The Conroe CPU Temp. Thread

[JWMc]

I'm just getting my feet wet with the overclocking. I'm doing it all in the BIOS for now... no memset or setfsb for me.

CPU: E6600 Retail OEM (not lapped)
Mobo: P5W DH Deluxe (BIOS 1101)
Ram: G.Skill 2GB HZ 4-4-4-12
CPU Cooling: Scythe Ninja+ with Coolaboratory Liquid Pro (not lapped)
Mobo Cooling: AS3 on SB with cover removed and AS3 on NB with Swiftech MCX-159CU. No extra fans anywhere.
Case: Lian Li v1200b (case closed, not using the card slot exhaust fan or cpu fan duct because neither fit. My SCSI card blocks the card slot cooler from going in and my ninja blocks the cpu fan duct.)

Right now I am running at 340x9, fsb:dram 4:5, DDR2-850mhz (425.6mhz)
for a CPU clock speed of 3064mhz. All BIOS settings are stock except for manual ram timings of 4-4-4-12 and FSB of 340.


Ambient: 23C

Orthos large FFT for 6 hours:

Load Temperatures:
Core Temp: 58C (fluctuates between 54 and 59 but mostly stays at 58)
PC Probe: 41C

Idle Temperatures 15 minutes or so after stopping Orthos:
Core Temp: 44C - 50C fluctuating
PC Probe: 32C


When booting up after leaving the machine off for 6+ hours PC Probe reads like 25 or 26C at idle. Bear in mind that I use stock BIOS settings which means that it throttles back voltage and multiplier at idle so I'm only running at like 2ghz at idle. My temperature readings go down VERY fast after a short stress test but after 6 hours of Orthos it takes a bit more time .

Last night I tried for a few hours to get 380x9 stable but I could not acheive it. I tried up to 1.475 vCore, 1.65 vMCH and 1:1 with SPD timings but Orthos large FFT would always throw an error after roughly 5 minutes. Increasing vMCH from stock to 1.60 took me from an almost instant error in Orthos to many minutes straight before an error. I think if I keep increasing vCore I may see more stability but I think I'm going to give up on 3.4. Tonight I'm going to try for 360x9 @ 1.45 vCore and other voltages stock with ram at 1:1 and 4-4-4-12 or better timings. I hope I can hit cas3 at this memory speed with a little vDIMM and vMCH .

Edit: I raised vCore in the BIOS to 1.500 (with 1.60 vMCH) and I am now 4 hour Orthos large fft stable at 380x9 and 4-4-4-12! My new core temp load temperature is 67-69C though . After seeing the stability increase of going to 1.500 vCore I tried going back to stock vMCH and it bombed out pretty fast .

Edit Edit: I reduced my FSB to 378 and vMCH to Auto and I am now 7.5 hour Orthos large fft stable super cerial.

-JWM

[HaZe303]

I just got my P5B deluxe WiFi motherboard working and started it up. First thing I did was go into BIOS to see temps. This is the BIOS shipped with the board (have not flashed yet), I believe it is 3020 or something. Also I've only got the stock heatsink on the E6400 for now to make sure everything is in line. With this setup my CPU temps in BIOS were 45C and I believe motherboard temps were in the 32C range. My room temperature is about 72F-74F which is 22C-23C. I'm doing this from memory right now since I'm currently installing windows.

Does anyone know if those temps seem about right for stock heatsink?

It seems like our MOBO has some tempbug, because i have 40c idle and 50c load with my e6600 using core temp app. But almost everyone with a i975x board has much lower temps in the 30-35c in idle. So either its a bug, or these 965 boards get higher temps with conroes?? Dont know, but i got a Zalman 9500 and its bout same temps as yours, so they seem about right!! Dont worry.

[JayG30]

It seems like our MOBO has some tempbug, because i have 40c idle and 50c load with my e6600 using core temp app. But almost everyone with a i975x board has much lower temps in the 30-35c in idle. So either its a bug, or these 965 boards get higher temps with conroes?? Dont know, but i got a Zalman 9500 and its bout same temps as yours, so they seem about right!! Dont worry.

Well after installing windows and waiting for the AS5 to cure I installed Core Temp, Asus Probe, and AiSuite. My temps at idle in Asus Probe and AiSuite dropped to 38C on the stock cooler. After messing with the QFan thing I got it as low as 36C. I'm now running it at about 37C at idle. Also Core Temp reports about 3C higher on every reading. I ran some tests and so far 50C-52C is the highest I've seen under full load.

Remember this is stock HSF and NO overclocking. Theres only 2 case fans as well, both 80mm. The front one is at a low 1300 RPM and the rear fan I'm not sure of because I can't monitor it. I defenitely need better case cooling and a better case but I'm impressed with the results so far especially since I turned down all fans so that it was running quiet.

[Fred_Pohl]

Core Temp reads directly of the digital diode on the core die so its VERY accurate.

Evil - those temps are about the same as mine when two cores under load ... 54-55C with intel stock HSF ....

I spoke ot another guy today with exactly same config as mine (p5W dh deluxe, e6400 and OCZ ram) and his temps are around 42C under load ... I was surprised and confused to see that.

Since all Intel cpu temps have been read from on-die thermal diodes since the P2, what makes Core Temp so much more accurate?

[diekaste]

CPU Model: e6600 ES Step 1 Rev A1 (as reported by CPU-Z 136 I couldn't read info on the spreader)
CPU Clockspeed: 3.050/1066/1354 FSB
CPU Voltage: 1.329 Set to 1.3625 in Bios
Motherboard: 975XBX
BIOS Version:1334
Heatsink/Block:Storm AS Ceramique TIM
Idle/Load: 50 Idle 59 Load Dual PI 32, 40 MCH
Ambient: 23
Utility Reading Temps: Intel IDU

[phile]

Hello all, first post here at XS. Seems like a great community to join if you want to learn a few things.

I understand that core temps as measured by CoreTemp, and other apps, are widely considered to be inaccurate. Nevertheless, I'm really unhappy with mine. Here goes (note that the following results are taken with my E6600 running at stock FSB, and undervolted to 1.25v):

CPU Model: E6600 ES Stepping 6 rev. B2
CPU Clockspeed: 2.4GHz
CPU Voltage: 1.25
Motherboard: Asus P5W DH
BIOS Version: 1301
Heatsink/Block: Zalman CNPS9500
Fans: 2 front intake, 1 rear exhaust, and 1 top exahaust, all Antec @ 2600 rpm

Idle Temps:
CPU mobo diode: 25C
CPU core1: 43C
CPU core2: 44C

Load Temps:
CPU mobo diode: 32C
CPU core1: 50C
CPU core2: 51C

Ambient: 20C - 21C

Mobo CPU diode readings: Everest Ultimate 3.01.652

Core temp readings: CoreTemp and Intel Thermal Analysis Tool (match to within 1C)

These core temp readings are very high, imo, for stock FSB and 1.25 vcore. And, what's particularly vexing is the extreme discrepancy between the mobo CPU diode readings and the core readings. It seems that most people don't have a 15C-20C delta between the two, like Im having with mine. Further, when I run FSB 333 w/ 1.35 vcore, my cores idle in the low 50s and shoot into the low 70s at load. I'm not seeing such temps most other E6600 users. My HSF is making decent contact with the CPU heatspreader, and I'm using AS5.

-phil

[nikhsub1]

Since all Intel cpu temps have been read from on-die thermal diodes since the P2, what makes Core Temp so much more accurate?

This is wrong. You are right that the capability to read on die has been there, but MOST mobo makers do not tap into this, and still dont. Also, the thermal monitor has been redone with the new core 2 chips which varied from the dual thermal diodes in P4's and before. The P5W DH STILL does not read on die.

[HaZe303]

These core temp readings are very high, imo, for stock FSB and 1.25 vcore. And, what's particularly vexing is the extreme discrepancy between the mobo CPU diode readings and the core readings. It seems that most people don't have a 15C-20C delta between the two, like Im having with mine. Further, when I run FSB 333 w/ 1.35 vcore, my cores idle in the low 50s and shoot into the low 70s at load. I'm not seeing such temps most other E6600 users. My HSF is making decent contact with the CPU heatspreader, and I'm using AS5.

-phil

Hi Phil, if you check my temps... You´ll see that i have almost identical temps as you. This must be wrong temps, i think there is some kind of bug/faulty part that read the temp wrong. Every review ive read about conroe´s say they are cooler than almost any amd cpu, that the 65nm does that they stay very cool. So something is off here, i got Zalman 9500 hs which is one of the best ones out there. My temps on my old Opteron 175 were 30c idle 40c load. And the Opteron is 90nm, so my E6600 should be cooler, and is cooler. If i touch my heatsink under load its cool, if i feel the air coming out back of the chassi, its not very hot(which it used to be with my opty)... So there is something fishy going on with these conroe temps!!

[Fred_Pohl]

This is wrong. You are right that the capability to read on die has been there, but MOST mobo makers do not tap into this, and still dont. Also, the thermal monitor has been redone with the new core 2 chips which varied from the dual thermal diodes in P4's and before. The P5W DH STILL does not read on die.

If modern Intel mobos don't read the on-die cpu thermal diode(s), what do they read? I haven't seen an in-socket thermistor since I built my last AMD rig with a T-Bird 1200 and I've never seen one on an Intel mobo.

[The Coolest]

Core Temp reads a DTS (Digital Thermal Sensor), not a diode.

[pogiman]

CPU Model: E6400
CPU Clockspeed: 3.6
CPU Voltage: 1.55
Motherboard: DS3
BIOS Version: F4c
Heatsink/Block: Apogee
Idle/Load: 50/62 (Core Temp) 29/37 (Speedfan and MB diodes)
Ambient: ~23
Utility Reading Temps: Core Temp, Speedfan, Intel Thermal Analysis Tool, EasyTune 5

I want to give this 1.6V since i have watercooling, but after seeing the core temps, im afraid. Whats a safe limit on the core temps? Im sure when intel gives the max temp, it will not be core temps, since majority of users and MB makers dont even read these.

[Helidoc]

CPU Model:X6800 Step 6 Rev B2
CPU Clockspeed:3.2
CPU Voltage:1.30
Motherboard:P5W DH
BIOS Version:1201
Heatsink/Block: Storm Rev2
Idle/Load:24/40
Ambient:23c
Utility Reading Temps:Speedfan


Don't know if the above temps are anything to be concerned with. This is my first time watercooling a rig and I was a bit concerned with the variance between idle and load. Didn't think I would get a 16c swing. Any comments would be helpful

[Fred_Pohl]

Core Temp reads a DTS (Digital Thermal Sensor), not a diode.

Thank you. That makes a lot more sense than what I've been told about Intel mobos not reading the on-die thermal diode since I know that they have ever since the P2.

So has it been confirmed that 965/975 mobos are not reading the same DTS that Core Temp does? C2D has both a thermal diode and a DTS in each core and the bios reads the diode?

[Donnie27]

Core Temp reads a DTS (Digital Thermal Sensor), not a diode.

But this is still inside the processor right?

[Squirrelz]

Mobo: Asus P5W DH Deluxe
CPU: E6600 @ 3.2ghz (356 x 9)
Bios vCore: 1.375
CPU-Z vCore: 1.35 (why does it undervolt so much?)

Temps
Ambient: 23C
Speedfan: 40-42C idle / 55-57C load
Coretemp: 58C idle / 69-71C load !! (is that even correct?!)

question, is it right to have such a high difference in temperature readings from speedfan and coretemp?

[Fred_Pohl]

Mobo: Asus P5W DH Deluxe
CPU: E6600 @ 3.2ghz (356 x 9)
Bios vCore: 1.375
CPU-Z vCore: 1.35 (why does it undervolt so much?)

Temps
Ambient: 23C
Speedfan: 40-42C idle / 55-57C load
Coretemp: 58C idle / 69-71C load !! (is that even correct?!)

question, is it right to have such a high difference in temperature readings from speedfan and coretemp?

I too get temps about 10C hotter with Core Temp vs bios reading software. Some claim that the bios is reading mobo sensors (not true). Others claim that only Core Temp reads the new DTS (Digital Thermal Sensor) on C2D. Others claim that Core Temp is misinterpreting the DTS data. Considering the lack of evidence to support any of these theories I can see only 2 possibilties:

1) Both the bios and Core Temp read the same DTS but calculate the temp differently. No telling which is correct but I'd trust the bios much more than some beta software.

2) Intel included both a thermal diode and a DTS on each core and the bios reads the diode while Core Temp reads the (presumably) more accurate DTS.

[GAM]

...
2) Intel included both a thermal diode and a DTS on each core and the bios reads the diode while Core Temp reads the (presumably) more accurate DTS.

Uuuurrrggghhhh. Spent 15mins writing a reply then had to log in again--writing efforts gone! I hate that.

I tend to agree with point 2. Can't verify it, but it does make some sense.

Before most hardware monitoring software was updated for p965/conroe/allendale etc, these programs were reporting what seemed to be the CPU (diode) temp. Temps fluctuated less and were typically lower under load.

From memory, Everest 3.01.692 beta (updated for 965/conroe etc), reported both CPU (diode) and CPU Core temps for E6400/DQ6. These were 10-20c different depending on what the CPU was doing.

With P945/D805 combo, only CPU temp is displayed.

3Ghz @ 1.375v (1.31v reported) normal load maxed at 47c / 22c ambient.
3.5Ghz @ 1.4v (1.3?v reported) normal load maxed at 57c / 22c ambient.
3.5Ghz @ 1.4v (1.3?v reported) dual-prime for hours maxed at 64c (avg 5c) / 19-23c ambient.

Less efficient HSF, same case, fans and other components except MB/CPU.

Sorry, can't remember all the other 'enlightening' things I wrote. So this will do for now.

[linflas]

Ok I know this may sound weird, but ever since I lapped my E6600 all my temps have matched with everything I use to monitor it, I am not sure if I lapped it so close to the die I am getting core temps, or what ( I am mostly joking there, but in reality I have no idea)

As you can see, everything adds up...



oh BTW link to newest beta everest for those who need it

http://www.lavalys.com/beta/everestu...9sdlkj2xzt.zip

[GAM]

Ok I know this may sound weird, but ever since I lapped my E6600 all my temps have matched with everything I use to monitor it, I am not sure if I lapped it so close to the die I am getting core temps, or what ( I am mostly joking there, but in reality I have no idea)

As you can see, everything adds up...

Can you try something for me?

Run dual prime or whatever 100% dual-core cpu stress test you like and let us know if the 'core' temps rise quickly compared to the 'cpu' temp. At some point I guess they will even out, but I would expect the 'core' temp to be more instantaneous.

(I'd do it myself, except I no longer have a p965/E6xxx to try it)

Cheers

[linflas]

hehe, already tested that, and no they all rise accordingly, at the same rate

[GAM]

hehe, already tested that, and no they all rise accordingly, at the same rate

Well, so much for that theory.
(I was sure CPU vs CPU Core showed different temps by >5c in Everest )

Still, the question remains:
Are we comparing apples to apples when comparing our old 'CPU Diode Sensor' ('CPU') temps against current 'CPU Digital Thermal Sensor' ('CPU Core #1/#2') temps?

[Fred_Pohl]

Uuuurrrggghhhh. Spent 15mins writing a reply then had to log in again--writing efforts gone! I hate that.

I tend to agree with point 2. Can't verify it, but it does make some sense.

Before most hardware monitoring software was updated for p965/conroe/allendale etc, these programs were reporting what seemed to be the CPU (diode) temp. Temps fluctuated less and were typically lower under load.

From memory, Everest 3.01.692 beta (updated for 965/conroe etc), reported both CPU (diode) and CPU Core temps for E6400/DQ6. These were 10-20c different depending on what the CPU was doing.

With P945/D805 combo, only CPU temp is displayed.

3Ghz @ 1.375v (1.31v reported) normal load maxed at 47c / 22c ambient.
3.5Ghz @ 1.4v (1.3?v reported) normal load maxed at 57c / 22c ambient.
3.5Ghz @ 1.4v (1.3?v reported) dual-prime for hours maxed at 64c (avg 5c) / 19-23c ambient.

Less efficient HSF, same case, fans and other components except MB/CPU.

Sorry, can't remember all the other 'enlightening' things I wrote. So this will do for now.

Been there, done that, got the T-shirt...

At first I was very skeptical of Core Temp readings but the more I think about it, the more I believe that C2D has a thermal diode and one DTS per core. The bios reads the thermal diode and Core Temp, etc. read the DTS's.

So if you wish to compare C2D temps to other cpus, you should use the bios (diode) temp. If you want to know what the actual core temp is, use a program that reads the DTS's.

[linflas]

well, in my isolated case, I would have to say yes, but from most of the post's here, my readings are quite different, in the sense they are all basically the same, so the conundrum continues...

BTW, my water cooling, the water temp is at 28C coming from the cpu, so that even confuses me more, as well as the water block being completely cool to the touch.

And yes I have tested my seal by reaseating the block more times than I care to admit, think I almost went through a syringe of AS5 in the last 10 days.

[nikhsub1]

I am beginning to think something is buggered somewhere with the Intel thermal tool and coretemp, but I can't confirm. I was reading intels data sheet found here: ftp://download.intel.com/design/proc...s/31327801.pdf . Tcase is defined as the temp at the top center of the IHS (see 5.1.2)... which NO ONE CAN MEASURE unless you mill a channel and insert a probe as per Intels specs on the TTV (Thermal Test Vehicle). The max value for Tcase is 60.1C, again this is something we can not measure. There are several thermal monitors, 2 of which can be enabled/disabled in bios. There are 2 more that will save the CPU regardless of bios settings, ie., the CPU will throttle and/or shutdown at or near critical temperature. The diode responsible for this function still has NOTHING to do with the diodes (DTS) that we can read, same as it was on the P4's...

What does all this mean? I'm not entirely sure, but I think the values set forth in coretemp may be off... my theory is that max Tdiode is ~85C and I know that max Tcase is 60.1C. Again, the DTS responsible for making sure that Tdiode does not breach 85C IS NOT READABLE!!!

Another discovery is that as people have mentioned Asus probe and Speed fan DO seem to be reading some sort of diode (DTS). Temps jump instantly from idle to load and viceversa. If I had to guess, it would SEEM like coretemp is somehow tapping into the DTS that is supposedly NOT READABLE and that the 'other' diode temps are tapping into the DTS that Intel intended for end users to monitor their temps with.

This certainly creates more questions than it answers - Chapter 5 in the pdf I linked to is worth a read.

[linflas]

FFS, @nikhsub1 are you saying all of our temps are off by 25C, that would make sense as to my internal water temp, but still, seems absurd that all the monitoring we have could be so flawed.

Though when faced with all the obvious being wrong, the absurd is what is left (very bad paraphrasing, my apologies, have been up for 30 hours)

Is their anyway to safely test the temp through a probe of some sort? I have been trying to come up with a resonable way to do it, I tried booting with my block off, at my increased voltage (stupid I know) was able to post, but then blamo, shut down, so I learned real quick, the damn things get hot, cause I have every thermal guard I could get at turned off.