Why is 90 degrees for i7 considered reasonable?

JonJ678

JonJ678

JonJ678

Soldato
Joined
22 Dec 2008
Posts
10,369
Location
England
As the title. I don't have a link at present, Ill search if I must. 80-90 seems to be considered 'all right' for the i7 chip on here and elsewhere.

Intel rate it for 68 degrees, so I'd like to know where this idea has come from.

This isn't directly relavent to me, as I don't have an i7 system and I use water cooling. However I'd like to know anyway, so what am I missing here please?
 
I don't really like the idea of running a CPU at 80-90c either - my first rig was an Athlon XP, and people recommended keeping the CPU under 60c (mine crashed every time it hit 68!)

The way temps are measured has changed though. With the AXP, the temps came from a sensor under the socket, which was always a lot cooler than the CPU core itself, though there wasn't really any way to tell. The Athlon's core temp was probably close to 90c as well - I just didn't know it.

I suppose GPUs often run at 90-100c, so why shouldn't CPUs?
 
Intel can rate whatever they like at 68 degrees.

Chip and core temp often are a full 10 degrees apart.

The 920 at stock with stock cooler under IBT reaches core temperatures well in excess of 68degrees, as for the actual CPU temp I don't recall what figures I got, but they must have been 68 degrees or damn close to it.
 
68C? My i7 gets hotter than that under load at stock settings with the stock cooler so that can't be referring to core temperature...

I'm happy with up to 80C under load - remember, your CPU will never actually get that hot in real usage so that leaves you some buffer room for hot weather.
 
Really high temps may be safe for a particular chip, whether it's CPU or GPU, but all it does is prove that your cooling isn't up to the job of running at higher clocks.

My Q9550 will happily run at higher clocks and higher temps but I keep it at 3.4ghz 24/7 with the stock voltage so that it doesn't have too much of an effect on the GPU temp which is in the same loop. That's a limitation of my loop that I'm happy with as it's very quiet.
 
Cheers guys. I'm not sure this has cleared the issue up though.

Am I right in summarising the above as until i7, software temperatures matched up to intel's spec. Since i7, they no longer bear any relation to each other?

If so, this seems a rather sudden change of heart by Intel. Does anyone know where this information comes from?
 
According to Core Temp and RealTemp my pre-clocked i7 920 @ 3.4 GHz seems to be idling at high 50s-60c, and when I tried testing it with Prime95 the core temps went to 100C within seconds! I stopped the test immediately of course...

Actually it was a pretty hot day yesterday, and the readings are a few degrees cooler today (not gonna risk testing again tho). But I am concerned about it being so hot. The case is a Coolermaster Elite 335, with an Akasa AK-967 CPU cooler - so I don't understand why the temps should get that high?

I know that different programs report different readings, and some aren't accurate etc. etc. SpeedFan reports the temps almost 10c lower than Core Temp and RealTemp, but I've read that those two are the most accurate and the fact that they both report exactly the same temps makes me think they're nearer the mark.
 
Rise ye deceased thread

I'm going to nominate your case as the cause of the temperatures, seeing as it's basically a very little sealed box. Next most likely is a bad heatsink mount, push pins are tricky and I guess you didn't take the board out to check they were all the way through.


Intel recommend the temperature reading in the bios, keep this below 67.9 and you're fine. Not sure I believe them on this one. I think it shuts down at 100 degrees or so, so yeah that's pretty hot
 
the 67.9c max quoted by intel is the cpu temperature at taken from the socket sensor and is the temp read by bios. there are also sensors that record the individual core temps which max out at thermal junction max of 100c, this is the temp read by Everest, Realtemp etc....

realtemp and coretemp work out the temp by deducting the reading from the TJmax and giving you a value. At 100c the i7 will throttle back or shut down depending on the heat.

On the UD5 mobo, easytune reads the socket or surface temp and this must stay below 67.9c, realtemp will show coretemp and this must stay below 100c, but 85c seems to be the safe limit.
 
Gaidin109 said:
the 67.9c max quoted by intel is the cpu temperature at taken from the socket sensor and is the temp read by bios. there are also sensors that record the individual core temps which max out at thermal junction max of 100c, this is the temp read by Everest, Realtemp etc....

realtemp and coretemp work out the temp by deducting the reading from the TJmax and giving you a value. At 100c the i7 will throttle back or shut down depending on the heat.

On the UD5 mobo, easytune reads the socket or surface temp and this must stay below 67.9c, realtemp will show coretemp and this must stay below 100c, but 85c seems to be the safe limit.
Click to expand...

Yep, that is correct. Intel say 67.9C is the temperature reading from the bios cpu temperature or via motherboard software of cpu temperature, (don't read 67.9C on either core temp or real temp by confusing. Anything up to 80C is the best safe side, about 20C degrees lower than the Tjmax set at 100. But, Intel told me that set the tjmax at 95 for i7 d0 stepping and 100 for i7 c0 cpu processor. Very confusing indeed. When I get to 85C, 85C, 85C 85C on all cores reading from real temp or core temp but the true temperature on the cpu temperature is at 57C That was 10 degrees lower than Intel told us to be under 67.9C. As the pc doesn't shutdown as I set the bios to shutdown when the cpu reach 70C ! but I know it safer anyway. I did try 90C on realtemp as the pc did shutdown after reach 70C on the cpu temperature. So, it work out 20C degrees higher on core temp or real temp. So anything up to 80C is best side.
 
Apparently, Intel CPU's are still covered by warranty as long as the temperature in the BIOS reads under 68c, even though that isn't under load. Someone on here had a conversation with Intel about it, and they confirmed it.
 
I don't understand what people find so difficult about all this.

Intel design the CPU. They then have to give the motherboard manufacturers a temperature the motherboard as to be able handle without dying (TCase). That figure is the 'maximum' the MOTHERBOARD can handle. Not the CPU. High temperature motherboard issues is one reason that Intel switched to metal retention mechanisms for S775 and S1366. Many motherboard manufacturers also now fit large heatsinks under the CPU socket to disperse heat. That's not to cool the CPU (although it also does this) it's to keep the TCase figure within specification.

To reduce warranty claims Intel then build a thermal protection circuit into their processors. They have been doing this for at least 6 years. It's called Thermal Throttling and it works in two stages. The first stage reduces the clockspeed of the CPU so the heat reduces and the second stage shuts the CPU down completely if it get's to dangerous levels.

Because of Thermal Throttling YOU CANNOT HEAT DAMAGE AN INTEL CPU. IT JUST SHUTS ITSELF DOWN TO PROTECT ITSELF. Caps for the hard of understanding. It's like the rev limiter on a car. If you put your foot down in first gear the rev limiter stops you destroying the engine. Thermal Throttling is the same thing. Anyone who is reducing an overclock because of high temperatures is running their engine at 75% revs because they're worried about over-revving.

I would be much more concerned about over-volting an Intel 45nm CPU with an Intel chipset motherboard because that will certainly damage it, and reduce it's lifespan to weeks or months in some cases.
 
Thanks guys, especially WJA96 - the reason it's hard to understand is because so many people post different/misguided advice about this! I did wonder if the CPU should automatically shut down if it really was at dangerous heat levels, like 100c. It didn't show any signs of this when being tested. However, the display has shut off automatically a couple of times in the middle of game-play, crashing the comp and requiring manual switch-off and restart (I dunno if this is a heat-related issue tho, coz one time I'd only just started playing).

The main things I want to know are a) should I be worrying about the effect of the heat on the lifespan of the system, and b) what can I change to stop this (not a techie so want to avoid opening the case and fiddling around inside).

The specs are:
Chipset = Intel i7 920 2.66 GHz @ 3.4 GHz
CPU Cooler = Akasa AK-967
Motherboard = Gigabyte EX58-UD3R
VCore = 1.23v
Maximum TCase = 67.9C
Graphics = EAH4870 1gb DDR5 + glaciator
RAM = Corsair XMS3 6GB

The graphics card idles around 40c and loads at late 50s-mid60s (according to ASUS SmartDoc)

The coolest it's ever been by any of the software readings is below:

tempsqwd.jpg
 
Last edited:
Myshkin said:
However, the display has shut off automatically a couple of times in the middle of game-play, crashing the comp and requiring manual switch-off and restart (I dunno if this is a heat-related issue tho, coz one time I'd only just started playing).
Click to expand...

That sounds like one core is unstable and if you were unlucky and got that core allocated to your game, it would just shut down. Also, it could be graphics related.

Myshkin said:
The main things I want to know are a) should I be worrying about the effect of the heat on the lifespan of the system, and b) what can I change to stop this (not a techie so want to avoid opening the case and fiddling around inside).
Click to expand...

Is it an overclocked bundle or did you overclock it yourself?

Myshkin said:
VCore = 1.63v
Click to expand...

I hope that's a mis-type or it would explain a great deal about your high temperatures;)

Myshkin said:
The graphics card idles around 40c and loads at late 50s-mid60s (according to ASUS SmartDoc)
Click to expand...

You might think about using standard tools like CPUz, REALTemp and CPUz as things like SmartDoc can give 'misleading' readings.
 
Lower that vcore! :eek:

You won't need more than 1.35v, even with a 4GHz overclock. I think max safe voltage is 1.5v, so I'd definitely lower it.

Edit: Never mind, I noticed speedfan says your CPU vcore is 1.23v, and RAM voltage is 1.63v. That's OK. :p
 
Last edited:
D D Danneh said:
Lower that vcore! :eek:

You won't need more than 1.35v, even with a 4GHz overclock. I think max safe voltage is 1.5v, so I'd definitely lower it.

Edit: Never mind, I noticed speedfan says your CPU vcore is 1.23v, and RAM voltage is 1.63v. That's OK. :p
Click to expand...


Vid range for i7 920 is 0.8v to 1.375v, with a D0, try to keep within these limits, DO NOT overvolt your i7 to 1.5v. You can disable thermal throttling in bios, but its not recommended.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom