Intel live chat just told my dad processor temps are irrelevant -.-

[Hyperlite]

I made another thread about this a week or so ago, but hes having an issue with the temp readouts on a e5200. One core is stuck at 43C, and the other idles at 27C and never moves much. Stock HSF. reseated and re-re-reseated. He got on live chat with intel to discuss it and this is what they said (some summary, some quotes)

got on a live chat with Intel... they say that most core temp tools do not work with the 45nm processors
furthermore, intel does not 'provide software that monitors core temps and considers that information irrelevant concerning processor operation'
only concern is total die temp, which must not exceed 74c

orly. So it seems they WILL NOT pursue warranty action over temperature issues...

 

[waffleironhead]

From their point of view.
Is the system experiencing instability from the one core being at a higher temperature than the other?
Is the processor guaranteed or warrantied to function at 27C? 30C? 43C?
I'm sure if you pursued it they would claim that the 43C core is perfectly safe and to consider the 27C core as a bonus.


FWIW I have no idea on the typical operating temperatures of the cores. But if they are within spec, I dont see that you have a claim against them.

 

[Hyperlite]

Alright, fair enough, but previous responses have led me to believe that this IS something to be concerned about. Is that false?

 

[BassBomb]

Originally posted by: Hyperlite
Alright, fair enough, but previous responses have led me to believe that this IS something to be concerned about. Is that false?

Stuck?

Have you loaded it up (both cores) to see what it does?

 

[Lemon law]

You may have uncovered one idiot intel live chat person willing to say damn near anything to get someone off their back, but I really really doubt that is the official intel position.

And besides its bullshit, my 45 nm E5200 intel processor has functioning temp sensors that certainly are not stuck, they react very predictably to load, but the question of total accuracy would take independent instrumentation.

Our OP either has a defective cpu, a motherboard with defective cpu heat sensors, or is using the wrong temperature monitoring programs.

To start out with, a cpu idle temp of 27 C is almost impossibly low unless the OP has a very cold house or a super cpu cooler, 45 C is probably high for a idle temp on a e5200, and a little low for a fully loaded temp, but if both sensors are stuck, you have no idea at what temp the cpu is running. It could be running at 85 C for all you know.

 

[pm]

I'm not an Intel spokesperson - I just work as a CPU designer and debugger in the server division - but I'm leaning towards what the Live Chat guy said. To the best of my knowledge, Intel doesn't provide the software to measure temps, and (at least on server systems) has internal hardware to measure temp that throttles or shuts down the CPU automatically. So it's not too big a deal what software says the value is at, or the motherboard, or anything else... all of the relevant circuitry to deal with high temperature issues is internal to the CPU (again, at least on the Intel server CPU's that I work on).

Edit:
I am reading the datasheet and the function does operate as I thought that it did - although it does appear that the PROCHOT# signal is bi-directional and that the chipset can drive it too - but if the chipset is somehow stuck reading a low value, that doesn't mean that the CPU internally is reading the same value. I didn't know that PROCHOT# was bi-directional though... learn a new thing every day.

For PROCHOT# (the pin on the CPU that indicates that the CPU is running too hot and that throttling is occuring)

"As an output, PROCHOT# (Processor Hot) will go active when the
processor temperature monitoring sensor detects that the
processor has reached its maximum safe operating temperature.
This indicates that the processor Thermal Control Circuit (TCC)
has been activated, if enabled. As an input, assertion of
PROCHOT# by the system will activate the TCC, if enabled. The
TCC will remain active until the system de-asserts PROCHOT#.
See Section 5.2.4 for more details."

for THERMTRIP# (the pin on the CPU that indicates that the CPU is running at the damage point)

In the event of a catastrophic cooling failure, the processor will
automatically shut down when the silicon has reached a
temperature approximately 20 °C above the maximum TC.
Assertion of THERMTRIP# (Thermal Trip) indicates the processor
junction temperature has reached a level beyond where
permanent silicon damage may occur. Upon assertion of
THERMTRIP#, the processor will shut off its internal clocks (thus,
halting program execution) in an attempt to reduce the processor
junction temperature.

Again, this is all automatic and internal the CPU. And there's no tool that I'm aware of to let a user know what the CPU thinks that it's seeing.


Patrick Mahoney
Senior Design Engineer
Enterprise Processor Division
Intel Corp.

* not a spokesperson or official representative of the company... just an engineer who participates on the forums here at AT. *

 

[Lemon law]

And I ask pm, irrespective of the accuracy at which a given motherboard or software will read the temps of the server side chips you work on, if the sensors in the chips do not function correctly, they could fail to shut down the server even if the cpu heat is way way way too high.

 

[KIAman]

You forgot the part where he mentions "total die temp is the main concern."

Intel support simply used the wrong wording to get the correct message out. It's like telling a patient, you will die within the next 60 years instead of saying you have 60 more years to live.

 

[pm]

Originally posted by: Lemon law
And I ask pm, irrespective of the accuracy at which a given motherboard or software will read the temps of the server side chips you work on, if the sensors in the chips do not function correctly, they could fail to shut down the server even if the cpu heat is way way way too high.

Yeah, I'll agree with that. Although on the servers I work on, we have five thermal diodes on the CPU die, they are scattered widely across the die, and they are all tested several times prior to shipping the CPU out the door.

My counter arguement to the idea that the thermal diode is bad to that is that you are dealing with a CPU with 582 million transistors on it, and the the idea that one diode out of these 582 million devices would go bad seems more than a bit unlikely. Possible, but unlikely. I'd be more inclined to look first at the way the measurements are being taken - are they really under a multicore/multithreaded load? If not then it's not too weird to see core 0 much, much higher than core 1. I'd also check the idle vs. load temperature on a single-core load and then a multi-core load to see how much variation there is. Then if things are still weird, I'd also at the software and it's compatibility with the motherboard. Thermal diodes are calibrated near thermtrip, so the closer one gets to higher temperatures, the more accurate the diodes will be, so then I'd dial the fan speed lower and see what happens.



... deleted my second edit... I get it now. I was right, not enough coffee.


For what it's worth, my 4th core on my i7 920 (stock) is running 28C right now according to "Real Temp".

 

[happy medium]

Originally posted by: Lemon law
You may have uncovered one idiot intel live chat person willing to say damn near anything to get someone off their back, but I really really doubt that is the official intel position.

And besides its bullshit, my 45 nm E5200 intel processor has functioning temp sensors that certainly are not stuck, they react very predictably to load, but the question of total accuracy would take independent instrumentation.

Our OP either has a defective cpu, a motherboard with defective cpu heat sensors, or is using the wrong temperature monitoring programs.

To start out with, a cpu idle temp of 27 C is almost impossibly low unless the OP has a very cold house or a super cpu cooler, 45 C is probably high for a idle temp on a e5200, and a little low for a fully loaded temp, but if both sensors are stuck, you have no idea at what temp the cpu is running. It could be running at 85 C for all you know.

I idle at 27c and dual prime at 48c. My temps fluctuate just fine. e5400 wolfdale

 

[Zap]

Originally posted by: waffleironhead
FWIW I have no idea on the typical operating temperatures of the cores. But if they are within spec, I dont see that you have a claim against them.

I don't think there is a published "typical" only "maximum," which is usually 80-90ºC for stock speeds. If power management is enabled, then the CPU will likely throttle when it gets too hot. If power management is disabled in BIOS, then the system will likely lock up when it overheats. If your CPU is not throttling under load and your system is not locking up, then it is fine.

Originally posted by: KIAman
Intel support simply used the wrong wording to get the correct message out. It's like telling a patient, you will die within the next 60 years instead of saying you have 60 more years to live.

LOL, that's a good way to put it.

 

[VirtualLarry]

Originally posted by: pm
Again, this is all automatic and internal the CPU. And there's no tool that I'm aware of to let a user know what the CPU thinks that it's seeing.

ProcHot is visible through MSRs I believe. RMClock can show you when and if you've throttled.

 

[pm]

Originally posted by: VirtualLarry
ProcHot is visible through MSRs I believe. RMClock can show you when and if you've throttled.

Thanks for mentioning that. I had no idea that MSR's were customer visible - I thought they were for internal-use at Intel only. I think I need to read our manuals better.

 

[Hyperlite]

First of all, thank you all for the informative discussion. Secondly, reading here about PROCHOT and THERMTRIP lead me to find this post, http://www.hardforum.com/showp...031827510&postcount=31 with some interesting info that basically mirrors what pm has been saying. Thanks guys! and pm, you were dead on about temps getting more accurate closer to PROCHOT or THERMTRIP...i didn't realize he hadn't been using a real stress program, and sure enough, both cores loaded up to 48C under load. Still a big difference in idle temps, which i'll chalk up to what you said.

 

[pm]

Glad it worked out, Hyperlite.

There's a big difference in temps on my idling Win7 system right now. On my Core i7 920, I have 36C on core 1, and 27C on core 4.

 

[Ben90]

If power management is enabled, then the CPU will likely throttle when it gets too hot. If power management is disabled in BIOS, then the system will likely lock up when it overheats.

Thats not how it works. The power management in BIOS throttles the CPU down when not under load to save power, this power management has nothing to do with temperature.

Every intel processor since the p4 has its own internal power management which will function independent of any user settings. If the system locks up from heat related issues there is usually a deeper problem. I havent read up on c2d power management, but i know for i7s/i5s the processor is allowed to run 20*C over Tjmax for up to 500ms before it shuts off.

That basically means you can staple a large piece of cheese to your cpu for a heatsink and it should function, albeit at a much much lower frequency, and skipping clock cycles, but i doubt it would lock up

 

[Idontcare]

If anyone is interested in reading reviews that delve into nitty-gritty of thermal trips and the likes you should really check out ixtbtlabs, they do it justice like no other review site I have seen.

Performance and Power Management Technologies in Intel Processors. Part 4: Intel Core 2 Processors

(That's just one I found for quick reference, peruse their site for more, they have quite a few)

 

[dug777]

thats not how it works. The power management in bios throttles the cpu down when not under load to save power, this power management has nothing to do with temperature.

Every intel processor since the p4 has its own internal power management which will function independent of any user settings. If the system locks up from heat related issues there is usually a deeper problem. I havent read up on c2d power management, but i know for i7s/i5s the processor is allowed to run 20*c over tjmax for up to 500ms before it shuts off.

that basically means you can staple a large piece of cheese to your cpu for a heatsink and it should function, albeit at a much much lower frequency, and skipping clock cycles, but i doubt it would lock up

lmfao

 

[Gunbuster]

Are you sure the CPU is not dead?

Cut and paste: "System does not post with CPU in board, System tests out OK with known good CPU"

 

[Fayd]

I made another thread about this a week or so ago, but hes having an issue with the temp readouts on a e5200. One core is stuck at 43C, and the other idles at 27C and never moves much. Stock HSF. reseated and re-re-reseated. He got on live chat with intel to discuss it and this is what they said (some summary, some quotes)




orly. So it seems they WILL NOT pursue warranty action over temperature issues...

e4400 here, and both cores idle at mid 40's.

it's stable, it doesnt matter.

 

[aigomorla]

guys intel assumes your not overclocking.

If under stock conditions, there is no way your chip will exceed tdp unless you disable the fan on the stock heat sink.

What intel is saying is that what your idle temps are reporting is moot.
What load temps report is also moot unless you notice a lag which then its called turbo throttling, on i7's and regular throttling on the other intel chips.

When you throttle you will have shutters and lags, then you can report to intel that something is wrong with your cpu.

To intel as long as your chip is below 100C that is fine, they only promised you 3 yrs, and at 100C there engineers predicted thats how long they would last.

In actuality is more because no one runs there cpu's at 100C.