at itle and load. For load you can just run and instance of a program like Orthos and something to stress the video, like ATI Tool. The CPU can be stressed harder, but no real application will stress things harder than Orthos.

One thing to remember is that the Kill-A-Watt measures the draw from the outlet, not how much power is being delivered by the PS. If you are testing with an 80+ PS, then multiply the Kill-A-Watt reading by 0.80. If it's not 80+ use 0.65 - 0.70. Power supplies are rated for how much power they can put out, not take in.

Once you have a reasonable idea of what the actual min and max power loads of the system are, then you can make some decisions on picking a PS. Here are some factors to consider: efficiency, reliability, noise and future upgrades.

Efficiency is a bigger factor if the computer spends most of its time at idle or running undemanding tasks. PS efficiency drops quite a bit below a 20% load so you wouldn't want to pick a PS that spends a lot of it's time putting out 5 - 10% of its max output.

Reliability is a big factor for most people and reliability is better if the PS isn't run near its rated capacity for extended periods. If you game for hours on end, then it's probably a good idea to get a PS that's rated at double your max draw. If it's an hour or less at a time, then it's probably safe to get a PS that's running at 75% of load.

One thing that may have an impact on sizing the PS is noise. Just about all PS's have active cooling fan control and there is a load point where they start speeding up the fan (varies by PS model). If having a quiet computer is a requirement, then you probably don't ever want the PS delivering more than 50% of its max output.

Computer upgrade reserve capacity is really more of a short term ( 1 - 2 year) factor than a long term one. Short term you're dealing with the same circuit technology, so if you get a video card that's twice as powerful then it will use twice as much power. Long term, circuits have gotten more efficient so in 5 years you can probably get twice the performance with a minimal power increase. If you know that you will probably upgrade factor that in, otherwise ignore it. In 5 years you may want a new PS for any number of reasons anyway. There's no point in paying for the electricity used by the inefficiency of a way over sized PS for years and years.
 
M4573R said:
If I got a whole new system here's what I planned out:

i7 920
3x2gigs ddr3 1600
3 WD 640GB caviar blacks
gtx285 ssc
1 dvd drive & 1 dvd burner
My left over x-fi extreme gamer
case with 6 120mm fans
Click to expand...

600W from a good brand will be more than enough.

I have an OCd E8400 and a 55nm GTX 260 and I draw at load AT THE WALL SOCKET around 290W. And that's running Orthos and Furmark at the same time, which is totally unrealistic as far as real world usage is. With my high efficiency PSU (Enermax Modu82+ 525W), that's roughly 245W DC, less than 50% usage. I'd add maybe 50W for your CPU, 40W for your GFX card, 20W for your HDDs (on top of what I have) and you peak at 370W, let's call it 400W to be on the safe side.
Best usage of a PSU is to have a peak (like a benchmark peak) somewhere between 60% and 80%, so 600W-ish from a good brand will be no problem, with at least 200W to spare.

Don't listen to the testosterone-fuelled call for a 1000W+ PSU, because you'll idle (which is a large part of your CPU time) way below its efficiency band.
 
NichoTL said:
I have an OCd E8400 and a 55nm GTX 260 and I draw at load AT THE WALL SOCKET around 290W.
Click to expand...
I'm guessing you're using something like a Kill-A-Watt to measure your power draw, and Kill-A-Watts are not accurate with APFC PSUs. Although it is possible that your power draw is around that level, I would estimate something a bit higher considering your overclocks.
NichoTL said:
Best usage of a PSU is to have a peak (like a benchmark peak) somewhere between 60% and 80%
Click to expand...
Between 40% and 60%, actually.
 
Zero82z said:
I'm guessing you're using something like a Kill-A-Watt to measure your power draw, and Kill-A-Watts are not accurate with APFC PSUs. Although it is possible that your power draw is around that level, I would estimate something a bit higher considering your overclocks.

Between 40% and 60%, actually.
Click to expand...

Okay I should change my statement by 50% to 70%.
My PSU is at its best efficiency at 50%-ish. So by aiming for a BENCHMARKING peak around the 50% to 70%, I'm looking for my real-world (read gaming) peak around the 50% mark. Maybe that's too convoluted, because I could just take my gaming peak and aim for 50% in the first place.

As for the accuracy of my power meter (Kill-A-Watts for you Yanks), it's kind of hard to measure. My calculation allow for a 10% error margin anyway.
 
Zero82z said:
I'm guessing you're using something like a Kill-A-Watt to measure your power draw, and Kill-A-Watts are not accurate with APFC PSUs.
Click to expand...

Active Power Factor Correction (AFPC) has nothing to do with efficiency. The Power Factor of a load relates to keeping the voltage and current draw in phase. This has an effect on the power company, so they factor this in for commercial customers, but has no effect on billing for residential customers. Just because the Kill-a-Watt isn't very accurate for measuring high power factors, doesn't mean that it's also inaccurate for the power being drawn; they are two different things. Obviously, for $20 you're not buying a lab grade instrument.
 
BFG800 so far has been no problem to me and it was cheap (I think I paid 130)

It is able to handle SLI + physic + another card, OC920 and whatever else I have in my machine.
 
If you are going to get top of the line stuff. Go for a good 700W power supply.

For my current system my old 500W (which had a total 12V rail of 456W) was not enough when I changed my 8800GTX to a 280GTX

My existing setup is as follows.
Core 2 Duo E6600
3 x 120mm fans
P965 Mother Board
280GTX
2 x SATA 7200 RPM Drives
 
blivit4 said:
Active Power Factor Correction (AFPC) has nothing to do with efficiency.
Click to expand...
I'm well aware of that. However, that doesn't change the fact that Kill-A-Watts often provide incorrect power draw readings when used with APFC PSUs.
Paul_Johnson said:
1) APFC can fool Kill-A-Watts into giving you abnormally low readings (some times giving better than 100% efficiency)
Click to expand...
 
Zero82z said:
I'm well aware of that. However, that doesn't change the fact that Kill-A-Watts often provide incorrect power draw readings when used with APFC PSUs.
Click to expand...

There is nothing in a Kill-a-watt that measures or calculates efficiency, because it can't be done by only taking measurements of what is going into the PS. Efficiency is the ratio of how much DC energy comes out to how much AC energy go in.

Take a look at how they test PS's for review on this site. It isn't done with a single instrument.

FWIW, I don't think I've seen any 80+ PS that didn't have APFC, so it stands to reason that more efficient PS's draw less energy than inefficient ones. It's the design of the power supply that makes it more efficient, not the APFC that throws off the measurement.
 
blivit4 said:
There is nothing in a Kill-a-watt that measures or calculates efficiency, because it can't be done by only taking measurements of what is going into the PS. Efficiency is the ratio of how much DC energy comes out to how much AC energy go in.
Click to expand...
I know what efficiency is, and I'm quite aware that Kill-A-Watts cannot measure it. However, Paul_Johnson (who you may know as the person who does the PSU reviews for HardOCP and is quite knowledgeable) has done testing with Kill-A-Watt meters and found that they can provide incorrect readings when used with APFC PSUs (hence the quote in my previous post). Presumably, he discerned this by placing a fixed load on a PSU and comparing the readings given by the Kill-A-Watt to the DC load value as well as to proper power consumption readings provided by a real power meter. I am inclined to believe this considering I have seen many reviews where power consumption was measured with a Kill-A-Watt and the actual consumption numbers were quite far off compared to values measured with proper equipment.
 
My primary system in sig pulls 700W under furmark, which loads only GPU. If I load both GPU and CPU, I think I'll break 800w... Granted though, I'll never see the kind of GPU load furmark generates, just like linx for CPU, but who knows ;)
 
TehQuick said:
My primary system in sig pulls 700W under furmark, which loads only GPU. If I load both GPU and CPU, I think I'll break 800w... Granted though, I'll never see the kind of GPU load furmark generates, just like linx for CPU, but who knows ;)
Click to expand...

800 watts? Wow and we have simular setups. I never knew I was even getting close to maxing out my PSU. Good to know.
 
Zero82z said:
I know what efficiency is, and I'm quite aware that Kill-A-Watts cannot measure it. However, Paul_Johnson (who you may know as the person who does the PSU reviews for HardOCP and is quite knowledgeable) has done testing with Kill-A-Watt meters and found that they can provide incorrect readings when used with APFC PSUs (hence the quote in my previous post). Presumably, he discerned this by placing a fixed load on a PSU and comparing the readings given by the Kill-A-Watt to the DC load value as well as to proper power consumption readings provided by a real power meter. I am inclined to believe this considering I have seen many reviews where power consumption was measured with a Kill-A-Watt and the actual consumption numbers were quite far off compared to values measured with proper equipment.
Click to expand...
QFT.

blivit4, Zero82z is pointing out a major technical problem with Kill-A-Watts: When used in conjunction with a PSU with APFC, you get highly inaccurate readings. He is not talking about the efficiency of a PSU whatsoever. Just that device commonly used to measure power draw can be very inaccurate.

EDIT: Here's Paul Johnson's post about the inaccuracy of the Kill-A-Watt (the proof to Zero82z's presumption):
http://hardforum.com/showpost.php?p=1032190998&postcount=7
 
Last edited:
Danny Bui said:
QFT.

blivit4, Zero82z is pointing out a major technical problem with Kill-A-Watts: When used in conjunction with a PSU with APFC, you get highly inaccurate readings. He is not talking about the efficiency of a PSU whatsoever. Just that device commonly used to measure power draw can be very inaccurate.

EDIT: Here's Paul Johnson's post about the inaccuracy of the Kill-A-Watt (the proof to Zero82z's presumption):
http://hardforum.com/showpost.php?p=1032190998&postcount=7
Click to expand...

Hmm it only proves the Kill-A-Watts problem. How can I tell if my power meter (different maker, different country, different everything) works the same?
 
Danny Bui said:
QFT.

EDIT: Here's Paul Johnson's post about the inaccuracy of the Kill-A-Watt (the proof to Zero82z's presumption):
http://hardforum.com/showpost.php?p=1032190998&postcount=7
Click to expand...

Danny - Your point would have been better served if your reference was to that entire thread which not only provides a context, but also some other useful posts.

Within that thread Paul makes no assertion that the inaccuracy of the Kill-a-Watt is due to APFC. In fact, his example refers to a PS with 62% efficincy that almost certainly didn't have APFC. If if a Kill-a-Watt isn't all that accurate, JonnyGuru still thinks it can be useful for sizing power supplies (post #14 in that thread).

Back on point: this current thread makes mention of efficiency and effects of APFC. A Kill-a-Watt makes no efficiency measurement. There is nothing in that other thread that connects APFC to inaccurate power measurements. Unless someone here can reference some valid data linking APFC and Kill-a-Watt inaccuracies, I'll just be of the opinion that the Kill-a-Watt isn't real precise, but useful anyway.
 
You must log in or register to reply here.
Back
Top