Ok? Or time for a new PSU?

Parja

Parja

[H]F Junkie
Joined
Oct 4, 2002
Messages
12,671
In my main rig (in sig) with an Antec TP3-550, I'm upgrading my HD 4870 to a GTX 280...that I'll overclock some. Am I ok with the Antec or should I be looking to upgrade to a bigger/better PSU? Also keep in mind that this rig will be Folding 24/7.

Rig specs:
Core 2 Quad Q6600 G0 overclocked to 3.5GHz (438x8) at 1.5V
CPU cooled with a TRUE sporting a fairly low speed 120MM fan
ASUS P5Q Pro motherboard
2x2GB OCZ Reaper PC2-6400 at 438MHz, 2.1V
Samsung 7200.10 500GB
Creative X-Fi Professional with front panel

All in an Antec Nine Hundred case with the 250mm fan on low and 5 low speed 120mm Yate Loon fans.

Other than that, nothing drawing any serious power over USB. Just charging the wife's Creative Zen Mosaic, the wireless dongle for my keyboard/mouse, and my Nostromo Speedpad.
 
Parja said:
In my main rig (in sig) with an Antec TP3-550, I'm upgrading my HD 4870 to a GTX 280...that I'll overclock some. Am I ok with the Antec or should I be looking to upgrade to a bigger/better PSU? Also keep in mind that this rig will be Folding 24/7.

Rig specs:
Core 2 Quad Q6600 G0 overclocked to 3.5GHz (438x8) at 1.5V
CPU cooled with a TRUE sporting a fairly low speed 120MM fan
ASUS P5Q Pro motherboard
2x2GB OCZ Reaper PC2-6400 at 438MHz, 2.1V
Samsung 7200.10 500GB
Creative X-Fi Professional with front panel

All in an Antec Nine Hundred case with the 250mm fan on low and 5 low speed 120mm Yate Loon fans.

Other than that, nothing drawing any serious power over USB. Just charging the wife's Creative Zen Mosaic, the wireless dongle for my keyboard/mouse, and my Nostromo Speedpad.
Click to expand...

I’d be inclined to move up to something like the Corsair 650/750 just for the extra headroom and cooler running you would get.

Some of these newer Vid cards are out of sight with their desire to suck power:)
 
Your going to need half of your available power to run that video card. You will want to isolate one of your 3 18A rails into the 8-pin connector on the vid card. Try to isolate your CPU on another rail. Try to get as many of your other devices on the third rail with the 6 pin PCIe connector.

If you have another place you can utilize the Antec then go ahead and upgrade. If you plan it out well you can be pretty good with your current PSU, but like BillR said it's nice to have some headroom especially if you are folding with it.


 
The Corsair TX650w/750w are nice. Many seem to get BFG's ES 800w for a bit more power.

There's also the Antec TruePower Quattro 850w also.

Think of any these four psu models will suffice.
 
flogge said:
Your going to need half of your available power to run that video card. You will want to isolate one of your 3 18A rails into the 8-pin connector on the vid card. Try to isolate your CPU on another rail. Try to get as many of your other devices on the third rail with the 6 pin PCIe connector.
Click to expand...

How would one go about the isolation you speak of? Is there a guide or something that spells out which strands on the PSU use which rails?
 
Depends on the documentation included with the PSU and the labeling on it. My Corsair HX1000W has excellent labeling but it only has 2 rails to deal with. Some of the better PSU reviews like the [H] ones include a chart spelling out what the rails are and what connectors are attached to them. I will try to find some more info on your particular PSU but can't do it right away.

You don't have to do a perfect job on this as the combined load rating is less than the sum of its parts. 18 amps x3 rails x 12volts is 648 Watts. This is higher than the total available output of the PSU so you don't need to balance them perfectly. It helps to get it as close as possible though. YMMV.
 
flogge said:
Depends on the documentation included with the PSU and the labeling on it.
Click to expand...

Nothing I've been able to find mentions anything about which connectors correspond to which rails. Antec's literature on this PSU seems to be very sparse. If you can find something, though, it would be much appreciated.
 
From what I've read, the TP3 series from Antec is kind of a pseudo-three rail PSU. Seems if you draw more than 18A from a rail, it just pulls power from another one of the rails to compensate.
 
That doesn't surprise me that much. There for a while everyone was marketing greater numbers of rails as a good feature. I don't see the benefit myself. I'd rather have one or 2 great big rails so I don't have to worry about balancing loads as much. I searched around also and came up with even less info than you did.
 
flogge said:
That doesn't surprise me that much. There for a while everyone was marketing greater numbers of rails as a good feature. I don't see the benefit myself. I'd rather have one or 2 great big rails so I don't have to worry about balancing loads as much. I searched around also and came up with even less info than you did.
Click to expand...

At one point the ATX spec required you to have multiple rails on the +12V as there was a limit of 20A max on each +12V rail, so manufacturers were forced to spec 18A max (peak 20A) to be compliant.

In ATX 2.2 spec, that was removed from being a requirement.
 
At one point the ATX spec required you to have multiple rails on the +12V as there was a limit of 20A max on each +12V rail, so manufacturers were forced to spec 18A max (peak 20A) to be compliant.

In ATX 2.2 spec, that was removed from being a requirement.
Click to expand...

Thanks for clarifying that. My old 430W Enermax has a single 32Amp rail I believe. I think I bought it in 2003. That PSU must have been compliant to the previous spec before the limit was imposed. The funny thing about efficiency though is that an HX1000W draws only 1amp more from the AC receptacle than my 430W from a few years ago.
 
Well, the PSU seems capable of handling the 280. According to my Kill-A-Watt, I'm pulling about 450W from the wall with Folding SMP and Furmark in Xtreme Burning Mode running. So assuming a pretty generous 82% efficiency, it's only putting out about 370W. From the reviews I've seen on this PSU, though, I'm getting up into the range where the efficiency drops off pretty quick, so I'll probably look into getting a new PSU...maybe this spring when the heat generation of the PSU isn't such a nice thing. ;)

The annoying thing is I found that my UPS isn't too happy about putting out 450W+. It's rated for 500W output, but a few minutes into running Furmark, she starts giving off the, "Hey, I'm gonna shut down!" beep. Grr...annoying.
 
Just a word of advice:
Paul_Johnson said:
Yes and a quick search would turn up this topic a million times over. Here is the recap:

1) APFC can fool Kill-A-Watts into giving you abnormally low readings (some times giving better than 100% efficiency)

2) Power supplies derate with temperature anywhere from 2w/c above a nominal rated at value to 10w/c.

3) Kill-A-Watt's and most power meters sample too slowly to catch transient loads (the Transient load from our tests is 117w and is COMPLETELY missed by Kill-A-Watts).

4) Power supplies last longer if you stay in the 40% to 60% range of their output.

5) power supplies are quieter if you stay in the 40% to 60% range of their output.

6) Power supplies are cooler if you stay in the 40% to 60% range of their output.

The power meters in UPS software are just as bad. You have to spend some change before you get anywhere near an accurate power meter when your PSU has APFC.
Click to expand...

So take your Kill-A-Watt readings with a grain of salt.
 
Parja said:
Interesting. Thanks for the info. So maybe I am hitting 500W draw and that's why my UPS is complaining.
Click to expand...

You have to keep in mind that a digital PSU by design pulls power as needed (on demand if you like) from your mains.

When your machine is running at idle as in you aren’t touching the keyboard or mouse your draw from the wall is one thing, but, if you do something as simple as click on IE or whatever browser you create what is called a “transient” and your PSU’s ability to respond is called a transient response.

Your Kill 0 Watt is incapable of seeing these transient responses and it really doesn’t even do an average reading of your power draw. The extra power from even opening your browser can be higher than 40 watts. Granted that power is only used for a few microseconds but it does count in the big equation.

Think of all the programs running in the background just doing the routine maintenance to keep your machine running and you aren’t doing anything yet. Start to play a game and all hell breaks loose and you start pulling a lot of power from the wall for fractions of a second to several seconds in a row. Add it all up and it can amount to a lot of power you aren’t aware you are using.

Something else to keep in mind is what you draw from the wall is after that 80% efficiency, that figure from the wall, or your mains is real, it’s what the power company charges you for.

So, if we factor in the approximate needs of the hardware those are static numbers. Add in say 30% or so to the first number you really do need a higher capacity PSU then you think. That is where Paul is getting those numbers in the above post and if you buy accordingly you will have a cool running longer lasting PSU.

Make sense?;)
 
BillR said:
You have to keep in mind that a digital PSU by design pulls power as needed (on demand if you like) from your mains.

When your machine is running at idle as in you aren’t touching the keyboard or mouse your draw from the wall is one thing, but, if you do something as simple as click on IE or whatever browser you create what is called a “transient” and your PSU’s ability to respond is called a transient response.

Your Kill 0 Watt is incapable of seeing these transient responses and it really doesn’t even do an average reading of your power draw. The extra power from even opening your browser can be higher than 40 watts. Granted that power is only used for a few microseconds but it does count in the big equation.

Think of all the programs running in the background just doing the routine maintenance to keep your machine running and you aren’t doing anything yet. Start to play a game and all hell breaks loose and you start pulling a lot of power from the wall for fractions of a second to several seconds in a row. Add it all up and it can amount to a lot of power you aren’t aware you are using.

Something else to keep in mind is what you draw from the wall is after that 80% efficiency, that figure from the wall, or your mains is real, it’s what the power company charges you for.

So, if we factor in the approximate needs of the hardware those are static numbers. Add in say 30% or so to the first number you really do need a higher capacity PSU then you think. That is where Paul is getting those numbers in the above post and if you buy accordingly you will have a cool running longer lasting PSU.

Make sense?;)
Click to expand...



Definately agree on this.
 
BillR said:
Your Kill 0 Watt is incapable of seeing these transient responses and it really doesn’t even do an average reading of your power draw. The extra power from even opening your browser can be higher than 40 watts. Granted that power is only used for a few microseconds but it does count in the big equation.
Click to expand...

In this case, since you are talking about power and not energy, 40W is basically 40W whether its for a microsecond or for a full second. Since wattage is energy over time, the time factor is calculated into the value. In other words though one can have a "pulse" or "transient" of energy that may last a microsecond or a second, when averaged over the a second, that pulse still gives an average 40 Joules over a second. So in this case, even if its over a short range of time, the power draw is still the same over the second. The enegy on the other hand can jump up to a tremendous amount for shorter pulses with similar wattage. So I guess the point is whether a PSU circuit can handle short loads of high energy in addition to its rated power.

One question I have regarding these values is the calculation of the power from the processor. When you make your PC more active by clicking on a command or booting up an application, I can understand the power requirements spike up for a temporary amount of time. However, this is basically since the processors start using a large portion of its circuitry that was idle to do the process (activate more transistors in the circuit). However, since the number of transistors and such are at a fixed design for any processor, you can't go beyond the circuit's design. Since the processor can also be pushed to run at a maximum load using software such as Orthos and the 3DMarks and such, one can force the PC to run these processors at near full load. Would the transient necessarily be at a power draw any larger than when the system is running at full load?

Another way to ask I guess is that these processors have a specific power rating (like the 65W rating for the A64 Brisbane BE processors). Are these power ratings given as max peak or as general average? Do they just run peak loads and then average down the value or do they take statistical values of power draw from a population of processors running at max load and then from the statistics, go say 3 sigma ABOVE the statistical values to ensure they have the max load of majority of the population under most circumstances? If its the second, then I would feel more comfortable knowing that the rated power draw from the various compoents of the PC can be used to determine the power requirements of the PSU running the PC.
 
Zenshi said:
In this case, since you are talking about power and not energy, 40W is basically 40W whether its for a microsecond or for a full second. Since wattage is energy over time, the time factor is calculated into the value. In other words though one can have a "pulse" or "transient" of energy that may last a microsecond or a second, when averaged over the a second, that pulse still gives an average 40 Joules over a second. So in this case, even if its over a short range of time, the power draw is still the same over the second. The enegy on the other hand can jump up to a tremendous amount for shorter pulses with similar wattage. So I guess the point is whether a PSU circuit can handle short loads of high energy in addition to its rated power.
Click to expand...

What you have to understand is that it isn't the energy draw itself that rises, it's the rate at which the PSU is required to supply energy. The wattage is a measurement of the rate at which energy is delivered, so a rise in wattage means that the components drawing power require a higher constant delivery of energy, even if that delivery is only required for a brief time.
 
Heh, now you want to play "Physics". I'm neither an engineer or a physicist and by choice I prefer not to refer to the energy that powers my car in terms of kilowatts.

I think what you are referencing would be this:

T = \frac{1}{2} m v^2 which is a very roundabout way of expressing Ohm's law, at least for our purposes.

If we look at the water in the pipe example, the common method of expressing our needs, it becomes a bit more simple.

One common mistake is to use the words energy and power interchangeably, while directly related they are very different yet both must exist in a frame of time to produce work.

A computer is a device that requires a given amount of power at a given time to operate. the more demanding the computer the more it pulls energy from the PSU. This factor can be defined two ways, watts and amps, one can be converted to the other, again by ohm's law.

With a CPU as far as I can recall the wattage given by a manufacturer is the wattage use to make the CPU operate at spec. Keep in mind that companies like Intel also give maximum power ratings to which we typically pay no attention when we over clock.

My point in my post was the electricity we use within the computer is DC and we get AC from the wall. To lower costs (of construction) we use digital power supplies that for all intent and purpose convert that AC to DC while simultaneously adjusting the power we need to run the computer and this all happens (for the sake of argument) in real time. As such the 120v mains in your home / office, being somewhat unlimited are called upon to respond to the computer's needs. Is there an average over time for current draw, yes. Can a Kill 0 watt measure this, no. The averaging over time for us normal folk is done by the power meter on the side of your house. Does that meter do a good job, no. Typically they are calibrated for a given number of unreadable spike or transients over a given period of time, call it a fudge factor. Your meter is a time based current flow device, a motor.

You must keep in mind that all components within your computer are reactive not static.

I don't know if that answers your question or not, like I said physics for me was back in the 60s, didn't like it then, don't much care for it now.;)
 
Agreed Zero82 regarding the rate. On the other hand when the wattage goes up, you to expect the energy to go up as well. If the units were some indefinite amount that can for example be in the microsecond, then the rate can increase for the decrease in the time constant at which you take the measurement. However for the unit Watts, it is defined as Joules/Sec. So one basically is fixed to one second and the transients can be averaged out. Now this is not quite true since one can always interpret the wattage from the transient as an integration of the peak energy over the entire second so that one assumes the transient peak is stretched out for the whole second. However, I think in most cases, the Wattage is determined from the energy that is observed over the whole second (base, peak, then base) divided by the second. Not sure on this.

In the case of a PSU and its capability requirements to the processor for example, I think the energy is just as important as the power. For example, if the processor during a transient needs a large jump in energy, though it may be averaged to 40W during the whole second, to perform the function, it needs to have whatever energy it needs (which is usually higher than the average energy computed from the wattage) during that instant. The PSU may be rated to provide the 40W but it may not be able to give that high an energy (or instantaneous power) in the small instant the processor needs it. This will depend on two things IMO. One is whether the circuitry can handle high draw for short times. The other is whether the circuits can withstand a high draw for a short time.

This part I think really comes from the quality of the PSU and not just the power rating. Since a power supply is in general an A/D rectifier, it takes the power available in the AC line at 120 60Hz and turns it into the necessary power at DC. On the average, this supply may be a steady rate of power for the most part. However, during transients, there will be a spike of power necessary for that short period of time. Can the power supply provide that power fast enough I would think depends on the speed of discharge of the cap banks as well as the speed ratings of protection diodes and inductors. If it cannot provide the power fast enough, then the processor may ask for the power but the supply will not be able to comply resulting in performance stability issues. As for whether the PSU can withstand the sudden transient, this depends on the robust design that can handle high pulses of current for short durations.

To me a lot of this depends on the limit to which the PSU needs to operate. If it is at the borderline limit of the PSU's capability, the transients will become more of an issue. Even far away from the limit, the quality of the PSU may not be able to supply power required by the components. I like the stupid simple approach since that keeps things easier for me...which is to minimize the power load of the PC to what you need (why put a 4870X2 on a HTPC? Do I really need 50 Hard Drives in my PC?) and then making sure a good quality PSU with decent wattage can accomodate the load. I am still learning on what makes a good PSU and so I really am grateful to people like JonnyGuru, Danny Bui, Paul Johnson, Oklahoma Wolf and others who review and rate PSUs and explains it in ways that even I can understand. I think my PSU choices are SLOWLY getting better...I think...I hope. :p
 
The energy is the 3.3, 5, 12v supply, the push behind that is the current or AMPS.

The better PSU will will have the ability to provide the current as needed while at the same time keeping the DC clean.
 
Zenshi said:
AHowever, I think in most cases, the Wattage is determined from the energy that is observed over the whole second (base, peak, then base) divided by the second. Not sure on this.
Click to expand...

Again, you're not getting it. The wattage is not measured over any period of time. It is a rate. Although the technical unit-based definition is joules per second, the actuality is that it is an instantaneous quantity that is in effect a measurement of the amount of joules per an infinitesimally small period of time. Therefore, an actual spike in wattage just means that the instantaneous rate of energy delivery at the point in time that the spike occurs becomes much higher. However, the length of time over which that spike happens can be extremely small. That doesn't mean it's negligible, because the wattage is an instantaneous rate. However, it does mean that for a device like a Kill-A-Watt, which measures the wattage over a long period of time, a very short spike will have a negligible effect on the average power draw over the period of time that the Kill-A-Watt measures.
 
Zenshi said:
I am still learning on what makes a good PSU and so I really am grateful to people like JonnyGuru, Danny Bui, Paul Johnson, Oklahoma Wolf and others who review and rate PSUs and explains it in ways that even I can understand. I think my PSU choices are SLOWLY getting better...I think...I hope. :p
Click to expand...

I really shouldn't be part of that list of names.:eek: Everyone you mentioned there are significantly more knowledgeable when it comes to PSUs than I am. In addition, most if not all of those guys are engineers of some sort whereas I'm just a enthusiast who knows just a bit more about PSUs than the average enthusiast.

Thanks for the accolade though :)
 
Oh I agree with you completely regarding the rate Zero82z. From a formal definition, the rate of the power is dE/dt and not DeltaE/Deltat. It is the derivative of the energy with the time and not just the ratio. Thus for the power within the temporal width of the only the pulse, the power is very high. And as you say, if the time constant of the pulse is very fast compared to the Kill A Watt (faster than the time constant of the detection circuitry) then the Kill A Watt will not even be able to see it. Actually, I have no idea what a Kill A Watt is but that is general for any measurement device.

So I am not disagreeing with either you or BillR at all. For me, however, I am questioning when PSU's may be rated at some power. What are the PSU manufacturers using to determine their ratings? Also, for companies who give power ratings of the components (such as the cpu, gpu, HDD, etc), what are they using to determine the power? One can always give it in terms of peak performance or one can determine it by averaging over the second. This is more the interpretation of a specification rather than the definition.

I apologize if I sounded like I was negating the issue. I just wanted to ask about the best ways to determine adequate power if one cannot use the data taken from measurement systems and if one has to incorporate transients and such into the equation. Again, for me, I just want to make sure the PSU is in a safe zone for my not too [H]ard rigs and not for pushing the PSU and compoents to their absolute limits. If this is not much of an importance for you or others, my apologies for bringing this up as well.


@Danny Bui. Sorry if I put you in that list but for me your comments have helped me out when reading these threads and that is why I wanted to say thanks. If not, I would still be looking for the cheapest PSU with the highest wattage and having 5 eggs. :p
 
Zenshi said:
For me, however, I am questioning when PSU's may be rated at some power, what is the manufacturer using to determine their ratings. Also, for companies who give power ratings of the components (such as the cpu, gpu, HDD, etc), what are they using to determine the power? One can always give it in terms of peak performance or one can determine it by averaging over the second. This is more the interpretation of a specification rather than the definition.
Click to expand...

The ratings are based on peak power draw, not average. Anything else would simply not be accurate.
 
Thanks Zero82z for clarification. So the peak is what is being given. To be clear for me I would like to ask a few more questions. If the processor happened to put out a transient draw of 4 J over 200 microseconds. Would this then put a rating to the processor as being 20W or of whatever is averaged over the time constant of the measurement device that measures the power spike?

If is considered 20W, then is a processor for example that is spec'd at 75W power draw saying that even during a peak transient situation, the instantaneous power would be about 75W max? If so and the components specifications are given from the peak draw, can I feel safe if the sum of the power specs of the various components falls within Paul Johnson's 40 - 60% power range of the PSU?
 
Zenshi said:
Thanks Zero82z for clarification. So the peak is what is being given. To be clear for me I would like to ask a few more questions. If the processor happened to put out a transient draw of 4 J over 200 microseconds. Would this then put a rating to the processor as being 20W or of whatever is averaged over the time constant of the measurement device that measures the power spike?

If is considered 20W, then is a processor for example that is spec'd at 75W power draw saying that even during a peak transient situation, the instantaneous power would be about 75W max?
Click to expand...

The TDP of a CPU is generally the maximum amount of power it will draw under any situation. If that ends up being 4J over 200 microseconds, then so be it. I don't know exactly how they measure it. If a CPU's TDP (especially an Intel CPU) is a certain wattage, then you can expect that CPU's power draw to remain at or below the TDP at all times. If a CPU doesn't fall within a certain TDP at a certain frequency, then it is binned lower until it hits a frequency at which it can stay within that TDP.
Zenshi said:
If so and the components specifications are given from the peak draw, can I feel safe if the sum of the power specs of the various components falls within Paul Johnson's 40 - 60% power range of the PSU?
Click to expand...
Yes.
 
Thanks again Zero82z. This is actually very important at least for me regarding the determination of the power requirement for a PC.

As always, you live and learn (at least I did today :D). Oh and I will stop hijacking this thread now. :p
 
You must log in or register to reply here.
Back
Top