Should be enough, even for SLI.

Damn, I wish silverstone would come up with a less cumbersome naming system..quite hard to be sure what are the specs of each model without googling up.

FSP Epsilon would definitely be better. The quality of FSP is top notch, though it doesn't like being overheated. Ironically, Cooler Master does a better job cooling its PSUs than FSP, they always use an above average fan (at the cost of noise) while FSP tends to go for quiet PSUs but at the risk of unstable voltages when it overheats. Keep the casing well ventilated and FSP will give very good performance for the price..though seriously you should consider getting a lower watt model from a better brand. Check the PSU calculator if you really need that many watts.

Yeah, like flubber said, its the speedstep mechanism (for intel) or Cool n Quiet (for AMD). The processor voltage and speed is reduced to save power and reduce heat when the processor is idle.

Ok, I'll say this again:

Fact 1
Motherboard sensors are usually off the mark by a bit. Especially vCore.

Fact 2
CPU voltage depends more on the board's voltage regulator than the PSU's voltage regulator. Even on catastrophically low 12v rail voltages (<11.7v) most boards can still keep the vCore at a operating level, though the ripple voltage will be psycho high causing the system to be unstable.



Btw shawty, I think the xtremesystems list is better than johnnyguru's list. His list is is basically all the ultra 'terror' PSU, definitely not for the budget strapped-when I say budget strapped, I mean those who can't afford to put out RM200 for a PSU. There are quite a number of <RM200-RM250 PSUs good enough to start out with for cheaper, low power systems (eg, moderately overclocked low end A64 X2 and C2D, using 7600/X1650 and below for GFX). In any case, any system better than that, budget should be no excuse since the system will already cost a fair bit.

Also, I can't understand how the hell he put Enermax on top of Fortron..but it could be that the Enermax Galaxy is much more stable than the rest of the Enermax's. Reviews seem to note that, ripple current is very much lower than the others in the family.

But I'm shocked that the Enermax Liberty 620W can even make the list..the ripple voltage is almost awful on that thing, when you consider how much cash you're paying for it.

This post has been edited by lohwenli: Jun 1 2007, 05:23 AM

Wow, thats a really long reply..I can see you've been doing a lot of research..even more than me..

On PSU reliability,

For the average power user/gamer (no overclocking), the pc will frequently take up at least 50-75% of the full power most of the time (which may range from 100-400W in most cases). On the other hand, the average office user's PC is mostly at idle most of the time which means it will rarely take up more than 80-100W except for short durations. So the important factor is, what is the effective lifespan of the PSU at under the usage conditions?

Generally, PSU lifespan depends on 2 things-power consumption and heat. Running a PSU close to or beyond its limits and high operating temperature will result in greatly shortened lifespan (which is usually expected to be 3-10 years). So what's important is to have a PSU that can actually live up to its promises, and most of the proper branded ones, even cheap ones(Acbel, Enlight, Cooler Master), can live up to that.

Many of the cheap no-name generic PSUs are actually over-rated or falsely rated and are actually designed to only handle around 100-150W continuously. Why these PSUs seem to be able to run even above their actual limit? Because most PC don't actually take up as much power as you people think, and most users' PC will never actually be run under full load. But it will still go down eventually, as the cheap components used will not have anywhere close to the lifespan of proper components.

However, for overclockers (about 'solid' PSUs)

Overclockers tend to do extensive stress testing (which will cause a PC to reach its max power consumption), not to mention the increased power consumption during normal overclocked use. So naturally, higher rated power supplies are a must. However, there is also another factor.

Overclockers run their hardware to the limits where even data signal integrity starts to become unreliable. As digital signals are based on a high and low voltage (vCore and 0V respectively in processors), it is important that the high voltage is stable so that the signals can be interpreted correctly. This is where ripple voltage comes in-minute variations in the voltage (too fast to be noticed except on a osciloscope) can cause signal integrity to worsen, limiting overclocks. And to make things worse, as psu's run close to their limits, ripple voltage increases significantly.

In this case, a better PSU with low ripple voltage will help. But question which is still left unanswered is-how much will it help? And also, the motherboard is actually more importantin when the vCore is in question as the vCore regulation is done entirely on the motherboard-the PSU has no rail which directly supplies the processor. The only way the PSU can affect is if the 12v rail supplying the CPU voltage regulator is unstable, the voltage regulator might not be able to regulate the vCore that well, but even then, its uncertain how much effect does this have.

Ripple voltage on the 3.3v and 5v will effect overclocks the same way a poorly regulated vCore does because the less complex circuits on the motherboard still use those rails in data signals. But the load on those rails is low, resulting in minimal ripple voltage. And even the chipset, ram and most of the high speed data paths no longer use direct supply from psu rails, but instead use motherboard regulated rails just like the processor.

This post has been edited by lohwenli: Jun 2 2007, 02:26 AM

Sorry if it was taken the wrong way, I was just adding on to what you said, not replying to any question in particular or criticizing. You really did one hell of a job reading that many psu reviews, I think I only read half of them at most. Thanks bro

Most of what I know is from proper psu reviews and the electronics portion of my physics course. I am not qualified to make any particular recommendations on PSUs, unless I have a proper review to refer to; hence I don't talk about those which I don't know about. I too find it particularly sad when someone just says a psu is good/bad due to their own (isolated) experience. Often its just luck really.

Dirt itself won't kill the psu under most circumstances, but it might kill the PSU fan. Then heat will kill the psu for sure. But usually the manufacturer would be smart enough to use a fan that would at least last as long as the warranty.

After warranty is over, heck, you can clean the fan or replace it altogether. Or more likely, you'll replace the whole PSU

Kind of knew that already, but I had no proof to show. Thanks bro. AMD procs and C2Ds take very little power even when heavily overclocked (less than 150W usually). I'm not so sure about the 8800 though, all I know is that it takes around 100-120W per card on full load, but in most cases it will never actually reach full load, especially if its in SLI. Not sure about overclocked power consumption for 8800. Rest of the system takes 50W at most.

However, the picture is very different with Pentium 4 and Pentium D. Those buggers suck juice big time. At least 150W for the processor alone for the 900 series, and the 800 series can even hit 250W at 4GHz.

Don't trust 100% the reading you get from programs that read the motherboard sensor. I've tested at least 5 boards with a digital multimeter, and 80-90% of the time the figure does not match or even come close. In fact, half the time the PSU was giving very nice readings on the DMM (11.95-12.05v) while the motherboard was giving me something like 11.86v, or even something outright moronic like 6.99v. And this also applies to the other rails (5v, 3.3v, ect).

Motherboard sensors can help give you a head's up that something is horribly wrong with the PSU (like if the voltage keeps fluctuating rapidly), but using it to judge a PSU's quality is not being accurate at all.

The 2 most important things to consider about a PSUs in terms of stability are-voltage regulation and ripple voltage. Better PSUs will have tigher voltage regulation, meaning that the voltages will be closer to the target values of 12.00v, 5.00v and 3.33v even under maximum load. Contrary to popular belief, higher voltage isn't necessarily better-it helps slightly in overclocking, but it also puts hardware at increased risk of failure. Most reliable PSUs will have good voltage regulation (within ATX spec of 5% variance) at any load within the PSU's rated specifications.

Ripple voltage is the often overlooked factor in PSUs, particularly because its can't be measured without the use of an osciloscope. This is where the best of the best stand out-those PSUs have very little ripple voltage, even at max load where most PSUs tend to have severe ripple voltage.

Generally PSU voltage regulation and ripple voltage gets worse as the PSU is loaded more and more. Technically, even under the worst cases PSUs should still at least keep to the ATX specifications. The best way to find out which PSUs are good are to read reviews where the PSU is tested under full load conditions. Most sites just hook up the PSU to a system and just give the motherboard monitor readings (which are crap most of the time). Also, most of the systems used rarely take up more than 300W, in fact most of the Athlon 64 systems used in reviews don't even take 200W. So even if the readings are taken by multimeter, the PSU's behaviour at full load is not known. However, there are a few sites which do make a proper review by loading the PSU to its full capacity-silentpcreview, jonnyguru, tomshardware are the few I'm familiar with.


5% higher or lower than the default values is tolerated in the ATX specification.
So that means

12V = 11.40-12.60v
5V = 4.75-5.25v
3.3v = 3.17-3.5v

is still acceptable, though for overclocking a 2-3% margin instead of the usual 5% is usually recommended to give an additional safe zone.

Probably died because it overheated. Heat is the no 1 cause of death in reliable PSUs. For those hopeless cap ayams, its overloading. Overloading doesn't happen to reliable PSUs because they are required to have

1. Current limiting- if demand is higher than the PSU spec due to overloading or short circuit, the PSU will shut off.
2. Voltage protection- if output voltages run above certain limits, the psu will shut off.
3. Voltage tolerance -even if high voltage (500-1000v) is applied to the PSU contacts, it must tolerate it without damage. Known as Hi-Pot test.

What is sometimes lacking in even reliable PSUs
1. Fan speed contol -adjusts fan speed to keep the PSU sufficiently cooled. Increasingly common these days, but sometimes questionable if they're set properly to pevent overheating.
2. Overheat protection -usually found only in very high power PSUs (>600W), and even then not all have it, which explains why sometimes even reliable PSUs can fail.

This post has been edited by lohwenli: Jun 7 2007, 06:41 PM

Only way to know is to check with digital multimeter. But anyway, silverstone PSUs have a good track record of voltage regulation, and the reading that you gave is still within safe ranges, so I wouldn't worry if I were you. The only time I'd worry would be when the voltage changes significantly from when the system is at idle and when its at full load-its a sure sign that the psu is of poor design or is being operated close to its limits.

I just recalled something-for those using 20-24 pin convertors, there is a slight voltage drop due to the resistance in the adapter (it may be worse if the board draws a lot of current-you can tell because the adaptor will get warm). Avoid having to use one if possible, even if it means plugging in only 20 pins into a 24 pin motherboard.

Read the previous posts-I've mentioned a lot about the 'accuracy' of motherboard monitoring chips. Also, software and bios readings don't always match because the calculations involved may not be the same. Best way is to check with a digital multimeter.

It depends on the quality of the convertor and how much current the board takes through the 20/24 pin connection.

Voltage drop due to adaptor = connection resistance x current used by board

You can get a multimeter from electronic component shops or hardware shops dealing with electrical items. Digital ones are recommended (they give a readout calculator style), even cheap ones (<RM30) are on average more accurate than most motherboard sensors. Analog ones (look like car speedometers) are useless for measuring PC voltages, as they are not accurate enough.

The problem is that the 'negligible resistance' isn't so negligible if you factor in the contact resistance. Although its still a fraction of an ohm, the current flowing through is high (a few amperes at least). So as an example even a resistance of 0.05 and current of 4A (most likely higher than this) will result in a voltage drop of 0.2v, which is very noticable. Under ideal contact conditions there should be no resistance, but that is unlikely. Its for this reason that CPUs use the 12v rail now instead of the 5v rail (that was before P4), as it would require far more current on 5v to power a processor than on 12v. 25-40A load on the 5v rail was likely even then, which could mean a very subtantial voltage drop if there was any resistance in the connection.

How the hell can a single PC use up 2000W?? Even if I peltier cool everything also still left over at least 700-1300W. I think if disconnect from PC can even use that thing's 12V rail for spark welding

Its refering to the VDimm for DDR. DDR uses 2.5v, DDR2 uses 1.8v and DDR3 uses 1.5v. Don't sweat too much about the voltage reading, its probably incorrrect. If you system can still boot, you can bet the actual voltage is nowhere that low-if it was, even DDR2 would not boot regardless of how low you underclock the ram.

Watts consumed by digital multimeter = less than 1 watt

I take it you mean use the digital multimeter to check PC wattage use. Yes, its possible, set to AC current measurement, then measure the current taken by the PSU and multiply it by 230 (or if you want to be so damn accurate, measure the voltage and use that instead). Perfect accuracy can be achieved if you're using a active PFC psu and a sine wave digital multimeter, but the accuracy of a normal digital multimeter is still pretty ok with an active PFC psu. With non active PFC psu, readings might not be accurate because you have to account for the phase shift in the current, which is pretty hard to measure using easily available equipment.

PS : to measure current you'll have to cut one of the AC power wires and put the multimeter in connected to both cut ends of the cable. DO NOT ever connect the multimeter directly to Live and Neutral on AC while in Ampere (current) measurement mode, if you're lucky you'll just blow the multimeter's fuse, if not you'll see smoke coming from the multimeter.

Should be less than RM100 for cheapo ones, better ones should cost RM100-200. You can get a decent digital multimeter for that money already.

Actually its not that hard, I'll put a photo to show how it can be one without any cable cutting (I'm borrowing a digicam next week to record a modding worklog). Only need an extension cord or a PSU cable which uses a 3 pin plug that can be opened.


Gigaabyte also want to main with PSU..haiz..wonder whose circuit design they will use..

But frankly, manufacturing PSUs isn't hard if you're ordering the components in bulk (otherwise its damn expensive), and you don't need anything other than a soldering iron and some crimping tools to put one together. No wonder all the companies want to produce PSUs, very good profit margin.

This post has been edited by lohwenli: Jun 16 2007, 07:44 PM

How come? Toughpower 700W is supposed to be pretty reliable. If capacitor leak/letup, got strong smell wan, anything else nothing much-just won't turn on.

Fan dead?!! Any PSU will die from overheating eventually, even a ultra-expensive PC Power and Cooling. The Toughpower 700W does have some degree of overheating protection, but it reacts too slow if the fan is already dead.

One thing I can't understand about PSU makers, can't they just implement proper overheating protection and jammed fan protection? PSU capacitors are even more vulnerable to overheating than any other component in a PC, with overworked power MOSFETs being second. Capacitors can blow at as low a temperature as 80C, and typical operating temperature in most PSUs is already 50-60C even with the fan running at moderate to full speed. With the fan dead, temperatures will easily get high enough to make popcorn out of the capacitors

Thats overkill..like using an atomic bomb to kill a ant.. I'll say 600W-650W would be the sweet spot if you intend to overclock that system. If not, then you can get away with 450-500W.

Non-overclocked
Processor -130W
Motherboard, RAM & drives -70W
Graphic card -150W

Total -350W (As it is, its actually a pretty generous allowance, considering that it the C2Q is actually rated only 105W, and the 8800 takes only 110-130w)

The ST56F will be enough for your future C2D set. It is a little excessive for your current barton and NF2 though..

Thanks for pointing that out, though I'll have to clarify a bit.

Power factor should have an effect on current readings taken with a digital multimeter. If the power factor is poor, the current (based on the apparent power) which is measured by the DMM should be higher for any given power intake by the PSU-AFAIK the PSU will only use the real power component. So for non-PFC psu, the watt reading will be higher, which will not reflect the power consumption of the PSU. Its bad enough that we don't have a decent method of measuring the PSU's real life efficiency, having the power factor further compound the problem of getting accurate power consumption readings is just (theoretically, its possible for the measurement to be as much as 2x off the actual value, assuming a power factor of 0.7 and efficiency of 75%, which are typical values for cheap PSUs)

Most affordable clampjaw meters don't have enough accuracy to measure the AC current used by a typical PC. Most PCs take up 1-2A only (even a fully loaded 1000W PSU will only take 5A at most), so a clamp meter must be able to read up to about 0.01A to be reasonably accurate to a few percent, and those are pretty expensive. If accuracy is any worse, might as well use an affordable DMM (which would have similar accuracy).

And yeah, a shunt works too, like you mentioned. But you still have to cut a cable to put it in.

---

Added on June 24, 2007, 3:40 pmJust remembered the key reason why I said using an active PFC PSU is important-active PFC PSUs typically have power factors of 0.99 or better, which will mean the current waveform will be sinusoidal (assuming the AC voltage is sinusoidal-not likely to be perfect sine wave, but should be ok unless your AC supply is outright shit). A sinusoidal current waveform will allow even non-RMS DMMs to give decently accurate readings (my estimate is it would be less than 30% off, which is better than most of the affordable clamp meters)

This post has been edited by lohwenli: Jun 24 2007, 03:40 PM