|
Guest
Posts: n/a
|
"Paul" <> wrote in message
news:jgrjug$ll4$...
> Allan wrote:
>
>>
>> Once again many many thanks for your detailed advice and help
>> I have a little bit or reading to do now :-)
>>
>> Could you please just clarify when I might use the 6 pin ATX
>> .................. are you saying IF the 4 pin was not powerful enough I
>> could use the 6 pin?
>>
>> Thank You
>
> The 2x4 ATX12V (four yellow on one side, four black on the other),
> is for extreme overclocking. for example, on Tomshardware, they
> cranked a D 805 to 4GHz, and the power draw is over 200 watts
> when you do that. A 2x2 connector would begin to overheat at that
> level of abuse. And then you'd want to use a 2x4 and ensure the
> 12V2 rating on the power supply is 200W/12V = 17 amps or more.
>
> In your case, you're not anywhere near that level (yet). I don't
> have a 2600K here, so can't measure the power for you. I measure
> the power on my processors here, by using a clamp-on ammeter to get
> a DC current reading, and that's how I can tell whether I'm in
> good shape or not.
>
> To give an example, on one 65W processor, power stays below 36W,
> even with Prime95. On another 65w processor, it might be around 45W.
>
> Using an older processor (P4), the measured power was actually
> a couple watts higher than the TDP value. But current generation
> processors, tend to be on the low side, rather than the high side.
> The high side behavior peaked around Prescott days.
>
> Note that there are a number of small connectors on the supply.
> There is the 2x2 ATX12V (two yellow, two black wires). But there
> is also the 2x2 section of the 24 pin main connector, but it has
> four different colored wires. Occasionally, someone tries to
> connect that four different color connector, to their ATX12V, with
> disastrous results.
>
> There are also 2x3 and 2x4 PCI Express connectors, but they shouldn't
> mate with the motherboard. (I haven't tested that though.)
> The shapes of the nylon shell around the pins, are to help
> prevent that sort of thing.
>
> Power supplies also come with a 2x4 style connector, that splits
> into two pieces. That type is handy for dealing with an old
> motherboard (with 2x2 connector only on the motherboard), or
> dealing with a newer motherboard with a 2x4. You split the
> connector in two, when you want just a 2x2.
>
> And some supplies, even come with one 2x2 as well as a fixed 2x4
> that doesn't come apart. Usually, they'll all be part of the same
> cable assembly, as a hint they're part of 12V2 output.
>
> This page shows pictures of many of these variations.
>
> http://www.playtool.com/pages/psucon...onnectors.html
>
> *******
>
> If you're overclocking 30%, that would be 95W * 1.3 = 123.5W.
> If the Vcore converter was 90% efficient, the input power needed
> to provide that power level would be 123.5W/0.90 = 137.2W.
>
> The most conservative rating for the 2x2 is 144W (12V @ 6A on two wires).
> The pins might be good for eight amps each, depending on the wire
> gauge used (fatter wire, removes more heat). So even at 30% overclock,
> you're still under the limit. Now, if you start cranking VCore, as
> the guy in that Newegg article was doing, then you might manage to
> get over 144W. But without a clamp-on ammeter, it's pretty hard to
> check.
>
> Yes, you can fit a multimeter in series with ATX12V, but your typical
> garden variety multimeter has a limit of 10 amps before the fuse blows.
> So to measure 12 amps, you'd need a better quality ammeter, or the
> services of a clamp-on ammeter.
>
> My ammeter can measure up to 400 amps. In fact, it doesn't really have
> good "low" ranges. But being a "non-contact" device, I can quickly take
> readings with it (you just clamp it around both yellow wires at the same
> time, and it takes the summation of the current flow). Too bad it costs
> so much more than a regular multimeter. It uses a Hall probe, to sense the
> current flow. Clamp-on ammeters come in AC only, or AC/DC, and for
> computer usage, you want DC measurement ranges. To measure DC
> currents, implies Hall probe technology.
>
> http://en.wikipedia.org/wiki/Hall_probe#Hall_probe
>
> My clamp-on ammeter.
>
> http://www.mitchellinstrument.com/me...8/380947_1.jpg
>
> Using a clamp-on ammeter. When measuring an AC cord, you have to
> split the wire cable between wires, so that only one wire goes
> into the jaws of the instrument (otherwise, sticking the jaws around
> the entire cord, the magnetic fields around the wires cancel out).
> I have a home made "breakout cable" with the wires spread apart, for
> that purpose, so I don't have to ruin any good cords.
>
> On an ATX supply, all the wires are exposed, so you can get the necessary
> wire groupings right within the jaws. On the output side of the ATX
> supply,
> you're making DC range measurements. On the input side (if you wanted to
> do
> a measurement there), it would be AC ranges. For AC inputs, a
> Kill-O-Watt meter is the instrument of choice, rather than one of
> these. Kill-O-Watt can take power factor into account.
>
> http://i.ytimg.com/vi/z-KfZvbjyBY/0.jpg
>
> *******
>
> So if you were doing that extreme an overclock, you'd want to
> switch to a 2x4 and a newer supply with a hefty 12V2 rating.
> If you're sticking to 30%, and not cranking the hell out of
> Vcore, you might still be fine. I pick the 144W number to
> be completely safe, but the pins probably can handle a bit
> more than that.
>
> Damage to pins, happens over a period of time. The metal on
> the pin, goes from shiny to dull colored. This is oxidation
> from heat, or a metallurgy change on the plating. Once
> oxidation occurs, the "resistance" when the two connectors
> mate goes up, which causes the connector to get hotter,
> which hastens the demise of the connector. Eventually, the
> connector will be a charred mess. Just as the computer
> will no longer boot, you take the side panel off and there
> is the connector with a pin or two completely burned off.
>
> I've only had that happen once here. I burned a Molex 1x4
> on an ATI video card. Apparently, the cheesy chinese connector
> on the power supply, wasn't gripping the pin very well. (The
> current flow before the incident, was only 5 amps.) Molex
> have gone downhill, since every Tom Dick & Harry has been able
> to make them. At one time, there were fewer sources of those
> connectors, and the quality was more consistent. If you
> feel a connector is not seating properly, keep an eye on it,
> as it might go like mine did.
>
> Also, the 2x2 and the 24 pin main connectors have latches.
> The latch keeps the connector seated, so it can't "walk out"
> and work loose. That's another way they get burned - a connector
> can work itself loose, and then the pins burn. Make sure the
> latch is engaged, when installing. The Molex 1x4 doesn't need
> a latch, because its insertion force and retention force are
> so much higher (or at least, they used to be that way).
>
> Paul
Many thanks for that Paul .........has made things a lot clearer, and I feel
happier.
Generally speaking (unless I am bored one day :-) I have no intention of
overclocking, to be quite honest there is no real need for me to do that,
but you know what us men are like :-)
I really do appreciate ALL your time and knowledge
Thank you
|
|
Similar Threads
Linear Mode
