Unstable A7N8X-DLX 2.0 system, Power Supply was culprit

Am writing this for the benefit of anyone that might take the time to
read this post

My trusty old A7N8X-DLX 2.0 based system has been behaving badly over
the last couple of weeks, primarily crash to desktop and BSOD crashes
while playing a CPU and graphics intensive game (COD4 to be precise),
and sometimes even a system reset. Researching each of the STOP codes
produced (OS is winxp) provided little comfort, something critical was
going wrong, bugchecks, minidumps, errors in critical system drivers,
etc. I even heard the voice error message "CPU System Failed", or
whatever that little witch says, while playing COD4 and I'd never
before heard BIOS messages when winxp or win2k was up and running
(system is dual-boot).

Normal desktop use did not cause problems, but intense gaming programs
did. I began to suspect my vid card, vid card drivers or some other
odd source for the crashes. I was just about to install a new video
card driver update when I decided to check system voltages. My PS is
a decent brand (Enermax 350W) that has a few years on it, and I was
quite surprised to see Speedfan's graph of the output voltages. The
5v rail was down to about 4.65-4.70v at startup, and during gaming the
voltage dropped a bit more, apparently causing CPU Vcore regulation
problems if not vid card problems (vid card is an ATI 9800 Pro). I
checked/unplugged/replugged all connections to the PS but none seemed
loose.

Much as I don't like to disassemble a system, I proceeded to pull out
the PS and open it up in the hope of finding a pot to adjust output
voltages. Much to my surprise, there were no pots to adjust that I
could find. So, I thought, I might as well clean up the dust/crud
inside the PS since I had it opened up (it was a bit dirty, not really
bad compared to some of the machines I've fixed for others) and do
some bench voltage probing on the PS. Cleaning dust and dirt from the
fans always makes me feel like the PS will run a bit cooler, even tho
this particular system lives in my basement where the ambient temps
are always cool.

After cleanup, the PS 5v unloaded output was 5.11v. Hmmm, me thinks
that looks good, so I reinstall the PS in my rig, and low and behold,
under heavy load, the 5v rail now regulates to no lower than 4.85v
during heavy loads. Moral of the story, crud/dust/fuzz inside the PS
apparently can affect regulation circuit performance sufficiently to
impact output voltages in a negative fasion.

I've seen dirt/dust/crud affect a mobo before (fixed IDE interface
problems once by brushing the dirt from a mobo, which fixed read/write
problems to a HD that checked good in other systems) but not seen such
with a PS before, so I thought I'd share this story.

Best regards,
--
Best regards,
Kyle

"Kyle" <> wrote in message
news. ..
> Am writing this for the benefit of anyone that might take the time to
> read this post
>
> My trusty old A7N8X-DLX 2.0 based system has been behaving badly over
> the last couple of weeks, primarily crash to desktop and BSOD crashes
> while playing a CPU and graphics intensive game (COD4 to be precise),
> and sometimes even a system reset. Researching each of the STOP codes
> produced (OS is winxp) provided little comfort, something critical was
> going wrong, bugchecks, minidumps, errors in critical system drivers,
> etc. I even heard the voice error message "CPU System Failed", or
> whatever that little witch says, while playing COD4 and I'd never
> before heard BIOS messages when winxp or win2k was up and running
> (system is dual-boot).
>
> Normal desktop use did not cause problems, but intense gaming programs
> did. I began to suspect my vid card, vid card drivers or some other
> odd source for the crashes. I was just about to install a new video
> card driver update when I decided to check system voltages. My PS is
> a decent brand (Enermax 350W) that has a few years on it, and I was
> quite surprised to see Speedfan's graph of the output voltages. The
> 5v rail was down to about 4.65-4.70v at startup, and during gaming the
> voltage dropped a bit more, apparently causing CPU Vcore regulation
> problems if not vid card problems (vid card is an ATI 9800 Pro). I
> checked/unplugged/replugged all connections to the PS but none seemed
> loose.
>
> Much as I don't like to disassemble a system, I proceeded to pull out
> the PS and open it up in the hope of finding a pot to adjust output
> voltages. Much to my surprise, there were no pots to adjust that I
> could find. So, I thought, I might as well clean up the dust/crud
> inside the PS since I had it opened up (it was a bit dirty, not really
> bad compared to some of the machines I've fixed for others) and do
> some bench voltage probing on the PS. Cleaning dust and dirt from the
> fans always makes me feel like the PS will run a bit cooler, even tho
> this particular system lives in my basement where the ambient temps
> are always cool.
>
> After cleanup, the PS 5v unloaded output was 5.11v. Hmmm, me thinks
> that looks good, so I reinstall the PS in my rig, and low and behold,
> under heavy load, the 5v rail now regulates to no lower than 4.85v
> during heavy loads. Moral of the story, crud/dust/fuzz inside the PS
> apparently can affect regulation circuit performance sufficiently to
> impact output voltages in a negative fasion.
>
> I've seen dirt/dust/crud affect a mobo before (fixed IDE interface
> problems once by brushing the dirt from a mobo, which fixed read/write
> problems to a HD that checked good in other systems) but not seen such
> with a PS before, so I thought I'd share this story.
>
> Best regards,
> --
> Best regards,
> Kyle
>
Glad you found/fixed problem, Kyle. IIRC, that mb also likes ps to
deliver 28 amps on the 3v rail, kinda high as compared to many mb's of its
era. I'm still using 2 of those--an -8-X/Rev 2.0 and an -8x-DeLuxe 2.0.
They've never given me a problem, and have spoiled me. Were it not for
fixing friends' and neighbors' (sometimes mutually exclusive!) boxes, I'd
have forgotten even the "basics" and missed out on bunches of fun.
ATB, sdlomi2

sdlomi2 wrote:
> "Kyle" <> wrote in message
> news. ..
>> Am writing this for the benefit of anyone that might take the time to
>> read this post
>>
>> My trusty old A7N8X-DLX 2.0 based system has been behaving badly over
>> the last couple of weeks, primarily crash to desktop and BSOD crashes
>> while playing a CPU and graphics intensive game (COD4 to be precise),
>> and sometimes even a system reset. Researching each of the STOP codes
>> produced (OS is winxp) provided little comfort, something critical was
>> going wrong, bugchecks, minidumps, errors in critical system drivers,
>> etc. I even heard the voice error message "CPU System Failed", or
>> whatever that little witch says, while playing COD4 and I'd never
>> before heard BIOS messages when winxp or win2k was up and running
>> (system is dual-boot).
>>
>> Normal desktop use did not cause problems, but intense gaming programs
>> did. I began to suspect my vid card, vid card drivers or some other
>> odd source for the crashes. I was just about to install a new video
>> card driver update when I decided to check system voltages. My PS is
>> a decent brand (Enermax 350W) that has a few years on it, and I was
>> quite surprised to see Speedfan's graph of the output voltages. The
>> 5v rail was down to about 4.65-4.70v at startup, and during gaming the
>> voltage dropped a bit more, apparently causing CPU Vcore regulation
>> problems if not vid card problems (vid card is an ATI 9800 Pro). I
>> checked/unplugged/replugged all connections to the PS but none seemed
>> loose.
>>
>> Much as I don't like to disassemble a system, I proceeded to pull out
>> the PS and open it up in the hope of finding a pot to adjust output
>> voltages. Much to my surprise, there were no pots to adjust that I
>> could find. So, I thought, I might as well clean up the dust/crud
>> inside the PS since I had it opened up (it was a bit dirty, not really
>> bad compared to some of the machines I've fixed for others) and do
>> some bench voltage probing on the PS. Cleaning dust and dirt from the
>> fans always makes me feel like the PS will run a bit cooler, even tho
>> this particular system lives in my basement where the ambient temps
>> are always cool.
>>
>> After cleanup, the PS 5v unloaded output was 5.11v. Hmmm, me thinks
>> that looks good, so I reinstall the PS in my rig, and low and behold,
>> under heavy load, the 5v rail now regulates to no lower than 4.85v
>> during heavy loads. Moral of the story, crud/dust/fuzz inside the PS
>> apparently can affect regulation circuit performance sufficiently to
>> impact output voltages in a negative fasion.
>>
>> I've seen dirt/dust/crud affect a mobo before (fixed IDE interface
>> problems once by brushing the dirt from a mobo, which fixed read/write
>> problems to a HD that checked good in other systems) but not seen such
>> with a PS before, so I thought I'd share this story.
>>
>> Best regards,
>> --
>> Best regards,
>> Kyle
>>
> Glad you found/fixed problem, Kyle. IIRC, that mb also likes ps to
> deliver 28 amps on the 3v rail, kinda high as compared to many mb's of its
> era. I'm still using 2 of those--an -8-X/Rev 2.0 and an -8x-DeLuxe 2.0.
> They've never given me a problem, and have spoiled me. Were it not for
> fixing friends' and neighbors' (sometimes mutually exclusive!) boxes, I'd
> have forgotten even the "basics" and missed out on bunches of fun.
> ATB, sdlomi2
>

I've measured the power on my A7N8X-E Deluxe, with a clamp-on DC ammeter,
and these are the numbers I got. The reason the 5V current is so high,
is because the processor draws power from it. (Some S462 motherboards
from that era, use a 2x2 ATX12V power connector for the processor, so
they wouldn't have a large +5V component.)

3.3V @ 5.18A
5V @ 16.6A
12V @ 0.53A

Test conditions were running nultiple copies of Prime95, in Linux.

I've also tested 3DMark2001SE in Windows, and the 3.3V on the A7N8X-E
rose to 6.4A or so. Which could be flowing to the video card. Since the
video slot power connections, come through the main power cable, there is
room for higher numbers. But the 3.3V rail on AGP slots, has an
upper limit for allowed current. So there are some limits as to
how much larger that consumption could get.

The video card I used, has a separate Aux connector, and that adds more
5V consumption (another 5 amps).

By comparison, my P4C800-E Deluxe draws 3.3V @ 14.4A, and that is
because the memory runs from 3.3V (seems to use what looks like linear
regulation). Power varies, depending on whether one or two memory channels
are active. On that system, the 5V consumption is very low, only 0.5A.

The rules on expected current consumption, change with the generation of
design. On modern motherboards, +12V is used for the processor, so the
lower rails won't see that loading. Modern memory has pretty low power
consumption. And that leaves a 10-20W Northbridge and lower power
Southbridge, to suck up power from the lower rails. For modern boards,
I'd assume 50W, coming from 3.3V or 5V, would be enough to cover it.

Is the 28A number a measured value ?

Paul

"Paul" <> wrote in message news:foiqad$5t1$...
> sdlomi2 wrote:
>> "Kyle" <> wrote in message
>> news. ..
>>> Am writing this for the benefit of anyone that might take the time to
>>> read this post
>>>
>>> My trusty old A7N8X-DLX 2.0 based system has been behaving badly over
>>> the last couple of weeks, primarily crash to desktop and BSOD crashes
>>> while playing a CPU and graphics intensive game (COD4 to be precise),
>>> and sometimes even a system reset. Researching each of the STOP codes
>>> produced (OS is winxp) provided little comfort, something critical was
>>> going wrong, bugchecks, minidumps, errors in critical system drivers,
>>> etc. I even heard the voice error message "CPU System Failed", or
>>> whatever that little witch says, while playing COD4 and I'd never
>>> before heard BIOS messages when winxp or win2k was up and running
>>> (system is dual-boot).
>>>
>>> Normal desktop use did not cause problems, but intense gaming programs
>>> did. I began to suspect my vid card, vid card drivers or some other
>>> odd source for the crashes. I was just about to install a new video
>>> card driver update when I decided to check system voltages. My PS is
>>> a decent brand (Enermax 350W) that has a few years on it, and I was
>>> quite surprised to see Speedfan's graph of the output voltages. The
>>> 5v rail was down to about 4.65-4.70v at startup, and during gaming the
>>> voltage dropped a bit more, apparently causing CPU Vcore regulation
>>> problems if not vid card problems (vid card is an ATI 9800 Pro). I
>>> checked/unplugged/replugged all connections to the PS but none seemed
>>> loose.
>>>
>>> Much as I don't like to disassemble a system, I proceeded to pull out
>>> the PS and open it up in the hope of finding a pot to adjust output
>>> voltages. Much to my surprise, there were no pots to adjust that I
>>> could find. So, I thought, I might as well clean up the dust/crud
>>> inside the PS since I had it opened up (it was a bit dirty, not really
>>> bad compared to some of the machines I've fixed for others) and do
>>> some bench voltage probing on the PS. Cleaning dust and dirt from the
>>> fans always makes me feel like the PS will run a bit cooler, even tho
>>> this particular system lives in my basement where the ambient temps
>>> are always cool.
>>>
>>> After cleanup, the PS 5v unloaded output was 5.11v. Hmmm, me thinks
>>> that looks good, so I reinstall the PS in my rig, and low and behold,
>>> under heavy load, the 5v rail now regulates to no lower than 4.85v
>>> during heavy loads. Moral of the story, crud/dust/fuzz inside the PS
>>> apparently can affect regulation circuit performance sufficiently to
>>> impact output voltages in a negative fasion.
>>>
>>> I've seen dirt/dust/crud affect a mobo before (fixed IDE interface
>>> problems once by brushing the dirt from a mobo, which fixed read/write
>>> problems to a HD that checked good in other systems) but not seen such
>>> with a PS before, so I thought I'd share this story.
>>>
>>> Best regards,
>>> --
>>> Best regards,
>>> Kyle
>>>
>> Glad you found/fixed problem, Kyle. IIRC, that mb also likes ps to
>> deliver 28 amps on the 3v rail, kinda high as compared to many mb's of
>> its era. I'm still using 2 of those--an -8-X/Rev 2.0 and an -8x-DeLuxe
>> 2.0. They've never given me a problem, and have spoiled me. Were it not
>> for fixing friends' and neighbors' (sometimes mutually exclusive!) boxes,
>> I'd have forgotten even the "basics" and missed out on bunches of fun.
>> ATB, sdlomi2
>
> I've measured the power on my A7N8X-E Deluxe, with a clamp-on DC ammeter,
> and these are the numbers I got. The reason the 5V current is so high,
> is because the processor draws power from it. (Some S462 motherboards
> from that era, use a 2x2 ATX12V power connector for the processor, so
> they wouldn't have a large +5V component.)
>
> 3.3V @ 5.18A
> 5V @ 16.6A
> 12V @ 0.53A
>
> Test conditions were running nultiple copies of Prime95, in Linux.
>
> I've also tested 3DMark2001SE in Windows, and the 3.3V on the A7N8X-E
> rose to 6.4A or so. Which could be flowing to the video card. Since the
> video slot power connections, come through the main power cable, there is
> room for higher numbers. But the 3.3V rail on AGP slots, has an
> upper limit for allowed current. So there are some limits as to
> how much larger that consumption could get.
>
> The video card I used, has a separate Aux connector, and that adds more
> 5V consumption (another 5 amps).
>
> By comparison, my P4C800-E Deluxe draws 3.3V @ 14.4A, and that is
> because the memory runs from 3.3V (seems to use what looks like linear
> regulation). Power varies, depending on whether one or two memory channels
> are active. On that system, the 5V consumption is very low, only 0.5A.
>
> The rules on expected current consumption, change with the generation of
> design. On modern motherboards, +12V is used for the processor, so the
> lower rails won't see that loading. Modern memory has pretty low power
> consumption. And that leaves a 10-20W Northbridge and lower power
> Southbridge, to suck up power from the lower rails. For modern boards,
> I'd assume 50W, coming from 3.3V or 5V, would be enough to cover it.
>
> Is the 28A number a measured value ?
>
> Paul

Hey Paul, as you notice, I prefaced my statement with "IIRC", and I
still am hiding behind it, as I know not where I read that. I do recall
that I saw it discussed several (as in 2 or more?) times along about the
time I was experimenting with the Athlons and the (in?)famous ECS / K7S5A
and overclocking using it with Athlons; also, it being the one that lost its
bios settings, etc. Again, I have nothing in writing to back it up--just I
seem to recall... As far as measured voltage, I never did know if it was
that or theoretical value as calculated somehow. Sorry. s

>> snip <<
>> Is the 28A number a measured value ?
>>
>> Paul
>
> Hey Paul, as you notice, I prefaced my statement with "IIRC", and I
> still am hiding behind it, as I know not where I read that. I do recall
> that I saw it discussed several (as in 2 or more?) times along about the
> time I was experimenting with the Athlons and the (in?)famous ECS / K7S5A
> and overclocking using it with Athlons; also, it being the one that lost
> its bios settings, etc. Again, I have nothing in writing to back it
> up--just I seem to recall... As far as measured voltage, I never did know
> if it was that or theoretical value as calculated somehow. Sorry. s
>
>
I found what I made the error on. Looking here,
>> http://www.geek.com/the-damnable-k7s5a-from-ecs/ <<<<, I found
>> one site that cleared up my error, I think. It WAS the ECS K7S5A that
>> called for such huge amps on the 3.3 volt rail. See "Mine Works Fine: I
>> installed all os and hardware under 100/100 conditions, then switched to
>> 133/133 and have not had a single crash in three weeks. I'm no expert,
>> but alot of the problems reported here sound like they are power related.
>> Keep in mind these minimum specs:
+3.3V need 28A
+5V need 30A
+12V need 15A or above".

Again, sorry for the confusion. My bad. s

sdlomi2 wrote:
>>> snip <<
>>> Is the 28A number a measured value ?
>>>
>>> Paul
>> Hey Paul, as you notice, I prefaced my statement with "IIRC", and I
>> still am hiding behind it, as I know not where I read that. I do recall
>> that I saw it discussed several (as in 2 or more?) times along about the
>> time I was experimenting with the Athlons and the (in?)famous ECS / K7S5A
>> and overclocking using it with Athlons; also, it being the one that lost
>> its bios settings, etc. Again, I have nothing in writing to back it
>> up--just I seem to recall... As far as measured voltage, I never did know
>> if it was that or theoretical value as calculated somehow. Sorry. s
>>
>>
> I found what I made the error on. Looking here,
>>> http://www.geek.com/the-damnable-k7s5a-from-ecs/ <<<<, I found
>>> one site that cleared up my error, I think. It WAS the ECS K7S5A that
>>> called for such huge amps on the 3.3 volt rail. See "Mine Works Fine: I
>>> installed all os and hardware under 100/100 conditions, then switched to
>>> 133/133 and have not had a single crash in three weeks. I'm no expert,
>>> but alot of the problems reported here sound like they are power related.
>>> Keep in mind these minimum specs:
> +3.3V need 28A
> +5V need 30A
> +12V need 15A or above".
>
> Again, sorry for the confusion. My bad. s
>

As a cross-check, there is another way you can look at the issue.

On a 20 pin connector, there are three 3.3V wires. There are four 5.0V wires.
On a 20 pin Mini-fit Jr. connector, a typical rating would be 6 amps per
pin. That means the motherboard should draw no more than 18A from 3.3V,
and no more than 24A from 5V. When supplies offer more than those amounts
of current, then you check to see if there is another way to draw current.
Older supplies had a 1x6 aux connector, which would be another potential path
for 3.3V. 5V can be drawn by disk drives, or via an AGP video card
aux connector (like my 9800Pro).

I've seen some of the older supplies, that had a 50A rating, but I
couldn't see a way that the wires and pins would allow it.

So, to draw the 28A from 3.3V, can be difficult to arrange in a safe
manner.

There is at least one motherboard, a dual Athlon, where too much current
is drawn from +5V, and some pins get burned. So sometimes the engineers
miss a detail. But for the most part, they stay within the bounds of
what the connectors will allow.

The current rating of the Mini-Fit Jr. pins, depends on a couple things.
It depends on the gauge of wire used (larger diameter wire means better
heat transfer down the wire). It also depends on the number of adjacent
pins. Which is why an ATX12V 2x2 can have a higher current rating than
the 24 pin connector - the 2x2 has no hot pins next to it, so the pins
are cooled a bit better. The pins in the center of a 24 pin connector,
have more hot neighbors.

I guess I'm a bit sensitive about some of the advice I see about
power supplies. For example, a few minutes ago, I was on an Nvidia
site, where an Nvidia employee was recommending a 1200W
power supply for a computer with three video cards. Somehow, I doubt
it'll use all that, because the box would probably incinerate if
it did :-) 1200W is a lot of heat to remove with a single good
fan in the back of the computer. Fortunately, the major consumers
would be drawing about 500W total, and even then may not make it
if the box is CPU bound.

Paul