Printable View
PCTwin: Thanks for bringing that problem to my attention. There were some advantages to the Core i7 averaging method that I was trying to use for Core 2 but if it's not 100% compatible with Core 2 QX processors then it had to go. I'm :) to see it's working for you with an 8X multi.
@uncle
Is a distanc to TjMax ~19 still safe for a G0 Q6600, thats 80c with 100c-105c TjMax (80,80,76,76) this is at 3.2Ghz with 1.355v bios to make it stable with LinX(Linpak64) under prime i dont go over 70c though.
In real life you'll never see Linpack temperatures. Anyway 20 distance to TJMax I'd say that it's perfect safe (as long as your CPU is stable).
QFT.....Only temps that matter to me would be running realtemp for several days of normal use, and checking max temps. If you fold at home, might equate to prime blend temps, but would never worry about linpack temps, even though I like it for testing.Quote:
Originally Posted by burebista
thanks
Demo: I think any temperature is safe as long as you're not thermal throttling and your computer is stable. Intel CPUs seem to run reliably at very high temperatures. The whole purpose of going to multiple on chip digital thermal sensors was so Intel could accurately monitor the hottest point on the core which allows users to run their chips hotter without thermal throttling kicking in.
When you read the documentation, Intel is proud of this fact. They're proud that they were able to boost TJMax by 10C from the original B2/B3 generation to the second gen G0 series like your Q6600. If doing this led to a lot of processors going ka-boom and the RMA numbers went way up then they would have lowered TJMax.
The Core i7 runs hot as hell when they are pushed hard but Intel left TJMax at 100C. I've seen lots of screen shots of people running them reliably at a core temperature of 80C to 90C and I haven't heard of a single report of a damaged CPU because of this. Intel doesn't seem to worry too much about high core temperatures so as long as I'm running stable, I don't worry either.
So what is the real TJMax on Q6600 G0 CPU? As far I have seen 95, 100, maybe other, how can I tell what is for my CPU? RT 2.70 shows 95, newer 2.90 beta 100 degrees C! :confused:
Want to set actual + calibrate it on box cooler and low speed/volt and then safely monitor temperatures when overclocking past 3.8 Ghz :up:
Given that graphics cards manage to run regularly at near 100C with the stock cooler, with TSMC processes and NVIDIA bumps.. ;) I have faith my Nehalem can take the sauna.Quote:
Originally Posted by unclewebb
Kashelz: My best guess for TJMax is 100C for the Q6600. rge showed me that my testing with my IR thermometer might provide an accurate surface temperature of the IHS but the actual core temperature is likely about 5C higher. The IHS does a great job of quickly transferring heat away from the cores so my original guess at TJMax was too conservative. That's why TJMax got bumped up from 95C to 100C between version 2.70 and version 2.90.
Even 100C is only an estimate. At the factory, setting TJMax does not seem to be an exact science. How much it varies from one CPU of the same model to the next is anyone's guess. With a Q6600, Intel might be using a higher TJMax for core2/core3 compared to core0/core1. It doesn't seem to be an accident or a coincidence since a very high percentage of Quads follow the same trend.
Use rge's guidelines and I think you'll be pretty close to an accurate temperature.
http://www.xtremesystems.org/forums/...postcount=2429
If you do a CPU Cool Down Test you might see the split TJMax that I've noticed. Post a screen shot of that test and I'll give you a second opinion of what your sensors are telling you.
bowman: I think AMD/ATI's release of the 4850/4870 series has made users a little less nervous about sky high silicon temperatures. With graphics cards or CPUs it all comes down to stability. If your GPU is at 90C and every game you play looks like it's snowing or is full of artifacts then that is too hot. If your game looks good and runs stable then you're probably not hurting your GPU too much.
So is a fair guess for Tj Max +10°C above Target Tj?
Got a bug for you.. Probably easy to fix, but hey, what do I know. My only programming experience is putting blocks together in Lego Mindstorms when I was, what, 9? :rofl: Due to the Hyper-Threading, RealTemp displays two temperatures for a single-core Intel Atom. They're both, obviously, the same temperature. Also, the sensor test seems to be cosmetically broken as well. (http://www.isarapix.org/pix61/1230858748.png) Temperatures look alright but for some reason the load % is 0.0 all the way down.
In the mean time, stand in awe at the immense computing power;
http://www.isarapix.org/pix71/1230857497.png
You ain't got nothing on this! :cool: (on the other hand, if the E8400 was the same frequency and this bench scales properly.. it'd score 533. Maybe not so far off?!)
Thanks for reply. I have done sensor test, will add screenshot later. Im useing intel box cooler fow Q6600 at the moment, because all computer is lied on small box and box un lied all over the table. Just do testings for new motherboard and voltage output from it. ;)Quote:
Originally Posted by unclewebb
The test showed that all cores looked fine, no stuck sensor, distance to TJ Max was 23/22/23/23 under full load with Prime95 and 8k ffts. Now testing Core Damage, pretty hot, 15/13/14/15 to TJ Max after 30mins of testing. The ambient temp is ~ 26 C!
Im interested to test each core when it gets into Thermal mode at what distance to TJ Max and if its 0 then I can calibrate manual seting of TJ Max, right? :shakes: And as I understand temp are more accuret after they are hotter 60C and when are more to Tj Max they are even more acurate? I mean fault of reading sensor on the core.
Ok, here is the screene, CPU is Q6600 @ 2.4Ghz at 400 x 6, dafault voltage 1.275v, just testing FSB speeds for longer usage. Few minits ago tested core3 and Throtling kicked in something like 97 degrees C, so maybe TJ Max for that core is not 100 but 97 for exact this CPU? Need more testing to see when really Thermal operation starts to work on other cores :)
Image of screenshot: http://www.xtremesystems.org/Forums/...1&d=1230859940
And here is the same testing but with Core Damage as burning software instead of Prime95 8K ffts.
Could anyone tell me why if I have a weird chip?
It's a E0 Q9550 OC'ed to 3.825GHz on 1.3125V on a 780i FTW and my temps don't look right
Basically at idle temps, it's at 31/31/39/37 while at full load on Prime95 the highest they go sometimes during high ambient temperature is like 49/49/47/48
http://img49.imageshack.us/img49/481...sortesthk8.jpg
Sorry for the uncropping of the image:D
SFaznSpEEdSTeR: It would have been nice if you would have let the Cool Down Test complete but even the partial test clearly shows that the sensors on core2 and core3 are stuck. Lots of these sensors stick and if you ran the test further you might find another one of your cores has a sticking sensor. Not all but most Quads have at least one good sensor. As for your uncropped image, why not do us all a favor and download a free program like Irfanview.
Kashelz: The TJMax on your 4 cores looks very well balanced. The sensors all have slightly different slopes. The RealTemp idle calibration factors can be used to balance the temperatures of your 4 cores at idle if you want. Just follow rge's recommendations for what a normal idle temperature is when testing at 1600 MHz and 1.10 volts. A Q6600 is about 1C higher than the E8400 he used during testing at these settings.
Thermal throttling starts to happen at a Distance to TJMax of 2 or 3 which is exactly what you found. It seems to do this because the goal is for the CPU to throttle itself and keep itself from ever reaching 100C.
I've posted this picture before but it does a great job of explaining exactly what happens so I'll post it again:
http://img387.imageshack.us/img387/8...8400fw5oj9.png
I ran Prime95 Small FFTs for 3 hours with the CPU fan turned off. The processor continuously throttled itself during this test. When it got too hot it would drop the multiplier to 6.0, cool down a little and then immediately the multiplier would go back to 9.0 until it got too hot. The multiplier cycles back and forth like this, likely hundreds or more times a second to keep the core temperature from going nuclear. It mostly sat at 98C. There were a few times where it would hit 99C but only once in 3 hours did it hit 100C as shown in the graph. I thought that was pretty amazing. You can be running a very stressful program like Prime95 and even if your CPU fan fails, your CPU will still take care of itself. When you see this, having TM1 / TM2 turned on in the bios isn't such a bad idea.
Sorry about that ;)
Anyways, the cool down won't matter anyways because after that, the distance from TJMax stays the same. It is at a default 100C so I have not changed that.
I did something similar last week , E8400@4GHZ stock cooler with Intelburn test but this time it will switch down when the temp reaches 95C , surprisingly it gave me 100% stable in 5 runs that lasted for 510SecondsQuote:
Originally Posted by unclewebb
+Can you explain a bit about sensor burning test , mine looks pretty bad when load gets higher Difference between 2 cores always get bigger Unless i am running some crazy volts and temps are well over 75C , is it sensor issues , bad cooling or uneven heatsink/IHS ?
http://i41.tinypic.com/vgiqlk.jpg
Edit:looks like sensor Core 1 is stuck
Please, I have a problem:
How can I set Real temp to start up when Vista start?
I put the .exe in automatic esecution folder, but it doesn't work.. an error message say: ....WinRing0.dll not found...
Thanks
You can use Task Scheduler, or (like I did) made an entry in registry:
HKEY_LOCAL_MACHINE\Software\Microsoft\Windows\Curr entVersion\Run right-click in right panel-New-String Value, name it RealTemp and in Value Data put your path to RealTemp executable (in my case "C:\Utile\RealTemp\RealTemp.exe")
Attachment 92011
Okay guys gotta quick question for you I have a stuck sensor on my Q9550 that reads 0 on the move test. but it does go up in temp just fine doesn't read below 35 thought so I went and adjusted the idle temp on it to roughtly what the other 4 where idling at and now it doesn't read higher then 32 when the rest of them are 40+ but the movement test now reads 3. kinda confused here...
Quote:
Originally Posted by burebista
Thanks :clap:
There isn't any formula for converting Intel's TJ Target specs to a TJMax number that can be used to give you reasonably accurate core temperatures. TJMax is not a fixed number that Intel is willing to commit to. It can vary from one CPU to the next of the same model. In some cases, it seems to vary from one core to the next within the same CPU.Quote:
Originally Posted by ptelles
TJ Target and TJ Max seem to be the same for many processors like most of the E8000 series but the TJ Target numbers that Intel released for the 65 nm processors do seem to be off by 10C for quite a few of them.
For the Xeon E5410, Intel lists the TJ Target as 85C. Run a CPU Cool Down Test to make sure your sensors aren't sticking at low temperatures. Set your CPU to 333 x 6.0 in the bios and drop the core voltage down to 1.10 volts if you can. If you enable C1E and SpeedStep and have Windows set correctly, when your computer is idle it should drop down to somewhere close to these numbers. Run CPU-Z to check your core voltage.
My guess is that TJMax is probably 95C or 100C. Open your case and compare your reported CPU temperature to your room temperature near your CPU.
http://www.xtremesystems.org/forums/...postcount=2429
Post or send me some screen shots and I'll give you my opinion of what I think it is.
bowman: I'll have to admit that RealTemp was not designed with the Atom N270 in mind. I'll make a few fixes to RealTemp to better support it but I won't be able to do anything with the simple Load meter I'm using. As for Atom performance, it looks pathetic! Here's how my Q6600 does when locked to 1600 MHz.
http://img49.imageshack.us/img49/691...1600mhzlh3.png
The single threaded XS Bench does a big calculation and scales very linearly with CPU MHz for pretty much any Core based processor. The Core i7 is about 20% faster at the same clock speed while the Atom looks like a good chip to check your e-mail with. :)
SFaznSpEEdSTeR: I was just interested in seeing the rest of your Cool Down Test so I could see if core0 and core1 were sticking.
kemo: If you go into the Settings window and click on the Defaults button, it should set your TJMax to 100C. It looks like it is presently set at 95C. The other problem you have is that TJMax is not consistent for both of your cores. I think you should set it to 100/109 for your two cores. As you noticed, core 1 is stuck so it won't be able to give you accurate idle temperatures but when Distance to TJMax is less than 55, your two cores should be well balanced.
colex: When using Vista, I prefer to use the Task Scheduler to add RealTemp to the Start Up area but burebista's method works. In XP you can copy a link to RealTemp.exe into your Startup folder. Edit: This also works in Vista. Go to All Programs -> Startup and copy a link into that folder.
Copying RealTemp.exe won't work since it won't be able to find the WinRing0 driver.
Demthios: You can't calibrate a stuck sensor. Post a screen shot of your CPU Cool Down Test. That's the easiest way for me to get a lot of information about your sensors without having to ask you a lot of questions.
unclewebb,
Can I trouble you to have a looky at my sensor test and tell me what I need/if I need to alter any TJmax or idle calib?
http://img90.imageshack.us/img90/3648/realtempjh0.jpg
Thanks matey
Figured as much. ;) But I also figured you'd want the crown for Intel compatibility so it's another notch in the belt, right? Aside from the two GUI quirks it seems to work fine.Quote:
Originally Posted by unclewebb
Check e-mail, browse the web, view videos, use Paint.NET.. It works just fine for my mobile purposes:)Quote:
Originally Posted by unclewebb