P
Paul
What was the extreme.outervision.com power number ?
A reasonable value ?
Paul
What was the extreme.outervision.com power number ?
A reasonable value ?
Paul
A
Adam
Paul said:
567W
But, I gave the system extra cushion by specifying more than actual.
P
Paul
Adam said:
Of the eight traces here, the bottom-most trace is 471W worst case,
and that's also using an overclocked Intel processor. I can't find one
of Xbitlabs old-style measurements, where they just measure the video
card itself.
http://www.xbitlabs.com/articles/graphics/display/radeon-hd6870-hd6850-crossfirex_12.html#sect0
The HD 6850 is rated "127W" here. These values are non-measured
ones and relatively sloppy. So it may not be 127W either,
but that's the only number I've got.
http://www.gpureview.com/Radeon-HD-6850-card-636.html
That, with your processor, should give you some idea what ballpark
you're in. I don't think the PhysX card was that hot. In any
case, you should have lots of headroom now, with a 750.
Paul
A
Adam
Paul said:
How about with the OCZ 600W (or 500W)?
P
Paul
Adam said:
I'm thinking along the lines of 125W CPU + 127W video + 25W Ageia
or about 275W baseline. You'd have another 100W for motherboard,
hard drive, that sort of thing, bringing it up around
375W, and only in a gaming situation. So I don't quite know
why the computed number was 567W.
I got the Ageia number from Xbitlabs, but they didn't measure
it and their estimate was a "shoot from the hip". I would have
guessed 35W based on the size of the cooler. I usually check
Xbitlabs, because they had a number of years where they used
to measure things.
Now, officially, all my information sources have dried up.
Absolutely nobody measures hardware properly any more.
Anandtech used a Hall probe for one article, and Tomshardware
has built custom instrumentation before, but these efforts
just don't stick around. And the GPUReview site, the operator
has given up on it, so even wild numbers aren't being added.
New cards aren't being added to the database. So for any
future video card efforts, the estimate information will be
thin indeed. Even the psucalculatorlite will be hard pressed
to provide the info. (I don't consider the NVidia/ATI info
to be that good, because they're unrealistic engineering
numbers intended for 3 sigma tails. Video cards now, I
think the high end power converters and control schemes,
are likely cutting off the distribution tails. The NVidia/ATI
info comes in second hand, and we have no idea what caveats
are involved. Even that 127W number could be too high.)
As another example, the HD 6870 received this comment:
http://www.xbitlabs.com/articles/graphics/display/asus-eah6870-directcu_4.html#sect0
"the ordinary Radeon HD 6870 consumes 124 watts in 3D mode"
And the 6870 is a higher clock than the 6850. So you know
the 127W number is too high. But I can't guess how much
lower, whether it's 100W or not.
And your system doesn't draw 375W when you first switch it on.
That HD 6870 draws 20W when idle in desktop. And the CPU won't
go to the wall in the BIOS either. As long as there are no
short circuits, the fans should at least come on. Even if
you had one of the gutless power supplies I have here,
it might only crash in the middle of a 3D game, and otherwise
be quite happy.
While a flat CMOS battery isn't always to blame, I'd give that
a check first, as it's an easy, low risk measurement you can do.
Set the multimeter to 20VDC fullscale, use the "volt/ohm" holes
on the face of the meter, touch black probe to chassis ground,
touch red probe to battery top surface (without sliding off
and hitting something else). If it reads 3.0, that's not the
problem. If it reads 2.3 or less, think about getting
another battery. As I said previously, all the PCs I have here,
an absolutely flat battery doesn't prevent them from
coming on. But there are motherboards out there, which have
that problem, so checking for the condition is a step in
debugging.
Paul
D
Dustin
DanS <[email protected]>
Sat, 20 Dec
2014 03:16:43 GMT in alt.comp.hardware.pc-homebuilt, wrote the
following message:
I've seen system builds that weren't done right. Things from loose ram
chips on the MB, to the MB being improperly installed in the case, and
shorting out against it. I've also seen various people improperly wire
the LEDs on the case as well as the buttons. IE: reset powered it
on/off heh. I've only seen two cases where the cpu was gooped with a
thick layer of thermalpaste which was, defeating the purpose...
I've seen a few newer system builds where the owner just didn't finish
connecting all of the power to the MB too. It varies...
Most of the time though, if it's not something as listed above, it's
usually a bad PS and/or mainboard, yep.
Sat, 20 Dec
2014 03:16:43 GMT in alt.comp.hardware.pc-homebuilt, wrote the
following message:
I've seen system builds that weren't done right. Things from loose ram
chips on the MB, to the MB being improperly installed in the case, and
shorting out against it. I've also seen various people improperly wire
the LEDs on the case as well as the buttons. IE: reset powered it
on/off heh. I've only seen two cases where the cpu was gooped with a
thick layer of thermalpaste which was, defeating the purpose...
I've seen a few newer system builds where the owner just didn't finish
connecting all of the power to the MB too. It varies...
Most of the time though, if it's not something as listed above, it's
usually a bad PS and/or mainboard, yep.
A
Adam
Paul said:
That's right, when "both" ATX power connectors (24-pin EATXPWR 2x12 and
8-pin EATX12V 2x4) are
connected to mobo (with CPU and RAM installed), PSU fan does not even
"twitch".
Mobo User Manual E6674 says ...
"The power supply plugs are designed to fit these connectors in only one
orientation."
One thing that prevents operation of PS_ON#, is the state
of the CMOS battery (CR2032 coin cell). With the multimeter set
to volts, you touch the top of that battery to get a reading.
It should be 3V. The minimum voltage is 2.0V + one schottky
diode drop or about 2.3V. If the battery is below 2.3V,
it might not start. It's when the battery is right down to
0.0V, that some motherboard logic simply refuses to turn
on the ATX Supply. Even though the green LED is on, the
motherboard is receiving +5VSB supervisory power, and in
theory, everything should work. I have not read of a reason
why some motherboards simply clam up, when the battery is
completely flat. I have plenty of PCs here, that start no
problem, with an absolutely flat CMOS battery. So I haven't
been able to reproduce the experience here.
Normally, I would say "take the CR2032 coin cell to Radio Shack
and have them test it". But at least one poster I gave the
advice to, managed to rip the entire CR2032 socket off the
board, ruining the motherboard. So now I have to include
the veiled warning "well, don't rip the battery socket
off the motherboard". The battery socket design is a damn
nuisance. Testing in place first, at least helps you decide
whether to attack and remove that battery
CMOS battery (CR2032 coin cell) multimeter voltage reading is ~2.8V.
Video card was removed to check CMOS battery.
Let me know when it should be connected again.
A
Adam
Paul said:
For my case, I'm thinking that 600W PSU should be sufficient.
Yep, I've never had to replace a CMOS battery before.
And, some of my computers are really-really old.
Maybe I've just been lucky or I don't work with enough computers.
But, it's definitely a good check.
P
Paul
Adam said:
The lock latches on the two power connectors, serve as a check
the orientation is correct as well. When installing, you can
visually check that the tab is going on the side with the bump.
You can check the physical 2x4 connector and compare
the pattern to the connector diagram in "10. ATX power connectors"
in the Asus manual. The four yellow wires should be on the latch
side. That's what the manual says. Verify your PSU is doing that.
If you had managed to reverse the connectors, you would at
least have got a twitch from the fan. And you don't have
even a twitch. So it's not likely you're being held
back by a short circuit, caused by ramming in the
connector the wrong way.
*******
Your CMOS battery is a little low. Repeat the test in
one month's time, to see if it has dropped below the knee (2.3V).
If takes about a month to slide down the knee, after it
starts to poop out.
The 2.8V value will not cause a problem with PS_ON#, so
this isn't the problem. (A value of 0.0V can cause problems,
on selected designs.)
*******
You can measure PS_ON# if you want, or measure PWR
as you push the button, and check the voltages.
You can use the screwdriver tip, instead of the
case front Power switch, to test whether it works or not.
It's already been suggested to you, it's a motherboard
problem. If I was working on it, I'd be checking the
various signals to see what they're doing, before
I'd spend the time on a warranty return. Nobody
wants to see a "no fault found" warranty cycle.
You can ground PS_ON# if you want, as a test case.
The only exposure by doing so, is the system then
has no opportunity to protect itself against
an overheating CPU. Because for as long as you've grounded
PS_ON#, the system is forced to stay powered. If the test
passes, the BIOS screen appears, then you're still left
with the problem of why the motherboard is not driving
PS_ON# to the logic low level like it should. Either
the PWR pulse isn't coming in the front panel pins,
or the open collector driver on the motherboard is toast,
and can't drive PS_ON# low enough. The multimeter, connected
to the PS_ON# while it is plugged into the motherboard,
can help you determine whether PS_ON# is 0.4V (good),
2.0V (PSU half on), or 5.0V (no effort was made to
switch on at all).
Paul
A
Adam
Paul said:
1) "both" ATX power connectors (24-pin EATXPWR 2x12 and 8-pin EATX12V
2x4) are
connected to mobo (with CPU and RAM installed, no video card)
2) +5VSB/purple pin9 of 24-pin EATXPWR measured at 5.16V when PSU=ON
3) PS_ON#/green pin16 of 24-pin EATXPWR measured at 4.54V when PSU=ON
So far, I have not attempted to short PWR-GND to turn on the PC.
The standby/SB_PWR green LED have never been a problem.
Shorting PWR-GND with a screwdriver does "not" cause
voltage drop of PS_ON# green at all. It remains high at 4.54V.
Looks like a faulty mobo? How do I send it back to ASUS ?
What about the CPU that is already installed?
So, no power because PS_ON# never drops. Geez, so simple.
A
Adam
Paul said:
Looks like the replacement mobo will take care of the low CMOS battery.
Should I still manually pull PS_ON# low or just RMA the mobo ?
M
mike
Switch your meter to AC volts and measure again.
For most any meter you've likely purchased, it should read approximately
zero VAC. Noise level, but less than a volt.
Not a sure test, oscilloscope is better, but if there's a lot of AC volts,
your power supply ain't right. I've seen that cause a computer to fail
to turn on.
If you're gonna ship it back, you have to remove the processor anyway.
Pull the processor and see if the power-on switch twitches the fans.
I recall you mentioning that the processor was hard to install.
If you were careful and got it all put together correctly, it should
have been easy.
P
Paul
Adam said:2) +5VSB/purple pin9 of 24-pin EATXPWR measured at 5.16V when PSU=ON
3) PS_ON#/green pin16 of 24-pin EATXPWR measured at 4.54V when PSU=ON
The standby/SB_PWR green LED have never been a problem.
Shorting PWR-GND with a screwdriver does "not" cause
voltage drop of PS_ON# green at all. It remains high at 4.54V.
Looks like a faulty mobo? How do I send it back to ASUS ?
What about the CPU that is already installed?
So, no power because PS_ON# never drops. Geez, so simple.
Yes, you have a faulty motherboard.
If you turn off the PSU via the switch at the rear, wait
30 seconds, turn on again, I would expect both purple and
green to read 5.16. Whether they both go to the same value,
depends on the open collector driver not drawing any current
at all (when it's not supposed to be doing that).
It could be, in measurement (3), the driver has only pulled
down a tiny bit (5.16 - 4.54 = 0.62). Which isn't very good
considering how weak the pullup on the PSU is.
You can ground the PS_ON# green if you want, and actually
power up the system, but there is really no point if the
motherboard "isn't a keeper". It would just be an unnecessary
risk.
If you were still in the 30 day or 7 day retail window, you
could get another board from the retailer. Any longer than that,
and it's warranty time.
I wish I knew why shit like this happens... Makes no sense.
*******
Is the motherboard still housed inside the computer case ?
Did you consider the other advice you got, such as "the motherboard
could be shorted to something" ? Note that, if you haven't
centered the motherboard properly, a screw head could touch where
it's not supposed to. The nine mounting holes have exposed copper
and are meant to be grounded. But the head of the screw should be
precisely centered over the "donut". The motherboard designer
is supposed to define "keepout zones", so the screw head won't
touch any analog components. The screw heads *are* supposed
to touch the donuts - it's part of the grounding strategy.
The screw heads aren't allowed to touch resistors, caps, or the
like.
*******
What's funny about the 0.62V voltage drop, is that is one
silicon diode drop below +5VSB. As if a semiconductor junction
is "holding up" the signal. Like this. Now, if you were to
ground PS_ON# at this very time, you'd burn out that phantom
diode
So don't do that, if you plan to RMA on warranty...-----+----------+ 5.16
| |
+ | pullup
diode resistor
- | |
| +------ (PS_ON#) 4.54V
+----------+
|
SUSC ___________|/ Open collector
|\ Driver
|
|
GND
HTH,
Paul
P
Paul
Adam said:
I might be tempted to do a "table test". Move the
motherboard outside the computer case, to rule out
a short circuit from something touching that isn't
supposed to be touching.
In my other post, I have a wild theory on the voltage
value, and don't recommend grounding PS_ON# at this
time.
I'd sooner try the table test instead, which is
to sit the motherboard on a piece of cardboard, and
test outside the case. You don't even need to leave
the video card installed in there. The objective is
to get those fans spinning, not bring the thing up
completely. It'll beep if there is no video.
No problem.
Paul
A
Adam
mike said:
2.5) +5VSB/purple pin9 of 24-pin EATXPWR measured at 0VAC when PSU=ON
Dropping the CPU into the socket was not the hard part.
The hard part was the fan assembly, which took a bit more muscle, sweat and
effort.
Not removing the CPU yet. Not until the very-very end.
3.5) PS_ON#/green pin16 of 24-pin EATXPWR measured at 0VAC when PSU=ON
A
Adam
Paul said:
Yes, the mobo is still in the Antec case. But, I will be removing it since
another one will be arriving to replace it in about a week. So,
I can do measurements/tests again when the mobo is out of the case.
I tried to be extremely careful about positioning/centering of everything
when
installing the mobo.
M
mike
Ditto if the case has a stud sticking up where it shouldn't.
Now that you have a meter, you can measure ohms from each of the
power supplies to ground.
I tried it on an old Dell Dimension. A short to ground would be obvious.
The nine mounting holes have exposed copper
D
Dustin
"Adam" <adam@no_thanks.com> Wed, 17
Dec 2014 21:30:18 GMT in alt.comp.hardware.pc-homebuilt, wrote the
following message:
Just a suggestion and feel free to take it however you like,
In the future, when seeking help on usenet, it wouldn't hurt to respond
to posters (you tend to respond, but vaguely and only to a few of us
asking you questions?) who ask you questions concerning the problem you
posted about having.
It would also be nice to keep us uptodate on what you have/haven't
tried. And finally, if you've decided to go with a particular posters
advice, you could let the rest of us monitoring the thread know.
Nobody likes wasting their time or repeating themselves in the event
someone else has already made one or more of the same suggestions. It
makes troubleshooting a bit of a hassle this way too.
Good luck with the box and I hope you get it figured out in short order.
Dec 2014 21:30:18 GMT in alt.comp.hardware.pc-homebuilt, wrote the
following message:
Just a suggestion and feel free to take it however you like,
In the future, when seeking help on usenet, it wouldn't hurt to respond
to posters (you tend to respond, but vaguely and only to a few of us
asking you questions?) who ask you questions concerning the problem you
posted about having.
It would also be nice to keep us uptodate on what you have/haven't
tried. And finally, if you've decided to go with a particular posters
advice, you could let the rest of us monitoring the thread know.
Nobody likes wasting their time or repeating themselves in the event
someone else has already made one or more of the same suggestions. It
makes troubleshooting a bit of a hassle this way too.
Good luck with the box and I hope you get it figured out in short order.
D
DecadentLinuxUserNumeroUno
It isn't a book, and if you followed the thread...
Well... anyway, he could have given some more feedback, but you could
read more than just the point at which you stepped into a thread at as
well. It isn't like it is the size of a novel.