Real Temp - New temp program for Intel Core processors

Printable View

Show 100 post(s) from this thread on one page

03-10-2008, 08:20 PM
Retrospekt

Thanks for the test Kain. I noticed one thing. You put the thermal sensor on the side of the IHS. While I know it wouldn't be possible to put it in the center, would it throw temperatures off so that they would be read wrong?

In almost every trial you reported that RealTemp reported 5-7c above what the temp probe was showing. Wouldn't the temperature be 5-7c higher in the middle of the ihs compared to the side of it? Just a thought.
03-10-2008, 08:21 PM
STEvil

Probe on the side of the IHS will be slightly inaccurate at low heat loads.

For proper testing a channel milled above each core into the IHS will give the best results. It will produce worse temperatures but they will at least be more accurate.
03-10-2008, 08:34 PM
Kain XS

Yes it should be.
Temps could be off load around 5 degrees and so Real Temp could be ok.
Anyway, temps reported by MBM are matched on Load.
Also, notice calibration on Real Temp always need to be -- because if not, temps are a lot different.
I might think on a way to run the probe on the cneter of the IHS, but that would be maybe with stock heatsink if I can dremel a path on the Heatsink base so I can put the Probe without interfering (completely) with the base and IHS contact.

Agian, notice when voltage cransks up, the Temps are to way high in mi opinion. The difference there can be more than 10 degrees.
Or maybe, taking off the IHS... Sorry I cant to that guys :(

But the channel on the heatsink base its a must I will try to do next week.

Kain
03-10-2008, 08:36 PM
Kain XS

Quote:

Originally Posted by STEvil

Probe on the side of the IHS will be slightly inaccurate at low heat loads.

For proper testing a channel milled above each core into the IHS will give the best results. It will produce worse temperatures but they will at least be more accurate.

Worst temps? actually, more accurate temps :) Worst because we dont want to see how high the CPU can go?

I will try to mill a channel over the stock heatsink so I can do better tests ;)
03-10-2008, 08:37 PM
Retrospekt

I have an e4300. I put it at stock and took some temperature readings. I have a Tuniq tower with great airflow and the fan on high.

Idle:
http://aycu14.webshots.com/image/472...2248700_th.jpg

My ambient is 23-24 degrees celsius. RealTemp reports my temperature as below ambient.

Load:
http://aycu05.webshots.com/image/485...5722176_th.jpg

CoreTemp says I load around 50c, which is completely ludicrous for a Tuniq Tower and e4300 at stock. RealTemp readings look good, but the idle reading is weird.

Any comments about this?
03-10-2008, 09:19 PM
Demo

Quote:

Originally Posted by Retrospekt

I have an e4300. I put it at stock and took some temperature readings. I have a Tuniq tower with great airflow and the fan on high.

Idle:
http://aycu14.webshots.com/image/472...2248700_th.jpg

My ambient is 23-24 degrees celsius. RealTemp reports my temperature as below ambient.

Load:
http://aycu05.webshots.com/image/485...5722176_th.jpg

CoreTemp says I load around 50c, which is completely ludicrous for a Tuniq Tower and e4300 at stock. RealTemp readings look good, but the idle reading is weird.

Any comments about this?

Calibrate your idle temp readings as posted on #1
03-10-2008, 09:35 PM
Retrospekt

Quote:

Originally Posted by Demo

Calibrate your idle temp readings as posted on #1

Ok so I put it at ++ and idle temperatures seem more reasonable. I am getting 33 and 36c load which is what I should be at...I think. I guess I'll trust RealTemp.
03-10-2008, 11:38 PM
Demo

Did a little experiment today;

Bios temps detected 40~48c ( randomly regardless of the weather/ambient temps )

Coretemp is almost always 9~11c lower than bios temps.
Realtemp is always 5c lower than coretemp reading.

Placed 70c on bios thermal shutdown, increased voltage from 1.335 to 1.515 ( to ensure that my processor hits 70c )

Coretemp at 69c = system shutdown ( take note the 9~11c offset from the bios temps )
Realtemp = 64c ( last seen before shutdown )

This is not to prove that realtemp is inaccurate but does this mean that the bios reading temps is similar to CT/RT when it comes to reading out idle temp ? ( needs calibration on idle but spot on when loaded )
03-11-2008, 02:11 AM
~aoe~

Quote:

Originally Posted by Kain XS

Worst temps? actually, more accurate temps :) Worst because we dont want to see how high the CPU can go?

I will try to mill a channel over the stock heatsink so I can do better tests ;)

I think what STEvil means is that, by milling a channel in the IHS, the contact area between CPU and heatsink/waterblock will be less and therefore cooling performance will be reduced (resulting in higher temps).
03-11-2008, 04:47 AM
rge

Quote:

Originally Posted by Kain XS

Yes it should be.
Temps could be off load around 5 degrees and so Real Temp could be ok.
Anyway, temps reported by MBM are matched on Load.
Also, notice calibration on Real Temp always need to be -- because if not, temps are a lot different.
I might think on a way to run the probe on the cneter of the IHS, but that would be maybe with stock heatsink if I can dremel a path on the Heatsink base so I can put the Probe without interfering (completely) with the base and IHS contact.

Agian, notice when voltage cransks up, the Temps are to way high in mi opinion. The difference there can be more than 10 degrees.
Or maybe, taking off the IHS... Sorry I cant to that guys :(

But the channel on the heatsink base its a must I will try to do next week.

Kain

If you do cut the groove, could you test with heatsink on and off. With heatsink off, use idle and slowly bring temp up, allowing to equilibrate and compare measured temp to distance from tjmax at 45, 40, 35, 30.

Once heatsink in place and actively cooling Tcase, you will be measuring Tcase accurately, but a more significant gradient will exist from core to tcase, while cooling tcase and loading core.

But would love to see it tested both ways for comparison, if you do it.
03-11-2008, 06:00 AM
KTE

I did a lot of testing and had a lengthy reply written back addressing many points. Sadly yesterday downloading UT3Demo legitimate from a legitimate file site at the same time, it had a relatively new virus and has messed my system up quite badly upon extracting. Now fixed it but lost all my desktop files including my pictures and notes which I had uploaded and posted in my reply. I'm not happy. :mad:

I'm not going in any detail now but just giving my results in a while when I can.

Quote:

Originally Posted by rge

If you do cut the groove, could you test with heatsink on and off. With heatsink off, use idle and slowly bring temp up, allowing to equilibrate and compare measured temp to distance from tjmax at 45, 40, 35, 30.

You can't slowly bring temp. up without a heatsink. It's not possible unless you have the environment temperature controlled very well at sub <0.7V <1.2GHz or if you're applying static loads to reach such temperature. Without a heatsink, temperature shoots up very fast above 10W TDP on an IC. Especially when you reach >80C, the temperature runs away. Hence any sub-$100 probe I've seen cannot even read, calculate and feedback the real-time IHS temp. before the system shuts down (THERMTRIP#). The delay is such that, they will end up reading temps. approximately easily 5-10C short of the real Tcase temp., depending on how quick the temperature shoots up (very quick).

Quote:

Once heatsink in place and actively cooling Tcase, you will be measuring Tcase accurately, but a more significant gradient will exist from core to tcase, while cooling tcase and loading core.

Yep.
03-11-2008, 06:02 AM
jas420221

Quote:

Originally Posted by Demo

Did a little experiment today;

Bios temps detected 40~48c ( randomly regardless of the weather/ambient temps )

Coretemp is almost always 9~11c lower than bios temps.
Realtemp is always 5c lower than coretemp reading.

Placed 70c on bios thermal shutdown, increased voltage from 1.335 to 1.515 ( to ensure that my processor hits 70c )

Coretemp at 69c = system shutdown ( take note the 9~11c offset from the bios temps )
Realtemp = 64c ( last seen before shutdown )

This is not to prove that realtemp is inaccurate but does this mean that the bios reading temps is similar to CT/RT when it comes to reading out idle temp ? ( needs calibration on idle but spot on when loaded )

Umm, the BIOS is not idle, like in windows. BIOS will always read a few C higher.
03-11-2008, 06:14 AM
KTE

BIOS is 100% load on core0, always. System Idle Timer cannot be executed in BIOS.
03-11-2008, 06:16 AM
jason4207

Quote:

Originally Posted by Retrospekt

I was still unsure about why RealTemp and CoreTemp are reading the TJmaxx differently (also causing a 15c change in temperatures), so I posted a thread at OCN for people to voice their opinions and talk about it. Here's the link, it's a pretty good read:
http://www.overclock.net/intel-cpus/...-realtemp.html
(If I'm not allowed to link to outside forums, please inform me so I can take it down.)

And another member just did a test to see where the cpu would actually throttle, and which program is correct. For now this test only applies to the e2xxx and e4xxx series that are the M0 stepping.
http://www.overclock.net/intel-cpus/...tjmax-dts.html

There is no program that 'reads' Tjmax, b/c there is no Tjmax to be read. It is all an assumption made by the programmer. That's why different programs have different values for Tjmax.
03-11-2008, 06:53 AM
Retrospekt

Quote:

Originally Posted by jason4207

There is no program that 'reads' Tjmax, b/c there is no Tjmax to be read. It is all an assumption made by the programmer. That's why different programs have different values for Tjmax.

Yeah I figured that out after a lot of reading and discussing. Unclewebb is trying to pick up a cheap e2xxx or e4xxx to test it.
03-11-2008, 09:08 AM
Demo

Quote:

Originally Posted by jas420221

Umm, the BIOS is not idle, like in windows. BIOS will always read a few C higher.

Figures, Makes me wonder how accurate the bios temps reading are as it is 1:1 with CT

Based on my guess, my ambient temp now is about 24c, RT has the lowest reading of 29c while CT is at 33c.
my room thermometer says 30c but no way its 30c ( night time )
03-11-2008, 09:28 AM
Retrospekt

I've just decided to trust RealTemp. CoreTemp gives me ludicrous temperatures and RealTemps are right around the range it should be.
03-11-2008, 11:46 AM
unclewebb

Quote:

Originally Posted by Retrospekt

I've just decided to trust RealTemp.

That's what I like to hear! :D

I was hoping to find an L2 today but all I could find was an E2160 M0 stepping. It should provide us with a few more numbers to ponder but probably nothing too exciting. RealTemp should assume a TjMax=95C for this new processor so I'll test that theory when I get a chance.

I really need an E4300 L2 so if anyone wants to trade then let me know. I might have to head to EBay but it was slim pickings last time I checked.

Testing with a thermocouple on the side of the IHS is not very accurate and I believe the delta to core temperature will change depending on the load. Next time I'm testing I'll try moving the IR thermometer further away from the middle of the IHS and see how much temps drop. I don't think I'll get an accurate reading when trying to read the edge of a CPU. The flat surface of the IHS is ideal.

I sometimes use a high speed hand held fan pointed at the CPU which helps stabilize core temperatures and keeps them from increasing too rapidly when the heatsink is not attached.
03-11-2008, 02:14 PM
ixtapalapaquetl

Good news: I have no IHS so direct die measurement is theoretically feasible
Good news: I have a thermocouple and the bead looks like it should fit under the heatsink
Bad news: my Fluke 179's temperature calibration is ridiculously off (showing current ambient as 2.2C!!!)

I kind of tabled this project while I have been working on something else, but that has just about run its course. Can anyone suggest how I would go about getting my Fluke calibrated?
03-11-2008, 02:27 PM
rge

Quote:

Originally Posted by KTE

You can't slowly bring temp. up without a heatsink.

done it several times on my e8400, very easy, but at underclock low v like u said
03-11-2008, 04:38 PM
Retrospekt

Quote:

Originally Posted by unclewebb

That's what I like to hear! :D

I was hoping to find an L2 today but all I could find was an E2160 M0 stepping. It should provide us with a few more numbers to ponder but probably nothing too exciting. RealTemp should assume a TjMax=95C for this new processor so I'll test that theory when I get a chance.

I really need an E4300 L2 so if anyone wants to trade then let me know. I might have to head to EBay but it was slim pickings last time I checked.

Testing with a thermocouple on the side of the IHS is not very accurate and I believe the delta to core temperature will change depending on the load. Next time I'm testing I'll try moving the IR thermometer further away from the middle of the IHS and see how much temps drop. I don't think I'll get an accurate reading when trying to read the edge of a CPU. The flat surface of the IHS is ideal.

I sometimes use a high speed hand held fan pointed at the CPU which helps stabilize core temperatures and keeps them from increasing too rapidly when the heatsink is not attached.

I might be interested in a trade...

I have an e4300 L2 stepping. How high do you think I could clock the e2160 on a Tuniq Tower? I'll read up a little and get back to you. PM me if you want to talk more. :)
03-11-2008, 04:48 PM
Kami.

Does this software read the north/south bridge temps on the maximus formula? For some reason I can't seem to find a version of speedfan/everest or anything else that will read them and I'd like to know while I'm overclocking.
03-11-2008, 05:24 PM
STEvil

Quote:

Originally Posted by ~aoe~

I think what STEvil means is that, by milling a channel in the IHS, the contact area between CPU and heatsink/waterblock will be less and therefore cooling performance will be reduced (resulting in higher temps).

Correct
03-11-2008, 06:27 PM
tet5uo

Quote:

Originally Posted by ixtapalapaquetl

Can anyone suggest how I would go about getting my Fluke calibrated?

http://www.flukecanada.ca/onLineServiceRequest/

This was a link I found on fluke... there must be one for each region they sell in.
03-12-2008, 01:39 AM
Kain XS

Quote:

Originally Posted by ~aoe~

I think what STEvil means is that, by milling a channel in the IHS, the contact area between CPU and heatsink/waterblock will be less and therefore cooling performance will be reduced (resulting in higher temps).

Youre right.
Didnt got it till now :)

Show 100 post(s) from this thread on one page