Real Temp - New temp program for Intel Core processors

[DOM.]

does this support X3350 ?

cuz my idle temps seem high and about the same at 2.66Mhz @1.082v and 3.6Mhz at 1.312 load

[Klimpen]

Hi unclewebb and all other of you in this thread.
I have like many others a E8400 and I want to use the RealTemp as accurate as I can.
I run my system with two goals in mind.
1. Absolute silence.
2. As high performance I can get without breaking rule 1.
I run a external watercooling with a quad radiator and I reach almost ambient temp in the water. I have some pretty neat control units attached so I can read realtime temps of almost everything.

So I have now tested the method you, unclewebb describes in your first post to calibrate the RealTemp to accurately show the idle temps.
I will start of with all possible data I have about my system:

CPU: E8400
MB: Gigabyte P35-DQ6
Mem: Crucial 1066MHz
VGA: ATI 3870
OS: Vista x64

Radiator: 2 x 120.2 with 4 x 120mm fan
CPU block: Swiftech Apogee GT
Pump: Laing DDC with Alphacool top
Tank: A 2 liter homemade chunk
GPU block: EK Waterblock (fullcover)

Measuring equipment is a T-Balancer with an analog hub connected.
I have all sorts of things to control and measure with it
It has a water temperature accurasy of .5 degrees.

Temperatures when doing the calibration.
Ambient in room: 19.5
Water FROM PC: 21.5
Water TO PC: 21.0

So what do I get out of this? I enter BIOS and set these numbers
vCore: 0.950 V
CPU Multi: 6x
FSB: 200
CPU: 1200 MHz
EIST: Enabled
C1E: Enabled

Settings in realtemp.ini
[RealTemp]
User=0
TjMax0=0
TjMax1=0
TjMax2=0
TjMax3=0
Idle0=-2
Idle1=-2
Idle2=0
Idle3=0
LogInterval=5
LogFile=0

So after all this data and info for you my question is simple:
Are my sensors stuck or broken since I get a readout of 30 degrees celcius after 30 minutes at absolute idle.





It does not make sence what so ever that I would have THAT high temps when running that much UNDER-clocked. I even have the ini set to -2 which is max downwards.
Am I doing it all wrong or missing some key function. RealTemp itself is a excellent piece of software but I feel my CPU is a bad one when it comes to the sensors.
On the other hand it will run stable up to 4 GHz at absolute stock settings and 4.2 GHz with additional .080 Volt

--------------------------------------

Edit for reference:

[rge]

Maybe unclewebb wont need the sensor adjustment for DTS that is supposed to be used in some 1st quarter and 2nd quarter 45nm intel cpus, if indeed it is as accurate as claimed, 1.5C between 40 and 120C, but that of course assumes the diodes it is reading is going to be linear over a wide range as well.

As many know, there are several transistors on each core that function as diodes/thermal sensors which are scanned by the DTS on chip which records the highest as core temp. (slide 22) http://download.intel.com/technology...forum_2006.pdf

National Semiconductor has illustrated some of problems of 45nm DTS reporting, using their competitors that apparently report significant temp inaccuracies http://www.national.com/appinfo/temp.../trutherm.html
While obviously biased, it does offer the point of more problematic 45nm temp reporting. Though they have supplied chips before on 65 and 90 nm, this is probably price/contract bids as to who supplies what, and this DTS for 45nm is supposedly relatively new and will report temp 11bit digitally.

Dont suppose anyone knows of a register that will report which DTS is employed on a cpu? I was trying to see if that DTS is going to be the solution or if it was already employed and not living up to its claim.

[gtj]

Here's one for you...

The temps on cores 0-1 are 10C higher than 2-3 at load and almost 20C higher at idle. Cores 2-3 are correct based on my water temperature and the CPU diode reading.

A few degrees I wouldn't mind bu there's something else going on here. The waterblock is a Swiftech Apogee GTX and the mount is simple...you screw it in until tight. Even so, I checked the TIM distribution and even rotated the waterblock 90 degrees. No difference.

My E8500's 2 cores are exactly like the Q9300's cores 0-1.

Left=load Right=idle

[unclewebb]

From my previous testing at idle with no heatsink I found that core0 and core1 track each other almost exactly and core3 is usually within 1C. Core2 is always a few degrees less so let's just ignore that core for the moment. With 3 out of the 4 cores reporting pretty much the same thing at a temperature where they are typically beyond any idle temp issues, I think we can conclude that they are all working and that TjMax has to be the same for all three.



The problem on my Q6600 is that when running Prime on all 4 cores, core0 and core1 track each other exactly still but now core3 is typically 5C less. I've seen other Quad owners, 45nm and 65nm, posting the same sort of thing. Knowing that at idle without a heatsink that these 3 sensors report similar data, I thought that this problem might have something to do with installation. It's possible that the heatsink wasn't square or I got carried away and used too much thermal paste and maybe even the heatsink was too tight and was screwing up the contact and full load temps.

I started by unscrewing the Tuniq Tower one turn at a time. After each turn I re-ran Prime small FFTs for about 5 minutes to bring up the heat. Here's the first picture after one turn out:



Here there is a 4C difference and as I backed out the screws, one turn at a time, the difference continued at 4C to 5C between core0/core1 and core3.

Finally at 8 full turns out, where there was virtually zero force being applied to the IHS, there was still this 5C difference at full load.



If the sensors are working and TjMax is the same, which it seems to be, then the only conclusion is that at full load it is normal for one core of a Quad to be running 5C cooler than the hottest core and it would be wrong to come up with any sort of correction factor to "cure" this because it is normal. I also wouldn't waste my time remounting my heatsink if I saw this difference.

I'm still not sure why this happens. Since the MHz is the same for all cores then the only thing I can think of is that at full load, core2 and core3 on my Quad might be getting slightly less core voltage which creates less heat. Maybe we should call this the internal vdroop theory which might someday be replaced by the s-h-i-t happens theory.

One thing I did learn from all this is that cooling performance only decreased slightly even when the screws were almost falling out of the heatsink. Three cheers for AS5! Tighten your heatsink until it is snug but there's no point in tightening it until the mounting plate starts to bend like someone I know used to do.

[unclewebb]

Dom7184: RealTemp works with the X3350 but it sounds like your sensors might be sticking at low idle temperatures like so many other 45nm users are reporting.

Klimpen: The -2 Idle adjustment was designed based on my E8400. My low MHz / low voltage test results were air cooled using a Tuniq with the fan on high. I have not done any testing with water to see if idle temps drop to the same relative values.

It's possible that your sensors are stuck at 30C but I think they are still working OK. You can try using ice to see if you can force the temperatures lower which will prove that they are not stuck at 30C.

I will include a -3 Idle adjustment in the next minor update release for you which will help a little bit at idle and at lower temps but without a full calibration with a IR gun or embedded probe, I think you'll just have to live with less than perfect core temperatures from that cpu. If you are Prime stable at 4200MHz with 1.3 volts, I certainly wouldn't be sending it back.

rge: Thanks for your continued research on this.

[DOM.]

I just dl the latest 2.4ver. and on the test sensors i got 3-4-3-4

are Q's just kind of hot ?

[NaeKuh]

I have and I can confirm (QX9650, C0).

The system is pretty much like yours, H2O with a *huge* Reserator 2 radiator.

Left the system unpowered overnight and the whole thing including 1 litre of coolant (+ 7kg radiator) was 20 C at the morning. Booted to idle: core 2 and 3 read 10 C in RealTemp. That is 10 grades below ambient.

Now I'll PM my log to uncle :-)

No sorry the reserator 2 is nowhere near my league. If you bought that thinkn its a good watercooling system i recomend you joining the liquid cooling section to learn some stuff.



System is on 3 x DDC-3.2' dual loop: PA120.3 on the cpu side, MCR320 on the GPU side. Tier 1 H20 cooling.


Anyhow Back OnTopic
Uncle bob if i do the calibration at max, its 1C under ambient.

List of hardware:
DFI LT X38
QX9650
8GB of DDR2 Ram PC6400
Raptors x 4 Raid.

Maybe a remount will fix it. However i dont see how that would matter because as i said, its more then 3C off.

Any other Yorkie owners have this problem?

[unclewebb]

are Q's just kind of hot ?

Quads are a little hotter at idle but probably not as hot as what's being reported. Cache memory is a large heat contributor and having 12MB of it on your X3350 will add some extra heat to your chip as well.

If you go into the RealTemp.ini file and if you set these factors all to -2

Idle0=-2
Idle1=-2
Idle2=-2
Idle3=-2

your idle temps will be a little closer to reality and in a day or two when I release an update that allows setting these to -3, your idle temps should be accurate enough even if they aren't perfect.

SpeedFan is using TjMax=100C so your reported temps will be 5C hotter than RealTemp and the other software you are using assumes that TjMax=105 so reported temps will be 10C higher than RealTemp. You can go into SpeedFan in the Advanced section and set offsets of -5 for each core to bring temps in line with RealTemp if you believe TjMax=95C like RealTemp uses.

[Mk]

Wich one is right i'm confused now!
19c on air is that even real?

btw this is stock temps 3Ghz QX9650

[unclewebb]

19c on air is that even real?

Maybe. Have a look at the chart in the first post that shows that idle temperatures with core processors, especially 45nm ones, might not be that accurate.

Enable C1E and SpeedStep and then post a screen shot of CPU-Z so I can see what core voltage it is using. If your motherboard properly reduces core voltage and MHz like it should then post a screen shot of your idle temps and a few more details like case open/closed, fan speed and most importantly your room temp.

If your idle temps are too low then add the following to the RealTemp.ini file

Idle0=1
Idle1=1
Idle2=2
Idle3=2

You should have idle temps somewhere around 5C above your room temperature. Post an update and we'll go from there.

[unclewebb]

Version 2.41 has a few minor bug fixes and the ability to use a -3 Idle calibration that will help some of the new 45nm chips that are reporting temperatures that are too high at idle. Give it a try Dom7184 and post another screen shot.

I also reworked the timer in the benchmark routine to try and get rid of the sky high scores that some users obtained. It works fine on my computer even if overclocked using SetFSB / ClockGen and repeatability is excellent. If you have problems, post a screen shot with any details you can think of.

It's available from the same place as usual:
http://www.fileden.com/files/2008/3/...7/RealTemp.zip

[axis]

Version 2.41 It's available from the same place as usual:
http://www.fileden.com/files/2008/3/...7/RealTemp.zip

Same price???

Axis

[DOM.]

2.4 -2



2.41 -3

[Richard Dower]

This is my Q9450 at 425x8=3.4GHz, C1E and EIST enabled:

[ChrisZ]

No sorry the reserator 2 is nowhere near my league. If you bought that thinkn its a good watercooling system i recomend you joining the liquid cooling section to learn some stuff.

Well, it is absolutely quiet (=fanless) and this is exactly what I looked for.

And let us not pollute this thread any further .

[unclewebb]

Dom7184: In your original post you showed a Minimum temperature of 37C which is equivalent to a Distance to TjMax of 58 with no calibration.

When using an Idle calibration of -3, this same DTS reading of 58 will now be reported as an absolute temperature of 29C which is getting pretty close to the truth. Far closer than CoreTemp is reporting.

The best way to find your minimum core temp is to turn your computer off at night, preferably in Stand-By or Hibernate mode with RealTemp running. Within a minute of reaching the Desktop, your computer should report its lowest core temperature and then even if it is just idle, it will probably go up a few degrees just sitting there. Don't run Test Sensors any time when trying to find your minimum.

Core processors heat up and change temperatures as soon as you push and release the power button, so you should never see a reported idle temperature less than or even equal to your room or water temperature. How do the reported core temperatures compare now to your water temp after your computer has been sitting for a while?

[DOM.]

From standby was on it for like 60min+

[Klimpen]

Klimpen: The -2 Idle adjustment was designed based on my E8400. My low MHz / low voltage test results were air cooled using a Tuniq with the fan on high. I have not done any testing with water to see if idle temps drop to the same relative values.

It's possible that your sensors are stuck at 30C but I think they are still working OK. You can try using ice to see if you can force the temperatures lower which will prove that they are not stuck at 30C.

I will include a -3 Idle adjustment in the next minor update release for you which will help a little bit at idle and at lower temps but without a full calibration with a IR gun or embedded probe, I think you'll just have to live with less than perfect core temperatures from that cpu. If you are Prime stable at 4200MHz with 1.3 volts, I certainly wouldn't be sending it back.

I will not use ICE but rather just open the doors and let the cold in and see if I can get below 30.
Should I set RealTemp to default regarding .ini or does it not matter?
I'm a bit confused about the TjMax. Let me ask you in a polite manner: How sure are you that it is 95C for a E8400? Did you find that out with the iR fluke?
I'm willing to do the same as I own the same equipment.
I'm not sure how to do it though and let me just say I'm not afraid to burn things
Is it right of me to do it this way:
Set everything as low as possible in BIOS - Volt, FSB, Multi and so on.
Remove heatsink (waterblock) and the power up windows from "sleepmode" with everything running.
When the iR hits 95C then RealTemp also should show 95C and at the same time the Status / History will light up?
Then I shut down with "rip powercord technique"

If Status / History lits up earlier then iR then I have that x-value as TjMax, right?
Also, how fast should I expect the temp to raise? Is it very slow or is it like several degrees per second and BAM I have 100 before I even get a chance to see what happend?
Can I have a fan blowing over the undressed CPU to "prolong" the event or will that change the outcome?

[unclewebb]

Here's an interesting test:



I decided to run Prime95 on my Quad and set it up like this:



By running it on 3 of the 4 cores, the load is at 75% but the processor is allowed to automatically and continuously move the processes around to whatever core it likes. The cpu seems to balance the load on the first pair of cores based on data coming from the DTS temperature sensors. It then balances the rest of the load on the other pair of cores based on temperature data coming from that pair of sensors. Because a Quad is two independent dual core processors, it's not designed to balance the load equally on all 4 cores based on equalizing the temperature. This partially explains why you can end up with different temperatures on different pairs of cores.

I compared power usage with Prime running on core0/core1 to Prime on core2/core3 and there wasn't any significant difference. I think the big reason why cores run cooler is because of the quality of the core to IHS solder job and the resulting heat transfer away from the processor.

The CPU Usage graph shows core0 and core1 to be well balanced which is reflected in their core temperatures always being close together from idle to full load. Core2 and core3 show a significant difference. In order for the cpu to balance the data coming out of the DTS for these two, it has to put most of the load on core2 and about half as much on core3.

The DTS data says that core2 is running cooler and core3 is running hotter so the processor automatically reduces the load to core3 when it can. If you run Prime on all 4 cores at the same time then the processor can no longer balance the load since all cores are at full load. The result for me is that at around 60C, core3 reports that it is running hotter than core2 when all 4 cores are at full load but this may not be the real truth.

At higher temperatures where these sensors are fully calibrated and accurate, core2 and core3 report the exact same temperature. Core0 and core1 also report the exact same temperature though the temperature they report is different than the other core pair. This is likely due to heat transfer issues.



If these Quad cores were built from scratch and the heat transfer was equal for each individual core then temperatures should be very similar for all 4 cores at load. The problem is they're not. Intel just joined two different dual cores together so you end up with different core temperatures on both sets of cores based on how well the heat is being transferred from the individual cores into the IHS. Heatsink and IHS squareness and mounting is also going to play a part in this.

You can run this same test on a dual core by only running one instance of Prime95. If the Processor Affinity is set up in the Task Manager so it is allowed to run on either core then you should see the same balancing based on temperature data coming from the DTS.

What if you combine the above variables with the fact that these sensors are not 100% accurate. You can see that there might be a problem or two with this automatic balancing mechanism especially if you're water cooled and your full load temperatures are far away from TjMax where these sensors are calibrated and most accurate. If a core reports that it is running cooler then it will get a bigger slice of the load whether it needs it or not. This helps explain why perfectly balanced temperatures on a Quad is impossible.

Just a theory, of course.

Edit: And theory 2 is that Intel deliberately sets a different TjMax in the second set of cores. 95C for core0/1 and 100C for core2/3 makes some sense. If you believe that is what's going on then you can use the latest versions of RealTemp and set:
TjMax2=1
TjMax3=1
within the RealTemp.ini file to give those two cores a boost up to 100C.

[unclewebb]

Klimpen: I have been looking forward to someone else with similar testing equipment to come forward and do their own independent testing to either confirm or deny the testing I've done.

You don't have to abuse your processor or bring it to TjMax with one hand on the cord ready to pull!

Start by setting RealTemp so it uses the default TjMax and has zero calibration. In the top line it reports Distance to TjMax which is the raw temperature data coming out of the processor.

I've found that from 60C onwards, the data is moving at the exact same rate as changes in core temperature. If you position an IR thermometer over top of the cpu within a couple of millimeters of the surface, you should get a pretty accurate core temperature as long as the processor is idle. If the IR says 60C and the Distance to TjMax (DTS) is reading 35 then you add the two together and that is your TjMax. At lower temperatures you may run into inaccuracies in the DTS sensors. At 40C there was already an error of 5C.

For my E8400, I got very consistent results all the way up until 95C but you don't have to go this high if your results at 60, 65, 70 are consistent. You'll know what I mean if the IR and the DTS data are changing at the exact same rate.

I use thin masking tape on the cpu to reduce glare which can cause inaccurate readings. I also set my E8400 to 266MHz X 6 and dropped the core voltage as low as it could go. It's good to have a small hand held fan close by so you can stabilize the core temperatures while testing. I used an old AMD Athlon noise maker which could be adjusted up to about 7000 rpm. I also had my Tuniq close by which I just placed on top of the cpu and temperatures would drop down instantly so I could test some more. I've never once had a re-boot during this kind of testing so you should have no problems.

[Klimpen]

unclewebb: Then I will do that, all that I said and you suggested.
I won't lie and say I'm not a bit nervous
I just want to make EVERYTHING clear before testing. So the last question is:
I imagine that I HAVE to reach 95C to se if that IS TjMax since my understanding is that TjMax is the same thing as the value that the cpu will throttle and that RealTemp will show that with Status / History.

So if I don't get the Status / History to lit up I still don't know FOR A FACT what the TjMax of my E8400 is? Right or wrong?

[unclewebb]

I found that the History flag will light up first and then the PROCHOT# Status will light up. This has happened for me generally just before the DTS reports zero which signals that you are at TjMax. About DTS=2 or DTS=3 and PROCHOT# will light up.

I saw over 101C on my IR thermometer before common sense returned. Afterwards I bolted on a heatsink and my E8400 didn't seem any worse off.



You don't HAVE to go right to 95C to determine TjMax. You should see a very clear relationship develop before you get that high.

IR Temps - DTS
60C - 35
70C - 25
80C - 15
90C - 5

For me it was impossible for TjMax to be 105C like CoreTemp and others are assuming. Let the testing begin. We're all very interested in your results.

I won't lie and say I'm not a bit nervous

Only a liar wouldn't be a little nervous! I wish I had some data from a heart rate monitor the first time I took my E6400 to 85C. Now that would be interesting.

PROCHOT# should light up during testing and you should also see DTS=0 if you go that high. It's a good idea to go into RealTemp and set the LogInterval=1 so it logs the data every 1 second. If you go higher than this the DTS data wraps around and starts over again. It is stored in 7 bits of data so after 0 comes 127, 126, 125 etc.

[Xilikon]

Don't worry about damaging them, they have thermal protection. Last year, I tested with unclewebb with a E4300 that I ran up to 90C before the computer shutdown itself. There is no damage at all

For testing for a good amount of time, you need to setup like unclewebb said.

[NaeKuh]

Well, it is absolutely quiet (=fanless) and this is exactly what I looked for.

And let us not pollute this thread any further .

Edit... okey i think i got it working for some wierd reason. Hmmmm... lemme play with it more..