useless, but fun facts.. effect of overclocking on power use

Jerboi

Jerboi

Waasssup????
#1
Hiya, this is my second power consumption related thread. I just wanted to know how power consumption is affected when you overclock a CPU by 5%.

system

CPU: Intel Pentium III 600MHz, Katmai. 2.05V core voltage
Mobo: Abit BE6
HDD: Maxtor 8.4GB 5400RPM
Video: ATI Rage Fury Pro Vivo
CD: one generic 24x10x40 CD-RW drive
mouse: Logitech Optical
KB: Microsoft Natural Elite
230W generic PSU
160MB of generic PC100 memory at default latency


At the official frequency of 600MHz I got these results.

when idling at Windows desktop with no disk activity:

power consumption: 84watts

more details for the tech inclined people:
line voltage 122.5V
VA=136VA
power factor=0.62
current=1.11A

CPU fully loaded with Sandra CPU Multimedia test

power consumption: 93W
Sandra result: (3246 Integer, 3971 floating points)


When overclocked to 630MHz
Idle consumption= 86W

details:
line voltage=122.6V
VA=139VA
power factor=0.62
current=1.13A


CPU loaded in same manner:
power consumption: 95W
Sandra result:( 3402 integer, 4163 floating point)

Section II:
Power supply capacity calculation. A generally accepted figure for the typical computer power supply unit is 70%. At 95W power consumption at the plug, I am using 7/10 of input power at the output which turns out to be 66.5W. Looks like this computer is only loading this particular 230W PSU to 30% leaving me with 163.5W of unused capacity. I can see how pre-built computers with 140W and 90W power supplies runs fun. If this computer was to be build with 140W PSU, you still have 73.5W of power available. 73.5W is enough for several HDD's, a CD-ROM drive and a graphic accelerator.

My Athlon machine only use 90W at idle. I don't see how an 180W power supply can't be used on it. People always say you need at least 300 and preferebly 350. I don't know what this hype is all about.
 
Todd a

Todd a

New Member
#2
A power supply generates 12v, 5v, and 3.3v at differant AMPs. The Power Supply is rated on all the wattages added together. Athlons require a power supply rated for 160w between the 5v and 12v (I think) to run stable. Most junk power supplies are difficiant in one of these, usually the 5v. You got lots of power for hard drives and such, but the system still runs like crap. :)
 
grover

grover

Ex pert
#4
Dr. Mass- check out Jerboi's other power thread in the general forum- he's using a quite sophisticated piece of electronics :)
 
Jerboi

Jerboi

Waasssup????
#5
Originally posted by Dr. MASS
how did you get the Wattage results ? A multimeter ? lol
Multimeter is quite useless here since it can't properly calculate power factor and accurate current readings for distorted current input like computer loads.

I used this. Of course it's too expensive to buy it, so I'm borrowing it from a friend's dad. They call it "low cost" even though it's like $800 :mad:
 
marv117

marv117

VIP Member
#7
wait...so is IT or is IT NOT necessary to have 300watts of power???

cuz running a 300watt+ psu all day long does a job on our bill.

:confused: :confused: :confused:
 
Todd a

Todd a

New Member
#8
Extra is always better. The higher the ratting on the Power Supply the less you will strain it. I ran a genaric 250w power supply on my Duron [email protected] (1.725v) 100% stable, but it burned out the power supply after about 3-4 months. It likely overheated and a better fan would have kept it alive for longer, but it came in a case that I only paid $25 for.
 
D

DuronClocker

.
#9
Originally posted by marv117
wait...so is IT or is IT NOT necessary to have 300watts of power???

cuz running a 300watt+ psu all day long does a job on our bill.

:confused: :confused: :confused:
No...when you have a PSU rated for 600watts, its not always consuming 600watts. It only consumes what your computer needs it to power...usually around 200-250w. Sometimes, much lower when using low-power or low-speed equipment. So having a 300w PSU does not always mean its pulling 300w and putting 300w worth of cash on your electric bill. That just means that it CAN pull that much power IF NEEDED by the computer... hope this was clear enough to understand... PM me if it wasn't
 
Jerboi

Jerboi

Waasssup????
#10
Originally posted by DuronClocker


No...when you have a PSU rated for 600watts, its not always consuming 600watts. It only consumes what your computer needs it to power...usually around 200-250w. Sometimes, much lower when using low-power or low-speed equipment. So having a 300w PSU does not always mean its pulling 300w and putting 300w worth of cash on your electric bill. That just means that it CAN pull that much power IF NEEDED by the computer... hope this was clear enough to understand... PM me if it wasn't
The wattage of power supply usually tells maximum total power you can draw from power supply, but sometimes they mean the peak wattage.

Total wattage is less important than so called the "combined power".

Even though it doesn't make sense in English language, "combined power" is the term used to describe the power of3.3V and 5V busses together rather than the whole PSU.

Keep in mind that often time [email protected][email protected][email protected] exceeds rated wattage. If you have lots of 3.3V and 12V load, the chances are, you'll have to back off on 12V somewhat. You can't exceed the maximum current of any given bus, but at the same time you the sum of wattage being drawn must not exceed the total wattage.

A power supply rated at 350W and combined power of 180W means you have 180W of power available from 5V and 3.3V. Other 170W is at 12V, which is predominantly used by drives and fans.

Total wattage has became the accepted way of shopping for a PSU, because it doesn't take much brain to look at. Unless you're running data center with lots of drives, a Delta 300W with 200W of power available at 5V+3.3V is better than 350W total with 180W at 3.3V+5V.

Higher total wattage power supply tend to be able to handle surge current(not to be mixed with line transient), but a properly engineered 200W power supply won't have any problem powering up 99% of consumer computers.

Such 200W PSU you cost more than crappy 430W power supply and it won't be made, because false belief that more wattage=better is too detrimental to its sale.

As for relation between PSU wattage and power consumption, it is rather unrelated. The wattage at plug is about 43% more than actual power used by the computer.

So, if you load a 300W PSU to full 300W, you're taking 430W from the plug.

600W PSU fully loaded means 860W at the plug. High wattage power supplies are made, because people buy them. They will sell anything if it sells. Practicalness doesn't matter.

My Athlon 1.4GHz Tbird ig with two 7200RPM 40GB HDD, two CD-ROM drive and ATI 8MB video card only use about 90W of power from power supply. Add a GF 4, you might use another 40W. I highly doubt that you'll see the average consumer computer that use more than 200W of power from power supply.
 
grover

grover

Ex pert
#12
That said, I'm happy I bought a 430W Antec- not because I need 430W, but because I sure as hell don't want to have another PSU fail on me for lack of capability! It's not all that much more expensive to manufacture a 500W PSU as a 250W PSU; and that's a lot of peace of mind for those of us that do have 7 drives, 12 fans, and every PCI slot used ;)
 
U

uart

New Member
#13
how did you get the Wattage results ? A multimeter ? lol
Actually I've made quite a few power measurements on PC's using only a multimeter, and quite a cheap one at that. The results that I obtained were quite consistent with those posted above by Jerboi (and have also been confirmed using some more sophisticated equipment).

The trick is that the front end of a normal computer PSU feeds straight into a full wave rectifier (see note) and a cheap multimeter only really measures “proportional rectified average” (pi/sqrt(8) times rectified average current) rather than true RMS current. The upshot of this is that if you measure the AC input current with a cheap multimeter and then divide the result by pi/sqrt(8) (approx 1.11) then you have a very good estimate of the average DC current supplied by the PSU front-end full wave rectifier. As you know the DC working voltage of this rectifier (approx 1.4 time the RMS line voltage) then you can deduce the PSU input power to a reasonable approximation.

Before you lampoon this technique remember that even if you use the most accurate equipment on the input power measurements that the figure of 70% efficiency is only a “ball park” figure and it could easily be somewhere between 65% and 80%, so there are inherent approximations no matter which way you do it. BTW I’m not criticizing the 70% figure as I think 70% to 75% is pretty right.. :).

Here are the results for the estimated PSU output power when running Windows for two systems I have here at the moment.:

System 1 k6-3+ @ 550MHz, One HD, one CDRW, basic video card and sound card. : 45 Watts idle, 55 Watts full cpu load.

2. 1.1GHz Duron, One HD, One CDRW, integrated video and sound, AMR modem : 65 Watts idle, 77 Watts full cpu load.


Note : All the PSU’s I use here in Australia have the (240 volt) mains input feeding directly into a fullwave rectifier, this is the topology on which I’ve based my calculations.. Some of our power supplies also have a dual voltage (120/240 volt) selector switch. When these units are switched for 120V operation then the input rectifier is re-configured to function as a capacitor/diode voltage doubler/rectifier circuit. I’m just pointing out that I’ve only used the above technique for the case of a simple rectifier front end and have not analyzed whether it can work in the voltage doubler case.
 
Last edited:
Jerboi

Jerboi

Waasssup????
#15
Originally posted by uart


Actually I've made quite a few power measurements on PC's using only a multimeter, and quite a cheap one at that.
Actually I have done this in the past also and got a pretty good result, but it's quite a pain in the butt. As you've said, the power supply runs on 340V DC. If you're on 240V AC, you simply rectify it and you have 340V(240V x sqrt2). If you're on 120V, the voltage, the voltage is doubled then rectified to yield 340v. (120x2xsqrt2). Voltage multiplier takes peak to peak from AC and it only works if input is AC.

So what I did was, shoot the voltage up to 240V AC using a transformer, then rectify it to 340V DC using a 600V rated bridge and 5,000µF worth of capacitor bank. You need to charge the capacitor through a resistor on initial power up or else you'll blow the fuse or worse, you'll blow the bridge rectifier.

Next, you set your computer to 230V(just in case) and connect the two prongs to 340V DC and connect a multimeter in series. You should place a jumper lead across meter probes, then remove it once connection has been established, so you won't push the inrush current through the meter. If you let the inrush current through your meter, it could cost you an expensive 600V CAT III meter fuse($5-10 each).


The trick is that the front end of a normal computer PSU feeds straight into a full wave rectifier (see note) and a cheap multimeter only really measures “proportional rectified average” (pi/sqrt(8) times rectified average current) rather than true RMS current. The upshot of this is that if you measure the AC input current with a cheap multimeter and then divide the result by pi/sqrt(8) (approx 1.11) then you have a very good estimate of the average DC current supplied by the PSU front-end full wave rectifier. As you know the DC working voltage of this rectifier (approx 1.4 time the RMS line voltage) then you can deduce the PSU input power to a reasonable approximation.


Wow that's new to me. I never really understood how non true-RMS DMM's measured voltage. True RMS ammeter and voltmeter both have a little computer that samples the waveform thounsands of times a second and calculate the true RMS current on the fly. Although, input power factor varies(meaning that waveform varies) depending on the load to an extent and I have a feeling that this method will fail to compensate for that.


Before you lampoon this technique remember that even if you use the most accurate equipment on the input power measurements that the figure of 70% efficiency is only a “ball park” figure and it could easily be somewhere between 65% and 80%, so there are inherent approximations no matter which way you do it. BTW I’m not criticizing the 70% figure as I think 70% to 75% is pretty right.. :).


I actually measured efficiency just a few minute ago, but don't count me on it.

I connected a dummy load(a long ass extension cord.. lol!) on 5V line and drew 17.8A from 5V bus and my PSU is rated at 33A on 5V.

The current was 17.80A based on 0.5% +3 digit DMM. I *assumed* voltage was 5.0V, because I only had one DMM and if I remove the DMM from current measurement, it alters the series resistance and varies current, so that was not an option.

That gives output of 89W

Line voltage=122.1V
Line current=1.56A
VA=189VA
power=121W
PF=0.64.

89/121=73.6% efficinecy at ~30% total load all on 5V bus.

During this measurement, power factor was 0.64, therefore input current was 1.58A(true RMS) and 189VA.

One thing to remember about multiple computer loads is wattage is sum of each computer's wattage, but not necessarily so for current.

Power factor is down to 0.64 due to harmonic distortion. 1.5A+1.5A+1.5A+1.5A will be less than 6A, because harmonics from multiple computers tends to cancel out the distortion and increase the power factor. The resulstant current is what really counts when you're trying to squeeze in many computers on the same circuit. You'll have to actually connect all the computers you intend to connect on the same circuit through power analyzer to see what kind of current they're actually putting on line. You shouldn't exceed 12A for computer loads on 15A circuit as computer loads are more or less continuous duty and National Electric Code specifies you can only load circuit to 80% of rated current if continuous use.
 
SexyMF

SexyMF

Multiphasic
#16
I've wanted to see such a thread for a while now. Great work Jerboi.

I hope this dispells so many misconceptions about PSU requirements. This also says to me that new PSUs need to be developed. The new CPU use increasingly small core voltages but consume a fair current.

As for the magic 1.1:

Pure resistive load,

Full bridge rectified (no cap) has Idc=2Ipk/pi
Irms supply = Ipk/sqrt(2) => Irms= Idc * pi/2 *1/sqrt(2) = 1.1

There is your relationship between average & RMS and hence the scewed scales on multimeters.

For my studies I have 1 year for research and will be looking at resonant voltage regulators and EMC compliance. It will be interesting to see if they would be viable for multirail computer supplies.



More magic numbers:

Idc = output current Irms=supply current

Half bridge: Irms=1.57Idc (resistive) 1.41Idc (inductive)

Full bridge: Irms=1.1Idc (resistive) 1.0Idc(inductive)
 
Jerboi

Jerboi

Waasssup????
#17
Originally posted by SexyMF
I've wanted to see such a thread for a while now. Great work Jerboi.

I hope this dispells so many misconceptions about PSU requirements. This also says to me that new PSUs need to be developed. The new CPU use increasingly small core voltages but consume a fair current.
Unfortunately marketing will continue to push 500W and such PSU. A PSU with properly balanced output in each bus and designed with low enough ripple will work just fine. 180W of power properly distributed is more than adequate.

You should have 42% extra power than necessary, so PSU is only loaded to 70% the rated power. Even then, 250W PSU is more than adequate for most consumer computers.
 
U

uart

New Member
#18
Yes, I agree about the power supply ratings turning into a bit of a marketing numbers game. I did my power measurements about a year ago (and posted the results in the cpu forum here). I was trying to dispel the myth that typical home computers actually use all the 300+ Watts that the PSU’s are rated for. Generally it is true that the higher rated your PSU then the larger the main capacitors and the more resilient it will be to load dump transients and the like, so bigger generally is better. But just the same I also firmly concur with Jerboi that a properly designed 200Watt PSU could (if one was specifically designed to ) handle just about any home PC in use today.

I sometimes think of the PSU ratings “numbers game” as being a bit like the old audio power ratings fiasco. You know when you’re buying and audio amplifier (or even a set of amplified computer speakers) and you see all those ratings like 1000 Watt PMPO or even 2000 Watt PMPO (peak music power output). Just as with computer PSU ratings this no doubt gives people the false impression that many hundred or even thousands of Watts are need to provide a reasonable sound level in the average living room. This is of course nonsense as it really only takes about 5 to 10 Watts (with reasonably efficient speakers) to do this. Hehe, I even pulled one of my friends “1000 Watt PMPO” amplified computer speakers apart the other day to show him that the actual driver (speaker) was rated at only 10 Watts, LOL. As many people know, when you go shopping for a stereo amplifier that a unit rated at something like “20 Watts RMS per channel at 0.005% THD” is often way better (louder and cleaner) than one rated “1000 Watt PMPO). Just another marketing numbers game LOL.

Power factor is down to 0.64 due to harmonic distortion. 1.5A+1.5A+1.5A+1.5A will be less than 6A, because harmonics from multiple computers tends to cancel out the distortion and increase the power factor. The resultant current is what really counts when you're trying to squeeze in many computers on the same circuit.
Sorry but I have to nit-pick the above just a little. Sure it is true in general that if multiple different types of load are connected in parallel that some of the current harmonics are likely to cancel out giving a results current with a somewhat lower RMS value than the sum of the individual RMS currents. You’re correct, indeed it’s a mathematically provable fact that the RMS of the sum of the currents can be no larger (only equal or less) than the sum of the individual RMS currents (triangle inequality). Unfortunately however if you connect multiple computers then they will all have very similar shaped waveforms on the input line current. In particular they will all tend to pull a positive/negative spike of current right at about the positive/negative peak of the AC mains voltage. In reality the scope for harmonic cancellation is quite limited with multiple loads that are so similar.

BTW, It’s nice to meet some others here with interest in electrical power. :).
 
Last edited:
Jerboi

Jerboi

Waasssup????
#19
Yes, I agree about the power supply ratings turning into a bit of a marketing numbers game. I did my power measurements about a year ago (and posted the results in the cpu forum here). I was trying to dispel the myth that typical home computers actually use all the 300+ Watts that the PSU’s are rated for.
Let's not forget that physically near identical power supply maybe rated 300W by manufacture one, but 340W by manufacture two. My power supply is supposedly 340W. The OEM manufacture limits safe load to 315W. Another seller says 340W max, but 5V+12+3.3V load shouldn't exceed 285W. One is playing more conservative.


Generally it is true that the higher rated your PSU then the larger the main capacitors and the more resilient it will be to load dump transients and the like, so bigger generally is better.
Larger main capacitor only means more energy stored in DC bus and it means more up-time in case of power black/brown out until UPS transfer. Large capacitor for small load means detrimental effect on input power factor. Larger the capacitor, worse it is for power factor. If you're running hundreds of computers with unnecessarily large input stage capacitors, this creates unnecessarily high neutral current on 208Y/120V distribution system.

Sorry but I have to nit-pick the above just a little. Sure it is true in general that if multiple different types of load are connected in parallel that some of the current harmonics are likely to cancel out giving a results current with a somewhat lower RMS value than the sum of the individual RMS currents.
No need to be sorry. This is a technical forums. I'm all happy if you can contradict my errors through technically valid arguments.

I guess you have a point here.
Since institutional computer facilities use multiples of identical power supplies, it is even more similar than home situation. Although, with slight difference in load, RMS current is slightly reduced.

Loads of obviously different character difinitely influences significantly. 30W CFL on electronic ballast has PF of 0.51, two lamp 40W magnetic ballast has PF of 0.6. When you use them together on same circuit, the resulting PF is about 0.75. One is inductive. Other is rectifier load.


Unfortunately however if you connect multiple computers then they will all have very similar shaped waveforms on the input line current. In particular they will all tend to pull a positive/negative spike of current right at about the positive/negative peak of the AC mains voltage. In reality the scope for harmonic cancellation is quite limited with multiple loads that are so similar.
Current crest factor is used to quantify such current pattern. Pure resistive load has a crest factor of sqrt 2. Crest factor is peak current/RMS current. As to cancellation... hmm.... I dunno.. I'll have to hook 'em up to my scope and see. I won't be able to tell exact distortion content though.. I don't have a Dranez-BMI power analyzer..


BTW, It’s nice to meet some others here with interest in electrical power. :).
]

I thought I was the only one.
 
SexyMF

SexyMF

Multiphasic
#20
Since institutional computer facilities use multiples of identical power supplies, it is even more similar than home situation. Although, with slight difference in load, RMS current is slightly reduced.
I have been looking at single & 3 phase mains (230V) in the power electronics lab at my Tech. With a couple hundred computers with SMPS in them then mains waveform is clipped by about 5 voltages with the 3rd hardmonic multilating the falling edges.



...that a properly designed 200Watt PSU could (if one was specifically designed to ) handle just about any home PC in use today.
To true.... and stop running peltiers off them!!! (badly at that)



BTW, It’s nice to meet some others here with interest in electrical power. .
These forums need such perspective.
 

Associates