A follow-up...
My friend and I worked on the computer case setup last Saturday (perfect
weather too because it was 85F/29C degrees in my room!).
That's not all that bad, my aunt didn't get A/C till
recently and the system I built for her ran in > 95F during
the summer.
This old case
really does suck due to its limitations. There were no ways to improve its
airflow.
I kinda like old cases... for one they're "free", at least
to me as I have stacks of them. "Often" they're sturdier
too. Airflow is certainly a problem though, you pretty much
HAVE to take a nibbler, sabre saw or
whatever-your-weapon-of-choice and knock out a few new
holes, which of course has to be done before a system is
built in the case, and that is the key- assess and prep a
case even before ordering parts for it, at the very worst it
will have too much airflow and you can just throtte back
fans till they barely spin.
For now, we replaced all the case fans and CPU fan with new ones.
They were much cooler than the old dusty ones. However, CPU still got hot
when tested (140+/60C with cpuburn).
60C is not that bad, especially in an 85F room running
CPUBurn, I'd say you did pretty good given the poor case.
Since my friend build computers for living and we don't have another full-
tower ATX case to switch to at that time, he tried lowering the CPU power
usage and memory timing.
You can test the memory for stability, but just reducing
it's timings is probably of no benefit if it was stable.
That doesn't reduce it's power usage or heat generation.
Lowering the CPU voltage is the most significant (and along
with that, the frequency as needed to keep it stable at the
lower voltage in addtion to the far lesser heat savings from
the lower frequency itself. However with it running at 60C
in aforementioned test, I'd consider it fine, it should be
completely stable if it only got up to 60C, that is, from
the CPU perspective.
He says it won't be that noticeable in performance
(didn't notice anything slower when tested). The CPU was much cooler like
up to 130F/54C degrees during stress tests like games and cpuburn.
Motherboard's temperature sensor still got warm: 105+F/40.5C degrees like
before.
Northbridge I presume. That's cool enough but you might
touch-test the southbridge... i dont' recall your system at
this moment but some didn't have heatsink on SB but could
use one in higher ambient temps. Since there may be no
mounting method you'd have to use frag tape or thermal
adhesive (epoxy), the latter voiding the board warranty if
applicable.
So far, I have not any blue screens from gaming/3D mode (e.g., screen
savers) as of this post. However, I did had one lock up when quitting
World of Warcraft after 2.5 hours of nonstop playing. Also, WoW frozed
for 15 seconds during gaming twice already. I don't know if this is where
blue screens tried to happen.
Which video card?
Some modern cards will dynamically switch to lower speed &
voltage when they overheat, or ironically enough, when the
voltage is too low! Funny how they designed it to almost be
self-defeating but sure enough I've carefully observed it
happening on an nVidia FX5900 while monitoring core voltage
on a multimeter, temp with a probe, and adjusting the core
with a POT. Don't try this at home! Risky but I couldn't
help myself but to find out what was going on.
Anyway, I suspect that your video card is too hot and that
causes the stutter ("frozed" as you put it).. It could just
be a game or driver bug though, and the error when exiting
the game would probably be a game or driver bug as well
(though probably unrelated to the stuttering bug).
If you have a spare fan (even a desk fan), you might try
leaving the side-panel off the case and pointing the desk
fan at the bottom side (face) of the card. Leving the
adjacent PCI slot empty and that slot-position's case
bracket cover off might help too, but beware that this could
reduce airflow in other passive areas, like through the hard
drive bay?
I will keep testing. I don't want to conclude that the blue screens are
gone because I have had blue screens that didn't show up for a week
nonstop. They are unpredictable. They show up in 3D mode like games and
screen savers (only once with an ATI screen saver).
Voltage and temp readings during these occurances might
help.
If this doesn't fix the problem, then I will research more like a new
case.
Try that fan idea- in a well-cooled case leaving the
side-cover off will often make cooling worse but in this
situation it could help to troubleshoot it.
Now, I need to fix my other Debian/Linux backup machine which its power
supply died (I think). Ugh! So many problems, so little time.
If all your systems are facing elevated ambient temps,
they're all more susceptible to problems. Power supplies
too, even a good one won't respond well after awhile if it
runs too hot.
If you still have comments and ideas, then please feel free to post. I
check my newsgroup threads from Google.
A temp sensor (thermometer) is still a good idea. The
person using the sysetm(s) might be happier with an air
conditioner though, and the systems will like that too.