Is this PNY Geforce 6600 GT graphics card defective?

[altcomphardware]

I have an AMD XP 2500+, 1GB DDR333 RAM and a PNY Geforce 6600GT.

When I test out some games with it, the system either reboots, crashes
to Windows or hangs. Occasionally artifacts are visible if a lot of
light effects are present. Unfortunately, the screenshot button doesn't
work in the game so I have not been able to save them. If you guys want
to see it then let me know, I'll try to get it with a digital camera.

When I tried updating with the latest drivers, I couldn't even get into
Windows - just a black screen. I had to use the "Last Known Working
Configuration" to get in.

Is there a stress test I can download which will show if the card is
faulty or not? IIRC although 3DMark stresses the graphics card, it
doesn't do so for long, and it doesn't detect if artifacts are present.
The results with 3D Mark were dismal, I got less than 1FPS for all the
tests. Is this to be expected?

 

[altcomphardware]

Hi guys. I have uploaded a picture of my 3D marks score:

http://rapidshare.de/files/27490726/3dmarks_results.JPG.html

My PSU is an Antec Smartpower 350W PSU.

The reboots tend to be preceeded by a freeze of around 10 seconds. One
thing strange I noticed about the reboots was that either the keyboard
or mouse wouldn't work after the reboot (sometimes both). A total PC
power off and restart would fix the problem.

The Geforce 6600GT is supposed to require a 300W PSU. I have 1 IDE HDD,
1 DVD+/-RW, 1 PCI/IDE card, 1 PCI Gigabit card and 1 SB Audigy.

Please let me know if you think the card is defective. I have a few
more days until I lose my right to return the card.

If you want me to do any other tests, please let me know too.

 

[kony]

I have an AMD XP 2500+, 1GB DDR333 RAM and a PNY Geforce 6600GT.

What motherboard and PSU make AND model?

When I test out some games with it, the system either reboots, crashes
to Windows or hangs.

These are three distinctly different things and probably
more than one problem causing them.

You need to not lump all games together, take ONE game that
is known to work properly with it's newest patches and a
6600GT card, and which video driver (generally I would
suggest trying a late 8x.xx series driver, no older and not
a 9x.xx series.

You might try one of the 3DMark benchmarks too or instead,
or looping some benchmark demo... but again it has to be
known stable with your hardware and driver, as games are
rushed in development and quite a few need patches to work
on a lot of hardware.

Reinstall DirectX9c. It doesn't matter if you already had
it installed... just do it.

If it's possible the sound is crashing it, disable the sound
or try other sound drivers.

Is the CPU or video card overheating?

Try to isolate each, run Prime95's Torture Test, large
in-place FFTs setting for at least a few hours to see if the
CPU produces errors. If it does you must rectify that
problem before you can determine whether the video card is
ok or not (as mentioned above, you may easily have more than
one problem).

Occasionally artifacts are visible if a lot of
light effects are present.

Could be overheating, the driver, power supply, or game
bugs. Could even be vaulty or ESD damaged memory on the
card but this seems less likely than any of the former
possibilities.

What video card did your system use previously?
Did you try playing any of the same games and did they have
any problems if so?

Unfortunately, the screenshot button doesn't
work in the game so I have not been able to save them. If you guys want
to see it then let me know, I'll try to get it with a digital camera.

Maybe if other things don't work, this could be useful but
for the time being, take a multimeter and measure the PSU
voltages under heavy load (like when gaming and running the
Prime95 Torture test), or if you don't have a multimeter nor
access to one, at least note the system temps reported by
software.

You might also leave the case open (does it have good
airflow??) and point a strong desk fan at the center to see
if it resolves anything.

When I tried updating with the latest drivers, I couldn't even get into
Windows - just a black screen. I had to use the "Last Known Working
Configuration" to get in.

you didn't even mention the games you're having trouble with
but that might be information better suited to a gaming
forum where far more people are familiar with problems any
particular game might have with or w/o patches, drivers,
etc.

Is there a stress test I can download which will show if the card is
faulty or not? IIRC although 3DMark stresses the graphics card, it
doesn't do so for long, and it doesn't detect if artifacts are present.
The results with 3D Mark were dismal, I got less than 1FPS for all the
tests. Is this to be expected?

That seems very odd, to get only 1FPS. I'd wonder if the
video card is overheating though instability from bad power
would also cause a similar pausing and downclocked speed ->
lower average performance.

Try an older 3DMark, 2001 and 2003. One of not both can be
set to loop mode, and by only setting it to run on high
detail, the demanding tests, you will keep as much load on
the video card as possible, minimize the CPU as a
bottleneck... or vice versa, choose low detail and low
resolution to put most load on CPU in a 3D environment.

 

[Sleepy]

I have an AMD XP 2500+, 1GB DDR333 RAM and a PNY Geforce 6600GT.

When I test out some games with it, the system either reboots, crashes
to Windows or hangs. Occasionally artifacts are visible if a lot of
light effects are present. Unfortunately, the screenshot button doesn't
work in the game so I have not been able to save them. If you guys want
to see it then let me know, I'll try to get it with a digital camera.

When I tried updating with the latest drivers, I couldn't even get into
Windows - just a black screen. I had to use the "Last Known Working
Configuration" to get in.

Is there a stress test I can download which will show if the card is
faulty or not? IIRC although 3DMark stresses the graphics card, it
doesn't do so for long, and it doesn't detect if artifacts are present.
The results with 3D Mark were dismal, I got less than 1FPS for all the
tests. Is this to be expected?

try the previous offical drivers (as kony suggest) 84.21. they're in the
archive section of www.nvidia.com and set AGP speed to 4x (I find 8x to show
up graphical glitches though it doesnt crash my machine. fast-writes OFF is
also best for stability. and if you mean you've recently changed cards then
thoroughly uninstalling the old drivers is a must and it helps to reinstall
the motherboard drivers (including the AGP port driver). google for
coolbits.reg and use it - it unlocks hidden parts of the display control
panel and check your temps - 50s when idling up to 65c when gaming is
normal.

good luck.

 

[Paul]

Hi guys. I have uploaded a picture of my 3D marks score:

http://rapidshare.de/files/27490726/3dmarks_results.JPG.html

My PSU is an Antec Smartpower 350W PSU.

The reboots tend to be preceeded by a freeze of around 10 seconds. One
thing strange I noticed about the reboots was that either the keyboard
or mouse wouldn't work after the reboot (sometimes both). A total PC
power off and restart would fix the problem.

The Geforce 6600GT is supposed to require a 300W PSU. I have 1 IDE HDD,
1 DVD+/-RW, 1 PCI/IDE card, 1 PCI Gigabit card and 1 SB Audigy.

Please let me know if you think the card is defective. I have a few
more days until I lose my right to return the card.

If you want me to do any other tests, please let me know too.

And from your other post, you have "AMD XP 2500+, 1GB DDR333 RAM".
One thing missing, is what motherboard you are using.

Entry #39 in the table here is beating your pants off
By 1356 points to your 269 points. And he is using an "XP 1900+".

http://www.techimo.com/forum/t161088.html

The very first question is, did you remember to uninstall the
drivers for the old video card ? That has caused me grief in
the past (so much grief, that in one case the only way I could
get a new video card to work right, was an OS reinstall). When
the old video card is still in the computer, you uninstall
the video driver. Then, you are ready to come up at 640x480
driverless with the new card. Followed by installing
whatever minimum DirectX version the new card needs, and the
new video card driver. (Assumes the chipset AGP driver is in
good shape and is up to date.) The best order of install is
AGP chipset driver first, video card driver second, and DirectX
can be either before or after or as many times as you like.

If the chipset and graphics card drivers were not installed
properly, I'd be surprised if 3DMark06 would run properly.
It could be that texture acceleration is not enabled. Run
Dxdiag from the directx9 install, and see what it reports.
(I'd recommend the taskbar options popup of Powerstrip from
entechtaiwan.com, but it has a limited trial period now. At
least I cannot run it any more.)

http://www.windowsresource.net/guides/hardware.php

I'll reserve comment on the artifacts, until you are
sure the drivers are all in place and everything is
working properly. Artifacts can mean the card is bad,
with say a lot of blocks or colored squares could mean
some bad video card RAM. If the problem is occasional,
it could be GPU temperature, and maybe you'd be seeing
it as the card gets warmed up by the 3DMark test run.
(Does the video card heatsink look solid to you ? Is
it tilted or does it look like it is out of place ?
I don't know anything about utils that can display
GPU temp, but that would be one thing worth checking
if you can find a util to do that.)

The freeze for 10 seconds, followed by the restart, doesn't
sound like a power problem. It sounds like the OS is
crashing because of a driver problem or something.
Is anything recorded in the Event Viewer ? If you
disable automatic restarts on a crash, what info is
displayed on the blue screen ?

In terms of your hardware power situation, I don't think
you are in trouble there. I found a picture of the label
on the side of a SmartPower 350W, and you can compare
my numbers to what is printed on the label on yours:

3.3@22A 5V@21A 12V1@10A 12V2@15A [email protected] +5VSB@2A (3.3V&5V power <130W)
http://www.endpcnoise.com/cgi-bin/e/std/sku=smartpower_350

I have an Nforce2 board with a 3200+ on it, and my measured
combined power is about 100W on 3.3V&5V. Which is less than
the 130W limit. And my board uses the 5V rail for CPU power.

AFAIK, the 6600GT draws 12V@4A in 3D mode, via the motherboard
power cable. I think that would be coming from the 10A output
(12V1). Again, I don't think you have enough other loads, to
exceed the 10A limit.

So, you could have a bad card. But first you have to clean uo
the software problems. In the interest of time, find a spare
hard drive. Disconnect all other hard drives in the
system (making notes of where they were connected), get out
your Windows install CD, and reinstall. Plop on the drivers
from the motherboard CD, which should give you chipset
drivers. Install the video card drivers from the video card
CD. Now, you can shut down and connect whatever disk has
the installer for 3DMark etc. Run Dxdiag and your other
tests, and see if things are any different.

By using a clean install, you may find that texture acceleration
is enabled and your score is better. If you are still seeing
artifacts and time is running out, return the card just in
case. Who knows how well video card memories are tested...

Paul

 

[altcomphardware]

Thanks for helping Paul

Don't know what alt.comp.hardware would be without you & kony :-D

And from your other post, you have "AMD XP 2500+, 1GB DDR333 RAM".
One thing missing, is what motherboard you are using.

Motherboard is an Abit NF7.

Entry #39 in the table here is beating your pants off
By 1356 points to your 269 points. And he is using an "XP 1900+".

http://www.techimo.com/forum/t161088.html

How embarassing! The result discrepancy *has* to be due to a problem,
right? Not variability between graphics card manufacturers, etc.

The very first question is, did you remember to uninstall the
drivers for the old video card ? That has caused me grief in
the past (so much grief, that in one case the only way I could
get a new video card to work right, was an OS reinstall). When
the old video card is still in the computer, you uninstall
the video driver. Then, you are ready to come up at 640x480
driverless with the new card. Followed by installing
whatever minimum DirectX version the new card needs, and the
new video card driver. (Assumes the chipset AGP driver is in
good shape and is up to date.) The best order of install is
AGP chipset driver first, video card driver second, and DirectX
can be either before or after or as many times as you like.

At first I didn't uninstall the drivers for my old Sapphire Ati Radeon
9800 Pro. It artifacts so badly I can't see anything even in Windows,
and my motherboard doesn't have onboard graphics.

I have uninstalled it through "Add/Remove Software", but as we all know
Windows tends to be quite sloppy in clearing up old files.

If the chipset and graphics card drivers were not installed
properly, I'd be surprised if 3DMark06 would run properly.
It could be that texture acceleration is not enabled. Run
Dxdiag from the directx9 install, and see what it reports.
(I'd recommend the taskbar options popup of Powerstrip from
entechtaiwan.com, but it has a limited trial period now. At
least I cannot run it any more.)

I will follow your advice later on to try it on a fresh Windows
install.

http://www.windowsresource.net/guides/hardware.php

I'll reserve comment on the artifacts, until you are
sure the drivers are all in place and everything is
working properly. Artifacts can mean the card is bad,
with say a lot of blocks or colored squares could mean
some bad video card RAM.

The patterns I saw was checkerboarding, with very distinct and well
separated squares.

If the problem is occasional,
it could be GPU temperature, and maybe you'd be seeing
it as the card gets warmed up by the 3DMark test run.
(Does the video card heatsink look solid to you ?

It is asymmetric so it is hard to tell. My Sapphire Radeon 9800 Pro's
heatsink moves a *bit* when you push it around, so isn't a bit of give
of the heatsink expected? Granted the Radeon was busted, which was why
I dared to push the heatsink around.

Is
it tilted or does it look like it is out of place ?
I don't know anything about utils that can display
GPU temp, but that would be one thing worth checking
if you can find a util to do that.)

The freeze for 10 seconds, followed by the restart, doesn't
sound like a power problem. It sounds like the OS is
crashing because of a driver problem or something.
Is anything recorded in the Event Viewer ? If you
disable automatic restarts on a crash, what info is
displayed on the blue screen ?

Faulting application oblivion.exe, version 0.1x, faulting module
oblivion.exe, version 0.1x, fault address 0x002c7c2c.

In terms of your hardware power situation, I don't think
you are in trouble there. I found a picture of the label
on the side of a SmartPower 350W, and you can compare
my numbers to what is printed on the label on yours:

3.3@22A 5V@21A 12V1@10A 12V2@15A [email protected] +5VSB@2A (3.3V&5V power <130W)
http://www.endpcnoise.com/cgi-bin/e/std/sku=smartpower_350

I have an Nforce2 board with a 3200+ on it, and my measured
combined power is about 100W on 3.3V&5V. Which is less than
the 130W limit. And my board uses the 5V rail for CPU power.

AFAIK, the 6600GT draws 12V@4A in 3D mode, via the motherboard
power cable. I think that would be coming from the 10A output
(12V1). Again, I don't think you have enough other loads, to
exceed the 10A limit.

So, you could have a bad card. But first you have to clean uo
the software problems. In the interest of time, find a spare
hard drive. Disconnect all other hard drives in the
system (making notes of where they were connected), get out
your Windows install CD, and reinstall. Plop on the drivers
from the motherboard CD, which should give you chipset
drivers. Install the video card drivers from the video card
CD. Now, you can shut down and connect whatever disk has
the installer for 3DMark etc. Run Dxdiag and your other
tests, and see if things are any different.

OK, I will try reinstalling Windows. I happen to have a spare 160GB HD
around.

By using a clean install, you may find that texture acceleration
is enabled and your score is better. If you are still seeing
artifacts and time is running out, return the card just in
case. Who knows how well video card memories are tested...

Paul

Thanks.

 

[altcomphardware]

What motherboard and PSU make AND model?

Motherboard is an Abit NF7, PSU Antec Smartpower 350W.

These are three distinctly different things and probably
more than one problem causing them.

One thing I thought of was that - my Sapphire Radeon 9800 Pro died
after I accidentally bumped it on a powered-on system. Could that have
triggered a surge that damaged other components?

Also, despite changing to a 3200+ rated CPU heatsink, my 2500+ still
overheats when I overclock it to 3200+. At the moment, everything is at
stock speed and I will be changing the thermal grease, but could a
partially damaged CPU due to overclocking be responsible, or would I
definitely notice if I had a damaged CPU?

Once, after an mobo overheat (>75 Celcius) inspired shutdown, the CPU
was identified as a 1100+ and not a 2500+!

You need to not lump all games together, take ONE game that
is known to work properly with it's newest patches and a
6600GT card, and which video driver (generally I would
suggest trying a late 8x.xx series driver, no older and not
a 9x.xx series.

OK, I will do what Paul says - reinstall Windows, then test games like
HL2 or Quake 4 or Doom 3 that have been out for a while and known to
work with the card.

You might try one of the 3DMark benchmarks too or instead,
or looping some benchmark demo... but again it has to be
known stable with your hardware and driver, as games are
rushed in development and quite a few need patches to work
on a lot of hardware.

Can you recommend any benchmark with a loop feature that I can run for
a few hours?

Reinstall DirectX9c. It doesn't matter if you already had
it installed... just do it.

OK, I will do that with my new Windows install.

If it's possible the sound is crashing it, disable the sound
or try other sound drivers.

Hmm, I don't think this is the cause. When my Radeon 9800 Pro was
working, Oblivion had the occasional crash but nothing as bad as this.

Is the CPU or video card overheating?

The CPU is at stock speeds, and my mobo should start beeping when it
gets too hot.

Try to isolate each, run Prime95's Torture Test, large
in-place FFTs setting for at least a few hours to see if the
CPU produces errors. If it does you must rectify that
problem before you can determine whether the video card is
ok or not (as mentioned above, you may easily have more than
one problem).

OK, will do.

Could be overheating, the driver, power supply, or game
bugs. Could even be vaulty or ESD damaged memory on the
card but this seems less likely than any of the former
possibilities.

I will test out other games on a fresh Windows install.

What video card did your system use previously?
Did you try playing any of the same games and did they have
any problems if so?

Sapphire Atlantis Radeon 9800 Pro 128MB. Oblivion had the occasional
rare crash (but I understand the game is buggy). Most other games were
stable with no problems.

Maybe if other things don't work, this could be useful but
for the time being, take a multimeter and measure the PSU
voltages under heavy load (like when gaming and running the
Prime95 Torture test), or if you don't have a multimeter nor
access to one, at least note the system temps reported by
software.

You might also leave the case open (does it have good
airflow??) and point a strong desk fan at the center to see
if it resolves anything.

Yup. My case has a 16" desk fan cooling it, which is how I bumped the
Radeon during operation :-/

That seems very odd, to get only 1FPS. I'd wonder if the
video card is overheating though instability from bad power
would also cause a similar pausing and downclocked speed ->
lower average performance.

Paul noted my 3DMark 2006 score of ~250 was very low compared to
another guy with a slower CPU (1600+ vs 2500+) who got ~1100! So it
seems something is definitely amiss.

Try an older 3DMark, 2001 and 2003. One of not both can be
set to loop mode, and by only setting it to run on high
detail, the demanding tests, you will keep as much load on
the video card as possible, minimize the CPU as a
bottleneck... or vice versa, choose low detail and low
resolution to put most load on CPU in a 3D environment.

OK. I'll try that. BRB after a couple of hours, it looks like I have my
afternoon mapped out :-(

 

[altcomphardware]

try the previous offical drivers (as kony suggest) 84.21. they're in the
archive section of www.nvidia.com and set AGP speed to 4x (I find 8x to show
up graphical glitches though it doesnt crash my machine. fast-writes OFF is
also best for stability. and if you mean you've recently changed cards then
thoroughly uninstalling the old drivers is a must and it helps to reinstall
the motherboard drivers (including the AGP port driver). google for
coolbits.reg and use it - it unlocks hidden parts of the display control
panel and check your temps - 50s when idling up to 65c when gaming is
normal.

good luck.

Thanks sleepy. I'm off to reinstall Windows. Luckily I have a backup
(if slow) computer so I won't be bored in the meantime

 

[kony]

Motherboard is an Abit NF7, PSU Antec Smartpower 350W.


One thing I thought of was that - my Sapphire Radeon 9800 Pro died
after I accidentally bumped it on a powered-on system. Could that have
triggered a surge that damaged other components?

To clarify, you mean that you bumped that system, and have
thus far only replaced the seemingly dead Radeon 9800 Pro?

If so, yes you may have damaged the AGP slot or something
else but I don't know what else. problem is we have too
many variables now, it would be good to test the card in
another system if possible, but even then it's not certain
whether you have a software (OS, driver or game) or hardware
(motherboard probably, or maybe PSU, the new card does use
more power, IIRC).

Also, despite changing to a 3200+ rated CPU heatsink,

What heatsink?
Lots of lower-end sinks (even some that aren't priced so
low) are rated a bit higher than realistically capable.
However, you're overclocking, there are more variables
involved. You'll just have to keep it at stock speed till
you can isolate the variable causing the instability.

my 2500+ still
overheats when I overclock it to 3200+.

The same rules for cooling still apply, decent sink, decent
thermal interface, decent chassis cooling... and not
overvolting it much unless you have a really good 'sink.

At the moment, everything is at
stock speed and I will be changing the thermal grease, but could a
partially damaged CPU due to overclocking be responsible, or would I
definitely notice if I had a damaged CPU?

How hot was it getting?
Generally, if it was only getting hot enough to crash for
awhile, it's probably fine. I've taken apart systems where
the junk silicone thermal grease had degraded till there was
a dark brown spot on the back of the CPU and after fixing
the cooling problem, it then ran fine... but I didn't try
overclocking it, this was someone else's system.

You can't really determine "definitely notice" it until you
first make sure it never gets too hot. The higher the
speed, (especially overclocking), the lower the stable temp
will be. You can raise the voltage slightly to offset this,
and indeed it could be necessary to raise voltage on a
Xp2500 to get it around 3200 and beyond, but with it
overheating already, it'd just run that much hotter... so
you have to keep it cool as-is first, and by doing so, you
also reduce the voltage it'll need for any given speed.

Once, after an mobo overheat (>75 Celcius) inspired shutdown, the CPU
was identified as a 1100+ and not a 2500+!

Not a big deal, the bios may do that - default to the lowest
FSB speed. it shouldn't be over 75C though, I'd have taken
the 'sink off long ago and reapplied thermal compound,
remounted it. Also confirm that it is running at the
correct voltage.

OK, I will do what Paul says - reinstall Windows, then test games like
HL2 or Quake 4 or Doom 3 that have been out for a while and known to
work with the card.

Considering the overheat issues, I'd probably open the case
and point a fan at it first, and remount the heatsink. We
know it's not overheating due to windows, so you might as
well take care of the known problem first then see what
other problems remain once it's running cool enough.
However, if it has been running too long long term, you
could have all kinds of data corruption but it seems
unlikely if you only have these crashes/reboots/etc in
games. That is the case, yes? Only problems while gaming,
never any other time?

Can you recommend any benchmark with a loop feature that I can run for
a few hours?

3Dmark2001 will, I don't recall about 2003.

OK, I will do that with my new Windows install.


Hmm, I don't think this is the cause. When my Radeon 9800 Pro was
working, Oblivion had the occasional crash but nothing as bad as this.


The CPU is at stock speeds, and my mobo should start beeping when it
gets too hot.

You should be able to keep that CPU under 60C (actually
significantly lower with a good heatsink) unless your
computer room is sweltering hot.

Yup. My case has a 16" desk fan cooling it, which is how I bumped the
Radeon during operation :-/

Pointing into the case? If so, you should definitely
remount the heatsink, the CPU should never be getting up to
75C. Now a stange thought- is it possible your motherboard
is reporting the wrong temp, and unnecessarily rebooting the
system due to this misreported temp though it might not
actually be that hot? It wouldn't account for all the
problems, but I still suspect you might have more than one
cause.

 

[altcomphardware]

Hi Kony. Looks like you all were right; old graphics card drivers were
part of the problem.

A fresh Windows install seemed to improve 3D Mark scores and remove the
hangs/reboots. However, my 3D Mark 2006 scores are still slightly
underperforming (another guy with a AMD XP 1600+ got ~1300 IIRC; see
Paul's post) and it is still artifacting. I will test it out with other
games tonight and see if it is just Oblivion that is causing problems.

Check out:
http://rapidshare.de/files/27578520/geforce6600gt_problems.JPG.html

1247 3D Marks with version 2006. I have taken some digital photos of
the artifacts in Oblivion in the bottom right corner. The right picture
has wierd-shaped artifacts because my shield arm went up and refreshed
the view :-/

To clarify, you mean that you bumped that system, and have
thus far only replaced the seemingly dead Radeon 9800 Pro?
Yes.

If so, yes you may have damaged the AGP slot or something
else but I don't know what else. problem is we have too
many variables now, it would be good to test the card in
another system if possible, but even then it's not certain
whether you have a software (OS, driver or game) or hardware
(motherboard probably, or maybe PSU, the new card does use
more power, IIRC).

OK it is possible there are other problems, or I might have gotten
lucky. A reinstall of Windows seemed to help.

I will test it out with other games tonight and see if there are any
other problems. Will also download 3D Mark 2001 and do a continuous
loop.

What heatsink?
Lots of lower-end sinks (even some that aren't priced so
low) are rated a bit higher than realistically capable.
However, you're overclocking, there are more variables
involved. You'll just have to keep it at stock speed till
you can isolate the variable causing the instability.

Crap! I didn't know coolers are like PSUs which rate higher than
reality. I thought WYSIWYG. This is a Speeze cooler.

http://www.ebuyer.com/UK/product/87736

The same rules for cooling still apply, decent sink, decent
thermal interface, decent chassis cooling... and not
overvolting it much unless you have a really good 'sink.

I'm guessing it is the thermal grease. It is a few years old (reused,
just scraped around the CPU die and slapped back on).

How hot was it getting?
Generally, if it was only getting hot enough to crash for
awhile, it's probably fine. I've taken apart systems where
the junk silicone thermal grease had degraded till there was
a dark brown spot on the back of the CPU and after fixing
the cooling problem, it then ran fine... but I didn't try
overclocking it, this was someone else's system.

You can't really determine "definitely notice" it until you
first make sure it never gets too hot. The higher the
speed, (especially overclocking), the lower the stable temp
will be. You can raise the voltage slightly to offset this,
and indeed it could be necessary to raise voltage on a
Xp2500 to get it around 3200 and beyond, but with it
overheating already, it'd just run that much hotter... so
you have to keep it cool as-is first, and by doing so, you
also reduce the voltage it'll need for any given speed.

The voltage to get it to work at 3200+ was 1.8V. Otherwise at 2500+ is
was the stock voltage for a Barton. (can't remember exactly)

Not a big deal, the bios may do that - default to the lowest
FSB speed. it shouldn't be over 75C though, I'd have taken
the 'sink off long ago and reapplied thermal compound,
remounted it. Also confirm that it is running at the
correct voltage.




Considering the overheat issues, I'd probably open the case
and point a fan at it first, and remount the heatsink. We
know it's not overheating due to windows, so you might as
well take care of the known problem first then see what
other problems remain once it's running cool enough.
However, if it has been running too long long term, you
could have all kinds of data corruption but it seems
unlikely if you only have these crashes/reboots/etc in
games. That is the case, yes? Only problems while gaming,
never any other time?

Yes. The problems were only in games. I only ever play FPS games. Now
and then when I still had the Radeon I'd get freezes in Doom 3 or
sudden exits in Oblivion.

Since reinstalling Windows and the drivers, and replacing the Radeon
9800 with the Geforce 6600, I get artifacts. I will test out other FPS
games for freezes.

3Dmark2001 will, I don't recall about 2003.

2003 doesn't at least with the demo version (just tested it).

You should be able to keep that CPU under 60C (actually
significantly lower with a good heatsink) unless your
computer room is sweltering hot.

Room temp is currently 26.8 Celcius.

Pointing into the case? If so, you should definitely
remount the heatsink, the CPU should never be getting up to
75C. Now a stange thought- is it possible your motherboard
is reporting the wrong temp, and unnecessarily rebooting the
system due to this misreported temp though it might not
actually be that hot? It wouldn't account for all the
problems, but I still suspect you might have more than one
cause.

Yeah, I suspect I have more than one cause myself. I'm OK not
overclocking my CPU for now, so I won't be getting some good new
thermal grease. When I do buy more stuff, I'll slap on some thermal
grease on the order and see how it goes.

 

[altcomphardware]

Incidentally, I am not sure if this is an important clue to the
remaining problem(s), but that annoying glitch of the mouse/keyboard
not working on a warm reboot but returning to normal on a power
off/power on reboot is still there, after a fresh Windows install.

Does this glitch mean anything to you guys?

 

[kony]

A fresh Windows install seemed to improve 3D Mark scores and remove the
hangs/reboots. However, my 3D Mark 2006 scores are still slightly
underperforming (another guy with a AMD XP 1600+ got ~1300 IIRC; see
Paul's post)

Don't worry about slight differences, those can be a matter
of tweaking later, after you determine if the card works
right at all or what the other system problem(s) were.

and it is still artifacting. I will test it out with other
games tonight and see if it is just Oblivion that is causing problems.

Check out:
http://rapidshare.de/files/27578520/geforce6600gt_problems.JPG.html

See if other things cause it, like 3Dmark. If 3Dmark is
also problematic, the card may be bad but we still haven't
ruled out the motherboard or PSU.

Crap! I didn't know coolers are like PSUs which rate higher than
reality. I thought WYSIWYG. This is a Speeze cooler.

http://www.ebuyer.com/UK/product/87736

That heatsink is not suitable for overclocking, it is
marginal for an XP2500 as it is. An appropriate heatsink
would have a solid copper baseplate, either a high
RPM/noisey 60-70mm fan or a quieter 80 x25mm fan.

If the base of it is rough, it might help to lap it, then
rub thermal compound hard into it and wipe it off (dry paper
towel only, leaving compound in the surface flaws) before
applying another tiny tiny speck of thermal grease and
reapplying to the CPU.

I'm guessing it is the thermal grease. It is a few years old (reused,
just scraped around the CPU die and slapped back on).

It's good to avoid doing that.
For an open flipchip like an Athlon XP, it is best to use a
thin coat of fresh synthetic based compound, not old and not
silicone based.

The voltage to get it to work at 3200+ was 1.8V.

No wonder it was overheating, you'll need a very good
heatsink to run it at 1.8V unless the room temp was very
low.

Also keep in mind that even if the heatsink were accurately
rated for an XP3200, that doesn't mean you can overvolt a
CPU to get there, it means an XP3200 at stock voltage.
Rasing the voltage increased the heat more than running at
XP3200 speed, quite a lot more. You might find that with
that heatsink lapped and good thermal compound, or a
different 'sink, it runs cooler, thus lower voltage is
needed, thus it runs ok at 3200 speed. Even so, it would be
as easy to just pick a lower value, something between XP2500
and XP3200 speeds if you are compelled to overclock.

One possibility if your board allows full adjustments in
bios is to pick the voltage, not the speed, that allows it
to stay cool enough, THEN see what speed it can do stabily
at that voltage... but of course minor readjustments and
stress testing will be necessary, you'll often find faults
in Prime95's Torture Test if the voltage is too low or heat
too high, or both.

Otherwise at 2500+ is
was the stock voltage for a Barton. (can't remember exactly)

Think it might be 1.6 or 1.65V, though I could be wrong, at
some point I got rid of all my regular Bartons and only have
mobile bartons left which are all overclocked. The label on
the CPU should allow searching for correct voltage but if
the motheboard is left at defaults, it should be right.

Since reinstalling Windows and the drivers, and replacing the Radeon
9800 with the Geforce 6600, I get artifacts. I will test out other FPS
games for freezes.

Your video card isn't wedged up against another card that
blocks good airflow?

 

[kony]

Incidentally, I am not sure if this is an important clue to the
remaining problem(s), but that annoying glitch of the mouse/keyboard
not working on a warm reboot but returning to normal on a power
off/power on reboot is still there, after a fresh Windows install.

Does this glitch mean anything to you guys?

No but you might see if there's a bios update for your
board.

 

[Paul]

Faulting application oblivion.exe, version 0.1x, faulting module
oblivion.exe, version 0.1x, fault address 0x002c7c2c.

When is the last time you ran memtest86+ (memtest.org)
on your memory ? (2 full passes error free is enough.)

I'd also try the Prime95 (mersenne.org) torture test option
while in Windows (or Linux even). It is also a good test that
the memory and CPU are good, even though Prime95 cannot
test every byte of the memory like memtest86+ can.

If you want to save time, just try Prime95, because it will be
testing the same chunk of memory that Oblivion was using. Use
both tests for more complete coverage.

If the error message mentioned an "nv" driver, that would make
a graphics problem more likely. If the error is in the application
itself, I'd want to test my memory just to eliminate that as
the source of the problem. As with all software, it could
also be a problem with Oblivion itself

Paul

 

[Paul]

At first I didn't uninstall the drivers for my old Sapphire Ati Radeon
9800 Pro. It artifacts so badly I can't see anything even in Windows,
and my motherboard doesn't have onboard graphics.

That is another hint for us. So both the old video card has
artifacts and the new video card has artifacts. Hmm. What
are the odds of that happening. Maybe the common factor
here is not the video card ? Like maybe the motherboard ?
Or maybe not enough computer case cooling ?

At least try the tests I suggested in the other post.

And visually inspect the motherboard, for anything
suspicious looking. Like leaking or bulging tops on
the capacitors (aluminum cylinders with plastic sleeving).

And remember that there is a limit to how much diagnosing
we can do over the Internet, so we cannot guarantee a solution.

Paul

 

[altcomphardware]

That is another hint for us. So both the old video card has
artifacts and the new video card has artifacts. Hmm. What
are the odds of that happening. Maybe the common factor
here is not the video card ? Like maybe the motherboard ?
Or maybe not enough computer case cooling ?

The old video card had artifacts right on bootup. The new video card
has artifacts so far only in one game and at random times in the game.
My guess is that the old video card was dead but the new one just has a
buggy game . I'm sick of Oblivion anyway; the story just keeps on
going on...and on...and on...

At least with Quake 4 or some other FPS you just get your hands dirty
and your daily irritations satisfied, then it is back to having your
day back

3DMark2003 looping for a few hours didn't show any artifacts, and the
results appear to be close to normal after a Windows reinstall (see
posts below).

At least try the tests I suggested in the other post.

And visually inspect the motherboard, for anything
suspicious looking. Like leaking or bulging tops on
the capacitors (aluminum cylinders with plastic sleeving).

And remember that there is a limit to how much diagnosing
we can do over the Internet, so we cannot guarantee a solution.

No problem Paul. You guys have been a great help already. I would never
have guessed a Windows reinstall was in order to fix the old driver
problem (I did uninstall the old drivers - as usual Windows is sloppy
about complete uninstallations).

 

[altcomphardware]

When is the last time you ran memtest86+ (memtest.org)
on your memory ? (2 full passes error free is enough.)

Probably about a year ago.

I'd also try the Prime95 (mersenne.org) torture test option
while in Windows (or Linux even). It is also a good test that
the memory and CPU are good, even though Prime95 cannot
test every byte of the memory like memtest86+ can.

OK, I'll run Prime95 overnight tonight.

If you want to save time, just try Prime95, because it will be
testing the same chunk of memory that Oblivion was using. Use
both tests for more complete coverage.

If the error message mentioned an "nv" driver, that would make
a graphics problem more likely. If the error is in the application
itself, I'd want to test my memory just to eliminate that as
the source of the problem. As with all software, it could
also be a problem with Oblivion itself

Yup. Software must be the only consumer good AFAIK that ships "sort of
working" and is patched along the way. It is just too expensive to
write a complex program and test all possibilities for debugging.

 

[altcomphardware]

Don't worry about slight differences, those can be a matter
of tweaking later, after you determine if the card works
right at all or what the other system problem(s) were.



See if other things cause it, like 3Dmark. If 3Dmark is
also problematic, the card may be bad but we still haven't
ruled out the motherboard or PSU.

*Phew* I didn't notice any problems with 3D Mark '03 looping for a few
hours (registered version has looping feature).

1600x1200 produced just a black screen though, although my monitor is
rated for up to 1600x1200. Not sure if it is important.

That heatsink is not suitable for overclocking, it is
marginal for an XP2500 as it is. An appropriate heatsink
would have a solid copper baseplate, either a high
RPM/noisey 60-70mm fan or a quieter 80 x25mm fan.

If the base of it is rough, it might help to lap it, then
rub thermal compound hard into it and wipe it off (dry paper
towel only, leaving compound in the surface flaws) before
applying another tiny tiny speck of thermal grease and
reapplying to the CPU.

I'm curious - has anyone developed a refridgerating case yet? Sort of
like an air conditioned, insulated case. That way all CPU/GPU/PSU/etc.
fans can be switched off if the case is internally kept at ~0 Celcius.
Use a large 12" fan or so inside the case to blow air around.

It's good to avoid doing that.
For an open flipchip like an Athlon XP, it is best to use a
thin coat of fresh synthetic based compound, not old and not
silicone based.

Yeah. Just about everyone I know expressed their horror the way I used
all 5g of the thermal paste on the CPU :-/. The way I saw it, just
about *anything* is a better conductor than air, so what the heck.

Is 99.9% pure isopropanol acceptable for cleaning out the gunk?

No wonder it was overheating, you'll need a very good
heatsink to run it at 1.8V unless the room temp was very
low.

Also keep in mind that even if the heatsink were accurately
rated for an XP3200, that doesn't mean you can overvolt a
CPU to get there, it means an XP3200 at stock voltage.
Rasing the voltage increased the heat more than running at
XP3200 speed, quite a lot more. You might find that with
that heatsink lapped and good thermal compound, or a
different 'sink, it runs cooler, thus lower voltage is
needed, thus it runs ok at 3200 speed. Even so, it would be
as easy to just pick a lower value, something between XP2500
and XP3200 speeds if you are compelled to overclock.

3200 was the magic number because it would match my memory's FSB of
400, IIRC.

One possibility if your board allows full adjustments in
bios is to pick the voltage, not the speed, that allows it
to stay cool enough, THEN see what speed it can do stabily
at that voltage... but of course minor readjustments and
stress testing will be necessary, you'll often find faults
in Prime95's Torture Test if the voltage is too low or heat
too high, or both.

Last winter, the 2500+ overclocked to 3200+ at 1.8V performed alright,
no problems with the Prime test. Now in summer a minute or so and it
crashes :-(

However I will be testing the comp overnight tonight at stock 2500+
speeds in case any spillover damage from my overclocking or knocked
graphics card occurred.

Think it might be 1.6 or 1.65V, though I could be wrong, at
some point I got rid of all my regular Bartons and only have
mobile bartons left which are all overclocked. The label on
the CPU should allow searching for correct voltage but if
the motheboard is left at defaults, it should be right.





Your video card isn't wedged up against another card that
blocks good airflow?

No, I have 2-3 empty PCI slots underneath the graphics card, for this
very reason.

Thanks for the help Kony. You and Paul are real contributors and assets
to the NG. Hats off

 

[kony]

I'm curious - has anyone developed a refridgerating case yet? Sort of
like an air conditioned, insulated case. That way all CPU/GPU/PSU/etc.
fans can be switched off if the case is internally kept at ~0 Celcius.
Use a large 12" fan or so inside the case to blow air around.

Small compressors usually aren't very quiet, who would pay
more for this?

Video card 'sinks are more difficult to find in HQ versions
as everyone seems taken to extremes of thin higher RPM fans
or those very loud squirrel cages but for the CPU, it's
quite easy to find quiet cooling. I can't hear the fan at
all on my o'c XP Barton Mobile, but it's a tad below 1.8V,
IIRC (don't recall exact voltage it uses at the moment).

Yeah. Just about everyone I know expressed their horror the way I used
all 5g of the thermal paste on the CPU :-/. The way I saw it, just
about *anything* is a better conductor than air, so what the heck.

If the compound isn't very thick, it'll squish out pretty
well from the core of an Athlon XP, but over the long term
it may still separate into silicone oil and solid islands
which dont' cool as well or as evenly.

Is 99.9% pure isopropanol acceptable for cleaning out the gunk?

Yes, though if it hasn't dried out and hardened, even a very
soft toothbrush, detergent, and warm tap water will suffice.
Just make sure that whatever you use, when you're done there
is minimal if any residue on the pins and it's completely
dry before reinstalling.

 

[altcomphardware]

OK, I'll run Prime95 overnight tonight.

Results of Prime95 overnight below. Everything seems OK!

http://rapidshare.de/files/27631288/prime95_test_2006_07_31.JPG.html

So maybe it was old drivers and a buggy game.