Re: 5v rail low

The message <13ibosiijr3j5b1@xxxxxxxxxxxxxxxxxx>
from "Roger Cain" <rogerspamtrap@xxxxxxxxxxxxxx> contains these words:

Thanks a lot, chaps, for all the hints and tips. I have had a tinker over
the weekend.

To clear up the config:
1.6G Athlon running on an iWill KK266 mobo.
2x 256Mb RAM
1 x HD; 2 x optical
Radeon 9500 video.
Sundry low consumption PCI, USB etc.

PSU calculators on the web tell me I need upwards of 284W to fire this up.
The new PSU is an el-cheapo Sumvision 400W.
I used 3 monitoring points; the mobo sensor, a spare molex and the pins I
could reach on the mobo side of the ATX power connector. The readings for
the +12v rail were consistent and slightly above spec. at all 3 points. As
the 12v never changed during the readings I won't mention it again. The 5v
line measured at the molex and on the mobo side of the connector also did
not change.

Assuming your voltmeter is in reasonable calibration, it looks as
though the PSU is set a little on the low side for the 5v rail. Most
smpsus manage to keep this within a tenth of a volt of the specified
voltage (5v rail). The 12v rails usually rely on the 5v regulation and
turns ratio which means below nominal voltage when the 5v power draw is
light and above nominal for heavy loading of the 5v supply rail. This
fine as long it stays within 10% of nominal since this rail isn't quite
as critical.

On a cold start the measured 5v was 4.7 and the mobo-sensed 4.6.

This isn't a huge discrepency but the MoBo sensor reading a little bit
below reality on a supply rail that's already on the low side doesn't
create a climate of confidence in the MoBo's monitoring software.

On removing AGP, PCI, drive components one by one the only significant
change was caused by removing the Radeon board which caused the measured to
rise to 4.9 although the sensed voltage did not change at all.

Looks like that Radeon takes a fair bit of juice but it's curious that
the sensor reading stays rock steady.

When the kit is fully warmed up the measured V is 4.8 but the sensed V
decreases to 4.5.

Doesn't give you much confidence in that MoBo sensor, does it?

I investigated the mobo, especially the banks of caps. around the CPU. No
signs of doming, weeping and all felt the same temperature to the touch.
Since the new PSU was installed there have been no involuntary reboots.

The PSU is specced for +/- 5% on the 5 and 3.3 volt rails to guarantee
that it will meet the +/- 10% specification of the MoBo chips supply
rail voltage tolerance. Provided the 5v rail stays above 4.5v (and the
3.3v rail stays above 3.0v), there _shouldn't_ be any problem (at least
in theory :-).

So ... the upshot is:
1. The Radeon (processor plus fan) takes a small bite at the 5v line.
2. There is a difference of about 0.35v between the metered and sensed 5v
levels. Other sensed levels are close enough to those measured.

Maybe the mobo 5v sensor is duff?

Not duff, exactly, just reading low from a point that's sensitive to
loading by the graphics adapter. Not the best of design layouts.

Maybe there really IS a current drain deep with in the mobo which occurs
near enough the sensor to be detectable but which is sufficiently decoupled
from the main 5v bus so is not detectable at my metering points.

Almost all socket A MoBo designs power the cpu vrm from the 5v rail,
you could be looking at circa 10 amps just for the CPU when fully
occupied.

Hmmm .... I could always prod around the mobo looking for under-spec 5V to
try to localise but, as I can visualise blowing an IC through clumsiness,
I'd better not try :-)

Unless you've got access to special chip leg meter probes, I'd
recommend against such an exercise. The risk of damage is exactly as
you've visualised.

I guess I will just have to tolerate this mystery until something else
happens and make a note-to-self to spend more than £9 next time I need a
PSU.

Yeah, an 18 quid PSU might have been a wiser choice. :-)

--
Regards, John.

Please remove the "ohggcyht" before replying.
The address has been munged to reject Spam-bots.

.

Quantcast