Heat probs reported for Intel E8400s

[informal]

Temperature rising as overclockers break into sweat

By INQUIRER staff: Friday, 01 February 2008, 9:34 AM

A HEAP of heat problems with Intel’s E8400 microprocessor are being reported on a host of hardware forums.

Users are claiming that the sensor diodes aren’t up to scratch and that’s causing the overclockers to get overheated themselves.

That’s not something Intel is admitting. A response from an Intel techie to a user over at OC Forums claims it has nothing to report on sensor problems but that just seems to be adding fuel to an already crackling blaze.

A search on Saint Google reveals there’s plenty of heated discussion raising people’s temperatures around the world. µ

Link

Some links for the threads:
XS thread
XS thread
XS thread
http://www.hardforum.com/showthread....post1032003123
http://forums.slickdeals.net/showthread.php?p=9946992

Anyone knows what's going on with temp. reports?Maybe a botched readings?

[STaRGaZeR]

OMFG bad sensors, the end of the world!

[Calmatory]

Sure intel just drops 10% from sensor reading, making it look good in reviews!!

[Eastcoasthandle]

Sure intel just drops 10% from sensor reading, making it look good in reviews!!

XS thread
XS thread
XS thread

[metro.cl]

The question is the CPU really get hot? or is just the sensor mis reporting...

[Calmatory]

XS thread
XS thread
XS thread

Oh boy,
Should have taken a look at the links posted in first post.

Sorry!

[informal]

The question is the CPU really get hot? or is just the sensor mis reporting...

Yep,that's a big question.Anyone knows if someone tried to measure the temps directly on the IHS?

[Eastcoasthandle]

Lets hope this temp problem gets sorted.

[[XC] gomeler]

Looks like when I re-bench my E8400 this weekend I'll stick a probe on the IHS and see what sort of temps I get. I know they overclock like beasts and that works for me.

[informal]

Looks like when I re-bench my E8400 this weekend I'll stick a probe on the IHS and see what sort of temps I get. I know they overclock like beasts and that works for me.

With your measurement on the very IHS,it should make it easier for us to know if the temps are just botched readings or the real deal(although the temp. of the core differs from the IHS temp. of course).

[stardust8750]

The question is the CPU really get hot? or is just the sensor mis reporting...

Wasn't that an issue when the e6x50's were released? Only on the other side of the scale?

[Shintai]

Since they use less power than the E6xxx series its easy to conclude they dont run hot, but rather colder.

Anyone tried with TAT instead to measure the thermals?

Sounds more like wrong numbers from the program(s) used.

[Bail_w]

There are certainly something wrong with the temperature of E8400. Ex, my Q6600 at 1.58v loaded at 65C when i touch the radiator it felt warm. On the other hand, my E8400 @ 1.4v loaded at 70C, i feel very cool on my radiator.

[brewer265]

its not just the e8400's, the qx9650's have the same problem. probably an issue on all the 45nm processors.

[GripS]

OMFG bad sensors, the end of the world!

Not the end of the world... But a huge annoyance nonetheless. Especially for those that are overclocking. Kinda tough to monitor the temps if they are incorrect now isn't it?

[Vapor]

Since they use less power than the E6xxx series its easy to conclude they dont run hot, but rather colder.

HUUUUGE misconception.

Okay, maybe not huge, but still a notable misconception I see at a lot of places.

Basically: yes, it may have a lower heat output per clock-volt, but because the die size is significantly smaller (and IHSs are not super functional), you cannot make generalizations about temps. Power density is no longer the same.

In fact, each HSF handles different heat-dump characteristics differently...some have very high load capacities but use it on a small chip at medium load and it can't move the heat away fast enough leading to high temps. The opposite is also true for some sinks.

And not to sound snobby or anything, but you go to phase/DI/LN2 and OC, you're largely flying blind for CPU temps--why do people rely on them so much on air/water? Heck, when B1 ES Conroes were around, we had nothing to know the temps and that didn't stop us from getting the exact same clocks we would eventually get with temp software. Practice your HSF mount, and just overclock.

[STaRGaZeR]

Not the end of the world... But a huge annoyance nonetheless. Especially for those that are overclocking. Kinda tough to monitor the temps if they are incorrect now isn't it?

Why is an annoyance? If my E8400 is rock stable with a moderate voltage but the reported temps are too high, I'm perfectly fine since it's stable.

[informal]

Why is an annoyance? If my E8400 is rock stable with a moderate voltage but the reported temps are too high, I'm perfectly fine since it's stable.

On a moderate voltage it's ok,BUT some people are considering putting more than 1.45V(which in and on itself could cause electromigration effect after long periods of time and damage the chip,even without taking the temperatures in consideration).So the temp. readings are important,but as Bail_w said in his post,the heatsink of his cooler felt cold so the chip probably has either the bad IHS contact and therefore bad heat transfer ,or the temps. are simply misreported(which could just annoy some people who are used to knowing every little detail in their precious system,and i personally can't blame them)

[Shintai]

HUUUUGE misconception.

Okay, maybe not huge, but still a notable misconception I see at a lot of places.

Basically: yes, it may have a lower heat output per clock-volt, but because the die size is significantly smaller (and IHSs are not super functional), you cannot make generalizations about temps. Power density is no longer the same.

In fact, each HSF handles different heat-dump characteristics differently...some have very high load capacities but use it on a small chip at medium load and it can't move the heat away fast enough leading to high temps. The opposite is also true for some sinks.

And not to sound snobby or anything, but you go to phase/DI/LN2 and OC, you're largely flying blind for CPU temps--why do people rely on them so much on air/water? Heck, when B1 ES Conroes were around, we had nothing to know the temps and that didn't stop us from getting the exact same clocks we would eventually get with temp software. Practice your HSF mount, and just overclock.

Ok if we stay at stock clocks for abit. 45nm uses about half the wattage from what we have seen in reviews. And the 45nm chips is about 74% the size of 65nm.

Also if you check the usual sensor reading programs. They all seem to show very different temperatures. In short, we just need updates to the programs with the new sensor information and base values.

[Xvys]

The version of TAT program floating around does not work with the Wolfdale's. Intel has responded that TAT was released for professional software developers only under a NDA (non disclosure agreement). It may or may not be updated and was not intended for general public use.

[Vapor]

Ok if we stay at stock clocks for abit. 45nm uses about half the wattage from what we have seen in reviews. And the 45nm chips is about 74% the size of 65nm.

Also if you check the usual sensor reading programs. They all seem to show very different temperatures. In short, we just need updates to the programs with the new sensor information and base values.

No argument there, but we don't buy expensive motherboards and HSFs to stay at stock settings.

1) they have lower stock voltages, which is huge for heat (considering user voltages haven't really gone down...), and
2) reach higher clocks

Power density is up for our purposes, therefore we should not expect OC'd temps to necessarily be better, even if heat output is lower.

Anyway, my main point is that temps are essentially useless for overclocking. Once you know you have good HSF/IHS contact, that's all you need (at least if you know what you're doing...and if you don't, don't overclock until you do).

[STaRGaZeR]

On a moderate voltage it's ok,BUT some people are considering putting more than 1.45V(which in and on itself could cause electromigration effect after long periods of time and damage the chip,even without taking the temperatures in consideration).So the temp. readings are important,but as Bail_w said in his post,the heatsink of his cooler felt cold so the chip probably has either the bad IHS contact and therefore bad heat transfer ,or the temps. are simply misreported(which could just annoy some people who are used to knowing every little detail in their precious system,and i personally can't blame them)

That's the point, some people are annoyed. As you usually say, % of overclockers and people that are actually worried about their temps is negligible. But for that %:

a) It's a bad contact between the IHS and the core itself. Intel must fix it.
b) The sensors are bad calibrated. I want them fixed, but I understand that most of the people don't care about it, so it won't have priority in Intel's problems-to-fix list.
c) They are really that hot. Intel can't do anything. This one is impossible though, because 45nm and low voltages used. The reason is a) or b).

[Shintai]

That's the point, some people are annoyed. As you usually say, % of overclockers and people that are actually worried about their temps is negligible. But for that %:

a) It's a bad contact between the IHS and the core itself. Intel must fix it.
b) The sensors are bad calibrated. I want them fixed, but I understand that most of the people don't care about it, so it won't have priority in Intel's problems-to-fix list.
c) They are really that hot. Intel can't do anything. This one is impossible though, because 45nm and low voltages used. The reason is a) or b).

You need a few more options in your list.

d) There is no problem, only with OC due to heat density.
e) Programs we use to read sensors are doing it wrongly. (Specially considering 3 programs gives 3 very different results).

[informal]

@STaRGaZeR

The c) part is not entirely impossible(look at Vapor's posts).Smaller die size coupled with higher voltages(than required for the 45nm process) and coupled with large increases in working frequencies(over 1Ghz overclocks) could lead to to problems(heat density).

[STaRGaZeR]

You need a few more options in your list.

d) There is no problem, only with OC due to heat density.
e) Programs we use to read sensors are doing it wrongly. (Specially considering 3 programs gives 3 very different results).

d) I don't think is that one, because it happens at stock too.
e) The only problem with monitoring programs is that the guy who actually writes the program (I'm talking about CoreTemp and Everest) has to adjust the Tj.Max temp for every processor model out there. Once it's adjusted, and for E8400 is already adjusted, the problem can only be a) or b).

@STaRGaZeR

The c) part is not entirely impossible(look at Vapor's posts).Smaller die size coupled with higher voltages(than required for the 45nm process) and coupled with large increases in working frequencies(over 1Ghz overclocks) could lead to to problems(heat density).

Yeah, but again it happens at stock too, so no high voltages. And at stock voltage the heat density of the 45nm parts alone is not enough to create the problem.

I think the reason is simply a bad calibrated sensor, and nothing else.