Is 55C too hot under load for an E8400?

[moose004]

I've just built my first new system in five years:

Core 2 Duo E8400
Arctic Cooling Freezer 7 Pro
Gigabyte GA-P35-DS3L
4 GB G.Skill DDR2/800
Antec Sonata 550 Plus case (Neo 550 power supply), currently no extra fans
Seagate Barracuda 7200.11

I am a little concerned because my E8400 idles, according to RealTemp and Speedfan, at 35C core, and pushes a steady 54-55C core when running Prime95 torture tests. And this is BEFORE I overclock it--it's running at the stock 9x333. I've tried getting a little more aggressive with the Gigabyte auto-fan timings, so that at 55C the fan on the FP7 is running about 1200 rpm (I think this is about 50% speed). I haven't tried the fan at 100% yet, but I might.

Now granted, the Sonata is not the best airflow case out there. And I do not have any front fans installed yet, they're coming in today, and I'm planning to install at least one 92mm fan in the drive bay. But I'm really wondering what the problem is, or even if there is one. 35 and 55 seem awful high from what I've read about the E8400 and the Freezer Pro 7. To make matters worse, Coretemp reports them 10 degrees higher, and Realtemp reports them five degrees LOWER, than what Speedfan does. So now I'm totally confused--I don't know exactly what my temps are!

I've read the sticky on C2D temperatures, and to be honest I don't understand half of it. I know roughly the difference between "Tcase" and "Tjunction", but the article says my difference should be no more than five degrees. In Speedfan, it's 13 to 15; when my core temps are 53-55, the corresponding "Temp1" or "Temp2" is 13 or so degrees lower (55C core, 42C case).

The whole system is working well and I'm not too worried about normal use, but I was hoping to be able to push 3.6GHz out of this puppy with a simple FSB tweak. Now I'm not so sure I want to try it. Does anybody have any advice?

 

[CompuTronix]

The CPU temperature, as well as the Core temperatures are "canned" values, and are seldom accurate out of the box. Don't get stuck on this. If you were to do the calibrations, then your Tcase to Tjunction Delta will be 5c at Load. Also, calibrations and temperatures are only valid with case covers removed, and all fans at 100% RPM, which is shown very clearly in the standardized "Test Setup". Since you don't mention your ambient, please measure it and check your temps again by following the Test Setup to the letter, so that you're not trying to compare apples to oranges.

Since the release of Real Temp, I don't recommend Core Temp to anyone. Although Real Temp will display more accurate Core temperatures, SpeedFan will give the best accuracy for CPU temperature AND Core temperatures only when calibrated to your particular rig by using the Temp Guide. If you don't want to do the calibrations or feel that it's over your head, then just use Real Temp.

Comp

 

[closed_deal]

The 45Nm chips are renowned for the DTS to be u/s. However, i'd say the temps your getting atm with bad airflow are about right. (*Edit* to keep CT happy ) "Tcase" 72C load is the max for 45Nm. Mine currently idles at 42C @ 4.0Ghz and 63C Load with a very minor vcore increase.

If i were you just increase the FSB to 400Mhz see how stable it is, I doubt temps will raise too much, only 5C maybe 10C tops.

 

[CompuTronix]

Let's be clear about temperatures; as I've said many times, there's a tendency for users to fling temperatures around like so much gorilla poo. This is very misleading and confusing, so it's important to be specific about ambient (room temperature), Tcase (CPU temperature) and Tjunction (Core temperatures).

Intel's thermal specification, which is the value shown in their Processor Spec Finder, is for CPU temperature, and NOT Core temperatures. When users don't observe and notate these simple temperature protocols, then the discussion is quickly reduced to apples and oranges thermal fruit salad in a blender. This is precisely how confusion over temperatures continues to spread through the community, and is how users continue to become confused.

Comp

 

[bimmerbill]

I'll have the same MOBO, CPU and HSF up and running as soon as my stuff is delivered. I have a different case, HDD and PSU tho.

 

[ocguy31]

I dont think anyone is confused by his question. The OP asked if 55C load is too high for the cores of a E8400. The answer: Not at all.

 

[CompuTronix]

ocguy31, that detail has not escaped me, however, I posted that explanation for closed_deal, as well as for the benefit of others.

Comp

 

[moose004]

OK. For now, I'll roll with it, install one of the 92mm fans in the front case (to cool my Barracuda if nothing else), push the throttle up on the FP7 fan, and do a full temperature calibration using Speedfan before attempting to shove the FSB up to 400 MHz. I got a bit worried because I was reading stuff in the FP7 reviews on Newegg about people saying "zomg my E8400 is running 4 GHz and it never goes over 45C!!!1!", and here I am sitting bone-stock and running somewhere between 50 and 65C depending on which utility I use to check it.

Don't get me wrong, even at 3.0 GHz this thing is literally 10x faster than what it replaced (an old Athlon XP 2500+ on an Asus A7N8X, with an BFG GF6600GT AGP card). But one of the reasons I went with the E8400 was to have a little headroom and be able to go longer between big upgrades.

Thank you all very much. This is definitely a very helpful place!

 

[CompuTronix]

After you complete the calibrations, you'll find that your CPU temperature and Core temperatures are quite reasonable, and will provide you with the overclocking headroom that you're expecting.

Comp

 

[Shadow703793]

Listen to what Comp said and you will get quite an accurate reading.

@OP: You should have gone with the Xigmatek HDT-S1283 for ~$15 more, it's a great cooler.

 

[kad]

moose004:
Do not get frustrated
You're not the only one suffering from temps
It seems E8xxx processors have temps issues
I have E8500 and was hoping to run at 4 GHz
Armed with Ultra 120 Extreme HS lapped + Sythe 120 Fan+AS5
Ambient temp 25
Core Temps At Stock speed idles 43 with no load
Under prime95 will reach 53
OC to 3.6 (9X400) under Prime95 Temps 63
OC to 3.8 (9.5X400) under Prime95 Temps 68
Temps are equally reported by Core Temp Version 0.99.0.3 and HWMonitor
RealTemp is always reporting 10 degrees less and I do not trust it
SpeedFan reporting more than 112 (I ignored it)
Have LianLi PC-A70 Case (2X120-1500 RPM-63CFM) Front fans and another same fans at rear in additin to the PSU 80MM fan
Temp inside the case equals ambient 25, so there is no problem with airflow
CPU voltage at stock 1.18
At 3.6 I set it in Bios to 1.2 but CPUZ and HWMonitor both say 1.18V
At 3.8 It needs 1.22 in Bios But CPUZ And HWMonitor will say1.2V

 

[fatty35]

Alot of The new 45nm chips have this problem & is going to eventuly be fixed via bios .etc.

 

[CompuTronix]

kad it seems that you have some homework to do in order to get up to speed on the temperature topic. Please read my Core 2 Quad and Duo Temperature Guide: http://www.tomshardware.com/forum/221745-29-core-quad-temperature-guide

Also read the documentation for Real Temp: http://www.techpowerup.com/realtemp/docs.php

Real Temp is the most accurate Core temperature monitoring utility currently available. When you read the documentation, you'll understand how Real Temp addresses sensor linearity problems that other popular utilities do not, and why the temperatures are correct at 10c cooler.

Also, my Guide explains your issues with SpeedFan, and Section 11 has the information to correct them. If you want your temperatures to be as accurate as possible, then SpeedFan can be used with my Guide to calibrate CPU temperature AND Core temperatures.


From the Guide:

Section 8: Tools

Note 2: SpeedFan - Very flexible and configurable, SpeedFan is the preferred temperature monitoring utility because Tcase and Tjunction can be Calibrated. SpeedFan detects and labels thermal sensors according to various motherboard, chipset and super I/O chip configurations, so the label for Tcase can be CPU, Temp 1, Temp 2, or Temp 3. Even if Tcase is labeled as CPU, it is still necessary to confirm the identity of Tcase prior to performing Calibrations.

■Repeatedly start and stop Prime95 Small FFT`s at 15 second intervals, while observing which SpeedFan temperature scales with an Idle to Load Delta similar to the Cores. This will identify the label corresponding to Tcase. Labels can later be renamed using the Configure button. See Section 11.

If a temperature shows a flame icon, this indicates alarm limits which require adjustment. Use the Configure button to set CPU and Core temp alarms to Warm Scale. If a temperature shows Aux 127, this is simply an unassigned input which can be disabled using the Configure button. See Section 11.


fatty35, the 45 nanometer temperature problem has nothing to do with BIOS. The problem is caused by faulty Digital Thermal Sensors (DTS). From the Guide:


Section 5: Findings

(A) Tcase is acquired on the CPU Die from the CPU Case Thermal Diode as an analog level, which is converted to a digital value by the super I/O chip on the motherboard. The digital value is BIOS Calibrated and displayed by temperature software. BIOS Calibration affects the accuracy of Tcase, or CPU temperature.

(B) Tjunction is acquired within the Cores from Thermal Diodes as analog levels, which are converted to digital values by the Digital Thermal Sensors (DTS) within each Core. The digital values are Factory Calibrated and displayed by temperature software. Factory Calibration affects the accuracy of Tjunction, or Core temperatures.

(C) Tcase and Tjunction are both acquired from Thermal Diodes. Tcase and Tjunction analog to digital (A to D) conversions are executed by separate devices in different locations. BIOS Calibrations from motherboard manufacturers, Factory Calibrations from Intel, and popular temperature utilities are frequently inaccurate.


I hope this helps to clear up any misconceptions.

Comp

 

[kpo6969]

Moose004
I have the almost identical set-up as you (except I have the Sonata III case with the 120mm front fan also).
stock 32c/53c
3.6ghz 37c/58c
Everest and CoreTemp TjMax adjusted to 95c

 

[kad]

CompuTroni x :

Thanks for the detailed answer
But this is a lot of homework to do
As a summary I can trust real temp

As I understand from techpowerup article
RealTemp can be calibrated also
The risk in calibrating is if someone got a faulty proc running at high temps and calibrated to his taste and this my worry for my proc
The article mentions the difference should be 5degrees between ambient temp and proc temp (It is not clear CPU temp or CORES temps) for 65n proc at idle with min core voltage, clock frequency and FSB bus

What do you think ??

 

[CompuTronix]

kad, prior to the introduction of Real Temp a few months ago, the only popular monitoring utility that could calibrate temperatures was SpeedFan. Since the developers of other temperature monitoring utilities have realized that Real Temp is more accurate, they've only recently begun to update their utilities to include calibrations.

Even so, the other monitoring utilities have a single calibration point, while Real Temp has 2 calibration points. The hotly debated and model specific Tjunction Max values, which Intel won't release, is the "guesstimated" calibration point used by utility developers to establish Core temperatures. Only Real Temp has a separate Idle calibration, which resolves sensor linearity problems. This is the feature that all the other temperature monitoring utilities do not have, and is what makes Real Temp unique.

Since Real Temp measures Core temperatures only, the approximate 5c difference at Idle that you questioned, is between Ambient temperature and Core temperature (Tjunction). SpeedFan can calibrate CPU temperature AND Core temperatures. Since Intel specifies and supports CPU temperature only, and CPU temperature and Core temperatures are always relative to one another, I have always featured and supported SpeedFan in my Temp Guide as the best utility to provide the best of both worlds, because both sets of temperatures can be used to cross check one another.

Although the test setup Real Temp uses is essentially the same as I use in my Guide, the calibration techniques I developed also use 2 calibration points, but differ from Real Temp in that I account for processor and cooler variables to calibrate CPU temperature (Tcase) at Idle, then calibrate Core temperatures (Tjunction) at Load, which closely replicates Intel's test methods. The typical result is that SpeedFan's CPU temperature is accurate from Idle to Load which satisfies Intel's thermal specifications, while SpeedFan's Core temperatures at Load will be just a few degrees higher than Real Temp, but a just a few degrees lower than the very popular utility, Core Temp.

Does this help to put the issues into perspective for you?

Comp

 

[kad]

SpeedFan's Core temperatures at
Load will be just a few degrees higher than Real Temp,
but a just a few degrees lower than the very popular utility, Core Temp.

Comp

So you mean the results of Calibrated SpeedFan are in the middle between
High results of Core Temp and Low results of Real temp, right ??

Did you ever compared the results of your calibrated SpeedFan
with results of Real Temp at idle ??
What was the result ??

Now SpeedFan4.34 is available and I think it's better than
previous versions but still needs translation to english

Of course I did not understand anything but the only
benefitial thing is no number is high
It was surprising to see some temps -48 as I do not live in North Pole!!!
Wondering why there are 2 CPU one with 48C and the
other 36 C and trying to undertand
I clicked on Configure trying to understand something

and another list of mysterious things appeared
The good thing I can understand is temps of cores

Sorry man But I need a simple program tht dosen't need reprogramming
This is the most frustrating program/utility I ever met after Win Vista
and Creative drivers and softwares
Your efforts are highly appreciated
Many thanks

 

[pcnazz]

nice quick run - down on setting up speed fan , thank you , I bought BE 6400 from Tiger Direct today and was bouncing around picking up some #'s for OC-ing it and bumped into this , I never adjusted the temps on Speed Fan , much better , BE 6400 was $163.99 free shipping to Boston Ma , I had TD match NewEgg on the shipping and they did , also got ZALMAN 9500 to cool it from NewEgg , might save someone a $ , thanks again !

 

[Reynod]

Intel needs a kick in the pants over this issue ...

It just isn't good enough.

How many 45nm temperature related posts is this now??

Cripes ... no wonder so many are sticking to the Conroes / Kenties - might be a bit hotter to run but at least the FSB is lower so it is easier to OC on standard gear.

Plus the temps read reasonably correct straight out of the box.

This whole issue is freaking out the n00bs ... and I don't blame them.

This deserves an article all on its own ... testing the temps on the Penryn's across a heap of mobos.

Or is THG still hell bent on bagging just AMD all of the time??

Tom would have gotten his teeth into this and made it front page news ...

 

[moose004]

Just thought I'd follow up...I did a calibration cycle according to CT's C2D Temp Guide, and in the end, RealTemp appears to have been within 1C of correct the entire time. With my covers off and my AC FP7 running wide open, I got 34C core at idle, and 47C core running Prime95. I have the fan back to a more conservative sliding scale using EasyTune 5 which runs it at 1200 rpm (60%) at idle, and I get 35C core, 30C Tcase. Under Prime95 small FFTs, it ramps the fan to 2000 rpm and gives me 48C on the cores, 44C on Tcase.

So my question is, are those temps too high at 3.0 GHz, considering I'll probably have to do voltage tweaks to run 3.6? I already tried 9x400 without voltage tweaks and Prime95 failed immediately...but I don't know whether I'll have to up the Vcore or just the voltage to the memory.

Thanks!

 

[CompuTronix]

moose004, thank you for the update, and for vailidating the accuracy of my Temp Guide / SpeedFan calibrations of 1c compared with Real Temp.

Your temps at 3.0Ghz for those fan settings are normal, but you're correct to anticipate that you'll have to increase Vcore to reach 3.6Ghz. Vdimm should be OK at stock settings.

Comp

 

[kad]

Just thought I'd follow up...I did a calibration cycle according to CT's C2D Temp Guide, and in the end, RealTemp appears to have been within 1C of correct the entire time. With my covers off and my AC FP7 running wide open, I got 34C core at idle, and 47C core running Prime95. I have the fan back to a more conservative sliding scale using EasyTune 5 which runs it at 1200 rpm (60%) at idle, and I get 35C core, 30C Tcase. Under Prime95 small FFTs, it ramps the fan to 2000 rpm and gives me 48C on the cores, 44C on Tcase.

So my question is, are those temps too high at 3.0 GHz, considering I'll probably have to do voltage tweaks to run 3.6? I already tried 9x400 without voltage tweaks and Prime95 failed immediately...but I don't know whether I'll have to up the Vcore or just the voltage to the memory.

Thanks!

Your temps are normal for E8xxx processoes
For sure you need to increase core voltage
If you have ( AUTO ) option for Vcore voltage and North Bridge
set both to auto and increase FSB to 400
If have an option in Bios named Loadline Calibration set it to enabled (It will reduce voltage drop)
Set also C1E and Speed CPU to disabled
Reboot and see the reported voltages of Core voltage and N.Bridge
Now reboot and in Bios set Core voltage to a value more than reportd by 0.02V
Also set N.Bridge voltage to a value near 1.4
i.e if reported core voltage in or CPU-Z is 1.18 change it in Bios from AUTO to 1.2
Test under Prime 95, if failed increase one notch and retest
And so on
Waiting your results

 

[bimmerbill]

I have the same CPU and heat sink and get almost the same temps as you.
Coretemp and Realtemp are exactly 10C different tho.
Realtemp shows 46/47C full load with Prime, while Coretemp shows 56/57C.

I set mine at 9x400 last night and let Prime run for a while. No issues at 3.6G. I didn't increase any voltages yet. I will Prime it a while longer over the weekend to see what happens.

I didn't get any higher temps with the OC than I did stock.