How much power do you really use?

[LxMxFxD4]

For kicks and giggles I recently purchased a Kill-A-Watt Electricity meter from newegg. It set me back $23 but I just tested how much electricity my computer (and everything else in my house) eats. Below are the results:

My Setup

Computer:
Antec Earthwatts 430, 430W PSU, 1 week old
AMD X2 3800+ @ 2.4ghz
A8N-SLI Premium Motherboard
2x1GB (2GB) PC3200 Ram
4x300GB Maxtor SATA Hard drives (old)
1x160GB EIDE Seagate HD (very old)
8800GT eVGA 512MB 600mhz/1800mhz, default
Envy24 Generic Videocard by Via
NEC DVD Burner
Big Typhoon Cooler
120mm case fan @ 7v
2x 80mm case fan @ 5v
Boston Accoustics BA-4800 4.1 (circa 2000)
Dell 24" 2405 Monitor
Samsung SyncMaster 191T (suuuper old)

Networking:
Time Warner Cable Modem
DLink 4100 Wireless router, 1000mbit
netgear switch fs105, 100mbit
Brother HL-5250 Networking laser printer
Uniden PowrMax Wireless phone & VoIP Adapter

Lighting:
40W Florescent bulb
Lamp with 60/100/150W bulb (multi point)


Results
Computer at idle (HDs spinning but 0% cpu, monitors off): 150W
Computer at 100% CPU (2 instances of Prime95, max power setting): 209W
Computer at 100% CPU with video card running at max (2 instances of prime95, 3dmark 2006): 251W
Computer during boot up, max spike: 260W

Dell 24" LCD: 50W
Samsung 19" LCD: 31W

Whole system setup @ 100% load (computer @ 100% w/bioshock running, monitors, networking, printer at idle, speakers, etc): 377W

So I think this solves the issue of whether or not you really need a 600W power supply, and also what kind of UPS you'll actually need in case of power outage.

Intel C2D may use more power, I don't know, but for an X2 system you could easily get by on a 300W (true) supply.

 

[Zap]

Originally posted by: LxMxFxD4
So I think this solves the issue of whether or not you really need a 600W power supply,

That's been settled a long time ago, just that many "enthusiasts" steadfastly refuse to believe it. Kind of like people believing 6" tailpipes automatically give more HP than 4".

See this recent thread.

 

[LxMxFxD4]

Originally posted by: Zap

Originally posted by: LxMxFxD4
So I think this solves the issue of whether or not you really need a 600W power supply,

That's been settled a long time ago, just that many "enthusiasts" steadfastly refuse to believe it. Kind of like people believing 6" tailpipes automatically give more HP than 4".

See this recent thread.

Zap, you seem to be quite helpful on these forums. Looks like this guy's power usage with the kill-a-watt jives with mine. If only I had seen that thread prior to ordering the device I'd be $23 richer. Oh well, its a fun toy.

As for whether i was revealing any esoteric, grandiose wisdom with my post, I didn't think I was. Hehe, but my conclusion has obviously been replicated!

 

[Zap]

Originally posted by: LxMxFxD4
my conclusion has obviously been replicated!

Obviously great minds think alike.

 

[MarcVenice]

But erm, how about I don't want to run my PSU at maximum capacity ? If it needs 250w to play games, then I'd like a 400w PSU. Not a 300w PSU.

 

[jonnyGURU]

Originally posted by: Zap

Originally posted by: LxMxFxD4
So I think this solves the issue of whether or not you really need a 600W power supply,

That's been settled a long time ago, just that many "enthusiasts" steadfastly refuse to believe it.

Sure, I can run a single 8800GT with a quad core on a 450W or 500W PSU no problem. I can run a pair of GTS 640MB with a quad core on a 600W PSU. No doubt. I can even run a quad FX with a pair of Ultra's on an 800W. Yes sir.

The EarthWatts is a good PSU with ample power on the +12V and a single 8800GT doesn't need much power, so it's not a good example to use for this "argument"... the OP has adequate power for his rig. But since you're using this as a stepping stone for opening up this can of worms, I'm going to play devil's advocate and throw my hat into the ring, because the fact of the matter is, there are still more pros to using a higher power power supply with ample overhead than there are to using a power supply that provides "just enough" power. Usually, is comes down to cost. Smaller PSU's are cheaper than big ones. But I don't like when people tell me how to spend my money, so that argument goes right out the window....

So....

1. Kill-A-Watt is inaccurate. We've seen many examples of where using a Kill-A-Watt as a gage yields efficiency number of > 95% when used the DC load is known.

2. Transient loads. Even very good data logging DC current measuring devices can't catch every transient, but when an accumulative sample of DC current is taken and compared against an accumulative AC sample, the DC sample is often greater than the AC sample and you and I both know that's not possible.

3. Rail distribution. Although the unit may only be pulling 260W from the wall (not including any un-measurable peaks for the sake of argument) without measuring the DC usage, do we know how much of that is on the +12V?

Yes... It's all been said before, by me and others, but if you're going to rail for the argument of using a PSU you believe to be well capable of powering the rig it's powering without taking into account the long term stress on that PSU, I'm going to have to challenge it each time.

Originally posted by: MarcVenice
But erm, how about I don't want to run my PSU at maximum capacity ? If it needs 250w to play games, then I'd like a 400w PSU. Not a 300w PSU.

Right.

What if I don't want my PSU to....

Have it's internal components running at the upper level of it's operating temperature?
Have it's fan running at the fastest, noisiest point of the fan curve?
Have it's DC output at it's greatest ripple and noise possible?

PSU's, even those rated at only 25°C, actually tend to run well within their components thermal properties. 85°C caps vs. 105°C caps? Pulll-eze... marketing. Caps don't even get CLOSE to that hot in a PSU. It's the long term exposure to higher temperatures, even if they're within spec, that eventually kill a PSU. Even high ripple and noise can kill the PSU itself and not just the components it's powering. Everyone publishes their MTBF at 25°C. Normal operating temperature cuts that MTBF in more than half and anything hotter even more so. So even if the unit and all of the components within are running within their specification, they're not going to last forever.

 

[gingerstewart55]

Doesn't matter, Jon. Despite posting numerous videos where you show the Kill-a-Watt showing horribly inaccurate current draws from the wall as compared to the measured DC output from the power supply....others will continue to wave Kill-a-Watt numbers as the be-all and end-all holy grail.

Personally, I want my power supply to run at no more than 50% of its rated output...if I can, anyway.....which is the sweet spot for almost all power supplies in efficiency, quietness, temperature, and noise and ripple production from the ps. So I've got a Kw unit in my computer......Asus Maximus Formula mb, 2 x 2GB memory, Q6600 cpu, 4 hd's (two Raptors in RAID 0, two Seagate 500GB hd's for storage), two 3870's in Crossfire, two Laing DDC2 pumps in a dual loop water cooling setup (mb and NB/SB on one loop, video cards on second loop).....and not to forget the requisite pair of optical drives or the six 120mm fans running on the two radiators and the three 120mm fans in the case, and the Antec Spot Cooler fans for memory cooling.

My system is in no way an extreme system or unique situation......and I shudder to think of trying to run it on a 400 or 500W power supply. It'd probably boot with a 500W ps, but it'd stretch that ps to its limits, and I don't want that.

The 1Kw unit I'm using runs very cool, has probably a better internal build than a comparable 500W ps, and allows me to upgrade any components inside my computer without even thinking about if I have enough power to get it running.

You want to run "just enough" of a power supply.....great. I'll run more than enough and not stress my ps into an early death.

Interesting, though, how your reply, Jon, was responded to by the "experts" here..........

 

[jonnyGURU]

Interesting thread:

http://forums.anandtech.com/me...=2147352&enterthread=y

 

[Yellowbeard]

There are so many daily comparisons that you can make to why people might "WANT" a larger PSU than some chart or friendly forum guy may recommend. We see it every day in other aspects of life but, some people take a militant stance about telling others what they should and should not buy.

For example, let's say you have some spare money and build a nice 2 seat muscle car with 500hp. Does anyone "need" 500hp on the street? No, but it's fun to have even when the car is parked in the garage. But, some guy in his Dodge K car will be sitting next to you at the light berating your decisions based on his selection criteria. Then, you'll have a family man in a minivan also shaking his head. But, the hot rod guy does not care what these 2 think.

Does a deer hunter "need" a 7mm magnum to kill the average white tailed deer? No, if you are a good shot you can do it with a .243 or even a .223 which = an M-16 caliber Nato 5.56mm round. But, some guys think size matters. Others will insist that you only hunt with a bow or knife to make it more sporting. I guess some guys would even hunt deer with a claymore mine if allowed. I guess it depends on your resources and how hungry you are.

I overclock so I want the biggest baddest PSU I can reasonably afford at the time of a system build. I don't care about noise, heat, or efficiency when OCing and I want to be sure I have more than enough power to get the job done when I am aiming for a higher clock. But, I still think like a consumer so I take budget into it at the same time. In a 24/7 system, I aim for a PSU that will max at about 65% of its rated output.

I think that there are some fundamental rules people should pay attention to and then make their own decisions. It's not a good idea to run a PSU at or above a certain percentage of its rated power. IMO, I don't EVER want to go above 70% but that's just me. 65% is ideal. 80% or less is smart, above 80% is asking for trouble.

Also, just because you can run a lower rated PSU, don't go cheap. The chances of catastrophic failure are greater when you are pushing a cheap PSU really hard vs a quality PSU under the same conditions.

Then, there is always brand loyalty, features, and bling to consider. IMO, just stay away from cheap crap and buy what you can afford. But, don't take shaky information and try to convince others that their opinion is wrong. Computing is a hobby for many people and every decision does not have to be purely practical or pragmatic in nature. It's really odd that PSUs get the most criticism when they are the backbone of the system.

EDITED for Tipograffical airer.

 

[HOOfan 1]

Originally posted by: Yellowbeard

Does a deer hunter "need" a 7mm magnum to kill the average white tailed deer? even a .223 which = an M-16 caliber Nato 5.56mm round.

that is illegal in many states

 

[Yellowbeard]

Originally posted by: HOOfan 1

Originally posted by: Yellowbeard

Does a deer hunter "need" a 7mm magnum to kill the average white tailed deer? even a .223 which = an M-16 caliber Nato 5.56mm round.

that is illegal in many states

So are claymores and frag nades

 

[HOOfan 1]

the .223 aren't illegal, but shooting deer with anything smaller than .243 caliber is.

 

[Zap]

Originally posted by: jonnyGURU
there are still more pros to using a higher power power supply with ample overhead than there are to using a power supply that provides "just enough" power.

Right. However, how much "ample overhead" is enough? If full load is 80% of capacity? 60%? 40%?

Originally posted by: jonnyGURU
Usually, is comes down to cost. Smaller PSU's are cheaper than big ones. But I don't like when people tell me how to spend my money, so that argument goes right out the window....

That goes the other way too, not just with PSUs. I often see people request recommendations on hardware within a certain budget. Without fail someone (often more than one) will tell that person they "need" to buy something above their budget.

Originally posted by: jonnyGURU
1. Kill-A-Watt is inaccurate. We've seen many examples of where using a Kill-A-Watt as a gage yields efficiency number of > 95% when used the DC load is known.

I remember your videos. These days I pretty much assume with APFC PSUs that the device is giving out "wrong" numbers and that the number may be closer to DC power draw, not AC after conversion loss. IMO for a $20 device and knowing the limitations it is close enough for rough estimating.

Originally posted by: jonnyGURU
3. Rail distribution. Although the unit may only be pulling 260W from the wall ... without measuring the DC usage, do we know how much of that is on the +12V?

That is a really excellent point! I personally just assume that the majority (80-90%?) of power draw in a PC using "modern" components will be +12v since that's what the video card, CPU and fans will depend on. Heck, maybe even just assume 100% goes on +12v. That way the power usage gets overestimated a hair but according to our local PSU manufacturer representatives that isn't necessarily a bad thing.

Originally posted by: jonnyGURU
Yes... It's all been said before, by me and others, but if you're going to rail for the argument of using a PSU you believe to be well capable of powering the rig it's powering without taking into account the long term stress on that PSU, I'm going to have to challenge it each time.
...
Normal operating temperature cuts that MTBF in more than half and anything hotter even more so. So even if the unit and all of the components within are running within their specification, they're not going to last forever.

Again, where's the empirical evidence? Has someone actually taken a number of otherwise identical PSUs and run them to death under varying conditions? Also, how long do people really expect PSUs to last? If someone bought a 500W PSU five years ago and it still functions like new, would... should they still be using it today? Likely it wouldn't have enough +12v, wouldn't have any PCI-E plugs, wouldn't have any SATA plugs and efficiency would be "bad" compared to more modern units.

Longetivity is fine, but to some people there comes a point when it no longer matters. I tend to overclock and some people who don't overclock believe that overclocking can shorten the life of the CPU and motherboard. Are they right? Maybe. Again, no empirical evidence. However, would I care that a CPU that would otherwise last 20 years dies in 2 years due to overclocking when I can't seem to keep parts that long?

Originally posted by: Yellowbeard
There are so many daily comparisons that you can make to why people might "WANT" a larger PSU than some chart or friendly forum guy may recommend. We see it every day in other aspects of life but, some people take a militant stance about telling others what they should and should not buy.

What, you discovered that Rollo's back? :Q Oh nevermind, this isn't the Video forum.

Originally posted by: Yellowbeard
I think that there are some fundamental rules people should pay attention to and then make their own decisions. It's not a good idea to run a PSU at or above a certain percentage of its rated power. IMO, I don't EVER want to go above 70% but that's just me. 65% is ideal. 80% or less is smart, above 80% is asking for trouble.

I took the liberty of bolding. Of course a lot of what we're talking about is "IMO" for all of us as there is very little empirical data backing it up. Power usage, efficiency, ripple... sure. A PSU not being able to last a "normal" lifecycle due to running above XX% of output? Where's the evidence. Of course it is entirely logical and reasonable to assume that running a PSU at 99% of output all the time is not good, and that running a PSU at 30% of output is perfectly fine, but where's the cutoff point? Well, there isn't one, is there? So, best thing "IMO" is for people to determine their own "comfort level." Yellowbeard, yours is at 65-70%. Someone else may be at 80%. A third person may be at 50%. Who's right and who's wrong?

Anyways

My original post in this thread was a poke at rabid enthusiasts. Rest assured that I do believe in using sufficient PSUs, be it in quality or output - think of me as a "moderate." I do also take into consideration the limitations and expectations that people have. If someone wants to buy a kilowatt PSU for his midrange system because he has the money and likes the way it looks, so be it. If someone wants to run a high end rig with a high end GPU on a crappy $20 PSU, then I'll tell them they're nuts.

If someone asks what they "need" for a specific system, then I'll tell them what I believe it needs - and what I tell them can be thought of as a minimum, if they want more and can afford it, then go for it!

We really can't put any blanket statements on requirements, besides "don't use crappy units and don't use at close to 100% output." Afterall, what works for a $300 computer with integrated video is not going to be the same as a monster Tri-SLI setup.

BTW JonnyGURU, I still :heart: you! When's the next LAN? That X-Qpack2 system I sported at the last one now has a 600W BFG PSU in it, and as soon as it gets here (already shipped) I'll have a single BFG 8800GTS 512MB G92 to go with it. Do you think I have enough "overhead" in my PSU to meet your stringent approval?

 

[Yellowbeard]

Peak efficiency in the testing I have seen falls at about the 65% point on the bell curve. That's why I like that point. Of course 60% or 70% are not huge deviations from that. So, it's not just a random "IMO".

I too have not hooked up a row of PSUs and run them at various loads for 5+ years to see what effect this will have. But, based on efficiency it is a solid number to work with as a starting point. I liken it to the RPM range and shift points in a car with a manual transmission.

 

[Zap]

Originally posted by: Yellowbeard
Peak efficiency in the testing I have seen falls at about the 65% point on the bell curve. That's why I like that point. Of course 60% or 70% are not huge deviations from that. So, it's not just a random "IMO".

Okay, that's a very good point. So if peak is at 65% usage, then idle would be... 45%? Depending on hardware?

 

[Fullmetal Chocobo]

270 watts idle, 342 watts loaded.

 

[Yellowbeard]

Originally posted by: Zap

Originally posted by: Yellowbeard
Peak efficiency in the testing I have seen falls at about the 65% point on the bell curve. That's why I like that point. Of course 60% or 70% are not huge deviations from that. So, it's not just a random "IMO".

Okay, that's a very good point. So if peak is at 65% usage, then idle would be... 45%? Depending on hardware?

LOL...I have not thought it thru that far but I would hope it would idle at a little less than 45%. Idling at 45% does not sound very "green" does it?