Temperature Readings...

[Ruger22C]

http://img147.imageshack.us/my.php?image=cpuxq8.jpg

Why am I getting so many different temperatures?

E8400, wolfdale (45nm)

What's the correct temperature!

 

[CurseTheSky]

IIRC, Real Temp used a TJ Max of 95C, and Core Temp was 105C. The correct TJ Max for Penryn is 100C.

So, Real Temp is 5C too low, and Core Temp is 5C too high. I believe you can change the TJ Max in both programs under one of the menus.

It sort of makes me wonder if the TJ Max on Conroes was wrong as well. My E6600 used to idle in the low 30s no problem. Now I'm lucky if I see low 40s.

 

[AmberClad]

RealTime 2.75 beta uses the correct TJunction Max.

 

[Ruger22C]

http://img355.imageshack.us/my.php?image=cpu2yj3.jpg

under load

So, I can't overclock? I thought that Wolfdales ran cool

 

[CurseTheSky]

Read my last note about my E6600 in my post above.

What cooler are you using? If it's the stock Intel cooler, I'd look for something a bit beefier. Otherwise, if you're not in a hot room and your case has sufficient airflow, I'd try removing the heatsink, cleaning it off, and reapplying your favorite TIM, and reseating it. Chances are it's making bad contact with the IHS (especially with push pins).

 

[Ruger22C]

I just (before I took those photos) reseated it, took off the stock paste and put on Silver5.
My room is very cool, I have no heating yet where I live ..

 

[Ruger22C]

Yes, Curse. 5C too high.. so that means mine went to 67C at highest.

 

[BonzaiDuck]

Originally posted by: Ruger22C
http://img355.imageshack.us/my.php?image=cpu2yj3.jpg

under load

So, I can't overclock? I thought that Wolfdales ran cool

Post your motherboard make, chipset and perhaps even the BIOS revision -- like CurseTheSky.

I updated to a 780i motherboard from a 680i, and exchanged an E8600 processor for an E8400 such as yours.

According to RealTemp, my sensors are either stuck, need an update to both RealTemp and Coretemp programs, or there is a BIOS glitch that requires a revision to read the sensors on these processors properly. Look for a March 2008 Anandtech article on the E8500 and changes in the sensors and how they're read.

The same problem comes up with Everest Ultimate's edition that was "new" before my E8600 was released. You're using several different programs including SpeedFan to read your temperatures. What appears telltale to me is that both cores are reading exactly the same.

Anyone else have experience with this problem? I'm guessing that it's the new E0 stepping and processor release, but the OP has an E8400. I had to upgrade my BIOS with the 680i board before it would read the sensors correctly. [and then, I blew the Northbridge on the 680i -- dead in the water -- and reason I moved up to the 780i motherboard.]

This could be more of the disconnect between nVidia and Intel. Some of the mobo manufacturers handled it better than others, but each time a new stepping or new processor appears, I expect to see these little troubles.

 

[DrewSG3]

I have that same problem too, except I have a Q9300 @ 2.92... Coretemp, Speedfan all say an average of 57-59C across all 4 cores, but when I go into the BIOS, it'll say CPU temp at 33C. This is also the temperature that Gigabyte's tuner application reports.

 

[Ruger22C]

Drew,

When I use software, as in the first photo? the CPU reports as 30's ºC. But, CORES report as 50's.

Bonzai,
GA-P35-DS3L .. BIOS F6 or F7.

 

[Ruger22C]

When I use the software that Gigabyte wants me to use, I'm getting 24C idle (lol)

Also,

Just found this on another forum

"TheKarmakazi
Quote:
Originally Posted by RangerXLT8
"Coretemp reads the actual die temp. Those temps are not dangerous at all. These chips have a tj of 105C. Coretemp is misleading because it measures core temps not the Intel spec temp that is measured right under the center of the heatspreader."

So the individual core temperatures are normally going to be higher (and it's acceptable for them to be so)? So Intel's recommended max temp of 70* is based on the CPU temperature, not core temperature?"

 

[Tempered81]

Originally posted by: Ruger22C
http://img355.imageshack.us/my.php?image=cpu2yj3.jpg

under load

So, I can't overclock? I thought that Wolfdales ran cool

Your core temperature is 66C in that screenshot.

 

[BonzaiDuck]

Originally posted by: Ruger22C
Drew,

When I use software, as in the first photo? the CPU reports as 30's ºC. But, CORES report as 50's.

Bonzai,
GA-P35-DS3L .. BIOS F6 or F7.

I have a pair of Gigabyte boards with the 610i chipset, but the BIOS revisions have the same names -- F6 and F7.

The key to all this -- beyond simply a bunch of flawed motherboards across several chipsets -- is "when were the BIOS' released in relation to when processors being used were released?"

It took ASUS about six months from the release of the first WolfDales to clear up irregularities in sensor readings with the 680i chipset.

Now, on the matter of what is reported in BIOS versus what is reported in software.

The BIOS reports "TCase" temperatures, or the temperature at the center of the IHS. The core temperatures on C2D Conroe chips will be about 10C more than that, per Intel specs, and for the C2Q Kentsfields -- about 15 to 18C more.

How TCase varies from the cores on the Wolfdales and Yorkfields -- I haven't yet looked it up or made a personal assessment. Part of the reason for my tardiness in doing so is the marvelously low temperatures I had with my E8400 -- when the BIOS had been flashed with a July '08 revision, and as read by RealTemp.

The fact is, for me, I decided to choose last summer's "second-best" heatpipe cooler over the Noctua NH-U12P for the re-build of this system after I blew the northbridge on my 680i board. I decided that CPU cooling was secondary to cooling the NB and other motherboard components, and I wanted more Lebensraum for ducting the board in order to achieve that. So I gave up 4C degrees-worth of cooling with the Ultima 90.

I could worry that I'm "flying blind" without readable temperatures from this E8600, but at something lik 3.75 Ghz, the over-clock is only about 13% over stock, and the voltage so far -- as read by the voltage sensor -- is about 1.5% above the retail-box maximum. In fact, I'm PRIME95 stress-testing right now at 3.9Ghz -- with the same voltage setting and same sensor reading.

I have to say that this is an amazingly over-clockable CPU, except that it's still only about 20% above the stock clock setting, and I've seen some Intel C2D's like the E2140 over-clock to as much as 66% above stock.

I just wish the temperature sensors could be accurately read. The voltage sensors seem to be working fine. Since the fan-speed control in BIOS is determined by the temperature sensors, this is a terrible shortcoming for this 780i board -- or for its BIOS -- at least temporarily -- and I'm hoping that it is just temporary and can be fixed with a BIOS revision.

 

[nyker96]

I think the reason for all this discrepancy is that Intel on-die sensor only shows the amount of temperature remain before auto shutdown. called delta to TJMax. TJMax been the max temperature before CPU will shut itself down. But to find your real temperature you need to know what TJMax is for each chip and their revisions to so you can calculate actual temperature. But until a month ago Intel never officially published this number so people used thermal guns, astrology etc to guess at this TJMax number for each chip. Some temp program guessed too high, others guessed too low. This is why you have to wait a little for these program to get updated then they will all have the same and correct readings.

 

[CurseTheSky]

I'm 99% sure the Penryn TJ Max is 100C. I'm sure someone has a like to the official documents. I'm running Core Temp 0.99.3, which correctly displays 100C TJ Max (really, it just saves you the step of subtracting 5C from the reported temp, so not a huge deal).

Your temperatures are about 5-8C higher than mine, and I'm using a 120mm HDT cooler (Sunbeam Core Contact Freezer). 66C under load is a bit higher than I try to go, but it's not going to fry your chip. I try to stay under 70C at all times (for constant use, I like to stay under 55C), but to each their own.

First, try running the computer with the case's side panel off. If your temperatures are more than a few degrees cooler, your case isn't getting enough air flow, or the air is going to / going out of the wrong places. When I originally set up my P182 SE, I had all of the fans on low (whisper quiet). I couldn't figure out why I was idling around 50C, until I took the side panel off (it made a 5-6C difference). Now that I run the fans on medium or high, taking the side panel off only makes things run 1-2C cooler at best.

Additionally, if you said what cooler you have, I missed it. If you're using the stock Intel cooler, consider paying another $25-35 for something like the Xigmatek HDT-S1283 or Freezer 7 Pro. For an additional $5 or so, you can get a bolt-through kit that will eliminate the stupid push pins and all the problem they cause.

 

[BonzaiDuck]

Originally posted by: CurseTheSky
I'm 99% sure the Penryn TJ Max is 100C. I'm sure someone has a like to the official documents. I'm running Core Temp 0.99.3, which correctly displays 100C TJ Max (really, it just saves you the step of subtracting 5C from the reported temp, so not a huge deal).

Your temperatures are about 5-8C higher than mine, and I'm using a 120mm HDT cooler (Sunbeam Core Contact Freezer). 66C under load is a bit higher than I try to go, but it's not going to fry your chip. I try to stay under 70C at all times (for constant use, I like to stay under 55C), but to each their own.

First, try running the computer with the case's side panel off. If your temperatures are more than a few degrees cooler, your case isn't getting enough air flow, or the air is going to / going out of the wrong places. When I originally set up my P182 SE, I had all of the fans on low (whisper quiet). I couldn't figure out why I was idling around 50C, until I took the side panel off (it made a 5-6C difference). Now that I run the fans on medium or high, taking the side panel off only makes things run 1-2C cooler at best.

Additionally, if you said what cooler you have, I missed it. If you're using the stock Intel cooler, consider paying another $25-35 for something like the Xigmatek HDT-S1283 or Freezer 7 Pro. For an additional $5 or so, you can get a bolt-through kit that will eliminate the stupid push pins and all the problem they cause.

I have the E8600 @ 4.0 Ghz as we speak, same settings of 1.32V Vcore, and these G.SKILL DDR2-1000's (DDR=800 @ 4,4,4,12) at the low-end of 2.0V. Each notch up (5 Mhz CPU_FSB or "external frequency") I've run PRIME95 for between 1 and 2 hours.

HERE'S WHAT I'VE NOTICED ABOUT THE TEMPERATURES. With Everest Ultimate 4.50, it shows a TCase temperature idle of around 37C, but this is a fluctuating temperature and it climbs to 50C under PRIME95 (Blend) load. Yet, the core temperatures under RealTemp are stuck at 56C idle or load.

Now -- given the thermal wattage of the E8600 and some extrapolation of the thermal wattage for a 20% over-clock under conditions of a voltage that shows a reading of 0.01V over retail box maximum 1.26V -- or about 1.27V -- I believe based on past experience that the idle TCase should be closer to 30C with a 76F degree room ambient.

In fact, it had been my experience that the TCase value should be somewhere between room temperature and the "motherboard" temperature at idle.

So the questions of the day:

"What does Intel say the spread between TCase and core values should be under idle or load?"

I can take a look at the Intel specification/update for this processor in chapter 5 (.1?) where it should be listed. I wonder if anyone has already done that, because I've been up all night counting hours and minutes to bump up my FSB and RAM speeds to restart PRIME95.

I could be mistaken, but I thought TCase was determined from the same sensor readings as the core temperatures. This indicates to me that there's nothing wrong with my motherboard and nothing wrong with the processor -- this is a BIOS and software problem. I'm more upset that my BIOS fan-control features don't work. And I'm thinking that they're based on a TCase value, and so they should work. But in more than one fan control parameter setting, I can't even set the threshold temperatures but that they simply reset before my eyes to default values.