E6600 OC Temps: Idle/Load delta slightly exceeds recommended

[aoe]

I have recently completed a new OC, and would just like to check the resulting temps (all temps in Celsuis):

Tjunction idle = 28 (max peak)
Tjunction load = 55 (max peak)
Ambient = 20/21

Chipset: Intel 975X
CPU: E6600 (B2)
Cooler: Scythe Infinity (installed with supplied TIM)
CPU Frequency: 3.4GHz (378MHz FSB)
Load Test Program: TAT
Motherboard: Asus P5W DH Deluxe (BIOS 1506)
Vcore: set to 1.3625 in BIOS; reporting 1.34 in BIOS; reporting 1.34 idle and 1.31 load with SpeedFan 4.32

FSB voltage was increased to 1.4V; SB voltage (vICH) at 1.20V; vMCH at 1.55V

Tcase = 50 (load) with SpeedFan during TAT load test, so requires offset change.

I've tested stability with Prime95 (2 hours) with slightly lower load temps than those reported than TAT. Memtest 86+ ran a complete cycle without errors.

Also, I can expect my ambient temp to increase by as much as 10C during the Summer, so I will need to monitor that when the time comes.

Now, I realise that my idle/load delta is slightly over the recommended 25C, but temps are well within the guidelines posted, so I was wondering how big an issue this really is.

Thanks in advance.

 

[CompuTronix]

Tcase Idle?

 

[aoe]

I knew I would miss one of the temps :roll:

As I have many different results recorded at different times, I ran another test. Below are the results:

Ambient: 21C

Idle:

Tcase: 26C (SpeedFan)
Tjunction: 24C (SpeedFan); 27C (TAT - occasionally peaked at 28C for 2 sec bursts)

Load (Prime95):

Tcase: 47C (SpeedFan)
Tjunction: 47C (SpeedFan); 46C (TAT)

Load (TAT):

Tcase: 51C (briefly peaked at 52C)
Tjunction: 56C (SpeedFan); 53C (TAT on average - occasionally peaked at 55C for 3 sec bursts)

SpeedFan also reports 2 other motherboard temps:

Sensor 1: 26C idle; 27C load
Sensor 2: 21C idle; 35C load

 

[knowsitall]

um well i would say as long as u dont see anything over 60degress on ur processor u should be fine, but i see u have temps in the 55, a little bit toasty, but i would not worry about it unleess u realy want to push ur system further, just make sure ur heatsink is on correctly.. if it is install correctly i wouldnt want u to push ur system any further, especially with the spring and summer coming around the corner, ur system will probaly jump up a few degrees unless u blast the central ac, i dunt know my 2 cents, see wut others have to suggest.

 

[CompuTronix]

Ambient = 21
Chipset: Intel 975X
CPU: E6600
Cooler: Scythe Infinity
CPU Frequency: 3.4GHz (378MHz FSB)
Load Test Program: TAT
Motherboard: Asus P5W DH Deluxe
Vcore: 1.34

Idle:

Tcase: 26C (SpeedFan)
Tjunction: 24C (SpeedFan); 27C (TAT)

Load:

Tcase: 51C (SpeedFan)
Tjunction: 56C (SpeedFan); 55C (TAT)

aoe, you have a rather odd combination of offsets, so here's what makes sense; since it's obvious that Tjunction Idle can not be near or below Tcase Idle, and Tcase Idle should be closer to Ambient on a high end cooler at moderate Vcore, then Tcase is offset by +5c, and Tjunction is offset by -5c. If we consider how this effects Load temps, then corrected temps would be as follows:

Idle:

Tcase: 21C (SpeedFan)
Tjunction: 29C (SpeedFan); 27C (TAT)

Load:

Tcase: 46C (SpeedFan)
Tjunction: 61C (SpeedFan); 60C (TAT)

As you can now see, Tcase + ~15c = Tjunction, so these temps are appropriate, and conform to specs and "Parameters". Use the "Offsets" section of the Guide to configure these corrections.

Hope this helps,

Comp 8)

 

[aoe]

Thanks for the suggestions. It was clear to me that I had to apply offsets to the temp readings in SpeedFan, and I have done some further analysis based on your proposals. I have made some additional data readings, as my ambient temps this morning are 18C. I am not sure how accurate this reading is, but it comes from the control unit for my central heating, and I have removed the unit and placed it next to the PC, so it's the best data I have. At this temp, I now have the following results (without offsets - all readings taken from SpeedFan):

Ambient = 18C

Tcase (idle) = 22C
Tjunction (idle) = 21C

Tcase (load - TAT) = 49C
Tjunction (load - TAT) = 53C

This gives me a delta between Tcase and Tjunction of -1C (idle) and 4C (load), providing a Tjunction/Tcase delta between idle and load of 5C (last night's data readings at a higher ambient are consistent with this). This seems to be causing the most problems, and suggests that the heat dissipation is better at lower temps, but I have no explanation for this.

Starting from our basic premises:

1. Tcase must be higher than Ambient (air cooling)
I think this is the most important and most irrefutable. From the data above, this suggests that the Tcase offset cannot be reduced by more than 3C (i.e. -3C), which would provide a Tcase of 1C above ambient. Taking an offset value of -3C is an assumption (the assumption being that the Tcase at idle is 1C above ambient). I can't prove this on the data available, and it may be higher, but I think we can agree it can't be lower.

2. Tcase should be ~15C below Tjunction
This is the one causing the problem, because my data doesn't support this. I would suggest that the best way to interpret this is that the delta should be ~15C on average between idle and load (although this again is an assumption). So, with a 5C delta between the Tcase and Tjunction delta at idle and load, and requiring an average Tjunction/Tcase delta of 15C, I will take a value of 13C as the difference between the Tcase and Tjunction offset values that need to be applied. This would then give me a required Tjunction offset of +10C.

Applying these offsets (-3/+10) to the above data would give me the following readings in SpeedFan:

Ambient = 18C

Tcase (idle) = 19C
Tjunction (idle) = 31C
Delta Tcase/Tjunction (idle) = 12C

Tcase (load - TAT) = 46C
Tjunction (load - TAT) = 63C
Delta Tcase/Tjunction (load - TAT) = 17C

Alternatively, we could assume that the max delta between Tjunction/Tcase should be 15C. This would then give me a difference between Tcase and Tjunction offset values of 11C, and require +8C offset to Tjunction.

Applying these offsets (-3/+8) to the above data would give me the following readings in SpeedFan:

Ambient = 18C

Tcase (idle) = 19C
Tjunction (idle) = 29C
Delta Tcase/Tjunction (idle) = 9C

Tcase (load - TAT) = 46C
Tjunction (load - TAT) = 61C
Delta Tcase/Tjunction (load - TAT) = 15C

I'd rather not try the 3rd option with the min delta between Tjunction/Tcase as 15C, as this would give me a Tjunction load temp of 66C (+13C offset to Tjunction).

All of these proposals leave me with the problem that TAT is also not reporting the Tjunction temps correctly:

Tjunction (idle) = 24C
Tjunction (load - TAT) = 52C

However, as stated in the guide, 'Any hardware and/or software may misreport Tcase and/or Tjunction temps'.

One other possible contributing factor is that I am running on the original BIOS supplied with the motherboard (1506). There have been several new revisions released since, and maybe one (or more) of them has a correction for sensor readings. However, I have been unable to find any detailed BIOS revision notes to back up this theory. I also read somewhere that revisions after 1506 removed the ability to adjust the multiplier for the CPU, so I am reluctant to change the BIOS (also considering that the system is currently running without problems).

I'd appreciate your thougths on the above analysis and assumptions.

Thanks again.

 

[CompuTronix]

It seems that you've acquired a solid understanding of the thermal relationships between Ambient, Tcase and Tjunction. :trophy: Just a few advanced pointers:

In a perfect world, Tcase Idle should not read at or below an accurate Ambient reading, however, remember to consider that the sensor accuracy is +/- 1c, and temp tables may not be linear, so with minimum Vcore and clock at Idle on a high end cooler with excellent computer case cooling, Tcase may read at, or even a degree or two below Ambient, as an exception to the parameters. The overall primary consideration for offsets are always Load temps. Additionally, offsets are typically in increments of 5. Also, keep in mind that the single most influencial factor on temperatures is Vcore.

I'm running 1305 BIOS because all the later revisions have "Stack Overflow" errors when running DOS utilities such as test programs and Norton Ghost, (which I use frequently and religiously for cloning my RAID 0). Additionally, 1305 has not required any offsets, so all temperature monitoring utilities work properly, and appear to be very accurate. I depend on this accuracy for solving temperature problems, since I'm continually altering my BIOS settings to duplicate the different configurations which users post. The 975 chipset is more mature than others, so I rarely see users with temperature problems on this platform. The #1 temperature offenders are the 965 boards.

I'm currently configured to your settings at 3.4 Ghz (378 X 9) with 1.3625 Vcore (BIOS), 1.35 Idle, 1.32 Load (SpeedFan). At 22c Ambient, Idle = 25c Tcase, 40c Tjunction. Load = 44 Tcase, 59 Tjunction (SpeedFan), 57 TAT.

Hope this helps,

Comp 8)

 

[bryan1010]

very detail.i understand.thank you.

 

[aoe]

Hope you don't mind me ressurecting this thread

I've recently switched cases (poorer airflow unfortunately), and as we're fast approaching summer, the ambient temps are also increasing. Consequently, my PC temps are increasing, and I'm trying to resolve some minor stability issues.

I'm working on improving the airflow in the case, and also looking to improve the chipset cooling (have some Thermalright HR-05 heatsinks ready to install). Got some additional fans on order also.

What I need to do is verify my temps, and this is where I am struggling. I'm getting various readings from different software and cannot establish which is correct.

Firstly, my specs again (changed cpu heatsink and PSU):

Chipset: Intel 975X
CPU: E6600 (B2) L639A265
Cooler: Thermalright Ultra 120 Extreme (fitted with supplied TIM)
PSU: Seasonic M12-700W
CPU Frequency: 3.4GHz (378MHz FSB)
Load Test Program: TAT
Motherboard: Asus P5W DH Deluxe (BIOS 1506)
Vcore: set to 1.3625 in BIOS; reporting 1.34 in BIOS; reporting 1.34 idle and 1.31 load with SpeedFan 4.32


Ambient = 24C


Temps [idle]

tCase readings:
PCProbeII = 30C [frequently issues warning with temp 255C - problems either reading sensor or with software - will investigate further]
PC Wizard 2007 = 34C [something wrong with this reading - this is probably SB reading - see below]
BIOS = 34C
SpeedFan 4.32 = 30C (without offset)

tJunction:
TAT = 33C
CoreTemp 0.95 = 31C
PC Wizard 2007 = 31C
SpeedFan = 31C (without offset)

System [SB]
SpeedFan 4.32 = 35C

Aux [no idea where this sensor is reading from]
SpeedFan 4.32 = 14C (clearly an offset is required here)


Temps [Load using TAT]:

tCase
PCProbeII = 55C [frequently issues warning with temp 255C - problems either reading sensor or with software - will investigate further]
PC Wizard 2007 = 35C [something wrong with this reading - this is probably SB reading - see below]
BIOS = No idea
SpeedFan 4.32 = 55C (without offset)

tJunction
TAT = 59C
CoreTemp 0.95 = 60C
PC Wizard 2007 = 59C
SpeedFan = 60C (without offset)

System [SB]
SpeedFan 4.32 = 36C

Aux [no idea where this sensor is reading from]
SpeedFan 4.32 = 29C


PC Wizard 2007 was reporting the following:

Mainboard = 31C (idle) / 54.5C (load)

Power/Aux = 13.5C (idle) / 28.5C (load)


I think PC Wizard is mixing the Mainboard and CPU temps, so I would take Mainboard values as tCase, and tCase values as System readings.

As I'm writing this, I am now getting the following readings at idle for tCase:

PCProbeII = 67C
PC Wizard 2007 (Mainboard reading) = 67.5C [something wrong with this reading - this is probably SB reading - see below]
SpeedFan 4.32 = 68C (without offset)

These readings are clearly incorrect as my tJunction readings remain around 30C on all apps. This is the 1st time I've noticed this behaviour.

I feel that the tCase sensor cannot be trusted. The heatsink feels firmly attached and the heatpipes are cool to the touch.

A couple more things I noticed whilst replacing the heatsink (which I did a couple of times):

The heatspreader on the cpu may be very slightly convex at the centre looking at the TIM pattern after removing the previous heatsink. I don't think this is serious, as the pattern suggested the complete heatspreader surface was still in contact with the heatsink.

The Thermalright Ultra 120 Extreme contact surface has not been finished well. One of the edges is clearly not level (reflected light distortion). However, I am fortunate that this part of the surface is not in contact with the cpu heatspreader.

The only consistent temp readings I see are the tJunction readings. With the case open, temps drop by around 2-3C, so an improvement in airflow is definitely required (I'll work on that).

To summarise, I have the following questions:

1. Would you agree that there is most likely a problem with the tCase sensor readings?

2. Can I rely on the tJunction readings alone?

3. There is a 30C difference between idle and load temps (tJunction) - am I pushing this system too hard? (recommendation in OC guide is 25C delta)

4. Any clues as to where the Aux sensor is reading from?

Thx in advance for any responses.

 

[disposablehero]

if im understanding your post correctly, im having a similar issue. unfortunately, it hasnt been responded to.

Tj is correct and ambient is correct, but Tc is not inline with the temp guide (~5 degrees lower than Tj, instead of ~15).

http://forumz.tomshardware.com/hardware/Help-sorting-correct-temp-readings-ftopict236415.html

for now, im running under the assumption that Tj is the most critical to keep in "safe" territory.