Am I thinking right about these temps?

I

IcarusSC

Gawd
Joined
Apr 19, 2006
Messages
614
I have an i7-920 running a low profile cooler (Evercool HPK) because it's built into an SFF. I recently decided to put my cores to work using Folding@Home SMP.

When not running FAH, I'm idling at about 30C, but the SMP client pushes my cores to 90-94C 24/7. Am I right in thinking that that's unhealthy for prolonged usage? Should I back off the SMP usage until I can get a bigger cooler, or is it OK to do this for six months or so?

Thanks for your suggestions -- I've read conflicting opinions on temps, so I just want to make sure my head is screwed on right about this.
 
Rising 60c (from idle) I would think is definitely a sign of bad cooling.
 
Stop right now.

Definitely get a better cooler, if you're going to be running 100% load 24/7.

In the meantime, if the chip is currently overclocked, drop it back down to stock speeds. If it's already at stock speeds (and you know how to overclock/undervolt), then you can look into seeing if you're able to undervolt the chip a bit to improve temps.
 
What kind of temp delta or temp max should I regard as acceptable, then? Just so I know I'm doing it right when I DO get a bigger cooler.
 
It is perfectly safe to run your CPU with load temps in the 90s. As long as they don't reach 100C, there is nothing to worry about.
 
Argh ... this is what I was talking about. Now it's the battle of the acceptable temperatures. I welcome advice, but if one person says one thing and another person says another, I don't really know what to do.

My temps aren't ever above 95 ...
 
The CPUs are designed to throttle at 100C and not below that. If Intel didn't think their CPUs could survive being loaded at temperatures like that, they would have set Tjmax at a lower temperature.
 
I totally understand your point. It does seem like there might be a difference between prolonged high temps and occasional high temps in throttle points. I'm not an Intel engineer, so I don't know what their thought process was, but I could certainly understand a decision to set a throttle point at 100C with the idea that prolonged 95C temps were also undesirable.

I'm just not sure :p
 
My 920 has hit the lower 90s on hotter days and it runs just fine. I have my CPU running Folding@home 24/7, so it is under constant load.
 
I will say this:

Ever since seeing the design quality of the Evercool WC-201 I had many years ago which was got my feet wet with watercooling, I have very little faith in Evercool products. They rely off fancy wording and blingage to sell their products. The performance is pretty basement level.
 
Zero82z said:
It is perfectly safe to run your CPU with load temps in the 90s. As long as they don't reach 100C, there is nothing to worry about.
Click to expand...

yup

took me awhile to find this pic

89419909.jpg
 
imho, try not to push it more than 80c on the smp.

try to get a better cooler.

It might prolong the proc.

my 920 D0 (1.42) HT ON 4ghz was around 75-80c.
 
It isn't OCd currently, mostly because of the weak cooler. I'm planning on switching up to a big cooler eventually, but it'll require some bodywork on the case.

I certainly don't want to appreciably shorten the lifespan of the machine by running it hot.
 
Based on what evidence?
Click to expand...
The Tj max is for stock speed & voltage only. As the CPU draws more current, there is more electromigration and Tj max goes down pretty rapidly.

There's no question that the CPU is operating outside of the safe limits described by Intel. However, a lot of people run their CPUs outside of those limits and, as far as I know, no rash of CPU failures has been reported. Some amount of degredation over time is pretty common, though. It's not uncommon for max overclock to slowly decrease over a, say, 2~3 year typical lifespan.

The only thing I would personally not do is to run an unstable overclock for any significant length of time, since that is much more likely to cause CPU damage.
 
The OP's CPU is not overclocked.
Click to expand...
Woah...you're right - I missed that message. His CPU should be safe, then, but he could probably get better cooling by trashing the heatsink and strapping a tin can on top of that thing.
 
Zero82z said:
It is perfectly safe to run your CPU with load temps in the 90s. As long as they don't reach 100C, there is nothing to worry about.
Click to expand...

Physically safe, more or less, though it's theoretically going to make the chip not last as long.

But keep in mind that TJmax is supposed to be the temperature where the chip is physically damaged by temperatures... short term. It is NOT meant to represent the threshold that a computer will get unstable during overheating, nor is it meant to represent long-term reliability.

If your CPU goes above 80c with ANY benchmark, even a difficult one like, say, Lynx, then your cooling needs to be improved. PERIOD.
 
You're not getting consistent answers on this because nobody really knows. There's a lot of hearsay and common beliefs about PC hardware which are probably safe but aren't necessarily correct.

You could make a reasonable deduction by following the hardware manufacturer's own guidelines. We can reasonably assume that Intel would design their processors and cooling to withstand normal usage, including running at 100% load 24/7 on stock clocks w/stock cooling. If the failure rate under warranty under those conditions were too high it would be expensive for them. Therefore we can probably say that the temps the CPU runs at at 100% with stock clocks/cooling is a good maximum for what is acceptable (figuring that it probably also includes a margin of safety).

So the answer is, do a test or find a test for those conditions (100% load, stock clocks/cooling) and use that as your reference point.
 
Yes, because Intel would never take the cheap way out of heatsinking.

Oh wait...



Bigger heatsink is necessary and I'd try to lower the Vcore as much as possible.
 
dandragonrage said:
Yes, because Intel would never take the cheap way out of heatsinking.

Oh wait...
Click to expand...

Not if it would hurt their bottom line. Do you have any data that supports the implication that their stock heatsinks are insufficient to the point of causing damage or instability under normal operating conditions?
 
Necere said:
Not if it would hurt their bottom line. Do you have any data that supports the implication that their stock heatsinks are insufficient to the point of causing damage or instability under normal operating conditions?
Click to expand...

I can all but guarantee you that I can take my CPU, stick the stock heatsink on with low quality TIM, run it at stock settings in a hot (maybe 90* F, and some people see even higher at some parts of the year if they don't have AC) room with Lynx and Furmark (to get case temps up - does Intel take that into account when giving us heatsinks?) for long enough to get the computer to crash due to heat. I'm not going to, though, because it would be too much work when winning this debate won't even mean anything to me.
 
dandragonrage said:
But keep in mind that TJmax is supposed to be the temperature where the chip is physically damaged by temperatures... short term. It is NOT meant to represent the threshold that a computer will get unstable during overheating, nor is it meant to represent long-term reliability.
Click to expand...

TJmax is the point throttling kicks in, my OC still holds to this day, I have even got it upto 200x21 for awhile but backed down for F@H stability.
 
dandragonrage said:
Physically safe, more or less, though it's theoretically going to make the chip not last as long.

But keep in mind that TJmax is supposed to be the temperature where the chip is physically damaged by temperatures... short term. It is NOT meant to represent the threshold that a computer will get unstable during overheating, nor is it meant to represent long-term reliability.

If your CPU goes above 80c with ANY benchmark, even a difficult one like, say, Lynx, then your cooling needs to be improved. PERIOD.
Click to expand...
Theoretically, yes. Practically, we're talking about a lifespan on the low end that will be longer than the CPU's useful lifetime. It'll become obsolete before it dies.

Tjmax has nothing to do with temperature damage. Intel does not allow their CPUs to reach a temperature where damage will occur. That's why the Tjmax and thermal shutdown systems exist in the first place.

Absolutely incorrect. If your CPU goes above 80C, that doesn't make a difference to anything. I fold on my PCs 24/7 and I have CPUs that continuously run in the low 70s and high 80s. There is nothing at all harmful about that.
dandragonrage said:
I can all but guarantee you that I can take my CPU, stick the stock heatsink on with low quality TIM, run it at stock settings in a hot (maybe 90* F, and some people see even higher at some parts of the year if they don't have AC) room with Lynx and Furmark (to get case temps up - does Intel take that into account when giving us heatsinks?) for long enough to get the computer to crash due to heat.
Click to expand...
Well, I can guarantee (without the "all but") that doing that with properly-functioning components will not cause the computer to crash, because I have done it numerous times in the past. I have run CPUs up to and past Tjmax while under full load, and they did not crash. They merely started to throttle once they reached it.
dandragonrage said:
I thought throttling was around 87-90c on these.
Click to expand...
Throttling is triggered when the DTSes report a distance to Tjmax of zero. In other words, when the CPU temperature reaches Tjmax.
 
It's entirely possible that I mis-mounted the heatsink. This is my first build ever, so I may have done it wrong, but what I don't know is how to be sure that you've mounted it right. I was very careful to tighten the pins in the correct order.
 
IcarusSC said:
It's entirely possible that I mis-mounted the heatsink. This is my first build ever, so I may have done it wrong, but what I don't know is how to be sure that you've mounted it right. I was very careful to tighten the pins in the correct order.
Click to expand...
If the pins are all the way through and the heatsink is mounted securely, it is installed correctly. If you aren't sure, push down hard on each of the pins and see if any of them go down any further.
 
I believe you're suppose to "twist" them to lock them down
 
86 5.0L said:
I believe you're suppose to "twist" them to lock them down
Click to expand...
No. Twisting a push pin in the direction of the arrow will release it. All the pins should be turned fully in the direction opposite to the arrow, and then simply pushed down to activate them.
 
Sorry, I misspoke -- they aren't pins, they're screws. And they're tightened as much as I dared -- didn't want to crack the board, but they're nice and tight; the cooler is on there solid.

@Zero: I had wondered about the theoretical vs. practical lifetime issue. So you're saying that even if I *do* wear it out sooner, it'll still be fully functional six or seven years from now, which is when I'll retire it?
 
IcarusSC said:
@Zero: I had wondered about the theoretical vs. practical lifetime issue. So you're saying that even if I *do* wear it out sooner, it'll still be fully functional six or seven years from now, which is when I'll retire it?
Click to expand...
Yes.
 
OK, so bit of an interesting fact (which I should have thought of earlier): I gave my case a thorough once-over with a can of air last night. Temps are now ~7C lower across the board.
 
You must log in or register to reply here.
Back
Top