Is this safe? Or how soon will my cpu fry?

[theJesus]

Been running Prime95 v25.7 build 3 for about an hour like that since it's apparently within the "tjmax", even if you take the highest reported temperature from any of those programs.

I don't know how accurate the core readings are though since they get stuck on idle.

 

[Mussels]

TJmax is how it 'guesses' the temperature, thusly why its less accurate the further from TJ it is.

short version: Hotter is it, the more accurate it is.

I prefer 20C to TJ, because my room can vary that much without air con.

 

[theJesus]

TJmax is how it 'guesses' the temperature, thusly why its less accurate the further from TJ it is.

short version: Hotter is it, the more accurate it is.

I prefer 20C to TJ, because my room can vary that much without air con.

Yeah, I wish Intel had made more accurate sensors and actually designed them for temperature monitoring. I read through the Realtemp guide and it seems like those measurements should be more accurate . . . maybe.

Anyway, do you think I'll be ok letting my temps get up that high under load, or should I tone it down a bit?

 

[Mussels]

IMO, as long as its 20C or more from TJmax, its ok.

30C gives you the room for the swings from winter to summer, if that helps you decide.

 

[FordGT90Concept]

Well, my two Xeon 5310 quad-cores were running at 80-84C for a good 9 months until I discovered they were that hot. I doubt they were damaged but I definitely do not recommend temps that high. Personally, I don't like anything higher than 60C on any processor.


TJ Max is probably the temperature from the power-off temp of the computer. My server is that way where tMax = 85C (never outright says it). If the temp ever hits or exceeds 85C, it orders the PSU to flip to cut off power supply.

 

[Mussels]

Personally, I don't like anything higher than 60C on any processor.

and how would you feel with my 8800GTX, that did 120C at stock?
They vary. TJmax is a way of *knowing* the safe range for your CPU.


TJ Max is probably the temperature from the power-off temp of the computer. My server is that way where tMax = 85C (never outright says it). If the temp ever hits or exceeds 85C, it orders the PSU to flip to cut off power supply.[/QUOTE]

TJmax is *not* what you think it is. TJmax is intels own tested maximum temperature, at which the chips crash at stock speeds.
TJmax can vary between CPU's (not just batches, but individual ones) which is why its hard to get reliable temperature readings - core 2 CPU's have a "degrees til crash' sensor, NOT a "degrees" sensor.

Basically, with core 2's you cant really say "oh its 60C, its 40C" - we dont know! its "X degrees until overheat/shutdown" instead.

 

[FordGT90Concept]

From what I've seen, the motherboard determines the cut off temperature, not the processor; however, the processor is responsible for reporting its temperature.

What I think really needs to happen is a concrete definition of where to take the CPU temperature and how to take it. Ya know, like how we tell the temperature of a person. We need more consistent means to monitor the health of major components.


And ya, I was talking CPUs there. GPUs run ridiculously hot...but do they? Or do they just use some wacky way to measure the temperature that exaggerates it.

 

[Mussels]

And ya, I was talking CPUs there. GPUs run ridiculously hot...but do they? Or do they just use some wacky way to measure the temperature that exaggerates it.

the stickers melted off my card. and my mobo, near the card. it also gave me a blister from touching it it was that hot alright.


Anyway, they *used* to have a temp sensore, with the core 2 line they moved to the TJmax way of doing it.

Setting cutoffs in the bios would be innaccurate - they all seem to assume TJmax is 100C (playing it safe? guessing?) so if you set it to 85C, thats 15C til TJmax.

If your CPU has a TJ of 100C, then its 85C. If you have 95C (Q6600) then its 80C. If you have a chip with 80C, then its 65C... and so on.

TJmax is a better method, however it does mess up some traditional temp reporting features/safety features.

 

[theJesus]

Well, I was gonna go ahead and keep it at those speeds, but they weren't stable enough at that voltage. Prime95 kept failing after 17~18 minutes, so I went back to 3.6ghz and started messing with my memory but these sticks are stubborn li'l buggers so some time I'll go back and try to get close to 4ghz again. Probably will do that after I finish working on my case and the weather gets a bit colder.

 

[cdnbum88]

What cooling are you doing?

Maybe you should look at better cooling options?

 

[theJesus]

What cooling are you doing?

Maybe you should look at better cooling options?

Look at my specs. That's nice cooling already, my room's just hot as hell.

However, all the airflow in my case is kinda messed up at the moment since I took off my entire rear panel to do some cutting and thus couldn't mount the cooler properly. Right now I've got it sitting on top of a drive cage exhausting out the front of my case with another fan in series to help it out. There's hardly any airflow over the motherboard since I don't have my side-panel on either. Just an 80mm on the northbridge and a small radial fan blowing across the vrm area.

I might try to modify that cooler though. I was thinking about mounting the pump somewhere else (it's mounted on the radiator by default to make it more compact) and putting a cpu heatsink there since it gets pretty hot at the base.

The other thing I want to try with it is taking one of the peltier's off and mounting it on a cpu heatsink.

 

[JC316]

Thats not too bad. My E5200 stays around there with air cooling. These 45nm chips are pretty damned tough anyway. Just don't let it peek over 70*C and you should be OK.