Printable View
They should have done their research before the first round of numbers. I mean c'mon, way below ambient on air? They should have picked up on what was going to happen, and stuck with the higher numbers to start with. Oh that's right, it was just... slope error... of course...Quote:
Originally Posted by unclewebb
The Core0 and Core1 in my Intel Xeon E5440 CPU were getting stuck at a DTS value of 67 at idle. Does this mean the real temperatures of these two cores are almost same with other two cores? Only when the real temperatures of these two cores are higher than 67, then can I see the change of temperatures of these two cores? Is the stuck a big issue and do I need to worry about this? Thanks.
Generally yes, but as you can see from jasjeet example it's possible to see a 5°C (more or less) delta between each cores group.Quote:
Originally Posted by ericjia
Yes indeed.Quote:
Originally Posted by ericjia
For me it's not a issue and to quote uncle:Quote:
Originally Posted by ericjia
Anyway, take latest Realtemp and make a CPU Cool Down test. Then you'll see how your sensors move.Quote:
Go use your computer and be happy. :)
http://www.fileden.com/files/2008/3/...alTempBeta.zip
Download the latest version and do a CPU Cool Down Test.
This will tell me a lot more about your sensors. It looks like two of them are stuck alright.
You're right. It's very likely the temperatures are similar for all cores. Lots of sensor issues with these things.
If your CPU runs good then there's no need to worry about this problem. It just won't report accurate idle temperatures. If you ever get core0/core1 above 67C then they will probably work fine.
I switched RealTemp to UNICODE and hope that I finally fixed this ugly problem:
http://img78.imageshack.us/img78/9458/uglygn6.png
Thanks Kevin for this amazing program.
I'm also a developer and all I can say is, the current 290 version of this program is almost perfect.
Keep up the good work,
Your welcome Metroid. Thank-you for showing off RealTemp with your speedy i7 920. The best way to spread RealTemp is by word of mouth and your 920 at 4.2GHz speaks volumes. :up:
I always have plans for a few more improvements to project RealTemp. Stay tuned.
Can I say this CPU is a defective product? Thanks.Quote:
Originally Posted by unclewebb
Here is the CPU Cool Down test.
1. Why do Real Temp 2.70 and 2.90 obtain different core temperatures at idle for the same system? (the right snap shot is from v2.7)
2. From CPU Cool Down test result, can I say the maximum core temperatures are around 72, which is same with stuck temperature (core0/core1)?
3. The new version solved the degree symbol display issue (for Chinese). Great!
4. One small bag: when "Always on Top" is selected, if you right click the title of the main window, you don't have any chance to select any item from the menu. The menu either moves back to the main window quickly or destroys the main window display.
Any input about my CPU cool down test is appreciated.
As long as the CPU throttles at an appropriate temperature, the sensors are working as they were designed to. Therefore Intel won't accept it as defective.Quote:
Originally Posted by ericjia
is it true that the tjmax for Q6600 G0 is 90c?
ericjia: 1) The problem with getting temperature data from Intel Core processors is that they do not have a typical thermometer inside. They work backwards. As the CPU heats up the Distance to TJMax decreases until it reaches zero. At this point, or about 2 degrees before this point, the processor will start to thermal throttle and try to cool itself down. The formula used looks like this:
Reported Temperature = TJMax - DTS
DTS is the raw data from the digital thermal sensors and TJMax is the temperature where the DTS will be reporting zero.
The problem has always been that Intel has been very secretive about TJMax and until just recently have refused to release this information to users. Intel is calling the numbers they did release recently TJ Target which may or may not be exactly equal to TJMax. For some processors it seems that the two are pretty much equal while for other processors TJMax might be 10C or more degrees higher than Intel's TJ Target numbers.
It doesn't take a rocket scientist to figure out that if you don't have the correct TJMax for your processor then using that formula above is not going to be able to give you accurate core temperatures.
If you open up the Settings window in both version of RealTemp you will notice that one version uses TJMax=95C and the latest version uses TJMax=100C. That's where the 5C difference in your reported temperatures is coming from.
As a programmer, there seem to be so many inconsistencies in the documentation released by Intel that I've decided to ignore it. Intel also admits that TJMax is not a fixed number and may vary a little from one processor to the next, even of the exact same model. TJMax also seems to vary from the left side to the right side of many Quad core processors. Because of all of these problems as well as the sticking sensors that your CPU has, Intel does not recommend using them to report core temperatures. They disagree with programs like RealTemp and they probably really dislike the latest Cool Down Test in RealTemp that shows users just how bad these sensors really are.
I looked up your E5440 and Intel says the TJ Target is 85C. I have no idea how accurate that is. Changing RealTemp and setting TJMax to 85C instead of the present 100C is going to instantly drop your reported temperatures by 15C. Is this accurate? I don't know. Maybe.
The best test that we've come up with to try and prove whether a TJMax value is possible or impossible is by running your CPU at very low MHz and core voltage. The plan is to reduce the heat output to a fixed amount so you can compare to what other users have found. The problem you'll have with this test is that your motherboard may not allow you to set your core voltage manually to a low enough value. For 45nm, I recommend about 2000 MHz and 1.10 volts. If your board doesn't allow manual voltage adjustment then make sure SpeedStep and C1E are enabled in the bios. After you boot up and when you are at idle, your computer should drop down the voltage and MHz to similar values. Use a program like CPU-Z to confirm this.
Try this test and compare your reported idle temperature to your room temperature near your open computer case. 8C to 10C higher is a typical gradient depending on what CPU cooler you're using. Let me know a few more details like that.
2) Holy smokes. So far I've only answered question 1. :)
At the moment, we can't say anything about your maximum core temperature during this test. Until we are more sure of TJMax, these are just a bunch of numbers.
3) You have no idea how happy I am to finally have got rid of the Chinese symbol. Switching a program after the fact to UNICODE was about an 8 hour pain in the butt just to get rid of that funny looking character. Do you know what it means? I hope I didn't offend anyone with that.
4) I don't think I can do anything to fix problem 4. I've also noticed that issue and will have another look. I might have to momentarily turn off the Always on Top feature when a menu is selected.
For your Cool Down Test, sensors for core0 and core1 are stuck and will be useless. The other two sensors look pretty good. If you can find the correct TJMax, they should give some reasonably accurate core temperatures. Give the low MHz / low voltage test a try.
DJSUB: Yes and no. Intel says the TJ Target for the Q6600 is 90C. My opinion is that TJMax is actually 100C for most Q6600 CPUs that I've seen and maybe even 105C on some / many of the core2 / core3 that I've seen screen shots of. That's why I'm full of doubt when it comes to Intel's release of TJ Target data that has already been quietly updated once. I'm waiting for their next update before I start to believe some of their numbers. ;)
ok thanks for the info master uncle :up:
i will wait for the intel next update & your opinion :D
The next Intel "update"... I wonder when that will be. Most of the lustre of their "updates" has been lost since the last two IDFs.Quote:
Originally Posted by DJSUB
hey, unclewebb. Couple questions.
I'm using v 2.84 for my Quad 6600. Is there any need for me to upgrade to latest version?
Lately my temps in 2.84 are jumping. Like every second, litteraly. It goes from 57 to 66, like clockwork every single second. 57,66,57,66,57,66 etc..etc..
I've been tweaking my system. Surely this is the result of something I have done???? OR, realtemp is correct and I'm spiking.
Thanks unclewebb for your detailed information.
I am not familiar with low MHz test, but will learn and find time to give a try.
concretefire: I like the improved CPU Cool Down Test in the latest version as well as two lines of information at the top that includes CPU Load. It's been more thumbs up than thumbs down recently so I think you should give it a try and decide for yourself.
http://www.fileden.com/files/2008/3/...alTempBeta.zip
The CPU Load display in RealTemp 2.90 is very accurate and repeatable. Just running Task Manager on my system creates 0.5% load and increasing the refresh rate doubles that. I never knew useless information like that until I used the new RT Load feature. :)
It might help you find out if you have any software running in the background causing your jumping temperatures. The sensors rarely jump around unless the load is changing. Disable useless background tasks until you get your Load down to 0.0% and then have a look to see how stable your temps are.
ericjia: I'd be very interested in what temperature your CPU idles at when at low core voltage and low MHz. That's what C1E is designed to do to save power but I found that doing this manually gives you a good baseline which you can compare to others. A wide variety of CPUs don't vary too much during this test. Plus or minus 1C covers the majority of Intel Core processors. Let me know when you have the time if you need any help or suggestions.
Sorry for the delay in getting some results posted. Haven't really had a chance to overclock my Q9450 until yesterday.
By the end of the day I was able to get to 8 x 425 Linpack stable (I know this doesn't mean it's completely stable but it gives me a baseline for fine-tuning later).
What I can tell you is that now that I've overclocked it some and I'm pushing about 1.35 volts (trying for 8 x 450) through vcc with LLC, the temps have increased and are now looking MUCH closer to each other - 5-7 degrees of each other. I'm sure some calibration is still needed, but it looks like what you had assumed earlier holds true.
Hopefully I'll be able to post screenshots of Real Temp cool down test by the end of this week with my final overclock.
Excellent program as I can tell thus far.
Quote:
Originally Posted by unclewebb
Damn me and my eternal search for ET. :gay: (Seti @ home)
Hey unclewebb....in no way did I mean for that question to come across as a problem or glitch with your temp program. I already knew the answer, and you were correct. I have no explanation for even asking about it other than the fact that I am a dumbass.
My Real question was would I benefit, or should I upgrade versions. I have been reading how you've been working on i7 alot, and somehow got it thru my head that the latest and greatest might not be ideal for my "old" Q6600.
In hindsight, another boneheaded question. :shrug:
Think I'll stop posting for a while. :D
I was going to say because I have my Q66 and my temps are nice and low in the upper twenties and when idle only move maybe one or two degrees. I love this program keep up the good work UncleWebb
No problems concretefire. I get a little carried away sometimes trying to make sure everyone is getting some decent temperature numbers out of these crappy sensors.
There have been a lot of tweaks to RT for better Core i7 support but I also added some features to keep the rest of us happy that can't afford or aren't interested in an advanced 8 threaded heater. Nice in the winter but I can just imagine the complaints next July. Time to invest in water cooling companies. Every Core i7 owner will be getting one next summer. :rofl:
No kidding about these i7's being heaters. I had to adjust the heating vents in my entertainment room, b/c my computer is raising the temp 2-3C in this room over the rest of house when running prime overnight.
My core i7 is even taxing my water cooling, I cant imagine it on air. I have a cpu only loop, PA120.3, 3 yates 1300 rpms, swiftech GTZ 1366, and my load temps are 73C with 1.35 vcore bios, 1.31v load for 4ghz stable (4.2 turbo). But all the new bioses with turbo enabled are set such that turbo is active on all 4 cores (instead of being TDP/2core limited on F3 bios), so have to find new settings for new bioses. Ambient is 26C.
4.2 ghz w/ 1.41 bios, 1.36 cpuz ran prime 4 hours before rebooting but load temps were 82C. I am just not into 80+ load temps especially on water.
Now trying for 4.14 (23x180) w/ 1.381 in bios, 1.34 load, temps are 76-77C on load, though apparently bounced up to 79 on core 0 once.
Not using faster fans as I like quiet. If next year the leakage or TDP does not go down on newer ones, I may be using dual rads just for cpu:eek:. Intel needs to get a better handle on leakage before going 8 core/16 thread.
BTW...loving the new realtemp layout:up:. mhz total and w/multi*bclk, load %, and timer all with easy glance, perfect for ocing. The only thing I need cpuz for is check vcore. Not to mention max temps recorded which are very helpful for o/n runs on these.
WTF is going on? My QX6800 cores heat up like this:
30-27-62-31 degrees? Is the processor heatspreader twisted or what? The "test sensors" gives like 6 sensor movements for that one core, the others are 14-15. Is the sensor malfunctioning? I've tried different air coolers and that didn't help :(
I like quiet air cooling so Core i7 in its present form is not for me. Maybe when they shrink the process down to 32nm next year or bring out a Dual Core version, I'll think about it. I'm kind of spoiled by the cool and quiet performance you can get out of an E8400.
xarot: Sounds like you've got a stuck sensor. Intel says these sensors are not designed to be accurate at lower temperatures and will become saturated as the CPU temperature drops. That's a fancy way to say that some of them stick and stop moving as the CPU cools down. They were designed to trigger thermal throttling and thermal shut down and even sticking sensors don't seem to have a problem with that when the CPU temperature goes higher.
Post a screen shot of the CPU Cool Down Test. It's an excellent tool for spotting stuck sensors and should show you exactly what's going on.
Hi, thanks for help.Quote:
Originally Posted by unclewebb
Here's a quick test: http://pumppu.info/~dakotha/kuvia/Hardis/qx6800.jpg
Not any of the sensors are totally stuck, or what do you think? The temps can't be right anyway? I just got this chip from EBay as a new one and I'm worrying do I have any overclocking chances if one of the cores is reporting crazy temps at default clocks. :(
I swapped the mobo for this test..now running on DFI X48 and Noctua NH-U12P. The previous test I did was with QX9650 stock cooler running at max speed.
Your Core 2 sensor is not the typical stuck sensor and it's definitely not working right. It's possible that you'll still be able to get a decent overclock out of this chip but the Core 2 sensor is likely going to be the first core to signal thermal throttling. When this happens the multiplier is going to drop down to 6.0 and is going to kill performance. On the main RealTemp screen keep an eye on the Thermal Status area where it presently says OK. If this ever changes to LOG then that means that Core 2 has logged a thermal throttling episode. If it shows HOT then that means that thermal throttling is in progress.
Overclock as much as you can. Thermal throttling usually occurs at about 98C. The slope of this sensor is very slow. Your other CPU cores might be able to get 20C or 30C hotter and this sensor might still not move enough to trigger thermal throttling. Let's see some numbers out of this QX6800.