E8400 Temperature problems

[Mav451]

Whats the packdate on your good chip? My packdate is 01/09/2008 and coretemp has a bit of flucuation (not as bad you what you describe) and Core0 is always cooler than Core1... which is strange.

It's all in the sig guys, 1/15/08 pack date.

 

[WhiteZero]

See what you guys think of this.
According to Abit uGuru my System Temp is 31C. SpeedFan reads my Ambient as about 41C
Reading from CoreTemp:
My E8400 idles at Core0: 27C, Core1: 26C
As soon as I turn on Orthos, Core0 jumps to 34C and Core1 jumps to 35C with in 1 second of Orthos starting.
Core0 maxes out around 44C and Core1 at 42C
Does this sound right?

I'm using the Thermalright Ultima-90 with a 120mm Scythe S-Flex fan and AS5.

 

[Forceman]

See what you guys think of this.
According to Abit uGuru my System Temp is 31C. SpeedFan reads my Ambient as about 41C
Reading from CoreTemp:
My E8400 idles at Core0: 27C, Core1: 26C
As soon as I turn on Orthos, Core0 jumps to 34C and Core1 jumps to 35C with in 1 second of Orthos starting.
Core0 maxes out around 44C and Core1 at 42C
Does this sound right?

I'm using the Thermalright Ultima-90 with a 120mm Scythe S-Flex fan and AS5.

Sounds too low to me, but who knows. Abit's are known for having issues with E8400 temps.

 

[Sabrewulf165]

Didn't bother reading the rest of the thread but I put together my Wolfdale build yesterday with the stock heatsink (for now) and am getting ~50c idle, 65c load (orthos) using Core Temp 0.96.1 (which correctly identifies my CPU).

E8400
Gigabyte GA-P35-DS3L w/ Bios F7
Shitty temporary case, 120mm exhaust @ 1200rpm, stock E8400 heatsink on autospeed

I've read that you "need" the F8a BIOS to support Wolfdales, but my board w/ F7 properly recognizes and runs the CPU just fine *shrug*

 

[WhiteZero]

Didn't bother reading the rest of the thread but I put together my Wolfdale build yesterday with the stock heatsink (for now) and am getting ~50c idle, 65c load (orthos) using Core Temp 0.96.1 (which correctly identifies my CPU).

Cool. If thats what you get with the stock heatsink, I think my temps might be accurate with CoreTemp. Looking at AnanTech's Ultima-90 review, the stock Intel heatsink got 21C higher temps under load. And your getting exactly 21C more than me under load as well.
Grated that was with a different Intel CPU and different stock cooler, but still.

 

[KamikazeCow]

Well, after double checking the heatsink, i found my problem. When i first installed it, only 2 of the tabs popped in, while the other 2 went into the motherboard, but never "clicked" in. I was a little afraid to push them any further before, because it was bending the motherboard a bit, and i didnt want to crack the thing.... Anyway, now the e8400 is idling at 26C (acording to the bios) but core temp says its 51c. Either way, its much cooler then before. For people who are able to OC their 8400 to 4ghz+, which do you use to monitor the CPU's temperature? Core temp, the bios, or something else?

 

[Nenu]

Well, after double checking the heatsink, i found my problem. When i first installed it, only 2 of the tabs popped in, while the other 2 went into the motherboard, but never "clicked" in. I was a little afraid to push them any further before, because it was bending the motherboard a bit, and i didnt want to crack the thing.... Anyway, now the e8400 is idling at 26C (acording to the bios) but core temp says its 51c. Either way, its much cooler then before. For people who are able to OC their 8400 to 4ghz+, which do you use to monitor the CPU's temperature? Core temp, the bios, or something else?

Use Coretemp, theres a setting in Options "Show delta to Tj max.", enable this.
Now the figure shown is how many degrees C you have until the CPU will auto shutdown.
You likely wont get anywhere near the figure so temperature shouldnt matter.
Intel Recommend going no higher than 72C using your motherboards sensor but we know how unreliable that can be.

 

[blade52x]

Add about 8-9C to your CPU temp and that's your Core Temp. That's how it pretty much is on my Gigabyte board with the Q6600 at least. That temperature reading is being pulled straight from the BIOS. The CPU temp always seems to be about 8-9C behind the average core temps.

I'm not sure about the accuracy of this sensor, I might run stock settings just to see how close to ambient is gets, because I'm pretty much sure it's not possible to get within 5C of ambient unless undervolting, but I'm always well above it anyway.

 

[Forceman]

Add about 8-9C to your CPU temp and that's your Core Temp. That's how it pretty much is on my Gigabyte board with the Q6600 at least. That temperature reading is being pulled straight from the BIOS. The CPU temp always seems to be about 8-9C behind the average core temps.

The problem is that you don't know where the "cpu" temp is coming from. The only on-chip sensor is the DTS, which coretemp reads, so the BIOS (and motherboard) are getting that CPU temp somewhere else - probably a thermal sensor around or maybe under the socket. But without knowing what that sensor is reading you don't really know what a safe temp is. Intel's Tcase, which is what their max temps are based on, is measured by putting a sensor right on the top of the heatsink, which is pretty much useless for any real world applications.

 

[unclewebb]

I pointed an IR thermometer at a bare E8400 core all the way up to 100C and discovered that TjMax for these processors is 95C and not 105C that CoreTemp is assuming. Here's the long winded version that I posted on XS:

http://www.xtremesystems.org/forums/showpost.php?p=2763280&postcount=1259

http://www.xtremesystems.org/forums/showpost.php?p=2765889&postcount=1308

http://www.xtremesystems.org/forums/showpost.php?p=2766465&postcount=1328

If you don't have time to read then use CoreTemp and subtract exactly 10C from the reported core temperatures. Your temps will now be 100% accurate from 60C to 95C. No software will be 100% accurate below 60C because the DTS is not now and never has been accurate for reporting idle temperatures. Intel did not design or calibrate it for that purpose. Air cooled idle temperatures might be off by +/- 5C or even more after the above correction. During testing 40C reported was 35C actual.

If your load temps never get up over 60C then your temps will never be 100% accurate if you use software like CoreTemp that reads data from the on chip digital thermal sensors. Only the mobile 45nm chips have an Intel documented TjMax=105C. My testing shows TjMax=95C for the desktop 45nm dual core processors.

If your E8x00 is suffering from the stuck sensor problem where temperatures don't move whether at idle or full load then you are screwed and there is nothing you can do. No software is going to work properly. Garbage in = garbage out.

 

[Vix]

Thanks for that info UW!

I noticed that my temp in BIOS and my CoreTemp temperatures were off by approx 10 degrees. That would explain the difference.

Gee, maybe my CoolIT Freezone isn't AFU like I thought it was...

THANKS AGAIN

 

[Forceman]

I pointed an IR thermometer at a bare E8400 core all the way up to 100C and discovered that TjMax for these processors is 95C and not 105C that CoreTemp is assuming. Here's the long winded version that I posted on XS:

http://www.xtremesystems.org/forums/showpost.php?p=2763280&postcount=1259

http://www.xtremesystems.org/forums/showpost.php?p=2765889&postcount=1308

http://www.xtremesystems.org/forums/showpost.php?p=2766465&postcount=1328

If you don't have time to read then use CoreTemp and subtract exactly 10C from the reported core temperatures. Your temps will now be 100% accurate from 60C to 95C. No software will be 100% accurate below 60C because the DTS is not now and never has been accurate for reporting idle temperatures. Intel did not design or calibrate it for that purpose. Air cooled idle temperatures might be off by +/- 5C or even more after the above correction. During testing 40C reported was 35C actual.

If your load temps never get up over 60C then your temps will never be 100% accurate if you use software like CoreTemp that reads data from the on chip digital thermal sensors. Only the mobile 45nm chips have an Intel documented TjMax=105C. My testing shows TjMax=95C for the desktop 45nm dual core processors.

If your E8x00 is suffering from the stuck sensor problem where temperatures don't move whether at idle or full load then you are screwed and there is nothing you can do. No software is going to work properly. Garbage in = garbage out.

That is good news - makes temps like mine seem a lot more reasonable. Lot of people have really low coretemps though (like 50s load) - I wonder if those are accurate as they seem too low to be true when you subtract 10 from them. I'm guessing your IR thermometer is correctly calibrated and all?

 

[blade52x]

The problem is that you don't know where the "cpu" temp is coming from. The only on-chip sensor is the DTS, which coretemp reads, so the BIOS (and motherboard) are getting that CPU temp somewhere else - probably a thermal sensor around or maybe under the socket. But without knowing what that sensor is reading you don't really know what a safe temp is. Intel's Tcase, which is what their max temps are based on, is measured by putting a sensor right on the top of the heatsink, which is pretty much useless for any real world applications.

But I do, because I've got functioning core temps. I'm just letting the OP know how the CPU temp may possibly relate to his real core temps. It was the same on an older E6600, the CPU temp being about 8-9C lower than the core temps. Whatever that sensor is reading, it's directly related to the temperature of the cores... so it could have it's use for those with broken sensors.

 

[Merauder|FX]

I as well built my Wolfdale yesterday and ran some test. I have a Zalman 9500A my case has 3 120mm fans 1 in 2 out. i let it run all night and my temps floated at 48ºC and a full load with COD4 was around 62ºC.

I have the latest updates on my Asus Striker Extreme. I will check the bios temps when i get home.

Asus Striker Extreme
Intel C2D Wolfdale 3ghz@ 4Ghz
3gb Corsair XMS PC3 10600 DDR3 1333mhz
2x EVGA 8800 GTX

The system sits on my desk, and my room is always cool so that may be a factor.

 

[unclewebb]

That is good news - makes temps like mine seem a lot more reasonable. Lot of people have really low coretemps though (like 50s load) - I wonder if those are accurate as they seem too low to be true when you subtract 10 from them. I'm guessing your IR thermometer is correctly calibrated and all?

I used a Fluke 62 Mini IR Thermometer for my testing and its accuracy is rated at +/- 1C.

As I said before, the DTS is calibrated by Intel and seems to be extremely accurate within a 35C range between 60C and 95C. If your load temps don't get up into this range then even after my TjMax=95C correction, your temperatures based on the DTS are not going to be 100% accurate.

Here's an interesting comparison between my previous E6400 and my new E8400:

I set both CPUs to 1600 MHz and core voltage to 1.080 volts which is the minimum for my P5B Dlx. I'm using a Tuniq and I set the fan to max in a very cold basement to try to get the idle temperatures as low as they can possibly go so I could do a fair, apples to apples, comparison.

The temp sensor in Core0 of my E8400 gets stuck at 38C but the sensor in Core1 works fine. It sat mainly at 35C but would occasionally drop to 34C showing that it wasn't getting stuck like the other one.

Anyone that believes that the DTS is accurate at low idle temperatures and that TjMax=105C for the E8400 needs to somehow explain the 25C difference in idle temperatures between two CPUs that should, in theory, be almost equal.

My E6400 is reporting a core temperature that is about 3C below the ambient temperature of 12C. That's impossible with an air cooler so obviously my E6400 is reporting too low when the idle temp is very low. My E8400 is the opposite. At low idle temperatures it reports too high. I measured an actual core temperature of 35C when it was reporting 40C which is 5C too high. At even lower idle temperatures, you would expect this margin of error to increase further.

To explain a 25C difference we can throw away 1C based on measurement error, room temperature changes, etc. If TjMax was properly at 95C that would explain 10 more degrees which leaves 14 left. Divide that number in two and that means my E6400 might be reading about 7C too low and my E8400 might be reading about 7C too high at idle. That is reasonable based on what I've observed but I'm open to any other explanation.

Intel never designed or calibrated these sensors to be accurate at very low core temperatures. They were calibrated to trigger thermal throttling at 95C. A margin of error that could be as high as +/- 7C at idle is huge so trying to compare idle temperatures to your neighbor is pointless. Reseating your heatsink 3 times based on inaccurate data goes beyond pointless to the point of being stupid.

 

[Forceman]

Good info. It's too bad Intel doesn't publish more data about the DTS and Tjunction of their chips.

 

[unclewebb]

After watching the performance of the DTS across the entire temperature range of the processor, I can fully understand why Intel doesn't document these things further.

 

[GrimR]

Can I ask a quick question? I'm looking at putting together a new rig in March and was looking at the E8400. I'm now a little worried about motherboard support and some of the other issues with this chip. Would it be worth it to pick up a cheaper CPU for now, like an E4500, for now and then just getting the 8400, or the Q9450, when some of the wrinkles are ironed out? Or should I just drop the additional $30 or so and pick up a Q6600? This would be mostly used for gaming with an occasional DVD rip.

I'm upgrading from an old Northwind 2.8 so even the low end part should still be a temporary boost. I was going to get an E6750 but I thought the ~$40 more for the E8400 would be worth the extra cash.

Thanks in advance.

 

[DaNotoriousDOM]

Can I ask a quick question? I'm looking at putting together a new rig in March and was looking at the E8400. I'm now a little worried about motherboard support and some of the other issues with this chip. Would it be worth it to pick up a cheaper CPU for now, like an E4500, for now and then just getting the 8400, or the Q9450, when some of the wrinkles are ironed out? Or should I just drop the additional $30 or so and pick up a Q6600? This would be mostly used for gaming with an occasional DVD rip.

I'm upgrading from an old Northwind 2.8 so even the low end part should still be a temporary boost. I was going to get an E6750 but I thought the ~$40 more for the E8400 would be worth the extra cash.

Thanks in advance.

Most definitely. I will sell you my 4400 that does 3.2 easy.

Some things that make me wonder.

Everywhere I look 8400s are sold out. 8500 isn't here yet, why? Possible temp issues. Could all those be related? Is Intel possibly working on a new rev and not shipping anymore 8400s and working on 8500s also?

 

[Forceman]

Can I ask a quick question? I'm looking at putting together a new rig in March and was looking at the E8400. I'm now a little worried about motherboard support and some of the other issues with this chip. Would it be worth it to pick up a cheaper CPU for now, like an E4500, for now and then just getting the 8400, or the Q9450, when some of the wrinkles are ironed out? Or should I just drop the additional $30 or so and pick up a Q6600? This would be mostly used for gaming with an occasional DVD rip.

I'm upgrading from an old Northwind 2.8 so even the low end part should still be a temporary boost. I was going to get an E6750 but I thought the ~$40 more for the E8400 would be worth the extra cash.

Thanks in advance.

I would hope that by March most of the BIOS issues would be worked out, at least for the major motherboard manufacturers. At this stage I would say the $40 for the E8400 is worth it. If you are primarily a gamer you are probably better off with the dual-core over the quad, unless you plan to keep the chip for a long time.

 

[Fritzz]

Right now according to CoreTemp Core 1 is running at 47* and core 2 is running at 52*. Should there be that much of a difference between the two cores? Could I have applied my thermalpaste wrong? Or is this normal?

This is my first dual core so sorry for the newbish question.

 

[Forceman]

Right now according to CoreTemp Core 1 is running at 47* and core 2 is running at 52*. Should there be that much of a difference between the two cores? Could I have applied my thermalpaste wrong? Or is this normal?

This is my first dual core so sorry for the newbish question.

It's fine for them to show a split, seems like a lot of the E8400s do it.

 

[WhiteZero]

Right now according to CoreTemp Core 1 is running at 47* and core 2 is running at 52*. Should there be that much of a difference between the two cores? Could I have applied my thermalpaste wrong? Or is this normal?

This is my first dual core so sorry for the newbish question.

Is that under load?
Running mine at 3600MHz, I get Core1 49C and Core2: 46C under load.

 

[Sabrewulf165]

It's fine for them to show a split, seems like a lot of the E8400s do it.

Lots of C2D/Q in general do it, I had a 6-7c range on my old Q6600 between the hottest and coolest cores, with the other two falling somewhere inbetween. Core 0 and 1 were usually pretty close though.

/anecdote

 

[unclewebb]

I agree that there are lots of Core 2 Duos that have two cores running at different temperatures but when everything is properly assembled they should be very close like this across the entire operating range of the processor:

Cores operating at different temperatures is usually a sign that either the IHS or heatsink or both are not square, or the IHS is not making equal contact with both cores or you did a lousy job with the paste when installing your heatsink. With the E8x00 at idle, it might also be a sign that at least one of the on chip digital thermal sensors are getting stuck. That's not good either.

 

[Fritzz]

Does the non-beta version of speedfan not support 2 cores? I downloaded v. 4.33 and I can't get both core readouts like what you posted(unclewebb). Also speedfan shows my temps to be right around 39*C, and nvidia system monitor along with Core Temp show my temps to be 47* and 52*.

Those readings are at idle, but I am not sure which one to believe.

Think I will take of my cooler and reapply the thermal paste and make sure everything lines up. I am running a Zalman 9700 and it seemed pretty hard to mess up the alignment. Is there such a thing as too much thermal paste?

 

[Fritzz]

Well I got the beta version, so I can see both cores now. Went into my bios and looked at the temps in there. I am assuming that is going to be the most accurate way to calibrate these other programs to the correct temperature. From what I can tell so far speedfan is the closest to bios temperatures.

 

[unclewebb]

SpeedFan uses a TjMax = 100C for the E8x00 series of desktop processors. After testing with an IR thermometer I believe that is wrong and it should only be 95C. This means you need to subtract 5C from what it reports for accurate core temperatures. You can make a permanent change by going into the Configuration area and setting an offset of -5 for both cores as I previously explained.

The latest beta version of SpeedFan supports the E8x00 Wolfdale but the previous v4.33 doesn't.

 

[kmuskrat]

I'm having the temperature problems too with my new E8400 and Gigabyte GA-P35-DS3L. Ive seen a few people reporting core temperatures stuck at ridiculously high numbers like 244 idle, but mine are around 35 (after using -5 offset suggested earlier) which sounds about right.

However, running Orthos for a few minutes puts Core Temp (for both #0 and #1) up to 95. I'm used to reading the older temperatures ("CPU temp") and 95 degrees would be ridiculous for that, but I'm not sure about within the cores.

I've tried to verify that the (stock) heatsink is seated properly. Also, when opening the case and feeling the heatsink blades after this, they don't feel hot.

What are normal numbers for each individual core under load?

(Thank you Unclewebb for the excellent 3 writeups, they cleared a lot of this up for me)

UPDATE: Looks like the heatsink wasn't completely seated. It felt secured but I took out the motherboard and checked the back, and it wasn't. After reseating it, I'm getting more reasonable temperatures of ~55C within each core under load. Sorry for the noise.

 

[unclewebb]

kmuskrat: Glad to hear you got things worked out.

Coming up with some "normal" idle and load temperatures causes nothing but problems. The lack of calibration of the DTS at low idle temperatures makes it useless for any sort of comparison to others.

Load temperatures should be accurate enough for comparison. Here's a list of variables you'll need to control and include in order to compare to someone else:

1) room temperature
2) case design, size and internal air flow ( how many fans, speed, how big)
3) case open or closed
4) core voltage and MHz at full load
5) heatsink type and fan speed / size or full water cooling details.
6) what software you're loading with. If Orthos are you using small or large FFTs.

I'm sure I've forgotten to include a few more relevant details. How often when you hear about a load temperature that you are comparing your system to is all that information included and how accurate is the included information? Basically never but people still spend their life comparing their temperatures!

For those that still need a number to compare to here is my temperature. My E8400 at a room temperature of 21C when running Orthos small FFTs at 4050 MHz and 1.368 volts in a fairly quiet, closed well ventilated case with two exhaust fans using a Tuniq tower with the original fan at its minimum speed of 1188 rpm will load at about 65C based on TjMax=95C which is 30C away from TjMax for those that don't believe my TjMax=95C findings.

 

[WalkedAirplane]

For those that still need a number to compare to here is my temperature. My E8400 at a room temperature of 21C when running Orthos small FFTs at 4050 MHz and 1.368 volts in a fairly quiet, closed well ventilated case with two exhaust fans using a Tuniq tower with the original fan at its minimum speed of 1188 rpm will load at about 65C based on TjMax=95C which is 30C away from TjMax for those that don't believe my TjMax=95C findings.

Let's see here.

E8400 @ 4.0ghz on a DS3L
CNPS9500
Seeing about 60C under Orthos load, assuming TjMax is 95C.

Satisfied with that personally. Would love to see it a bit cooler, but I'm happy where that stands.

 

[tenax]

try to google the exact thread title "E8400 Temperature problems " u'll find alot of usefull links (I hope yorkfield will not have the same problem)

hate to tell you, but yorkfield does have the problem..in my bios, my 9450 quad reads as 212C. in speedtemp and core 96, reads as 126C on all cores..and the temp never varies idle or load

 

[baleedit]

I won't be upgrading any of my pc's to 45nm until I can get some accurate and consistent temperature readings. Say what you will, but this is not something that should be hard to accomplish, and I do think that Intel will be correcting this soon.

I mean, if some of the temps I've been seeing are even close to accurate, then the whole bit about 45nm running sooooo much cooler is a farce so far. I guess it's a great chip but damn, it's just a temp sensor, it's been spot-on for a while now. Why with this new generation is it so screwy?

 

[~th@fonz~]

I was wondering if these screwy temp readings are screwy at default clock speeds, as they are with oc speeds ?

 

[tenax]

i absolutely agree with what you're saying..and hopefully, mobo manufacturers can address cause it's too late for current wolfdate owners and could turn out to be quite a fiasco if it can't be fixed in the bios..in my case, it's an es9450..i may be shit out of luck, frankly..the risk i took it's not a farce in that if the core temp is lower, it's got to be cooler, but no fun if you can't be rewarded by seeing the cool temps

 

[tenax]

I was wondering if these screwy temp readings are screwy at default clock speeds, as they are with oc speeds ?

from what i've heard, there doesn't seem to be a correlation to overclocking or stock..to enabling cie and eist or not, etc.

and definitely with my 9450, doesn't make a difference if it's overclocked or not

 

[unclewebb]

I won't be upgrading any of my pc's to 45nm until I can get some accurate and consistent temperature readings. Say what you will, but this is not something that should be hard to accomplish, and I do think that Intel will be correcting this soon.

Good luck with that. Intel does not support reporting core temperatures on their desktop processors so there isn't going to be any significant changes. As the 45nm process gets more mature there should be less chips produced with stuck sensors but even stuck sensors isn't an issue to 95% of Intel's customers.

In normal operation, especially when overclocked and over volted, these new 45nm cpus get nowhere near the throttling point so having an accurate DTS sensor in the chip is of no importance to its ability to operate.

I mean, if some of the temps I've been seeing are even close to accurate, then the whole bit about 45nm running sooooo much cooler is a farce so far.

At the same MHz, 45nm does run much cooler. At 3600 MHz my old E6400 needed 1.50 volts while my new E8400 only needs 1.20 volts to run that speed. That makes a big difference in temperatures. People are making comparisons of idle temperatures which have never been accurate and the CoreTemp reported temperatures have also been inflated by 10C. This may lead people to believe that 45nm runs hotter but it doesn't.

...it's just a temp sensor, it's been spot-on for a while now.

Actually it's never been spot-on. Not one core chip has been produced that is 100% accurate through the entire temperature range of the chip using the DTS sensors.

 

[baleedit]

Thanks for all your input unclewebb, but they'll fix it eventually. Every processor I've owned has had a working temp sensor, they may not be 100%, but nothing ever is in life.

I also never said they don't run cooler, that's an obvious fact, but I just want to be able to see it. Not because it's fun to see, but when I upgrade coolers, fans, etc I want a somewhat accurate point of reference.

This reminds me of tons of hardware launches where people jump on it, and then later it's fixed, without anyone admitting it was rushed or wrong, or whatever. Take the 8800GT for example. It was almost understood that you needed an aftermarket cooler just to get one to run even close to cool. Then they updated the cooler and all is well.

I'm sure not all my data is confirmed fact, but history doesn't lie, and so far it's been pretty consistent.

 

[unclewebb]

Most of the E8x00 series of processors that have stuck sensors are only stuck at idle or lower temperatures. All of these processors are calibrated before leaving the factory at higher temperatures so I think even stuck DTS sensors should become unstuck and start operating properly if the core temperature gets up over 60C. The DTS should certainly start to work before 95C which is the throttling point for these new chips.

If even one of these sensors on a dual core is working properly then if the chip gets too hot it will be able to throttle itself starting at 95C and will be able to shut itself down about 20C after that and prevent itself from becoming a melted mess.

These stuck sensors may be a problem to extreme users but for the majority of Intel customers, sensors that are designed to control thermal throttling that happen to get stuck at lower temperatures isn't likely to be a top priority problem for Intel to fix.

These chips run rock solid stable at 4 GHz and a little beyond for 24/7 use so accurate temps from 20C to 100C isn't that important.

 

[centvalny]

As long as HSF installed properly then you don't have to worry....

Heres a few different readings from asus prob,everest and core temp.... cpu @ load with stock HSF