CoreTemp vs. RealTemp - Page 5 - Overclock.net

GigaByte · 03-09-08

Quote:

Originally Posted by JeremyFr

All I know is using a TJ max of 100 on my system I get Idle temps of 38 and load temps around 55 or so.

Now thats on an L2 E4400 at 3.0Ghz (300x10) 1.325V, with a Zalman 7700CPS. Chip is lapped as well. Also Ambient in my office is around 75F at all times (due to 3 computers and a laser printer in a 10x14 room) Now these temp's would seem to be about right for my setup and cooling, If I was to assume an 85 TJmax for my chip that would put my idles at 23 and my load at around 40, which would be unheard with a 50% overclock if you ask me.

L2 are 100% Tj max 100c, so are M0 but thats where the battle begins.

**Retrospekt** · 03-09-08

Quote:

Originally Posted by iewgnem

Does it really matter which one is correct? Both reads the correct, or close to correct "temp to tjmax", so really which ever TJmax you use, you will still overheat at a certain point regardless if it reads 50C or 60C. Not to mention with these 45nm chips temps are no longer an issue anyway.

Yeah, it matters a lot. I am getting a 15c temperature difference. One of the temperatures is limiting my overclock, the other would let me go MUCH higher. How does that not matter?

Gollie · 03-09-08

I knew that Anandtech article was going to bring this debate up. The truth is that until Intel release some real TjMax numbers we won't know what our exact temps are. As long as i'm not grossly underestimating my temps and frying my CPU as we speak then i'm fine.

What is both good and bad for q6600 owners is that our temps are pretty consistent across all the programs (except speedfan). I sure would love to crank this baby a few Mhz higher but i'm more than happy with 3.6Ghz.

iewgnem · 03-09-08

Quote:

Originally Posted by JeremyFr

What are you talking about ?!?!?! I mean granted Voltage will be a bigger issue due to the smaller process but temp is always gonna be an issue with any processor.

You do know 45nm chips risks dying at above 1.4V and at 1.4V unless you are running passive even a Zalman can keep the temps well below TJmax.

GigaByte · 03-09-08

Quote:

Originally Posted by Retrospekt

Yeah, it matters a lot. I am getting a 15c temperature difference. One of the temperatures is limiting my overclock, the other would let me go MUCH higher. How does that not matter?

When overclocking the temp you should really look at is Tcase. Go to processorfinder.intel.com and look up your CPU and stepping, where it lists thermal spec click on it.

opt33 · 03-09-08

Read last several posts regarding reason at least for E8400 to have tjmax of 95 instead of coretemps 105, I am having same conversation on another thread.
http://www.overclock.net/intel-cpus/...utronix-6.html

But bottom line is temp reading of coretemp, realtemp or any other program is:
temp reading (guess based on tjmax guess) = tjmax (guess) - DTS (accurate relative temperature to throttling read directly from cpu)

For E8400,
Coretemp assumes tjmax of 105 because mobile cpu 45nm is tjmax 105, though intel has said that is wrong and said why. But coretemp uses it anyways, as a best guess.

Realtemp assumes tjmax of 95, because at idle steady state with heatsink removed, accurate temp reading at Tcase reveals a temp of ~95C, when DTS=0. The gradient from core to IHS is less than 1C at idle, thus realtemp is a more accurate guess. Others at xtremesystems have same measurement.

It is likely that all E8200, E8400, E8500 currently out have similar tjmaxs. But in time, as production processes improve, it would not be surprising if tjmax changes.

So for older cpu's that have undergone silent revisions from batch to batch, unless you measure Tcase and compare to DTS value at undervolted, underclocked steady idle....well, you are guessing.

However if you look at distance to Tjmax using realtemp, coretemp, or whatever program you want, that number is always exact, accurate, and read directly from cpu, and of course will be same program to program. So if you want to compare to each other and get exact values, you can express temp in distance from throttling, and just keep that temp ~ 25-35 or so from throttling

GigaByte · 03-09-08

Quote:

Originally Posted by opt33

Read last several posts regarding reason at least for E8400 to have tjmax of 95 instead of coretemps 105, I am having same conversation on another thread.
http://www.overclock.net/intel-cpus/...utronix-6.html

But bottom line is temp reading of coretemp, realtemp or any other program is:
temp reading (guess based on tjmax guess) = tjmax (guess) - DTS (accurate relative temperature to throttling read directly from cpu)

For E8400,
Coretemp assumes tjmax of 105 because mobile cpu 45nm is tjmax 105, though intel has said that is wrong and said why. But coretemp uses it anyways, as a best guess.

Realtemp assumes tjmax of 95, because at idle steady state with heatsink removed, accurate temp reading at Tcase reveals a temp of ~95C, when DTS=0. The gradient from core to IHS is less than 1C at idle, thus realtemp is a more accurate guess. Others at xtremesystems have same measurement.

It is likely that all E8200, E8400, E8500 currently out have similar tjmaxs. But in time, as production processes improve, it would not be surprising if tjmax changes.

So for older cpu's that have undergone silent revisions from batch to batch, unless you measure to case and compare to DTS value at undervolted, underclocked steady idle....well, you are guessing.

So Intel gave some word on this battle? Mind posting some

please.

opt33 · 03-09-08

Quote:

Originally Posted by GigaByte

So Intel gave some word on this battle? Mind posting some

please.

Already posted in other thread which is linked to. But what specifically are you looking for, and I will post here.

**Retrospekt** · 03-09-08

I'm confused as ever. Is all of this Intels fault for not releasing accurate tjmax?

GigaByte · 03-09-08

Quote:

Originally Posted by opt33

For E8400,
Coretemp assumes tjmax of 105 because mobile cpu 45nm is tjmax 105, though intel has said that is wrong and said why. But coretemp uses it anyways, as a best guess.

Realtemp assumes tjmax of 95, because at idle steady state with heatsink removed, accurate temp reading at Tcase reveals a temp of ~95C, when DTS=0. The gradient from core to IHS is less than 1C at idle, thus realtemp is a more accurate guess. Others at xtremesystems have same measurement.

That, you say CoreTemp's assumption of 105c Tj max is incorrect for E8400 and is 95c instead. Right up to this point I still believe Tj max is 100c for M0 E4xxx/E2xxx but CoreTemp went back to 85c while RealTemp shows 95c which no program has shown me yet. With what you said about E8400 could be very close or the same for M0 steppings.

		Overclock.net - Overclocking.net > Intel > Intel CPUs
CoreTemp vs. RealTemp

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)