Overclocked my Retail E8400...WOW!

[SteveLord]

Nice work so far guys.

[Power5]

Funny how a very poorly applied IHS is not that big of a deal, but anything wrong with an AMD is HUGE and cant be tolerated, or is expected from such a terrible company. I would say that core temperature is a huge thing when it comes to chip longevity and an errata bug is much less of a reliability issue. I know that my 65nm e4300 is almost always dead even on cores, and at most 2c different. Never seen anything more than that except during unequal stressing of the cores. Wont keep me from getting an 8400, but I may just wait to find out what the issue is with them. Also, starting to really like the AMD price drops. 5000+ BE for only $99. Wont do anywhere near as good as an 8400 ofcourse.

[aznsniper911]

Yep, its about 1 hour for me, so i am not going to drive there.

If your in San Francisco, Central Computers on 4th and Howard has it for $220. Man is it me or does it seem these things are doing 4ghz without any issue? I guess this is like the Conore in which 3ghz was obtainable very easily.

[Forsaken1]

4Ghz(450x9) appears to be the 24/7 setting for my E8400.To many volts after that then i would like.

[unclewebb]

CoreTemp reads the sensors built directly into the Intel CPU so an updated bios won't change your core temp readings any. All Intel Core based processors store core temperature data in the same register.

Lots of good info in this old thread concerning Intel temps and monitoring software. Check out post #210 to learn how to test temperature monitoring software.

When two cores at idle are reporting significantly different temperatures then either you did a lousy job applying the paste or attaching your heat sink or more likely, the IHS is not making equal contact with both cores. Unfortunately Intel doesn't have to replace these kind of chips because they run fine when not overvolted and overclocked.

I did a lot of testing with my E6400 Conroe and it also could report core temps a couple of degrees below ambient which is impossible with an air cooled processor. My conclusion was that the temp sensors in the Intel chips are not 100% accurate at very low idle temps. These on chip sensors are designed to signal processor over heating and shut down and for that purpose they are extremely accurate.

Looks like you guys are having fun with these new E8400 chips. Might have to go pick one up myself tomorrow and upgrade my E6400.

[jcniest5]

Funny how a very poorly applied IHS is not that big of a deal, but anything wrong with an AMD is HUGE and cant be tolerated, or is expected from such a terrible company. I would say that core temperature is a huge thing when it comes to chip longevity and an errata bug is much less of a reliability issue. I know that my 65nm e4300 is almost always dead even on cores, and at most 2c different. Never seen anything more than that except during unequal stressing of the cores.

That is exactly my issue right now, trying to resolve directly with Intel. So far they are playing it off, saying don't look at the core temp, just the CPU temp and I should be happy. Yeah, right, like I'm a child and don't know anything.

When two cores at idle are reporting significantly different temperatures then either you did a lousy job applying the paste or attaching your heat sink or more likely, the IHS is not making equal contact with both cores. Unfortunately Intel doesn't have to replace these kind of chips because they run fine when not overvolted and overclocked.

The latter is exactly what I think the culprit is. Very unlikely the first since I did like 5 times of reapplying thermal paste and reseating the heatsink. Same exact result each time, tells me it's poor contact between one of the cores with the IHS.

[C0ncrete]

The 8400 I had had about 6 deg differance in core. Was not happy.

[Xvys]

I will have to try reseating my heatsink as one core is still reading 17C at idle, the other 34C, with room temp. 23C. In spite of this I have fine tuned the settings at 4.0Ghz and it seems stable so far. I will Prime it overnight to confirm longterm stability. The ram is at default settings.

Cpu: 1.3875v
Cpu PLL: 1.6v
FSB term: 1.3v
Ram: 2.1v
N.B: 1.4v
Transaction Booster: auto
Overcharge Voltage: auto
Load line Calibration: auto
All other settings: auto
---------------------------------------------
Asus P5K Deluxe/E8400 / 2G Crucial 10th Anniv. DDR2-5300 / ATI X1950Pro / Ultra120 Extreme/ CM690 case / FSP FX600-E

[jcniest5]

The 8400 I had had about 6 deg differance in core. Was not happy.

I'm in the same boat, buddy. Hope something good can come out of this.

[Nosfer@tu]

What are the difference in 3dmark06 when comparing 6 vs 12 Mb cach?

[Xvys]

In my case, if it was a misaligned heatsink, why would one core temp not increase under load, while the other core increases 17C? Once the lower core reaches the other's 34C, then they both go up together in unison?? Seems more likely the sensor imbedded in the core is faulty or the mb's aren't reading it correctly.

Another thought is the Penryn processor inside the cpu is also amazingly small, not even filling up half the case. Perhaps with a humped or concave heatsink surface it might not be making good contact with much the actual processor. This might account for a 6 degree discrepancy, but in my case of 17C difference and a stuck core temp, there must be other issues.

[Power5]

Yeah, I may wait till this gets sorted out. I have to build a computer for my wife, so maybe I will just distract myself with that project so I don't go out and get an 8400 too early.

[vcp-cur]

i wonder if these chips E8400's will do just as good, atleaaaaast 4ghz on air on a gigabyte p35 ds3l

[flopper]

When two cores at idle are reporting significantly different temperatures then either you did a lousy job applying the paste or attaching your heat sink or more likely, the IHS is not making equal contact with both cores. Unfortunately Intel doesn't have to replace these kind of chips because they run fine when not overvolted and overclocked.

Not true.

I had similiar temps with my core 2 duo, flashed bios, now the temps are of on 2 core by 10c.
Most likely a bios issue.

[unclewebb]

Not true.

What software are you using? I'm strictly talking about the CoreTemp program which reads directly from the Intel chip and should completely bypass any bios settings. In theory at least!

[flopper]

What software are you using? I'm strictly talking about the CoreTemp program which reads directly from the Intel chip and should completely bypass any bios settings. In theory at least!

everest.

after u asked I checked with core temp also,
that program shows the same with 2c difference.

seems the biosflash screwed up the temps using everest.
core temp works as what I can see.

[unclewebb]

core temp works as what I can see.

That needs repeating. I did some extensive testing of temperature monitoring programs and CoreTemp was one of the very few that is correctly reading the Intel on chip DTS sensors exactly, in real time.

Xvys: If CoreTemp consistently reports one core at 17C and the other stuck at 34C at idle then that problem is likely with one of the on chip sensors. You can always use this method to read the individual sensors directly.

I did some testing with Orthos where I forced it to run on only one core at a time. The idle core immediately increased in temperature to within 6C of the other core that was at full load. If both cores are idle or both cores are at full load, the individual core temps should be within a degree or two of each other.

[Ace-a-Rue]

In my case, if it was a misaligned heatsink, why would one core temp not increase under load, while the other core increases 17C? Once the lower core reaches the other's 34C, then they both go up together in unison?? Seems more likely the sensor imbedded in the core is faulty or the mb's aren't reading it correctly.

Another thought is the Penryn processor inside the cpu is also amazingly small, not even filling up half the case. Perhaps with a humped or concave heatsink surface it might not be making good contact with much the actual processor. This might account for a 6 degree discrepancy, but in my case of 17C difference and a stuck core temp, there must be other issues.

it's not a misaligned heatsink...intel has QC problems with temp sensing from the cores...i had a 9650 where the 1st core (core #0) was locked on to 51C even at stock speed in idle condition...the core would go up when overclocking and loading all 4 cores.

read this short article which explains about the latest rumor: http://www.guru3d.com/newsitem.php?id=6317

this kind of ticks me off...i sent back a good overclocker to the retailer and got a POS in return for overclocking.

[flopper]

I guess these temps sensors are small and picky to get in there by the small transistors without disturbing the electron flow

I used core temp in the past, just that everest and other programs showed the same, flashed the bios the other day and forgot to check with core temp.

[Power5]

How do we know for sure? Its Intel's claim that its just a sensor issue. It could be a quality issue on the IHS but that sounds worse than a faulty temp sensor, which really does nothing except shut down a chip when it reaches its thermal threshold. Need someone to pull the IHS off their chip and see if there is any problems with contact.

[unclewebb]

I just ran a test on CPUID Hardware Monitor and it too is not tracking the Intel on chip digital thermal sensors (DTS) exactly in real time. It's close but not perfect.

The only explanation for a 17C difference at idle as reported by CoreTemp is a bad on chip sensor. It would be tough to hit over 4 GHz reliably and have the temps equal on both cores above 34C if there was a serious problem with the IHS.

[Xvys]

That's about what I'm thinking unclewebb, if it was a IHS problem it would of shown up when under load @ 4400Mhz, but both cores were showing 50C in CoreTemp. It is only under idle where the discrepancies are.

I tried the CrystalCPUID monitor as you suggested. At idle, I get a reading of 8854 on Core1 and 8847 on Core2...but under 100% load both Cores show 8840, with CoreTemp showing 45C in both Cores.

I suspect the both sensors are faulty. One reads 7C under room temp at idle, the other is stuck at 34C until the load increases the temp above that reading. Other than that she works perfectly!

[yenclas]

Hello I'm Spanish.

Sorry by my bad English.

Today arrived my new 8400 and set to this clock stable.



My problem is temp

Idle: 43-44 with my Thermalright Ultra 120

It's normal ?

[unclewebb]

I tried the CrystalCPUID monitor as you suggested. At idle, I get a reading of 8854 on Core1 and 8847 on Core2...but under 100% load both Cores show 8840, with CoreTemp showing 45C in both Cores.

Here's some quick background info. You can ignore the first two digits of the number returned by CrystalCPUID. The second two digits contain the temperature info directly from the sensors within the Intel CPU. You can set CrystalCPUID to read the sensor from either core or any of the 4 sensors on a quad core processor. The 40 in your reading of 8840 is a hexadecimal value so that translates to 4X16 + 0 or 64 degrees. That number is an offset to the maximum temperature that Intel rates your processor to be able to operate at safely and reliably. As your processor heats up the offset decreases towards zero. When it hits zero your processor will be throttling back the MHz to try to get control of itself. I took my E6400 about 12C higher and for reference, the dual core mobile chips are documented by Intel not to shut down until 20C after TjMax which for them is at about 120C.

If you go into the CoreTemp settings you can check off the "Show Delta to Tjunction max temp" option and it should exactly equal the numbers you are getting from CrystalCPUID.

CoreTemp then uses a formula like this to convert that offset to an absolute temperature:

Absolute Temp = TjMax - DTS

If TjMax is 105C and the direct reading of the digital thermal sensors say you are 64C away from that maximum then your absolute temperature is 41C. If CoreTemp is displaying 45C at this exact moment then it must be using a TjMax value equal to 109C. ( 64C + 45C )

You might want to go back and take CoreTemp and CrystalCPUID readings at the exact same time and compare the numbers.

With all previous desktop core processors, TjMax was never publicly documented by Intel. CoreTemp used to guess at that value since there was no documented way to read that value from a processor. I'm not sure if this new 105C TjMax number for the E8400 series is publicly documented by Intel or just another guess so the reported temps look believable.

One reads 7C under room temp at idle...

I noticed this problem with my original Conroe core processor but I was only a couple of degrees below ambient then. It's a tough crowd at XS and I wasn't ever able to convince anyone about this problem.

People need to keep in mind that this sensor that CoreTemp is using to calculate an absolute temperature with is not documented by Intel to be used for that purpose. In testing on a bare naked core with an IR thermometer I found that this sensor was very accurate for reporting core temps from about 45C to 85C and beyond but it wasn't accurate at lower temperatures. Reported temperatures were too low at lower values. This new series of processors consume even less watts at idle which seems to only magnify this issue.

Bottom Line: If you are running at 4400 MHz and you have piles of head room before throttling then who cares what temperature your chip is at!

[Ace-a-Rue]

How do we know for sure? Its Intel's claim that its just a sensor issue. It could be a quality issue on the IHS but that sounds worse than a faulty temp sensor, which really does nothing except shut down a chip when it reaches its thermal threshold. Need someone to pull the IHS off their chip and see if there is any problems with contact.

pulling the IHS is almost a sure way to kill the processor.