Gabriel said:
Hi all
I have a tower case with a warning temp display and sound that Ive just
figured out how to use and set, I know from previous question that a GPU can
be safe at 70 degrees and have a melt down at 85 degrees (is this right?),
but I need to know for my CPU how high to set the alarm temp - it is a :
CPU - Q9550 , Core 2 Quad Processor , FSB 1333 MHz , socket LGA775 , L2 Cach
12mb , Freq. 2.83GHz
And last
How hot can and should I set the temp warning for my OS HDD? Sata 160gig
HDD.
(my 3d card is a Nvidia Geforce 9800 GTX+ 512mb 16x pcie)
Thanks
GK
http://seagate.custkb.com/seagate/crm/selfservice/search.jsp?DocId=193771&Hilite=temperature
"The operating temperature range for most Seagate hard drives is 5 to 50 degrees Celsius"
Presumably that is "case temperature", so would correlate with sticking a
flat temp sensor on the drive casing.
The CPU and GPU will be harder to instrument, because of the difficulty of
making a good thermal connection to the case. The Intel documents on thermal design,
recommend milling a channel in the heatsink, so a sensor can be laid down and
then put in contact with the top of the processor. Trying to make a side connection
to the processor would be difficult, and could risk getting thermal
compound onto socket pins or the like. If the sensor gets jammed between
heatsink and CPU top, the air gap will reduce cooling efficiency.
The CPU is protected against overheat anyway. The throttle temperature is the
first response point. Effective compute rate drops when you hit the throttle
temperature (set to Tcase_max). A program like RMClock can be used for observing
whether throttling is occurring. The second temperature point is THERMTRIP, and
that is normally wired into the PSU PS_ON# signal. The computer, without warning,
will shut off, if the THERMTRIP temperature is surpassed. So if your heatsink
falls off, because all the pushpins fail, you're protected.
Tcase_max is 71.4C (throttling would start there). THERMTRIP is probably 20C more than that.
http://processorfinder.intel.com/details.aspx?sSpec=SLAWQ
The GPU likely has a throttling feature as well. At least some recent GPUs
have the ability to drop the clock rate, in the event of an overheat.
And sticking your sensor on the fins of the card, won't give a very
meaningful temperature. The fins will run much cooler than the core of
the GPU.
So
1) Hard drive is easy to instrument. The hard drive cannot protect itself.
An alarm makes sense.
2) CPU is well protected. CPU is hard to instrument. An alarm
might not really do much for you.
3) GPU probably doesn't have absolute protection against temperature.
If the heatsink falls off or the fan fails, even dropping the clock
speed may not be enough. An alarm makes sense. The silicon itself probably
fries (suffers damage) at 135C. The organic packaging of the chip, limits
the allowed temperature to a lower number. The body of the chip may be
degraded by operation at high temperatures. The underfill may also only
be able to handle stresses at lower temperatures. Maybe a figure like 95C-105C
makes sense ? You really need to get that from a datasheet (which Nvidia or ATI
won't be giving you). Who knows - maybe a utility that gives the current GPU
temp, will also be able to list the max allowed ?
An example of a number for a GPU here.
http://benchmarkreviews.com/index.p...k=view&id=289&Itemid=72&limit=1&limitstart=12
"But here's a little-known fact: the GT200 GPU is designed operate safely up to
its 105°C thermal threshold. What happens after that? Believe it or not, if
the GPU exceeds this temperature the clock speed will automatically be dialed
down to avoid damage."
I have heard of video cards frying (such as the guy who didn't have his
water pump turned on, for a GPU with water block). So if the clock throttling
feature exists, it doesn't dial the clock to zero Hertz. It would make
more sense for them to just disable the two power converters on the
video card instead, as an effective protection feature.
HTH,
Paul
