CPU? continued/possibly figured out

  • Thread starter Thread starter philo
  • Start date Start date
P

philo

the other day i posted about a guy who changed the cpu cooler on an XP2200
machine and thought he damaged the cpu because the machine was unstable
unless clocked down to about 1.3mhz (running XP)
a shop put in a new cpu and said the old one was "damaged"

well after a few days of experiments with the harddrives...
whenever i ran chkdsk (without the /f) parameter i'd generally get
*different* results... so i decided to lower the bus frequency and finally
got the machine working well at about 1.4ghz which shows that the shop
really missed the boat.

then i recalled that about a year ago (also on an XP2200) i had a similar
occurrence. the machine got very unstable after going with a larger
heatsink.
(it was quite repeatable as the error stopped entirely just by putting back
the original heat sink )

what i've surmised was that the lager heatsink must have been acting
as an antenna and radiating a great deal of "rf" into the interior of the
machine.

(or possibly just the reverse...that is picking up internal "rf" and
interfering with the cpu)

philo
 
the other day i posted about a guy who changed the cpu cooler on an XP2200
machine and thought he damaged the cpu because the machine was unstable
unless clocked down to about 1.3mhz (running XP)
a shop put in a new cpu and said the old one was "damaged"

well after a few days of experiments with the harddrives...
whenever i ran chkdsk (without the /f) parameter i'd generally get
*different* results... so i decided to lower the bus frequency and finally
got the machine working well at about 1.4ghz which shows that the shop
really missed the boat.

then i recalled that about a year ago (also on an XP2200) i had a similar
occurrence. the machine got very unstable after going with a larger
heatsink.
(it was quite repeatable as the error stopped entirely just by putting back
the original heat sink )

what i've surmised was that the lager heatsink must have been acting
as an antenna and radiating a great deal of "rf" into the interior of the
machine.

(or possibly just the reverse...that is picking up internal "rf" and
interfering with the cpu)

philo
Click to expand...

Seems incredibly unlikely, bordering on impossible. Most heatsinks have
no electrical connection to anything, at least not the metal part. Even
through-board 'sink mounting holes that're plated, usually aren't
ground(ed).

Possibly the board has a bad solder joint and the extra weight of the
'sink is flexing the board a bit more?
Possibly the memory slots have a poor tolerance and the flexing breaks a
contact there... you need to use the heavier 'sink, put the system back in
instable mode and see exactly what's failing, for example you might run
memtest86 or Prime95.

It seems curious though that reducing the bus speed helped... if the board
has a CPU multiplier adjustment you might raise that to 18X, so the board
is again running the CPU at similar to spe'c speed. Also considering you
report it works at 1.4GHz, which for the XP2200 is putting it down around
100MHz FSB, I'm wondering if this is a chipset issue, like a KT133 (non-A)
board... perhaps you mentioned the board/chipset already and I forgot.
 
Seems incredibly unlikely, bordering on impossible. Most heatsinks have
no electrical connection to anything, at least not the metal part. Even
through-board 'sink mounting holes that're plated, usually aren't
ground(ed).
Click to expand...

Yes i agree that it sounds very unlikely but i wnet back and forth many
times
and tried a number of different heat sinks and only had the problem with one
particularly large heatsink...it was aluminum and about the same weight as a
smaller copper unit that worked ok.

With radio frequency...you need no D.C. connection...
at the high frequency's used inside a computer...there would be plenty of
capacitive coupling between the heatsink and cpu
Possibly the board has a bad solder joint and the extra weight of the
'sink is flexing the board a bit more?
Possibly the memory slots have a poor tolerance and the flexing breaks a
contact there... you need to use the heavier 'sink, put the system back in
instable mode and see exactly what's failing, for example you might run
memtest86 or Prime95.
Click to expand...

I ran memtest86 for about 6 hours last night and it came up with no
problems.
the board is an ASUS A7V333 with the VIA KT333 chipset.
(the cpu is actaully an XP2500 btw)
It seems curious though that reducing the bus speed helped... if the board
has a CPU multiplier adjustment you might raise that to 18X, so the board
is again running the CPU at similar to spe'c speed. Also considering you
report it works at 1.4GHz, which for the XP2200 is putting it down around
100MHz FSB, I'm wondering if this is a chipset issue, like a KT133 (non-A)
board... perhaps you mentioned the board/chipset already and I forgot.
Click to expand...

no, i neglected to mention the board...and it certainly may be a chipset
issue.
i still need to play with the FSB settings a bit...but i did find that by
putting one stick of RAM in different slots...i get different results which
is leading me to believe that the motherboard board may just plain be
defective!
 
Yes i agree that it sounds very unlikely but i wnet back and forth many
times
and tried a number of different heat sinks and only had the problem with one
particularly large heatsink...it was aluminum and about the same weight as a
smaller copper unit that worked ok.
Click to expand...

But was it the same height? It's not merely the weight that's a problem,
it's the leveraged weight. Simply taking a light 'sink that's tall and
strapping a large fan on top will flex a board a lot more than a compact
yet heavier copper 'sink... usually the majority of the copper weight is
right at the base.

With radio frequency...you need no D.C. connection...
at the high frequency's used inside a computer...there would be plenty of
capacitive coupling between the heatsink and cpu
Click to expand...

You might not "need" a connection, but the strength of it should be
negligable. There are plenty of people running huge 'sinks on
same/similar CPUs without this problem. A CPU is meant to run in a
computer, it's going to have enough tolerance for the RF generated inside
a system.
I ran memtest86 for about 6 hours last night and it came up with no
problems.
the board is an ASUS A7V333 with the VIA KT333 chipset.
(the cpu is actaully an XP2500 btw)
Click to expand...

Well now I have some info I can "play with"... right now I have two
different revisions of that board sitting on a bench... one is rev
1.02(?) and one rev. 1.4. From first glance the revision differences are
that the 1.4 has Rubycon MBZ (Purple/gold) caps on the left of the socket.
Right now that one is running at XP1700 T'Bred (actually a pre-Thorton
name-release, Thorton) @ around 2.3GHz... might be 13.5 x 175, approx.

What BIOS version are you running? Supposedly v1016 is most stable,
though IIRC the two here have a "lumberjack" RAID-mod bios installed,
v1016 or 1017. Both boards are running at DDR333, have been used with
PAL8045 'sinks, Thermalright AX-7, both relatively large 'sinks but one of
the PAL had a custom made Al shroud and ductwork that makes it gigantic so
far as metalwork goes, the duct is only usable when the other end is
supported by mount to a rear fan as it was intended.

These boards have the socket lugs on the sides of the socket, is it
possible the 'sink you used had too little clip pressure so it wasn't
sitting well on the CPU? If you haven't, you might try running with the
system on it's side, see if that makes a difference having the 'sink and
CPU with a horizontal plane of interface.
no, i neglected to mention the board...and it certainly may be a chipset
issue.
i still need to play with the FSB settings a bit...but i did find that by
putting one stick of RAM in different slots...i get different results which
is leading me to believe that the motherboard board may just plain be
defective!
Click to expand...

The two I have here run fine up to about 185 MHz FSB, at which point the
AGP goes far enough out of spec that some cards (Geforce) start
complaining but the Radeons can tolerate another 10MHz or so, maybe
more... the boards I have are running resued, old low spec memory so it
could be the memory crapping out towards 200MHz FSB, but IIRC the USB
ports won't work much past 185MHz FSB either... though I could be wrong
about that, been a long time since I did any testing of max FSB on these
boards, but with two decent memory modules they're definitely stable way
past 133MHz FSB... doesn't mean the board you have doens't have some
problem though, not relating to the board's design or chipset but a
specific failure on that particular board.

The two boards like memory in first and third slot, are quite picky about
memory when all 3 slots are used, though this is at 166MHz, don't recall
having the boards down at 133MHz for long enough to test with 3 modules.

If you have the memory set to asynchronous, +33 mode, try synchronoous,
and the (regular) memory timings, not the BIOS "turbo" setting.

One of the boards is sitting a foot in front of me right now, I was using
it to test o'c of a CPU, which is just about done... if there's anything I
can test on the board while it's still together, LMK.
 
But was it the same height? It's not merely the weight that's a problem,
it's the leveraged weight. Simply taking a light 'sink that's tall and
strapping a large fan on top will flex a board a lot more than a compact
yet heavier copper 'sink... usually the majority of the copper weight is
right at the base.
Click to expand...

well it certainly could have been a pressure thing...
at any rate...that machine has been working perfectly for quite some time
now...
but eventually i'm sure i'll have a chance to try out that "problem"
heatsink
You might not "need" a connection, but the strength of it should be
negligable. There are plenty of people running huge 'sinks on
same/similar CPUs without this problem. A CPU is meant to run in a
computer, it's going to have enough tolerance for the RF generated inside
a system.


Well now I have some info I can "play with"... right now I have two
different revisions of that board sitting on a bench... one is rev
1.02(?) and one rev. 1.4. From first glance the revision differences are
that the 1.4 has Rubycon MBZ (Purple/gold) caps on the left of the socket.
Right now that one is running at XP1700 T'Bred (actually a pre-Thorton
name-release, Thorton) @ around 2.3GHz... might be 13.5 x 175, approx.

What BIOS version are you running? Supposedly v1016 is most stable,
though IIRC the two here have a "lumberjack" RAID-mod bios installed,
v1016 or 1017. Both boards are running at DDR333, have been used with
PAL8045 'sinks, Thermalright AX-7, both relatively large 'sinks but one of
the PAL had a custom made Al shroud and ductwork that makes it gigantic so
far as metalwork goes, the duct is only usable when the other end is
supported by mount to a rear fan as it was intended.
Click to expand...

the bios version is 1017
and it's using two, Crucial DDR 333

These boards have the socket lugs on the sides of the socket, is it
possible the 'sink you used had too little clip pressure so it wasn't
sitting well on the CPU? If you haven't, you might try running with the
system on it's side, see if that makes a difference having the 'sink and
CPU with a horizontal plane of interface.


The two I have here run fine up to about 185 MHz FSB, at which point the
AGP goes far enough out of spec that some cards (Geforce) start
complaining but the Radeons can tolerate another 10MHz or so, maybe
more... the boards I have are running resued, old low spec memory so it
could be the memory crapping out towards 200MHz FSB, but IIRC the USB
ports won't work much past 185MHz FSB either... though I could be wrong
about that, been a long time since I did any testing of max FSB on these
boards, but with two decent memory modules they're definitely stable way
past 133MHz FSB... doesn't mean the board you have doens't have some
problem though, not relating to the board's design or chipset but a
specific failure on that particular board.

The two boards like memory in first and third slot, are quite picky about
memory when all 3 slots are used, though this is at 166MHz, don't recall
having the boards down at 133MHz for long enough to test with 3 modules.

If you have the memory set to asynchronous, +33 mode, try synchronoous,
and the (regular) memory timings, not the BIOS "turbo" setting.
Click to expand...


I've got to run out of the house and repair a machine now...
so i may not have time to go through all the permutations you've mentioned
until this weekend...but you've certainly given me enough ideas to keep me
busy for a while

thanks!

philo
 
I've got to run out of the house and repair a machine now...
so i may not have time to go through all the permutations you've mentioned
until this weekend...but you've certainly given me enough ideas to keep me
busy for a while
Click to expand...


i got a chance to test the unit some more.
i boot up and run chkdsk and get different errors each time...
it does not make a difference whether the machine is oriented horizontally
or vertically...
and no matter how i play with different multiplier setttings...
the problem seems to get worse as i clock the machine higher...
and though it finds chkdsk errors less often...as the freq is lowered...
the problems are still there.
i will run memtest again and let it run overnight...
and if no errors are found i may suggest a new mobo to the owner...
i definately don't trust it

philo
 
Clocked machine up to 2ghz and ran memtest for 12 hours.
no problems.

tried a different cpu cooler
tried putting the drive on a pci controller
tried various combinations on clock settings

even if clocked down to 1.2 ghz there would still occasionally be chkdsk
errors

when drive is chkdsk'ed in another machine there are never any errors

so it looks like it's time for a new motherboard
 
Back
Top