Recommended power supply wattage.

[David W. Hodgins]

I've been having some trouble with a new system (about a month old),
that spontaneously reboots, with no error messages in /var/log/syslog,
or other log files. The system will run fine for a couple of days,
and reboot 4 times the next day, sometimes rebooting again while still
running the startup scripts.

I believe the cause is two small of a power supply, having found today,
that adding another usb device prevents the system from even running
the memory check. Removing any one of the usb devices allows it to
boot, but with all present, it fails.

Primary System config ..
Mb Gigabyte EP45T-UD3LR
Cpu Intel(R) Core(TM)2 Quad CPU Q9550 @ 2.83GHz
Ram 2x2GB ddr3 DIMM 400 MHz (2.5 ns) (Total 4GB)
Ide
- HD Maxtor 6Y080L0 80GB drive
- DVD LG GH22NP20
Sata
- 2x Seagate ST31500341AS 1500GB drives (Total 3TB)

self powered usb hub with parallel port connected to printer, usb card
reader, and usb speakers

With the above config, the system reboots at unpredictable times.

Replacing the ps2 keyboard and mouse, with a usb cordless keyboard/mouse,
the system will not boot, unless I remove the usb hub, or speakers
(after which the cordless keyboard/mouse work fine).

The system currently has a power supply rated at 500W output.

Should that be enough? Is this a bad power supply, or is it simply too
small? Would 750W be enough?

If more detailed info is needed, the full lshw output is
at http://www.ody.ca/~dwhodgins/lshw.txt

Thanks in advance.

 

[Paul]

I've been having some trouble with a new system (about a month old),
that spontaneously reboots, with no error messages in /var/log/syslog,
or other log files. The system will run fine for a couple of days,
and reboot 4 times the next day, sometimes rebooting again while still
running the startup scripts.

I believe the cause is two small of a power supply, having found today,
that adding another usb device prevents the system from even running
the memory check. Removing any one of the usb devices allows it to
boot, but with all present, it fails.

Primary System config ..
Mb Gigabyte EP45T-UD3LR
Cpu Intel(R) Core(TM)2 Quad CPU Q9550 @ 2.83GHz
Ram 2x2GB ddr3 DIMM 400 MHz (2.5 ns) (Total 4GB)
Ide
- HD Maxtor 6Y080L0 80GB drive
- DVD LG GH22NP20
Sata
- 2x Seagate ST31500341AS 1500GB drives (Total 3TB)

self powered usb hub with parallel port connected to printer, usb card
reader, and usb speakers

With the above config, the system reboots at unpredictable times.

Replacing the ps2 keyboard and mouse, with a usb cordless keyboard/mouse,
the system will not boot, unless I remove the usb hub, or speakers
(after which the cordless keyboard/mouse work fine).

The system currently has a power supply rated at 500W output.

Should that be enough? Is this a bad power supply, or is it simply too
small? Would 750W be enough?

If more detailed info is needed, the full lshw output is
at http://www.ody.ca/~dwhodgins/lshw.txt

Thanks in advance.

Can you give enough info, so we can see a picture of the label on the
side of the supply. There is lots of info on there. At the very
least, the PSU make and model number. Or ideally, a URL that points
to a picture of the power supply label. Newegg has pictures that
can help with that, if the supply is new enough.

If the computer just stopped, rather than rebooting, that could mean
something different. So, you're saying the computer abruptly reboots,
regardless of what the system is doing ? Or is the system doing
"heavy computing" just before the event ? If you run a copy of
Prime95, does it happen soon after the code starts to run ? Or
are your reboots totally random, and happening even when the
system is idle ?

On a Windows system, you can set the machine to blue screen and stop
on an error. I don't know if Linux offers an option to leave a
trail of breadcrumbs or not. On Windows, you'd collect all the
crash events, and see if they were randomly distributed or
clustered around one driver. (A random distribution might suggest
a memory problem.) Running a memory test and running Prime95 later,
are examples of vetting a system, before using the computer for
serious work. (While ECC makes the detection of memory problems
easier, it isn't exactly an easy to find feature.) Prime95 is
available for both Windows and Linux, so can be run in either
environment.

(Sometimes this site is hard to reach.)

http://www.mersenne.org/freesoft

Also, in your hardware list above, I don't see a video card listed.
Your lshw shows a 9600 GT. That is 25.5W idle, and 59.7W at max 3D.
I would think a 500W power supply could handle that load. But I
really want to read all the numbers off the label. There have been
some absolutely dreadful 500W supplies (ones with such weak 12V
output, that a good quality 350W beats them in real performance).

http://www.xbitlabs.com/articles/video/display/gainward-bliss9600gt-512gs_7.html#sect0

Paul

 

[David W. Hodgins]

Can you give enough info, so we can see a picture of the label on the
side of the supply. There is lots of info on there. At the very
least, the PSU make and model number. Or ideally, a URL that points
to a picture of the power supply label. Newegg has pictures that
can help with that, if the supply is new enough.

The computer is at my sister's house, and I won't be back there till
Monday or Tuesday.

If the computer just stopped, rather than rebooting, that could mean
something different. So, you're saying the computer abruptly reboots,
regardless of what the system is doing ? Or is the system doing
"heavy computing" just before the event ? If you run a copy of

/var/log/syslog shows it rebooted at 0627 today, during which time
it would have been idle. All of the cron jobs finished by 0410.

It didn't reboot at all yesterday. The day before ...
Sep 24 12:29:22 drew643 syslogd 1.5.0: restart.
Sep 24 17:35:26 drew643 syslogd 1.5.0: restart.
Sep 24 21:43:28 drew643 syslogd 1.5.0: restart.
Sep 24 21:59:49 drew643 syslogd 1.5.0: restart.
Sep 24 22:01:32 drew643 syslogd 1.5.0: restart.

It has happened while a user was playing solitare, and when
completely idle. At present, the heaviest cpu usage is the
snapshot making a backup at 1 minute after each hour. So
far the longest the backup has taken, is 2 minutes.

On a Windows system, you can set the machine to blue screen and stop
on an error. I don't know if Linux offers an option to leave a
trail of breadcrumbs or not. On Windows, you'd collect all the
crash events, and see if they were randomly distributed or

If the problem was a software problem, I'd expect there to be a message
in /var/log/syslog, unless the system locked up in a device wait, in
which case it would not reboot without someone pressing the reset key.

clustered around one driver. (A random distribution might suggest
a memory problem.) Running a memory test and running Prime95 later,

I ran memtest for 24hours, and badblocks on both of the new drives.

I'd expect ram or drive errors to generate error messages in syslog.
smart is enabled on all disk drives, and isn't showing any errors
detected.

http://www.mersenne.org/freesoft

I'll get a copy of prime95, and try it out.

Also, in your hardware list above, I don't see a video card listed.
Your lshw shows a 9600 GT. That is 25.5W idle, and 59.7W at max 3D.

Oversite on my part. Yes, it's a nvidia 9600GT hooked up to an lcd
monitor.

I would think a 500W power supply could handle that load. But I
really want to read all the numbers off the label. There have been

I'll get those numbers off of the power supply itself on Mon. or Tues.

some absolutely dreadful 500W supplies (ones with such weak 12V
output, that a good quality 350W beats them in real performance).
http://www.xbitlabs.com/articles/video/display/gainward-bliss9600gt-512gs_7.html#sect0

Thanks, Dave Hodgins

 

[ShadowTek]

I ran memtest for 24hours, and badblocks on both of the new drives.

I once ran memtest for 24 hours, recieved no errors, thought I had fixed
the problem, had more crashes, then ran memtest for 48 eours, and finally saw
some errors. So, I've learned that sometimes even a whole day is not
enough.

Oversite on my part. Yes, it's a nvidia 9600GT hooked up to an lcd
monitor.

I also have a 9600GT with 500w Antec PSU. I've never had any problems with it.

I think the nVida site recommended a 500w PSU for this particular card,
but you can check and see.

 

[David W. Hodgins]

The real question is as Paul mentioned, how good the PSU is
(asuming a modern design with bias towards more 12V current
rather than maximum 5V current that old generations of PSU
were designed towards), as a decent, accurately rated 400W
PSU with plenty of current on the 12V rail should power the
system fine.

Interesting. When I first got the system, it had a sata dvd
drive, which died after about 2 weeks. It was replaced with
an ide dvd drive (as the source only had ide drives that had
connectors for the audio card, which linux requires). The
reboot problems only started after replacing the sata dvd drive,
which I gather runs off of the 12V rail, with an ide one, which
I understand runs off of the 5V rail.

My current guess, is that this brand new power supply doesn't
have enough output on the 5V rail, to handle an ide hard drive,
an ide dvd drive, plus usb devices.

Given the times of day when it tends to fail, my guess is the
input voltage is dropping too low. In this area, I've seen the
input AC voltage vary between 112VAC, and 128VAC.

I think I will take the system back to the provider, and demand
a higher wattage power supply, particularly on the 5V rail.

Regards, Dave Hodgins

 

[David W. Hodgins]

I also have a 9600GT with 500w Antec PSU. I've never had any problems with it.

Thanks for the info.

I've had a lot of problems with it. I'm hoping it turns out the power
supply turns out to be the cause. Right now, in order to try and get
the system stable, I'm running a 32bit desktop kernel, with the vesa
video driver, resulting in 1.5GB of ram not being used, and any programs
requiring advanced graphic capabilities, not being usable.

This system should be able to run a 64bit kernel, with the latest
nvidia drivers, but currently it is just not stable.

I've decided to replace to power supply. If that makes it stop
randomly rebooting, I'll try the latest nvidia driver. If that
works, I'll switch to a 32 bit kernel with highmem support. If
that works, I'll install a 64 bit system kernel.

Regards, Dave Hodgins

 

[David W. Hodgins]

Can you give enough info, so we can see a picture of the label on the

I had another look at the system today.

I can't read the brand. It looks like MIOS, in a large script, that's
mostly covered by the edge of the case. The info on the label is ...

MIOS switching power supply
Model No. KY-500ATX
AC INPUT 115/230V~7A/4A,50-60Hz
OUTPUT:500W
+3.3V +5V +12V +5VSB
30.0A 32.0A 0.5A 2.0A

Strange that it says on the label it's 500W, but all of the KY-500ATX
models I find on google say that model number is a 400W power supply.

With three usb devices connected, turning the machine on causes it
to cycle through the post up to the point where it should be checking
the ram, and then it reboots.

Removing any one of the three devices allows it to boot.

I measured the voltages on wires going to an unused power connector
both with the three usb devices connected, and without any usb devices.

Colour WithUsb WithoutUsb
Yellow 12.11 12.2
Red 5.21 5.2
Orange 3.40 3.40

Any suggestions appreciated.

Thanks, Dave Hodgins

 

[David W. Hodgins]

MIOS switching power supply
Model No. KY-500ATX
AC INPUT 115/230V~7A/4A,50-60Hz
OUTPUT:500W
+3.3V +5V +12V +5VSB
30.0A 32.0A 0.5A 2.0A

Think I've found the problem. According to
http://forums.nvidia.com/lofiversion/index.php?t97431.html
the nvidia geforce 9600gt requires 26amps on the +12v rail.

This power supply doesn't even come close. I'm surprised it
boots at all.

Regards, Dave Hodgins

 

[Paul]

I had another look at the system today.

I can't read the brand. It looks like MIOS, in a large script, that's
mostly covered by the edge of the case. The info on the label is ...

MIOS switching power supply
Model No. KY-500ATX
AC INPUT 115/230V~7A/4A,50-60Hz
OUTPUT:500W
+3.3V +5V +12V +5VSB
30.0A 32.0A 0.5A 2.0A

Strange that it says on the label it's 500W, but all of the KY-500ATX
models I find on google say that model number is a 400W power supply.

With three usb devices connected, turning the machine on causes it
to cycle through the post up to the point where it should be checking
the ram, and then it reboots.

Removing any one of the three devices allows it to boot.

I measured the voltages on wires going to an unused power connector
both with the three usb devices connected, and without any usb devices.

Colour WithUsb WithoutUsb
Yellow 12.11 12.2
Red 5.21 5.2
Orange 3.40 3.40

Any suggestions appreciated.

Thanks, Dave Hodgins

I can find one other report for the MIOS KY-500ATX.
His computer restarts as well. Video card 9600 GT.

http://www.tomshardware.com/forum/264622-28-amps

The one interesting rail you missed in your readings, is the voltage on +5VSB.
If +5VSB drops, the system should stop. USB may be powered from that rail,
as your motherboard is pretty new.

The article archived in Toms (Google won't give it to me),
says the MIOS is rated 12V @ 24 amps.

Q9550 = 95W, 8.8 amps at 90% Vcore efficiency at max, from +12V.
Hard drive 0.6 amps at idle (times x 3)
Fans 0.5 amps
Video, say 5 amps max, 2 amps in the BIOS
CDROM 1.5 amps (zero, if no media in the tray)

Even if they all maxed, you're still short of 24 amps
by a good margin.

The processor won't run at max power at startup.
In the events you've listed, I'm guessing there isn't
a CDROM in the tray. That leaves a lot less than 24 amps
of total load. The video card won't go to 5 amps, unless
you're in a 3D game.

Since I can find at least three different brands using
the KY-500ATX model number, I'd have to conclude it
is an "OEM wonder", where they'll write virtually anything
they feel like on the label.

I'm not sure a multimeter is going to be enough to
catch the thing screwing up. You could include +5VSB in
your measurements, but if +5VSB dropped out, the system
would shut down. It likely wouldn't restart. So the symptoms
don't match. That leave the +12V being a load of bollocks,
and not an honest rating.

I think I'd just replace it. I;m not sure what else you
can do at this point. It is always good to have a
spare around anyway, so buy something nice that you
can reuse in a new build later.

If you want, you could try removing enough hardware
to get a good boot going, then run memtest/Prime95 or
whatever other stability checks you want. (Disconnect
the hard drives, boot from a LiveCD.) If the computer
passes those, and yet won't when extra load is added to
the supply, then your testing is done (and your PSU
goes in the trash). That would prove the other key
components in the system are working well.

Prime95 is available for Linux and Windows, and supports
loading all cores of a multicore processor. I like to see
a four hour run with no errors detected. Other people like
a slightly longer run.

Paul

 

[larry moe 'n curly]

Cpu Intel(R) Core(TM)2 Quad CPU Q9550 @ 2.83GHz
HD Maxtor 6Y080L0 80GB drive
DVD LG GH22NP20
Sata - 2x Seagate ST31500341AS 1500GB drives (Total 3TB)
Asus 9600GT
MIOS switching power supply
Model No. KY-500ATX
OUTPUT:500W
+3.3V +5V +12V +5VSB
30.0A 32.0A [18A - corrected] 2.0A

TweakTown says a 9600GT graphics card uses 140-250 watts, or 12-21
amps:

http://images.tweaktown.com/content/1/3/1373_11.gif

So considering that the CPU probably draws another 5-10 amps and the
PSU is rated for just 18A @ +12V, you're probably right that it's too
weak for your system. But I have doubts this particular model PSU can
continuously put out 18A @ 12V or 500W, especially in a computer case
running at 30-40 Celcius (common) because the UL number, E210743, says
it's made by Sun Pro, which isn't exactly in high regard. I couldn't
find a review of any Sun Pro PSU, but if your PSU resembles this one
internally:

http://tw.pic.bysources.com/sample/6089.jpg

then it's not really a 500W unit but more like a 300W one.

Stick with PSUs manufactured by good companies, like Zippy-Emacs,
Enhance, Etasis, Lite-On, Delta, Fortron-Source, or Seasonic. The
last companies supply Antec, while OCZ uses Seasonic, and Corsair uses
Fortron-Source. JonnyGuru.com, HardOCP.com, HardwareSecrets.com, and
XbitLabs.com are some websites that do thorough reviews of PSUs.

 

[David W. Hodgins]

should be considered the primary suspect. If you were to
open it, take a high-res picture and post that somewhere
like http://www.imageshack.us , someone might recognize who
made it, what it really is.

I've taken the system back to the shop that I purchased it from,
last month. They're in the process of replacing the power supply
with one that has a higher +12v amperage.

The tech I talked to, understood the problem, but was surprised by
it, as they haven't seen the problem with the same power supply and
video card, on other systems they've built, but those systems didn't
have a quad core, 4GB of ddr3 ram, with two 1.5TB 7200rpm sata
drives.

The combination of hardware in this system was just enough to
overtax the power supply, although it could run for a couple of days
and appear to be ok, but would then spontaneously reboot.

Being able to consistently have the system fail to boot, simply by
adding more usb devices, was enough to prove to their satisfaction,
that the power supply was not adequate.

I hate intermittent problems like this, as it's difficult to know
whether or not you've fixed the problem, until the system has been
running for quite a while.

Regards, Dave Hodgins

 

[David W. Hodgins]

Maybe, but if they are using bottom-end generic PSU do you
reasonably expect them to concede it's junk?

Just a followup, to confirm replacing the power supply fixed the
problem of spontaneous reboots, and occasional Xorg lockups,
on a system with a nvidia geforce 9600gt video card, 4gb ddr3 ram,
quad core cpu, two 1.5TB 7200rpm sata drives, and an 80GB ide hard
drive.

I'd misread the label on the original power supply. It did have
24 amps on the +12v rail, but that wasn't enough for all of the
stuff I had in the system.

The power supply was replaced with a Thermaltake 650W supply,
with up to 52 amps on the +12v rail.

The system has been running steadily for about 2 weeks now. Yeah!


Regards, Dave Hodgins