Bios temp vs core temp

[thegooddieyoung]

I recently built my sandy bridge system and all seems well with two exceptions

one question is regarding raid, which I asked in another forum here

the reason for this post is a question about reported cpu temps on my non overclocked 2600k.

My msi p67a-gd65 motherboard with 'clickbios' is showing a 'cpu temperature' of as high as 60c and never lower than 51c - these temps are reported after a cold boot, first thing in the morning.

'core temp' is reporting temps of 30,31,31,31

This is with the stock intel sink/fan combo and ambient room temp of 20c

 

[Leaps-from-Shadows]

Your CPU is still loaded (active) in the BIOS, meaning it can't down-clock to idle. Once it gets to Windows, it can down-clock to 1.6GHz and give you the 30sºC. Plus, Core Temp reads the digital temp sensors in each individual core, so is more accurate.

 

[thegooddieyoung]

thanks for the quick reply.

Am I to understand that these 60c+ temps in bios are nothing to be concerned about and I should not suspect that my H/S mounting needs some attention?

 

[Leaps-from-Shadows]

Correct. The BIOS is reading from a different sensor that is not always reliable or accurate. Keep your individual core temps below 75ºC when loaded.

 

[RetiredChief]

^ concur, that the readings are from diff sensors. But I find that Bios reading is generally more accurate than 3rd party software. Also (For Older system (E6400 and I5-750) my Bios readings are very close to Idle temps, Not equal to individual core temps, but close), Not sure about for SB, But 60 C in Bios is either Inaccurate, or High. Maybe some one who has the same setup could chime in. I could be wrong on this, but I thought that Processor speed (Speedtep) was determined at the BIOS/CPU (Based On CPU utilization). CPU utilization should be low by the time you get to the Health page in BIOS.

As Leaps-from-shadows indicated - Should check temps under load, ie run Prime95 while monitoring temps. Might also try HWMonitor to see if both programs aggree.

 

[thegooddieyoung]

I assumed that cpu useage would be low by the time I got to the setup screen in bios as well.

I was actually watch the cpu temp in bios go from 50-60c just by sitting at the bios screen. I'm really not convinced this is 'normal' behavior.

I ran 3d mark vantage this morning with core temp and saw max temps between 60-65c across all the cores during the run

 

[thegooddieyoung]

At your suggestion, I just ran prime85 thru 4 loops and core temp reported temps as high as 78c across two cores, the other cores maxxed at 75c

 

[RetiredChief]

I would suspect the Stock HSF. A number of people have problems with this. Mostly because the Push pins are not seated well, or thermo paste not properly applied. That said, the Intel HSF is on the bottom of the totom pole interms of performance. On my last two builds I just ordered a 3rd party HSF and did not even try the stock HSF. The stock HSF should have acceptable idle temps and unted load they will be higher than most 3rd party HSF, but still acceptable - I just consider them JUNK.
I prefer one with a backing plate.

 

[Leaps-from-Shadows]

Sorry guys, but Speedstep doesn't activate until the system is fully booted up and into Windows. The CPU cores are still pretty loaded in the BIOS (I have no idea why, but they are), and increasing temps while in the BIOS are normal. AMD's Cool N Quiet works the same way.

The CPU alone decides when to idle, when to do regular speed, and when to go full Turbo. The BIOS can turn the switch for that adjustment on or off, but it's actually controlled by a special power-management area of the CPU.

 

[Leaps-from-Shadows]

Yeah, with those temps I wouldn't overclock at all until you get an aftermarket cooler. The $30 CoolerMaster Hyper 212 Plus would be a good start.

 

[thegooddieyoung]

I've done some more thinking on this and here is what I've decided

My case is a cooler master scout http://www.newegg.com/Product/Product.aspx?Item=N82E16811119196

This case has a 140mm intake at the bottom front of the case, a 120mm exhaust on the top of the case and a 120mm exhaust at the top rear. Seems fine in practice

However, directly behind the 140mm, I have a bank of 4 hard drives - at the same level I have a 6850 video card.

The bank of hard drives certainly reduce the flow and then what air flow makes it past them is totally blocked from the cpu area by the long video card.

The case has two positions in the side wall for 120mm fans. I'm going to order two fans and an aftermarket HS/fan today. I'll set the fans for intake. One will blow directly onto the cpu and the other will blow onto the gpu.

I'll report back with new numbers in a few days. I'll test first with just the additional fans to see how much difference that makes, then more numbers with the aftermarket cooler.

Any recommendations for a good air cooler ? I bought a MASSIVE cooler for my q6600 a few years back, it had a 120 or 140mm fan that attached to the HS and blew air arcoss it. Cant remember the name, but it was highly regarded at the time. My case door actually touched the tips of the copper heat tubes.

 

[RetiredChief]

@ Leaps-from-Shadows.I'm surprised that there would be that much diff between i5-750 and i5-2500k as to how it operates while in Bios. My I5-750 Set to 3.2 is @ 30 C in bios.

Started to look at some posts/reviews on the SB systems. Hard to find posts that state CPU temp in Bios, but did find one stating 55 C. BUT also found:
Ref: http://discuss.extremetech.com/forums/thread/1004443177.aspx
Screen shot of Bios, P67A-UD3, HSFLooked up the Mega Shadow deluxe CPU temp = 29C
And this:
http://forums.anandtech.com/showthread.php?t=2141757
Quote: "For reference, my overclocked 2600k shows ~40C in BIOS and ~30C idling in Windows. Did you apply any thermal paste on your own? The stock cooler comes with it preapplied I think."

Any basis for "high CPU utilization when in BIOS" SHould be lower than when in windows @ idle which is normally 10% or less. Reason I ask this is that when in bios you have all the loops running that monitor the Health functions plus I'm sure there are others, BUT all of these are still there when in windows @ idle + the some 50 programs windows starts running by default.

Not faulting your comments, just questioning, as I have not gone to 2500K yet, Waiting for the Z68 MBs (plus grace period to iron out/find any bugs like in the first ed of P/H68 MBs

Edited - added:
looked up the Mega Shadow deluxe - High priced, poor performance - Hyper 212 better at half the cost.

 

[thegooddieyoung]

removed, cleaned and reseated stock HS with artic silver 5. If anything temps slightly higher.

Installed 120mm xigmatek fan in my case door directly over the cpu. This fan is moving ALOT of air. Noisy as hell but pumping so much air into the case it will hold a sheet of notebook paper over the intake with vaccum.

MB only dropped the speed on the big fan BRIEFLY, it's constantly running it at full speed and idle temps in windows via CORE TEMP are 35c on all cores as I type this.

 

[Leaps-from-Shadows]

I've had the higher BIOS temp behavior in at least my last four systems: one Sandy Bridge, one Athlon II X4, one original Phenom quad, and one original Phenom triple. It may go back farther, but I don't recall off-hand.

The CPU cannot go into low-power mode until fully loads and hits Windows. The BIOS is at least lightly loading it. I've read this in reviews before, and it's confirmed by my experience at least. If you're using the stock cooler, temps will slowly increase in the BIOS, up to a certain point of course.

If you're using a good aftermarket cooler, the difference between BIOS and Windows temps may not be as noticeable. Maybe that's where the confusion is.

 

[RetiredChief]

Quote
If you're using a good aftermarket cooler, the difference between BIOS and Windows temps may not be as noticeable. Maybe that's where the confusion is.
End quote

That may be the answer. None of my past 3 builds showed a higher temp in bios vs windows. E6400 @32. I3-540 stock (wife's) and I5-750 although stable @ 3.8 I've set it to 3.2. Both the I3 and the I5, the Stock HSF did not even come out of the Box. The E6400 I did install the stockHSF , but can not honestly say if there was much of a difference at idle between bios and win - can say it was below 50 C and probably closer to 40 C

The 6 so systems prior to these - Never paid any mind to temps and never had a temp problem.

 

[thegooddieyoung]

I've seen references on other places that the cpu is still loaded in bios, so I believe Leaps-from-shadows is correct in this.

I was hoping that getting more air in the case would help this problem, but my experiments have shown no meaningful reductions in temps, even with he case totally open

I've ordered a Hyper 212 from the egg. Will update here once it's working.

 

[RetiredChief]

While this defied my logic, I think I found the answer on utilization. It appears that the load while in bios is low, However; windows cycles the CPU to an effective "OFF" state about 90% of the time when (@ idle. This is simular to using PWM to control power. While utilization is higher when in windows, the "Average" utilization is lower. Example say 5% utilization in bios 100% vs 10% for 10% of time and near 0% for 90% of time when in windows @ idle.

Took some searching but found this - old, but still valid.
Quote. Under normal circumstances the CPU isn't always active but spends much time waiting for the keyboard, harddisk or CD-ROM. What would be more logical than to turn off the CPU for that period? That's exactly what the HLT machine instruction (Opcode F4) does. Whenever the CPU encounters a HLT instruction the clock is halted and the CPU enters suspend mode until an interrupt, NMI or reset happens. With the advent of power saving microprocessors like the Cyrix Cx486S the instruction acquired an additional property. When "Suspend on HLT" is enabled in the configuration register the processor not only stops on HLT but also enters the power saving suspend mode. End quote
Ref: http://www.hardware-one.com/reviews.asp?aid=103&page=1

From what I have seen that should only result in a 2 -> 7 C increase in bios vs windows @ idle.

This leaves the Operating freq/speedstep as the main driver as leap-from-Shadows stated, coupled with using the stock HSF These two factors are probably the biggest reasons for a greater diff in the newer 2x00 cpus than the older ones.

The factors affecting CPU temp are basiclly: (a) operating freq (b) vcore (c) CPU utilization and (d) HSF and case air flow - not in order of effect.
Biggest effect @ idle is probably a and d

Great choice is replacing the HSF, and 212 a great low cost choice.

 

[thegooddieyoung]

Great choice is replacing the HSF, and 212 a great low cost choice.

While up late last night, I did some more research and decided to get a different cooler. Since my order has not shipped I was able to change it

Tomorrow, UPS will be bringing a Noctua NH-D14

For a bit more money, I get better cooling and MUCH more 'quiet'

I'll be updating once I get it installed and then again when I O/C this bad boy

 

[RetiredChief]

Yes, even better HSF.

 

[thegooddieyoung]

Just installed the NOCTUA DH-14

I'm gobsmacked

Only 2-3 degrees difference at idle, but when I loaded it up with Prime 95 I was BLOWN AWAY

If you remember, I was hitting highs of 75-78c running prime.

I'm happy to show these new results


A high temp of 55c - thats a 20 degree drop.

Not really sure why my bus speed is .2 low either, also seeing my VID fluctuate a bit now. Was planning a new PS anyway. Will get a good one in here int he next few days and start overclocking this puppy. Going to go for 5ghz with these nice new temps

 

[RetiredChief]

Great

On the idle temps, There is less of a diff at idle between HSFs, it is at the "load" state that "good ones" shine. With the greater temp diff (CPU verse ambient) the better HSF have a increased ability to transfer the ambient temp to the HSF.

Enjoy

 

[Leaps-from-Shadows]

Good luck! Hope yours is one of the 1-2% that can make 5.0GHz or higher overclocks.