help with temps on E8600 (inconsistent)

Hi, im a bit of a noobie overclocker that would welcome some help.

Basically i've so far overclocked my E8600 to 4.2ghz stable but am unsure about going further because i am getting different temp readings with different programs and im not sure which are correct.

When OCCT has been running for a while @ 4.3ghz i get the following temps from each of these programs:

Realtemp: 60c
Everest: 64c
Coretemp: 70c
OCCT: 70c




if OCCT and coretemp are correct then i dont really want to take this further, but im not sure they are. Someone told me that realtemp is more accurate in which case I can keep going with the o/c but i wanna be sure which is the correct reading before i carry on.

I think its something to do with the TJmax that is causing these inconsistencies but as i said im a noob so not sure what that is.

other info:

VID = 1.25
vcore = 1.38 (for 4.3ghz), loadline calibration enabled, all other voltage settings on auto
im using a thermalright ultra 120 extreme with a sharkoon 2000rpm fan, asus maximus formula (bios rev. 1207).

would be very grateful if anyone can point out which is correct/incorrect and what I am doing wrong, thanks

If you notice in your background your FSB VTT = 1.39. This is notorious for the Formula series boards, which means you need to go into BIOS and set the FSB Term voltage manually. I'd start with 1.28 but not much more than that on FSB Term, this can cause a really quick death with 45nm chips. 60 c loaded is pretty consistent with mine at 4.264ghz.

The latest versions of coretemp/realtemp should report it accurately.

The problem in the calculation of temperatures is that what programs can measure is the distance between the max tjunction temps and the current temps. Thus, the distance will always be the same between measurement software, while the absolute value will be different because of assumed max tjunction temps on older software.

Intel has recently confirmed a 100C tjunction max temp for the E8xxx, and that has been reflected in the latest versions of both realtemp and coretemp. Can't say about the accuracy of the others.

mdamda said:

The latest versions of coretemp/realtemp should report it accurately.

The problem in the calculation of temperatures is that what programs can measure is the distance between the max tjunction temps and the current temps. Thus, the distance will always be the same between measurement software, while the absolute value will be different because of assumed max tjunction temps on older software.

Intel has recently confirmed a 100C tjunction max temp for the E8xxx, and that has been reflected in the latest versions of both realtemp and coretemp. Can't say about the accuracy of the others.

Click to expand...

so the TJmax value will differ from cpu to cpu? and therefore you have to take that into accout when u check temps? ]

ok so ive noticed i can change the tjmax in the options in coretemp, i think its @ 105 in that pic so if i put it at 100 then the correct temp would be 65c ?

is there a way to change the tjmax in realtemp? and should i just ignore what OCCT says? cheers

prime81 said:

If you notice in your background your FSB VTT = 1.39. This is notorious for the Formula series boards, which means you need to go into BIOS and set the FSB Term voltage manually. I'd start with 1.28 but not much more than that on FSB Term, this can cause a really quick death with 45nm chips. 60 c loaded is pretty consistent with mine at 4.264ghz.

Click to expand...

ok thx i will lower that, am i ok to have the loadline calibration enabled? i read somewhere that its not good for 45nm cpus, but when I disable it i find it impossible to even bootup at 4.2ghz even with 1.4v vcore :S

Yes, you are correct with changing it to 100C. --> confirmed by intel to be the correct value.

The TJMax value will be the same for CPUs of the same type (ex. all E8600s) but what I meant is that tjmax is assumed differently per program.

In both realtemp and coretemp, there's an offset function to do so. Downloading the most recent versions will correct this automatically. I'd say OCCT really isn't meant to be accurate with regards to temp.

mdamda said:

Yes, you are correct with changing it to 100C. --> confirmed by intel to be the correct value.

The TJMax value will be the same for CPUs of the same type (ex. all E8600s) but what I meant is that tjmax is assumed differently per program.

In both realtemp and coretemp, there's an offset function to do so. Downloading the most recent versions will correct this automatically. I'd say OCCT really isn't meant to be accurate with regards to temp.

Click to expand...

ok I've set the tjmax to 100c in realtemP and coretemp and now they all agree that the max temp under load is 65c. ill go by that then, shame cause I was hoping for 4.5ghz but don't wanna let it go past 70c :/. Thx for help.

Ps. I downloaded all the programs today so pretty sure they're the most recent . ( coretemp v.0.99.2 and realtemp v.2.70 )

elliotuk said:

ok thx i will lower that, am i ok to have the loadline calibration enabled? i read somewhere that its not good for 45nm cpus, but when I disable it i find it impossible to even bootup at 4.2ghz even with 1.4v vcore :S

Click to expand...

Loadline is perfectly safe so long as u disable speedstepping/C1E. The problem comes when the cpu throttles down or up and that sustained loadline voltage jumps to enormous levels for short periods of time. Anandtech posted a detailed review of this feature.

elliotuk said:

Ps. I downloaded all the programs today so pretty sure they're the most recent . ( coretemp v.0.99.2 and realtemp v.2.70 )

Click to expand...

Well you must have downloaded an older version.

Here's CoreTemp 0.99.3

Here's RealTemp 2.79.8

prime81 said:

Loadline is perfectly safe so long as u disable speedstepping/C1E. The problem comes when the cpu throttles down or up and that sustained loadline voltage jumps to enormous levels for short periods of time. Anandtech posted a detailed review of this feature.

Click to expand...

On Asus boards C1E doesn't work when you OC (or maybe its just if you have LLC enabled). At any rate you don't need to worry about disabling C1E b/c the voltage won't drop when speedstep kicks in. You can leave speedstep enabled, and the multi will drop, but the voltage remains constant despite the C1E setting.

ok ive got it running 12 hours orthos stable @ 4.3ghz now with 1.38v vcore, cant seem to get it much further unless i pump silly amounts of volts into it so will prob leave it there.

Im wondering now though whether too much voltage can make an overclock unstable? i have the cpu pll and NB set to auto and its automatically setting both those voltages to 1.65v, is that too much?

Also the ram is Ocz platinum PC2-8500 but is now only running @ 860mhz, i cant seem to get it higher. Have tried different FSB:RAM ratios but it refuses to boot at any other ratio other than 1:1. Have also tried lowering the multiplier to 9x and increasing the fsb to compensate but that is unstable too :S Should I need to adjust any voltages if I change the multi and up the fsb, even though the resultant speed is still the same? I've also tried changing the timings to 4-4-4-12 @ 860mhz but that is unstable too. Arghhhhhhhh what am i doing wrong :/

Take PLL and NB off of auto. You shouldn't need that much on them...especially the NB. 1.6 on PLL might help, but you should be able to run the NB at min voltage or just 1 tick up.

Fix these voltages, and then re-try your RAM. What voltage do you have your RAM set to?

I also had problems with RAM being unstable at anything other than 1:1, but fixed it by setting these options:

Load Line Calibration: Disabled
CPU Spread Spectrum: Disabled
PCIE Spread Spectrum: Disabled
both Clock Skew options: Normal

Setting any of those to Auto or Enabled would make my system unstable. But I'm on a P5Q-Pro board so your results might vary. As for LLC being off making it not boot even at 1.4v, it's because your vdrop and vdroop are increased when LLC is off, so you need to try higher than 1.4v, go up one notch at a time until it's stable. Your board seems to have more vdrop and vdroop than mine so maybe you'd have to try like 1.42v in BIOS for it to show 1.38v in CPU-Z... don't worry about BIOS voltage being high, because your real voltage in OS is lower.

jason4207 said:

Take PLL and NB off of auto. You shouldn't need that much on them...especially the NB. 1.6 on PLL might help, but you should be able to run the NB at min voltage or just 1 tick up.

Fix these voltages, and then re-try your RAM. What voltage do you have your RAM set to?

Click to expand...

the lowest I can set the CPU PLL in bios is 1.5 and that comes out as an actual value of 1.6 when I check in windows, but at least it's a bit better, have lowered the NB to 1.5. Ram voltage is set to 2.0v for 860mhz cl5. The 5:6 ratio seems to work ok so now mem is at 1032mhz cl5 with 2.16v wwhich is better. I read somewhere that these ram sticks can do 900mhz cl4 with just 1.9v but I can't get it working with even 2.1v at those timings :/

don'tknow said:

I also had problems with RAM being unstable at anything other than 1:1, but fixed it by setting these options:

Load Line Calibration: Disabled
CPU Spread Spectrum: Disabled
PCIE Spread Spectrum: Disabled
both Clock Skew options: Normal

Setting any of those to Auto or Enabled would make my system unstable. But I'm on a P5Q-Pro board so your results might vary. As for LLC being off making it not boot even at 1.4v, it's because your vdrop and vdroop are increased when LLC is off, so you need to try higher than 1.4v, go up one notch at a time until it's stable. Your board seems to have more vdrop and vdroop than mine so maybe you'd have to try like 1.42v in BIOS for it to show 1.38v in CPU-Z... don't worry about BIOS voltage being high, because your real voltage in OS is lower.

Click to expand...

I've got all the spread spectrums disabled, I tried turnng LLC off but it made no difference it won't post at the ratio I really want :/

Not sure what u mean by clock skew options.

EDIT: ahha I just disabled clock twister and now it posts @ ratio 4:5 now to stress it, it was set to "light", didn't think it would cause any probe but ye cheers! Should I have DRAM static read control
On or off?

elliotuk said:

I've got all the spread spectrums disabled, I tried turnng LLC off but it made no difference it won't post at the ratio I really want :/

Not sure what u mean by clock skew options.

EDIT: ahha I just disabled clock twister and now it posts @ ratio 4:5 now to stress it, it was set to "light", didn't think it would cause any probe but ye cheers! Should I have DRAM static read control
On or off?

Click to expand...

Static read control doesn't seem to make a difference for me; it's on default and I'm running 1066 RAM @ 1186 in 4:5 ratio.

I would turn off all the boosters, twisters, static read control for now. Focus on the CPU. Once that's 100% stable, you can try to tweak out those other things for better memory bandwidth.

The RAM I think should be running at 5-5-5-15 2T at anything over 800Mhz, as that is what mine does...

ive tried everything now to get it to run with 4:5 ratio, that would be 4.3ghz cpu and 1077mhz cl5 on the ram but everytime it fails after 8 hours on orthos. So wierd, ive tried upping the voltage on everything a little bit and it will run for 8 hours straight before failing :S

Anyway I got it working with a 5:6 ratio, so now my cpu is @ 4.3ghz and the ram is @ 1033mhz cl5. It was still going after 12 hours orthos so think im just gonna stick with that its good enough

Seems to want a lot of voltage though, requires 1.385v on cpu for 4.3ghz. Was wondering if you guys think these temps sound right?
Idle @ 40c
Highest recorded temp after 12 hours orthos = 70c
used realtemp and coretemp to record temps with tjmax @ 100c

thats with a thermalright ultra 120 extreme and a sharkoon 2000rpm fan.
Seems a bit hot to me :S

Badbonji said:

The RAM I think should be running at 5-5-5-15 2T at anything over 800Mhz, as that is what mine does...

Click to expand...

this review here has got me thinking otherwise though:

http://www.nordichardware.com/Reviews/?page=7&skrivelse=503

looks like there they can easily get 4-4-4-12 out of them at way past 800mhz, correct me if im wrong though