Real Temp - New temp program for Intel Core processors

[unclewebb]

Sweet Jesus, XS is back in business. And I thought crack was a bad addiction!

http://www.fileden.com/files/2008/3/...alTempBeta.zip

The above feature has been added and I also increased the maximum amount of negative calibration from -9.9 to -19.9. Some guys with 45nm Quad cores using the Intel specified TjMax are still getting idle temps that appear to be way too high during low MHz, low voltage, idle testing. If you can't get your idle temps looking good with this much correction then it's time to head to EBay and dump it, unless it's a good clocker!

Default TjMax for the early 65nm CPUs has been adjusted upward based on IR testing and Intel's release of TjMax for 45nm. For 65nm I'm still only guessing because Intel refuses to release TjMax information.

[gtj]

Sweet Jesus, XS is back in business. And I thought crack was a bad addiction!

http://www.fileden.com/files/2008/3/...alTempBeta.zip

The above feature has been added and I also increased the maximum amount of negative calibration from -9.9 to -19.9. Some guys with 45nm Quad cores using the Intel specified TjMax are still getting idle temps that appear to be way too high during low MHz, low voltage, idle testing. If you can't get your idle temps looking good with this much correction then it's time to head to EBay and dump it, unless it's a good clocker!

Default TjMax for the early 65nm CPUs has been adjusted upward based on IR testing and Intel's release of TjMax for 45nm. For 65nm I'm still only guessing because Intel refuses to release TjMax information.

I was just about to ask for the extra calibration window for my Q9300, thanks.

[Puttz]

Sweet Jesus, XS is back in business. And I thought crack was a bad addiction!

http://www.fileden.com/files/2008/3/...alTempBeta.zip

Looking as great as ever Uncle!!!

Many thanks for all your hard work, this is by far the best utility out there!!!!

[unclewebb]

I was working with a couple of guys on the TechPowerUp site today and their 45nm Quads appear to have at least 15C of slope error. For users with sensors that are reading too high, the official release of TjMax=100C hasn't done them any good. It's just made things worse. With this much error in some of these sensors at idle, it's no surprise that Intel didn't want to tell us too many details at the IDF.

Burebista on TPU posted some results from X-Bit Labs showing that an E8400 consumes approximately 3 watts at idle. With a big air cooler or water, my assumption that idle temps are only about 5C over the surrounding air temperature still seems reasonable. The new E0 series allows the processor to enter C4 power saving mode for even more power savings and low idle temperatures.

[psychok9]

Sweet Jesus, XS is back in business. And I thought crack was a bad addiction!

http://www.fileden.com/files/2008/3/...alTempBeta.zip

The above feature has been added and I also increased the maximum amount of negative calibration from -9.9 to -19.9. Some guys with 45nm Quad cores using the Intel specified TjMax are still getting idle temps that appear to be way too high during low MHz, low voltage, idle testing. If you can't get your idle temps looking good with this much correction then it's time to head to EBay and dump it, unless it's a good clocker!

Default TjMax for the early 65nm CPUs has been adjusted upward based on IR testing and Intel's release of TjMax for 45nm. For 65nm I'm still only guessing because Intel refuses to release TjMax information.

If I understand good... the difference from real temperature of the cores and IHS (measured from you) is -5°c, and the correct TjMax for Q6600 is 100°c?
E8400 = Q6600?
Sorry for my eng

[Puttz]

Uncle, hope you don't mind my asking a stupid question.

Since I have lapped my CPU & the D-tech block was already perfect (believe it or not) & half a half a grain of rice liquid metal, my temps at idle have dropped 5-7 & load 12-15 degrees.

Now water temps on the hot side are 27.6, ambient is 26.2, Tcase is at 29.6. Would it be safe to assume that my cores would be in the 32-33 range?

This seems to make sense to me!

Thanks

P.S. - All temps are taken with a laser thermal sensor. And if your curious my full load temps don't go over 43-44. RealTemp sensors show 7,7,7,7.

Again, Thanks.

[CompuTronix]

Burebista on TPU posted some results from X-Bit Labs showing that an E8400 consumes approximately 3 watts at idle. With a big air cooler or water, my assumption that idle temps are only about 5C over the surrounding air temperature still seems reasonable. The new E0 series allows the processor to enter C4 power saving mode for even more power savings and low idle temperatures.

Unc,

During which power state? I'd like to take a look at the X-Bit Labs results please.

[unclewebb]

CompuTronix: I'm assuming C1E but the article is not clear.

http://www.xbitlabs.com/articles/cpu...lfdale_11.html

It's clear that 45nm uses less power at idle and at full load than 65nm. Less power equals less heat to dissipate so sky high idle temps don't make any sense. Looks like 15C of slope error at idle is not unusual for some 45nm Quad processors.

Puttz: Your logic looks good and I agree with your idle temps.

psychok9: Intel says TjMax=100C for E8400. IR thermometer shows 80C when RealTemp reports 85C. Swap in Q6600 and set TjMax=100C and try the same experiment. IR shows 80C, RealTemp shows 85C. 5C gradient exists according to Intel.

Swap in E6400 - B2. Use TjMax=85C like every program has assumed for 2 years. IR temp equals RealTemp so TjMax=85C must be wrong. It needs to go up by 5C so the Intel specified gradient exists. B2 and L2 get adjusted to TjMax=90C to account for this.

Let the arguing and fighting over TjMax begin again!

[i43]

C4 state ? i think - they refer to to the (extra low power) - whatever its called .. this could be? that EPU chip + driver thing? ..in my guess

in present past i do have had only asus boards, so far ( p5K3 dlx , maximus . rampage )

so with latest ROG boards.. ME . RE and on base their offered options - i guess this could be that - ! on board EPU chip !- + driver + software ( EPU ? Ai suite )
could be they they call this C4 state ?
once on Maximus Extreme board i did tha whole '' wizard'' base install from mobo setup/driver cd/dvd and i got all this "crap" - sorry - loaded ..
Ai SUITE + with it's overclock/downclock utilities,
and all the following .. when it booted to xp asus ME board / qx9650

got loaded that '' Ai thing '' on taskbar - it had choosable options, and when select (very)Low power .. it downclocked my qx to ~1800mhz .. something to 0.98x volts .. actually it was stable .. i even run some prime there it take ~ 35-40W
will upload this funny screen ''polar bear playing with ball ''globe'' LOL - i did NOT get 3Wstts really here but ~11w
so could be 3w possible inn ''sleep' or Ai NAP state activated -// may be - Vista can handle these things better ?


btw new RE software disk gives a WARNING
'' do not install (Ai-Suite, EPU-Six Engine, PC Probe ) if You will use Tweakit func .. Operating both func. simultaneously might damage your motherboard

[unclewebb]

Here's Burebista's 4.24 watt screen shot for his E8400. He likes low power.
http://forums.techpowerup.com/showpo...2&postcount=31

[rge]

Actually the low watts and the below intel doc may be close to making sense with the 5C gradient we measure...if the chart below from intel is correct, which implies the gradient from core to casing increases (doubles) when removing the heatsink.
http://www.intel.com/support/process.../CS-011039.htm

Tjmax-tcasemax = theta (jc) x TDP
Solving for theta with Tjmax 100 = .43 C/W

If the thermal resistance from core to casing (theta) does increase/double without a heatsink as intel states, then if an actual 1.6-2.5C gradient does exist we would measure it as 3.2 to 5C with heatsink removed.

Or, stated another way, instead of actual gradient (heatsink on) = .43 X TDP, gradient we measure with heatsink off = .86 X TDP. At idle, extended halt intel lists 8W, deep sleep 6w, and at idle undervolted, underclocked probably more in 6W or less range...clearly this will be variable cpu to cpu and some are testing even lower. But at 6W (max?), the gradient would be 2.5C, we would measure 5C, assuming the intel doc is correct. If true, that would suggest tjmax is 100, when we measure the IHS with heatsink off we would measure 95C, if we could measure the IHS with heatsink on we would measure 97.5, which would make sense with the actual ~2C gradient. But other than a calibrated thermocouple embedded in IHS...dont know how to test that.

[unclewebb]

It only took a couple of years but this is all starting to make sense now, even for 65nm.
Thanks rge for that info.

[jason4207]

Actually the low watts and the below intel doc may be close to making sense with the 5C gradient we measure...if the chart below from intel is correct, which implies the gradient from core to casing increases (doubles) when removing the heatsink.
http://www.intel.com/support/process.../CS-011039.htm

Tjmax-tcasemax = theta (jc) x TDP
Solving for theta with Tjmax 100 = .43 C/W

If the thermal resistance from core to casing (theta) does increase/double without a heatsink as intel states, then if an actual 1.6-2.5C gradient does exist we would measure it as 3.2 to 5C with heatsink removed.

Or, stated another way, instead of actual gradient (heatsink on) = .43 X TDP, gradient we measure with heatsink off = .86 X TDP. At idle, extended halt intel lists 8W, deep sleep 6w, and at idle undervolted, underclocked probably more in 6W or less range...clearly this will be variable cpu to cpu and some are testing even lower. But at 6W (max?), the gradient would be 2.5C, we would measure 5C, assuming the intel doc is correct. If true, that would suggest tjmax is 100, when we measure the IHS with heatsink off we would measure 95C, if we could measure the IHS with heatsink on we would measure 97.5, which would make sense with the actual ~2C gradient. But other than a calibrated thermocouple embedded in IHS...dont know how to test that.

[genec57]

Here's one for Unclewebb or anyone else who can enlighten me.
System is per sig with the addition of TEC cooling. Real Temp is reporting 11c on both cores at 4500 with 1.28125 vCore. The weird part is that Everest and speedfan show a cpu temp of 30c.
The core temps are the important ones but it is strange and I would love to know what is going on. Stuck sensor?


Edit: just took the system down to 3.33 with vCore of 1.000 the cores are still showing 11c and the cpu at 30c

[loonym]

This is just a guess but the 30c reading might be a board sensor at or around the socket that's not directly influenced by the TEC.

[unclewebb]

genec57: The sensor used to report CPU temperature is different than the sensors used to report core temperatures. I'm not sure what the lower limit is for the CPU temp sensor.

The core temperature sensors have a theoretical lower limit of -27C but I've yet to see one get this low without getting stuck. Intel has told us that some of these sensors might start to "bottom out" up around 50C. If yours get down to 11C then you're doing better than most.

The bottom line is that I don't think any of these sensors are going to give you accurate temperature readings when using a TEC. When you were using just water did you calibrate RealTemp and were you able to get some reasonable looking idle and load temps?

[genec57]

Under water alone the core temps were almost always the same and were quite believable under both load and idle. Now, the cores are reasonable under load.
For a while yesterday (first day under TEC) the cpu reported -2c then up then down and now it wont go below 30 but will go up under load, always staying about 20c higher than the cores. The cores wont go below 11c but will go much higher under load.

That really just means I will have to adjust the trigger cpu temp to shutdown the system. I have speedfan set to shutdown at 70c which brings up a suggestion I have. I would love to see a high temp shutdown in real temp. Then I could use it for ALL monitoring.
Thanks,
Gene

[unclewebb]

Just a minor update so that the System Tray status gets saved on exit. The System Tray Current, Min, Max, Avg, Distance to TJMax options are saved in the INI file so you'll get to look at the same info next time you start up RealTemp.

The other minor change is to the XS Bench feature. I've changed it to use 2 megabytes of memory during the bench which should fit in the L2 cache of most processors. There's talk of the new E0 processors having increased cache latency so I thought a benchmark that easily fits inside and targets the L2 cache might give users a way to compare E8x00 C0 vs E0. Do this quick bench on a C0 and an E0 at the same multi and FSB and maybe it will tell us something. This simple bench has always been consistent and very repeatable so maybe it's capable of detecting an L2 cache latency issue with E0 if there is one.

http://www.fileden.com/files/2008/3/...alTempBeta.zip

genec57: I was afraid to allow RealTemp to run a .bat file when an alarm was triggered because I thought some anti-virus / anti-spyware type program would flag RealTemp as being dangerous and a threat to man kind. By using a .bat file, a user would have full control over what they want to happen when a high temp alarm is registered. You're the second person to ask for a high temp shut down feature for RealTemp so I'll move this to the top of the things to do list.

The lower point where the core temperature sensors "bottom out" or get stuck seems to stay at one consistent temperature. If your cores are getting stuck at 11C then they will very likely continue to do that for the life of your processor. I'll have to wait until winter to see if my E8400 still gets stuck at the same point. Canadian winters are an overclockers best friend.

I'm not sure what happened to your other CPU temp sensor. I've heard of freezing warts off but I didn't know you could freeze a sensor to death!

[genec57]

Thanks for the replies. The other CPU temp sensor isn't dead. Like the core sensors it is stuck at ~30 but it will go up under load.
I will look forward to the high temp shutdown feature. I really prefer to use and trust Realtemp so it will be great to be able to use it exclusively.
Thanks again,
Gene

[unclewebb]

I really prefer to use and trust Realtemp

That's all the motivation I need!

[spiro]

Thanks for the software. I wonder is it possible to have it setup for 8 core system? I tried it on Skulltrail but it reports only 4 sensors. Also I noticed that it reports about 5c different than Core Temp ... is that ok? Thanks again for your efforts!

[Marios]

http://softwareblogs.intel.com/author/taylor-kidd/
Great Intel blog about CPU power management, C and P states.
There is also a wish list going on there.
Enjoy reading.

[unclewebb]

Thanks for the software. I wonder is it possible to have it setup for 8 core system?

It would be possible but it's not really high priority at the moment since very few users have a SkullTrail system. The easiest way for me would be to try and create a second version of RealTemp. Regular RealTemp could read cores0..3 and the second version could read cores 4..7 for you. Not a perfect solution but then I wouldn't have to worry about bloating up the main RT code to cover a system that is in the minority. If interested in that then let me know. It might take a couple of weeks to get around to it.

Also I noticed that it reports about 5c different than Core Temp ... is that ok?

When you report something like this I need to know the version of RealTemp and CoreTemp you're using as well as the processor you're using. Both programs have recently made some changes. With the latest versions of each program, 45nm processors should be using the same TjMax since Intel finally disclosed that information. For older 65nm processors, it's still a guessing game since two years after the fact, Intel still hasn't released any TjMax documentation for their Desktop processors. Post some more info and I will explain further.

Here's the latest version of RealTemp:
http://www.fileden.com/files/2008/3/...alTempBeta.zip

Marios: Thanks for the link. With that being on the Intel website, I better read fast before all that information mysteriously disappears and Intel pretends that it never really existed!

[genec57]

Unclewebb,
A follow up to the stuck CPU sensor. I have found that when I am changing settings, particularly Vcore will come unstuck (dropping to about1c) for a while and then will revert back to its stick point of 30c.
Is this typical of a sticking sensor? My intuition says so but I have no experience on point. Right now it is at 0 and the cores are both at 11c.

[radaja]

both my sensors are stuck at 33 core0 and 28 core1 at idle once i put them under load they scale evenly. ive tried to calibrate them so theyre even at idle but under load they are way off