whats more accurate, core temp or bios temp?

[rothchilds]

just swapped out my e4300 for an e8400. Running it at 3.6 at stock volts. I had the e4300 at 3.0 with voltage up to 1.2875.

core temp shows my e8400 at 50c idle, which seems pretty unreal. the e4300 was around 38c idle in core temp. Going into my bios, gigabyte p35, temp shows 29c. Why is there such a discrepency, and which is more accurate? The heatsink is definately on there solid, used AS5 as I always do.

 

[Rubycon]

Something is amiss as the BIOS temp is usually higher than the idle temp in Windows.

 

[bryanW1995]

what's your distance to tjunction? if it's only about 50c then you need to reseat your hsf. also, what's it at load? you might have a stuck diode, which seems to be very common on these 45nm cpus.

 

[rothchilds]

goes up to 58 on orthos load. Maybe the heatsink just isnt on quite right. I'll pull it apart again tommorow I suppose

 

[rothchilds]

reseated/re-applied AS5 and heatsink, still core temp shows idle at 49 (using a coolermaster cooler, not stock)

Bios still shows in the 29 range. Are there other temp montoring programs I can use to compare, maybe core temp is just borked for me on this one?

 

[Foxery]

Load temperatures are more informative, and you can't load the system from the BIOS.

What exactly is your voltage set to? 3.6ghz on "stock" seems like a stretch.

 

[rge]

Bios temp is cpu temp, which is the diode between the cores, and if you are at stock settings that is about right at 29C bios (which is a small load), assuming your ambient is average 23-26C range. I have same E8400 and GB P35 DQ6, and bios does good job of calibrating cpu temp, so that is more accurate currently on yours for absolute temps.

Coretemp, everest, and HWM all use tjmax of 105 (mobile 45nm tjmaxes for desktops cpus) which is 10C too high, and thus your core temps will be reported 10C too high. For the 45nm chips I would go to http://www.xtremesystems.org/f...howthread.php?t=179044 and download realtemp, your load temps will be then be more accurate and idle at least better at 39C.

Also, all DTS on cores are only accurate and linear once you reach about 35 units/C from tjmax, ie load temps. So even realtemp though more accurate at load, will still read too high at idle. Either ignore the idle temps on core and use cpu temps, or you can calibrate cores using real temp.

Because cpu temps are dependent on bios calibration, and some mobo's either have not yet updated and supported e8400's or have done a poor calibration, cpu temps sometimes are crazy off and hence get a bad rap.
However, if cpu temp has been properly calibrated, and GB seems to do a good job, they are going to be more accurate in idle range, given all the DTS sensor problems on 45nm and the pronounced nonlinear/inaccurate readings at idle.

 

[rothchilds]

Originally posted by: rge
Bios temp is cpu temp, which is the diode between the cores, and if you are at stock settings that is about right at 29C bios (which is a small load), assuming your ambient is average 23-26C range. I have same E8400 and GB P35 DQ6, and bios does good job of calibrating cpu temp, so that is more accurate currently on yours for absolute temps.

Coretemp, everest, and HWM all use tjmax of 105 (mobile 45nm tjmaxes for desktops cpus) which is 10C too high, and thus your core temps will be reported 10C too high. For the 45nm chips I would go to http://www.xtremesystems.org/f...howthread.php?t=179044 and download realtemp, your load temps will be then be more accurate and idle at least better at 39C.

Also, all DTS on cores are only accurate and linear once you reach about 35 units/C from tjmax, ie load temps. So even realtemp though more accurate at load, will still read too high at idle. Either ignore the idle temps on core and use cpu temps, or you can calibrate cores using real temp.

Because cpu temps are dependent on bios calibration, and some mobo's either have not yet updated and supported e8400's or have done a poor calibration, cpu temps sometimes are crazy off and hence get a bad rap.
However, if cpu temp has been properly calibrated, and GB seems to do a good job, they are going to be more accurate in idle range, given all the DTS sensor problems on 45nm and the pronounced nonlinear/inaccurate readings at idle.

awesome info thanks!

 

[rothchilds]

Originally posted by: Foxery
Load temperatures are more informative, and you can't load the system from the BIOS.

What exactly is your voltage set to? 3.6ghz on "stock" seems like a stretch.

my voltage is set to normal, aka stock rating for that chip, also verfied in cpuz. It's not a stretch with the e8400 to go from 3.0 to 3.6 on stock volts. All I had to do was change the fsb to 400 and set ram to 2:2. These chips are beasts.

 

[MKvolunteer]

[/quote] my voltage is set to normal, aka stock rating for that chip, also verfied in cpuz. It's not a stretch with the e8400 to go from 3.0 to 3.6 on stock volts. All I had to do was change the fsb to 400 and set ram to 2:2. These chips are beasts.[/quote]

Exactly! At list one I have is

____________________________________________
Intel E8400 @ 4.05 GHz (450x9), 1.295v, w/EK, 3 x rad
Asus P5E3 X48 Premium
2GB Corsair Dominator 1800 MHz DDR3
NV 8800 GT 512MB
Samsung 1TB SP F1

 

[taihuia13]

Latest speedfan will show cpu temp ( bios ) and core temps . Core temps same as core temp. if not offsets can be applied.

 

[lenjack]

Also, keep in mind that CoreTemp is with your operating system running, while Bios temp is before the OS loads.

 

[Billy Idol]

Not to drag up an old thread but there was some decent info here and I thought I could tag on to the end.

I just built a rig (e8400, DS3L (e version), Xigmatech S1283) and my idle temps are also high. After reading about the base of the HDT coolers I think my method of a single dab of TIM in the center of the CPU was less than ideal. That said which temps do I believe? In BIOS after having the system off it reads the case at 29C and CPU 24C (are these reversed?). Which seems pretty reasonable to me considering the room is probably in the area of 22-3C. Once in Windows a couple minutes later RealTemp reads 37C, Speedfan 42C, and CoreTemp 47C. Gigabyte's EasyTune 5 software reports 21C CPU (lol) and 32C case. These numbers are all at 3.0 with C1E and EIST both enabled. After 15mins of Orthos: RealTemp 46, Speedfan 51, CoreTemp 56.

What do I believe? I think I've decided to reseat / lap the cooler regardless but which one do I go by? I haven't tried overclocking because of my concerns. What should I do next? Overclock anyway and see what the temps do? Reapply the paste and seat it again? Play some games and enjoy it at 3.0? Honestly the performance in the current state is plenty fine but I'd rather not run it hot needlessly.. especially during the summer. Oh yeah.. VID is 1.225v :/

 

[ShadowFlareX]

I was using Core Temp too and it read my Q9450 with an average of 56oC across all 4 cores at idle, 10oC higher than the Q6600 it replaced even at load. Reseated the HSF didn't make a difference, and i noticed the HSF was barely even warm, it was warmer with Q6600.

BIOS temp for Q9450 was showing cooler CPU Temp vs Q6600, that's when I finally tried Real Temp and convinced of its 10oC lower temps. In the end, I still used Core Temp but reduced the TJMax to 95oC just like Real Temp's.

 

[bryanW1995]

95c is most likely the correct tjmax for q6600 and q9450 both. I don't know much about the 45nm dualies, but I suspect that they're about the same. Fortunately for all of us, realtemp allows you to set it to display distance to tjmax. This completely removes the guesswork from temperature monitoring programs. keep your distance to tjmax at 25c+ and you'll have absolutely no problems. I've heard markfw900 say that one of his quads got as high as 77c (18c to tjmax) and never displayed any problems, but that territory is getting a little dangerous for the average quad user imho.

 

[rge]

Originally posted by: Billy Idol
Not to drag up an old thread but there was some decent info here and I thought I could tag on to the end.

I just built a rig (e8400, DS3L (e version), Xigmatech S1283) and my idle temps are also high. After reading about the base of the HDT coolers I think my method of a single dab of TIM in the center of the CPU was less than ideal. That said which temps do I believe? In BIOS after having the system off it reads the case at 29C and CPU 24C (are these reversed?). Which seems pretty reasonable to me considering the room is probably in the area of 22-3C. Once in Windows a couple minutes later RealTemp reads 37C, Speedfan 42C, and CoreTemp 47C. Gigabyte's EasyTune 5 software reports 21C CPU (lol) and 32C case. These numbers are all at 3.0 with C1E and EIST both enabled. After 15mins of Orthos: RealTemp 46, Speedfan 51, CoreTemp 56.

What do I believe? I think I've decided to reseat / lap the cooler regardless but which one do I go by? I haven't tried overclocking because of my concerns. What should I do next? Overclock anyway and see what the temps do? Reapply the paste and seat it again? Play some games and enjoy it at 3.0? Honestly the performance in the current state is plenty fine but I'd rather not run it hot needlessly.. especially during the summer. Oh yeah.. VID is 1.225v :/

When you are in bios, cpu temp (diode between cores of E8400) is under a very small load, and 24C is reasonable given 22C ambient. The "case" temp is either mislabled or GB installed a cpu socket sensor; just out of curiosity does it say "case" temp in bios or system temp (system temp is NB)? Regardless it is not actually measuring Tcase, as no true Tcase sensor exists.

When in windows, the cpu temp will drop couple-few degrees C, because the bios is a slight cpu load compared to idle in windows. In other words bios cpu temp and Easy tune are reading same diode, just bios is under slight load.

E8400 tjmax is not 105 as coretemp uses, you can show that mathematically, see post here
http://www.xtremesystems.org/f...=2962586&postcount=916

Or if you do what bryanW1995 suggested, which makes the program you use irrelevant as delta to tjmax is accurate on all, you are 49 from Tjmax on load, so regardless you have plenty of headroom to overclock. Not only are you guaranteed accuracy by doing this, but leaving 25C from tjmax (max possible gradient Tcase to core at max TDP) guarantees you are always within Tcase max specs.