How hot is hot?

[Mr. Fri]

How hot is too hot?

A few months ago I build a system with an Athlon XP 2000 (1667 MHz). I have 5 fans. Two on the PS, one 80mm blowing in (from the bottom), one 80mm blowing out (top - right next to the CPU fan), and the CPU fan. I use a Thermaltake Volcano 9 to cool the CPU - it has the original 80mm fan on it. (Perhaps the Volcano 9 is overkill but I wanted to make sure the CPU doesn't overheat.)

Since I live in Texas, the room temperature is often 27-28 deg.C (80-82F). The temp inside the box is around 40-42. The CPU is usually 10-12 degrees hotter at around 50-55.

With all those fans, the system is quite loud. The inlet fan and CPU fan can both be adjusted. I have the inlet fan at max to cool the case the best it can. What I've found is that most of the noise comes from the CPU fan. I'd like to turn it down but that will make the CPU temp rise. So, the question is how hot can I run the Athlon without damaging it? Is 60 degrees too hot? What about 65 or 70? I wouldn't want to fry my CPU.

Thanks,
><>Mr. Fri

 

[Yuriman]

The volcano9 Is definantly not overkill, in fact, if you paid over $20 for it you got ripped off. Just buy an slk-800, and a smartfan2, it will be almost siled and lover your cpu temp as much as 10c. Of course, thats in optimal conditions, you should probably expect a 5c drop and a 70% noise drop.

 

[L337 M33P]

My setup is almost the same as yours - but my CPU runs @ 67C on a stock HSF - rock stable though, so far I have had 1 crash with XP in 12 months. Palominos are renowned for their space-heater-style heat output, so it would be pretty pointless if they fell over at lower temperatures. (chocolate teapot anyone?) Of course, this only applies to stock situations - when overclocking the CPU will definitely fall over at those sorts of temperatures, so the lower the better.

As for the big Tornado9 - it probably is big enough to keep the Pally HeatProducingUnit at a low enough temperature for mild overclocking.

My case temps are 38-39 in hot weather (room 29C) and the CPU is 67 full load. Stock speed

 

[Mr. Fri]

So, is it safe to keep running up the temp until it becomes unstable then back off a bit? Won't that damage my Athlon?

Mr. Fri

 

[Buhammot]

Well i have 38C idle temps, and about 42C load with a 29C ambient temp, or room temp. I just made sure that my case is super clean, (free of exposed wires), and good airflow. 2 intakes in the front, and 3 exhaust, which are less than the intakes. But airflow can be super critical.

If you dont have a PCI slot fan thing, then you might want to expell some air around there. Vid card gets hot too...

Also, i read somewhere that the absolute max cpu temp (dont remember what type) specified by amd was 95C, but that is way up there..

 

[L337 M33P]

Your CPU will die if left on well before 95C - as hitechjb has investigated, the leakage current of the transistors will cause thermal runaway at high temperatures - ever heard of secondary breakdown with bipolar ampifiers? The same principle applies but in less severity because the transistors are CMOS. Mine runs stable, but above 73C (if I run TOAST and leave the COP off) the CPU falls over.

As for damage, the CPU will not instantly die. There is a process known as Electromigration that slowly destroys the gates of the transistors on the CPU, making them become conductive and killing the transistor over time. The effect is much more pronounced at higher temperatures, but should not kill the CPU in the time frame between the next upgrade. If you plan on running it for more than 5 years, get a bigger heatsink.

 

[Mr. Fri]

I hear you folks saying "at higher temperatures" but what does that mean?

I realize there's no hard and fast rule like "at 75C the CPU will last for 5 yearts and at 80C it lasts for 2 years" but there has to be some guide lines that overclockers use.

I know that cooler is better, but what's considered "normal" (no OC) operating ranges? Is 40-70C what most systems run at or should I try to keep it at 60C or less?

 

[L337 M33P]

Yes, try keeping below 60C for non-OC systems

 

[DAGO]

While the CPU may well "survive" temps between 60-70 Celsius or above, even on a prolonged basis, those temps tend to make me think about better cooling. Adding a better Heatsink, Fans and improving airflow in your case can knock those temps down to the more comfortable levels BELOW 50° C which are more generally thought of as "acceptable" hereabouts...especially for "stock" rigs...

When peace of mind can be bought so cheaply, why take a chance??

"Keep it below 50° C" is my motto...

 

[breez]

You have to remember that 50C reported from mobo diode can be worse than 60C from core diode. Comparing between different mobos is problematic IMO.

 

[altec]

I try to stay below 50C also...but sometimes it cant be done with a .4 increase in voltage.

 

[Gautam]

My advice: Don't even give a crap about temps. Have to be blunt about it. As long as you have adequate cooling, and your system is stable, there is nothing to worry about. No set temperature. 50°C has become a nice arbitrary number, but you can in most cases operate above 60°C and still be safe. Many of you are, and don't realize it. Altec, sorry to use you as an example, but its too perfect. With your setup, its inevitable that your temps will be in the mid to high 60's on air, and low 50's with water, despite whatever the mobo may report. Your system will be unstable 99.99999% before you reach dangerous temps. If a system's stable, its safe. If a temp is dangerous, you won't even be able to keep the system stable for five seconds. Using the stock AMD heatsink, the higher end processors would probably run in the high 60's at stock, and this is what AMD approves as safe. Electromigration is always happening period, and you can't stop it. Higher temps are not going to cause any discernable damage to the processor unless you run it for over a decade, and even then its questionable.

A Volcano9 is more than enough cooling for a 2000+. Its good to up to 2.1ghz or so in most cases. Many have pushed them still farther.

 

[GoRm]

My V9 is coolin a 1700+ at 2312 vcore at 1.6, temps 34°c idle 42°c,load running prime MB Temps at 28°-30°c . Basement stays cool,room temps between 19°-21°c
What prog. are ya reading the temps with?

 

[Gautam]

It doesn't matter. Your temps are way off too, by about 10°C. There's no point in even bothering with temps.

 

[Mr. Fri]

I'm not sure what is the best programs to measure temps. I tried "Hardware Monitor" and am now tring "CPU Cool." They both give about the same temperature values.

I assume they're measuring the core temp, is that right?

Mr. Fri

 

[GoRm]

Fri. If there reading from the core diode your fine,like Gautam pointed out If your system is stable theres nothing to worry about. My old XP2000 Pally used to hover in the 50°-55°c also,may want try motherboard monitor also

 

[Gautam]

No matter what software you use, they all utilize the motherboard's sensors, and will as such read identically. If your motherboard uses an in-socket thermistor, no matter what program you use, it will read off the in-socket. Some motherboards do read off the die, in which case, are more accurate, but are far higher than what most people are accustomed to. Sorry if I was blunt, but they are going to be far off, and they usually shouldn't be taken seriously.

 

[adelphia83]

A Volcano 9 is more than enough for even a moderately clocked Athlon.

I wouldn't recommend replacing the cooler, unless you are experiencing instability, or are planning to overclock.

A Volcano 9 in my current system was adequate (and you can see from my sig, my CPU gets MUCH hotter than a 2000+ pally). By upgrading to an SLK-800, my temps dropped about 5C. Not sure if that 5 degrees is worth the 40$ I spent on it though.

Truth is, unless your motherboard reads temps right off the CPU die itself (very few do), you never really know what temp it's running at.

What many people don't know is that their temps are actually being read from a pocket of air underneath the CPU, which hardly correlates to the actual temperature of the die itself.

For example, the external CPU temp (probe underneath the socket, which is read by the majority of boards), reads in the 50C range at full load for my CPU on my board. The on-die temperature however reads about 65C full-load.

I believe max core temperatures for most Athlon XP's are in the 85C range. That's probably correct. To some, that figure seems outrageously high, because they're used to temps being read from the under socket temp probe (which reads much lower). If a CPU was to ever hit 85C using that probe, the real temp would be FAR higher, and you'd end up with a dead CPU.

 

[Gautam]

I believe max core temperatures for most Athlon XP's are in the 85C range. That's probably correct. To some, that figure seems outrageously high, because they're used to temps being read from the under socket temp probe (which reads much lower). If a CPU was to ever hit 85C using that probe, the real temp would be FAR higher, and you'd end up with a dead CPU.

This is an excellent point. The 85°C rating is for the core, not the in-socket. Therefore one can't draw a comparison between their temps and the maximum ratings.

 

[hitechjb1]

Temperature measurement and importance of HSF
...

the absolute temp measured is not as important as getting a "feeling and monitoring" of how the temperature is changing during Vcore increase at each step of the overclocking at max load. Since temperature increase will indicate/track how the internal Icore current (cannot be measured directly at this time), leaking current, and power pumped into the CPU for heavy computation.

If temperature increases (delta_temp/delta_Vcore) at a much faster pace than Vcore, the system would become unstable and crash during boot or even during idle time. It indicates that overclocking has reached a wall or limit for the given setup of HSF and PSU. I hit such wall a few times during these experiments, if you read the history of this thread, from PSU limitation, to fan limit, to HS limit, etc, etc.

I use the sensitivities of temperature to Vcore (delta temp/delta Vcore) and clock frequency to Vcore (delta frequency/delta Vccore) as the indicator rather than just the absolute temperature from temperature sensor. It would be even better this way than just based on a perceived, fixed limit on CPU temp. The temperature reading can be off by 10-20 C, and there is no way to tell when to stop if just based on temp reading alone. The system would become unstable way before reaching the 85-90 max die temp limit.

I find that for this type high curent, high power CPU, the die temperature has to be kept under certain temperature (say 40 C) in order for effective overclocking, i.e. delta frequency/delta Vcore > constant1 (say 100-140 MHz/100mV) and delta temp/delta Vcore < constant2, and max temp has to be under certain temperature (say 50 C) for stability at maximum overclocking. Otherwise the increase in die temperature per unit Vcore increase (due to leakage current and active power) would be increased at a faster rate than the frequency increase per unit of Vcore increase, and the temperature at some point will eventually slow the chip down to a point that any attempt to increase frequency by increasing Vcore will lead to more temperature rise and system instability. Such temperature should be much lower than the max die temperature limit from the CPU spec (85-90 C).
...

Links:

Temperature measurement and importance of HSF (page 5)

How much voltage can be applied to a CPU (page 5)