SCSI resets caused by weird PSU problem

[[H]omer]

In the followup to last weeks saga of the exploding PSU...

This one really has me stumped:

I've got a lot of equipment in one of my full tower PCs, mainly U160 SCSI
hard disks, and so I figured I needed a fairly beefy PSU.

I bought a 550W Enermax PSU at the same time as upgrading the disks, but
for reasons I couldn't explain, it failed to power up the system. It got
as far as the POST and then I could hear the SCSI disks repeatedly trying
(and failing) to spin up. I gave up with the Enermax and switched back to
an old 235W PSU I had lying around.

Although this configuration worked without problem, I was really nervous
about running so much equipment on just 235W, so eventually I tried yet
another 550W PSU - this time an Antec. Well this worked for about a 3
months, until, well - see the "My PC Exploded" post of 26th June 2004:
in alt.comp.hardware.

I switched back to the Enermax, which seemed to work for a few days,
until today. The first symptom was occasional SCSI resets, which became
worse and worse. I had heard that turning off Domain Validation on the
Adaptec 29160N can sometimes prevent this. It did for a couple of hours,
but then things got worse from there. Finally the PC wouldn't even boot,
once again struggling to power up the SCSI disks.

I've now switched back to the puny 235W PSU, and everything is working
fine, but I'm still not keen on using such a small PSU. I should add, that
I've used this 235W PSU in this configuration before, for more than 2
years, without problem (still makes me nervous though).

I can't understand why two different monster 550W PSUs, from two different
manufacturers, cannot cope with the same system that a tiny 235W PSU can
easily cope with, especially such "high quality" brand name products.

I've had all three PSUs from new, and the exploding Antec is still under
warranty (not the other two though).

I must have build around 200 systems in my time, and I've never
encountered a weird problem like this. I've double checked for floating
joints, all cables and connections - including data cables, and verified
the integrity of all the disks at BIOS level. All voltages and electrical
pathways pass OK, but these 550W PSUs just aren't playing the game.

Any clues?

-
[H]omer

 

[Juhan Leemet]

I bought a 550W Enermax PSU at the same time as upgrading the disks, but
for reasons I couldn't explain, it failed to power up the system...

That should be more than enough. Maybe that's the problem? Some
(switching) supplies don't like it if they don't have enough load.

...first symptom was occasional SCSI resets, which
became worse and worse...

In another post on another newsgroup (or was it an e-mail?), someone
pointed out that "power blip" can cause a SCSI bus RESET (e.g. through
passive terminator at the end? depending on where it gets its power).

I can't understand why two different monster 550W PSUs, from two
different manufacturers, cannot cope with the same system that a tiny
235W PSU can easily cope with, especially such "high quality" brand name
products.

Suggests to me that you don't have enough load? How many disks? Each disk
is something like 10 or 15 watts, unless really busy. CPUs use the most
(about 50 watts? more?). Rest doesn't take that much, unless you have
lots. If you can run on 235W then you don't need 550W. Maybe you're only
using a little over 100W? Do you have a semi-smart UPS? Those can measure
the actual power consumption (on the AC side), er, more or less.

 

[VWWall]

In the followup to last weeks saga of the exploding PSU...

This one really has me stumped:

I've got a lot of equipment in one of my full tower PCs, mainly U160 SCSI
hard disks, and so I figured I needed a fairly beefy PSU.

I bought a 550W Enermax PSU at the same time as upgrading the disks, but
for reasons I couldn't explain, it failed to power up the system. It got
as far as the POST and then I could hear the SCSI disks repeatedly trying
(and failing) to spin up. I gave up with the Enermax and switched back to
an old 235W PSU I had lying around.

Although this configuration worked without problem, I was really nervous
about running so much equipment on just 235W, so eventually I tried yet
another 550W PSU - this time an Antec. Well this worked for about a 3
months, until, well - see the "My PC Exploded" post of 26th June 2004:
in alt.comp.hardware.

I switched back to the Enermax, which seemed to work for a few days,
until today. The first symptom was occasional SCSI resets, which became
worse and worse. I had heard that turning off Domain Validation on the
Adaptec 29160N can sometimes prevent this. It did for a couple of hours,
but then things got worse from there. Finally the PC wouldn't even boot,
once again struggling to power up the SCSI disks.

I've now switched back to the puny 235W PSU, and everything is working
fine, but I'm still not keen on using such a small PSU. I should add, that
I've used this 235W PSU in this configuration before, for more than 2
years, without problem (still makes me nervous though).

I can't understand why two different monster 550W PSUs, from two different
manufacturers, cannot cope with the same system that a tiny 235W PSU can
easily cope with, especially such "high quality" brand name products.

I've had all three PSUs from new, and the exploding Antec is still under
warranty (not the other two though).

I must have build around 200 systems in my time, and I've never
encountered a weird problem like this. I've double checked for floating
joints, all cables and connections - including data cables, and verified
the integrity of all the disks at BIOS level. All voltages and electrical
pathways pass OK, but these 550W PSUs just aren't playing the game.

Any clues?

Did you check the voltages at start-up, especially the 12V rail? The
higher wattage supplies my regulate differently from the 235W one. In
this case a D'Arsonval, (moving coil), voltmeter might be better than a
digital one, since some of these are not very good at reading rapid
voltage changes. Trying to hold the 12V rail constant might put a blip
on the 5V to the SCSI electronics, so check that too. A scope would be
best if you can find one to try. Just WAGs on my part but easy to try.
Let us know what happens.

Virg Wall

 

[[H]omer]

[H]omer wrote:

I bought a 550W Enermax PSU at the same time as upgrading the disks, but
for reasons I couldn't explain, it failed to power up the system.
I gave up with the Enermax and switched back to an old 235W PSU
Although this configuration worked without problem, I was really
nervous about running so much equipment on just 235W, so eventually I
tried yet another 550W PSU - this time an Antec. Well this worked for
about a 3 months
I switched back to the Enermax, which seemed to work for a few days,
until today.
I've now switched back to the puny 235W PSU, and everything is working
fine
I can't understand why two different monster 550W PSUs, from two
different manufacturers, cannot cope

Did you check the voltages at start-up, especially the 12V rail?

The startup values are normal, if you mean the first few microseconds of
current, it soon levels out to the following:

The 5V rail is lower than I'm happy with, although barely within spec, at
around 4.65V +/- 0.1V.

The 12V rail is a healthy 12.03V, but varies more by around +0.02/-0.18V.

Switching to the 235W PSU, I get a much better reading on the 5V rail,
5.03V +/- 0.02, but a poorer 12V reading at 11.84V +/- 0.07V.

The higher wattage supplies my regulate differently from the 235W one.

I need a high rating PSU with good 5V regulation, none of the 550W
PSUs I've tried seems to fit that spec. Damn. Is there some kind of
special PSU (e.g. for servers) that I could try?

In this case a D'Arsonval, (moving coil), voltmeter might be better than
a digital one, since some of these are not very good at reading rapid
voltage changes.

The analogue readings (from a cheap Protek) seem to correlate to the
digital readings on my other test equipment.

Trying to hold the 12V rail constant might put a blip on the 5V to the
SCSI electronics

Seems the mostly likely explanation, but what's the solution?

-
[H]omer

 

[Pen]

For whatever it's worth, your 550W supplies are out
of spec on the 5V rail. The 235 is in spec on both
rails. The 5V rail should never go below 4.75V.

[H]omer wrote:

I bought a 550W Enermax PSU at the same time as upgrading the disks, but
for reasons I couldn't explain, it failed to power up the

system. PSU

Did you check the voltages at start-up, especially the 12V

rail?

The startup values are normal, if you mean the first few microseconds of
current, it soon levels out to the following:

The 5V rail is lower than I'm happy with, although barely within spec, at
around 4.65V +/- 0.1V.

The 12V rail is a healthy 12.03V, but varies more by around +0.02/-0.18V.

Switching to the 235W PSU, I get a much better reading on the 5V rail,
5.03V +/- 0.02, but a poorer 12V reading at 11.84V +/- 0.07V.

The higher wattage supplies my regulate differently from the

235W one.

I need a high rating PSU with good 5V regulation, none of the 550W
PSUs I've tried seems to fit that spec. Damn. Is there some kind of
special PSU (e.g. for servers) that I could try?

In this case a D'Arsonval, (moving coil), voltmeter might be better than
a digital one, since some of these are not very good at reading rapid
voltage changes.

The analogue readings (from a cheap Protek) seem to correlate to the
digital readings on my other test equipment.

Trying to hold the 12V rail constant might put a blip on the 5V to the
SCSI electronics

Seems the mostly likely explanation, but what's the solution?

-
[H]omer

 

[kony]

For whatever it's worth, your 550W supplies are out
of spec on the 5V rail. The 235 is in spec on both
rails. The 5V rail should never go below 4.75V.

True, and even 4.75, is a sign of a serious problem. Many
m'boards these days, running modern CPUs, will further drop the
5V level by the time it reaches the VR circuit.

 

[Michael Hawes]

In the followup to last weeks saga of the exploding PSU...

This one really has me stumped:

I've got a lot of equipment in one of my full tower PCs, mainly U160 SCSI
hard disks, and so I figured I needed a fairly beefy PSU.

I bought a 550W Enermax PSU at the same time as upgrading the disks, but
for reasons I couldn't explain, it failed to power up the system. It got
as far as the POST and then I could hear the SCSI disks repeatedly trying
(and failing) to spin up. I gave up with the Enermax and switched back to
an old 235W PSU I had lying around.

Although this configuration worked without problem, I was really nervous
about running so much equipment on just 235W, so eventually I tried yet
another 550W PSU - this time an Antec. Well this worked for about a 3
months, until, well - see the "My PC Exploded" post of 26th June 2004:
in alt.comp.hardware.

I switched back to the Enermax, which seemed to work for a few days,
until today. The first symptom was occasional SCSI resets, which became
worse and worse. I had heard that turning off Domain Validation on the
Adaptec 29160N can sometimes prevent this. It did for a couple of hours,
but then things got worse from there. Finally the PC wouldn't even boot,
once again struggling to power up the SCSI disks.

I've now switched back to the puny 235W PSU, and everything is working
fine, but I'm still not keen on using such a small PSU. I should add, that
I've used this 235W PSU in this configuration before, for more than 2
years, without problem (still makes me nervous though).

I can't understand why two different monster 550W PSUs, from two different
manufacturers, cannot cope with the same system that a tiny 235W PSU can
easily cope with, especially such "high quality" brand name products.

I've had all three PSUs from new, and the exploding Antec is still under
warranty (not the other two though).

I must have build around 200 systems in my time, and I've never
encountered a weird problem like this. I've double checked for floating
joints, all cables and connections - including data cables, and verified
the integrity of all the disks at BIOS level. All voltages and electrical
pathways pass OK, but these 550W PSUs just aren't playing the game.

Any clues?

-
[H]omer

It's not just the max power rating, it depends on the various supply
lines. Many modern PSU have all the rating on the 3.3V for CPU, RAM and
video adapter. Hard drives use 12V and 5V. Compare the power ratings for the
different lines on the different supplies. If you have a lot of SCSI drives
you should set some of them on Delay Start to reduce the surge at startup.
Mike.

 

[[H]omer]

It's not just the max power rating, it depends on the various supply
lines. Many modern PSU have all the rating on the 3.3V for CPU, RAM and
video adapter. Hard drives use 12V and 5V. Compare the power ratings for the
different lines on the different supplies. If you have a lot of SCSI drives
you should set some of them on Delay Start to reduce the surge at startup.

Sure, but the problems manifest *during* normal operation too, not just at
startup. So this isn't just about sudden surges in demand, but also about
maintaining a steady 5V - which the larger PSUs ironically don't seem to
able to handle on this system.

It definitely seems to come down to poor regulation on the 5V rail ... on
two of the industry's leading PSUs???

-
[H]omer

 

[[H]omer]

True, and even 4.75, is a sign of a serious problem. Many
m'boards these days, running modern CPUs, will further drop the
5V level by the time it reaches the VR circuit.

Well that definitely identifies the problem.

I'm going to try a "PC Power & Cooling 510 Deluxe" and see if that makes
any difference. Expensive, but I have to try something.

-
[H]omer

 

[Dennis E Strausser Jr]

In the followup to last weeks saga of the exploding PSU...

This one really has me stumped:

I've got a lot of equipment in one of my full tower PCs, mainly U160 SCSI
hard disks, and so I figured I needed a fairly beefy PSU.

I bought a 550W Enermax PSU at the same time as upgrading the disks, but
for reasons I couldn't explain, it failed to power up the system. It got
as far as the POST and then I could hear the SCSI disks repeatedly trying
(and failing) to spin up. I gave up with the Enermax and switched back to
an old 235W PSU I had lying around.

Although this configuration worked without problem, I was really nervous
about running so much equipment on just 235W, so eventually I tried yet
another 550W PSU - this time an Antec. Well this worked for about a 3
months, until, well - see the "My PC Exploded" post of 26th June 2004:
in alt.comp.hardware.

I switched back to the Enermax, which seemed to work for a few days,
until today. The first symptom was occasional SCSI resets, which became
worse and worse. I had heard that turning off Domain Validation on the
Adaptec 29160N can sometimes prevent this. It did for a couple of hours,
but then things got worse from there. Finally the PC wouldn't even boot,
once again struggling to power up the SCSI disks.

I've now switched back to the puny 235W PSU, and everything is working
fine, but I'm still not keen on using such a small PSU. I should add, that
I've used this 235W PSU in this configuration before, for more than 2
years, without problem (still makes me nervous though).

I can't understand why two different monster 550W PSUs, from two different
manufacturers, cannot cope with the same system that a tiny 235W PSU can
easily cope with, especially such "high quality" brand name products.

I've had all three PSUs from new, and the exploding Antec is still under
warranty (not the other two though).

I must have build around 200 systems in my time, and I've never
encountered a weird problem like this. I've double checked for floating
joints, all cables and connections - including data cables, and verified
the integrity of all the disks at BIOS level. All voltages and electrical
pathways pass OK, but these 550W PSUs just aren't playing the game.

Any clues?

-
[H]omer

Just a suggestion, why not try a 660W Server PSU by ah,
the nice one is by one of those two companies you listed.
But for some reason I don't remember.
How would I know about it?
Cause I want one for my Dual Xeon Board. (2.66 GHz @ 3.1 or 3.2
I hope...)
Denny.

 

[[H]omer]

That should be more than enough. Maybe that's the problem? Some
(switching) supplies don't like it if they don't have enough load.

Hmm, I hadn't considered that. I always just assumed that the supply will
provide whatever the demand, up to it's max rating.

In another post on another newsgroup (or was it an e-mail?), someone
pointed out that "power blip" can cause a SCSI bus RESET (e.g. through
passive terminator at the end? depending on where it gets its power).

Yup, I think I've defiantly established that it's the 5V signal.

Suggests to me that you don't have enough load? How many disks? Each disk
is something like 10 or 15 watts, unless really busy. CPUs use the most
(about 50 watts? more?). Rest doesn't take that much, unless you have
lots. If you can run on 235W then you don't need 550W. Maybe you're only
using a little over 100W? Do you have a semi-smart UPS? Those can measure
the actual power consumption (on the AC side), er, more or less.

HD = 2 x U160 15K/RPM Seagate X15s 35GB - 160MB/s negotiation
1 x U320 10K/RPM Fujitsu MAP 147GB - 160MB/s negotiation
Tape = 1 x OnStream ADR50 - 80MB/s negotiation
Optic = 2 (DVD, CD-RW) - 40MB/s negotiation
Audio = SBLive! Platinum
VGA = TnT2/Ultra 32MB
Mobo = SuperMicro P6SBA v2.00 (Bios 3.2 I think)
CPU = P3 1GHz 100Mhz
Ram = 3x 128MB PC100 ECC
Ctrl = Digidoc5 + Nexus
Fans = 4x 80mm Case, 1x 40mm VGA, 1x Golden Orb (P3), 2x 80mm (PSU)
Floppy= 2x
PSU = Various 550W + 235W
UPS = APC SmartUPS 1500UI (not currently connected)
SCSI = Adaptec 29160N + Adaptec cables + Passive terminator on LVD.
NIC = Netgear FA312
In a Lian Li PC70 case.

That's it I think. So you think 235W should cope with this OK?

-
[H]omer

 

[VWWall]

[H]omer wrote:

HD = 2 x U160 15K/RPM Seagate X15s 35GB - 160MB/s negotiation
1 x U320 10K/RPM Fujitsu MAP 147GB - 160MB/s negotiation
Tape = 1 x OnStream ADR50 - 80MB/s negotiation
Optic = 2 (DVD, CD-RW) - 40MB/s negotiation
Audio = SBLive! Platinum
VGA = TnT2/Ultra 32MB
Mobo = SuperMicro P6SBA v2.00 (Bios 3.2 I think)
CPU = P3 1GHz 100Mhz
Ram = 3x 128MB PC100 ECC
Ctrl = Digidoc5 + Nexus
Fans = 4x 80mm Case, 1x 40mm VGA, 1x Golden Orb (P3), 2x 80mm (PSU)
Floppy= 2x
PSU = Various 550W + 235W
UPS = APC SmartUPS 1500UI (not currently connected)
SCSI = Adaptec 29160N + Adaptec cables + Passive terminator on LVD.
NIC = Netgear FA312
In a Lian Li PC70 case.

That's it I think. So you think 235W should cope with this OK?

You don't have the major power users, the CPU and AGP fast video.
You can stick a ~1 ohm resistor in series with the hot AC line and
measure the voltage across it to give a rough idea of input power
consumption. (I know PF and non RMS measurement will introduce some
errors.) Figure on about 60% efficiency to get the output power used.
You could put an ammeter in series with each output lead and actually
compare the current to the PS rating. Have you checked the current
ratings for each rail on the old and new PSUs?

Some PSUs have adjustment on the output voltage. I'm not sure which
ones are adjustable, but you might tweak the 5V into a working range.

If the 235 W one works, and is not overheating, stick with it.

Virg Wall

 

[Overlord]

I was running 8 or 9 U80/160 10k and 15k drives off an Enermax 431w PS.
I run them off your AHA 29160N too. Run the CDRWs off an older AHA 2940UW
as one of the CDROMs is a 68pin Ultraplex wide.
I trust your drives are both jumpered and the 29160 card configured to spin
up the drives sequentially. Might also help to let Winders power options
spin 'em down for inactivity. Would help with your amperage draws.
Eventually my 431watt PS started drooping on the +12V rail tho, even with all
the drives run from an old AT PS in an identical case with the cables run across
to a handful of drives in the ATX case.
I killed LEDs, drive current draw, CDRW current draw, went back to stock CPU
speeds, the whole 9 yards but the system would still spike down to 10.8V under
load.
I replaced the 431watt Enermax with a 550watt Enermax and it's running 2 15k
drives, uh... 5 10k drives, pair of CDRWs, LED fans, EL wire in my keyboard,
gigabit LAN, OCing 3.06 to 3.3xx, with no problems.
The 550 Enermax was running almost ¼V over the 12V spec at stock CPU speed.
Overclocking brought it down to like 12.16 by Asus Probe.

I've run up to 8 drives and an old K6-2 450 system before with a PC Power &
Cooling 220watt Silencer before. I am curious as to what sort of system you
are running that will even posf off a 235watt PS. Can't be anything recent...

In the followup to last weeks saga of the exploding PSU...

This one really has me stumped:

I've got a lot of equipment in one of my full tower PCs, mainly U160 SCSI
hard disks, and so I figured I needed a fairly beefy PSU.

I bought a 550W Enermax PSU at the same time as upgrading the disks, but
for reasons I couldn't explain, it failed to power up the system. It got
as far as the POST and then I could hear the SCSI disks repeatedly trying
(and failing) to spin up. I gave up with the Enermax and switched back to
an old 235W PSU I had lying around.

Although this configuration worked without problem, I was really nervous
about running so much equipment on just 235W, so eventually I tried yet
another 550W PSU - this time an Antec. Well this worked for about a 3
months, until, well - see the "My PC Exploded" post of 26th June 2004:
in alt.comp.hardware.

I switched back to the Enermax, which seemed to work for a few days,
until today. The first symptom was occasional SCSI resets, which became
worse and worse. I had heard that turning off Domain Validation on the
Adaptec 29160N can sometimes prevent this. It did for a couple of hours,
but then things got worse from there. Finally the PC wouldn't even boot,
once again struggling to power up the SCSI disks.

I've now switched back to the puny 235W PSU, and everything is working
fine, but I'm still not keen on using such a small PSU. I should add, that
I've used this 235W PSU in this configuration before, for more than 2
years, without problem (still makes me nervous though).

I can't understand why two different monster 550W PSUs, from two different
manufacturers, cannot cope with the same system that a tiny 235W PSU can
easily cope with, especially such "high quality" brand name products.

I've had all three PSUs from new, and the exploding Antec is still under
warranty (not the other two though).

I must have build around 200 systems in my time, and I've never
encountered a weird problem like this. I've double checked for floating
joints, all cables and connections - including data cables, and verified
the integrity of all the disks at BIOS level. All voltages and electrical
pathways pass OK, but these 550W PSUs just aren't playing the game.

Any clues?

-
[H]omer

~~~~~~
Bait for spammers:
root@localhost
postmaster@localhost
admin@localhost
abuse@localhost
postmaster@[127.0.0.1]
(e-mail address removed)
~~~~~~
Remove "spamless" to email me.

 

[kony]

It definitely seems to come down to poor regulation on the 5V rail ... on
two of the industry's leading PSUs???

Enermax really isn't very good. They have a lot of eye candy and
would run a P3 box with typical components fine (but so would any
good 180W PSU) so during that era they were mistakenly given
"good" ratings. Now that systems are using ever more power,
Enermax's offering are showing some weakness compared to other
makes of same wattage. Antec is a fair quality and easy to
find, so they are popular. I am surprised the Antec died on you.
Sparkle/Fortron are a little better in same wattage. Delta even
better. PCPower&Cooling, about the best you can get before
jumping up to a larger form-factor.

The curious part is that your system shouldn't be using THAT much
5V current, further evidenced by the 235W PSU working. Perhaps
these other PSU you tried are regulating on the 12V rail,
designed more towards a system using 12V power for CPU. I've had
pretty good experiences with Fortron 400W, if you had one already
that's what I'd suggest you try, but it's hard to recommend
continually buying more PSU just to find one that works so I
refrain from making any specific recommendation.

Oddly enough you might find that the higher wattage PSU you
already have, works better by putting a dummy load on the 12V
rail, so it ends up raising voltage on 5V rail. Otherwise you
might check the PSU interior for a voltage-adjustment POT(s).