CPU Temp 'Issues'...HELP!

[Yellowhello]

Originally Posted by ST!X

Ambient temps where? In the room or in the case?

Both I guess...? Because if you have the heat cranked in your house that could be it... lack of airflow in the case... along with what Rich mentioned.

[Capper]

well thats the reason they have a safety shutdown point.......before your processor allows itself to be damaged, its going to scale back, and then shut down.......If you keep it under 70C, you'll be fine.

[[Dr. V]]

Hmm...fair enough.

I'll play around a bit.

Well, I can get a good clock and Windows boot and use for a while with 1.35 @ 2.8GHz. As soon as I drop it to 1.34375 (hardly lower) I can only go up to 2.4GHz if I wanna boot into Windows. The thing is, at 1.35, I get program readings of 70 and sometimes 71. With 1.34375 I don't. But it's so utterly annoying to lose 400MHz because of that tiny increment of difference in the voltage.

I think my motherboard is shit. Either that or I got a bad O/Cing CPU. Or both.

[sbohdan]

why don't you try a real temp sensor with a digital readout that would phisically read the actual temps of your CPU. those things only cost couple of $ and you could always check before using software sensors to see which one is the most accurate. I use this one to check temps even on GPU's when OC'ing them

[[Dr. V]]

Hmmm...very good idea. I've gotta look in to one of those.

Thanks sbohdan.

Shame...I though this system would do me well. It's actually running kind of annoyingly "strained" I need a new system. I'm trying to sell mine for about $700-750 CAD and build a brand new system with an E6550, a new better P35 board (probably the GA-P31-DS3R/L, or maybe ASUS P5K), and a X1650 Pro for dirt cheap that can still run COD4 (which I beat a few days ago - great game, just a BIT short).

In any case, I've NEVER, EVER had ANY luck with overclocking before. My E6300 wouldn't run stable past 2200MHz, my Q6600 (G0 STEPPING - I still wish I never got rid of this) never ran stable past 2800MHz...and now this. This is my biggest success...an 800MHz overclock on a Core 2 Duo system with LOAD temps inching towards 70*C with a GeminII with 2 ARCTIC 12 fans on it...how does this work? Something has to be wrong here...every test no matter what showed this cooler cooling C2D's to sub-30 IDLE and sub-50 max on load. This is stupid.

[Capper]

Honestly, doing something like that simply doesn't work. First, you're creating a gap between the CPU and heat sink......its not going to seat correctly, which means the temperature you get from it is going to be inaccurate. You'll either have gaps where there is no contact, raising the temps, or you are going to put too much thermal compound on the IHS, raising the temps.....either way, its a bad idea.....unless you do something like cut a groove in the IHS (another bad idea, for obvious reasons).

The reason the CPU makers don't care about temperature sensors is that if they are accurate, then people are going to OC, which they don't support, and by pushing CPUs more and more, there will be more failures, which means more RMAs. By giving a general temperature, thats always within a safe range, they don't really have to worry.....because people are going to be weary of pushing their CPUs too far......make sense?

If you are that concerned, get some better cooling, or lap the IHS on your processor.....personally, I think you just drew the bad overclocker. Then look at your supporting hardware.....that board is not meant for overclocking.

I think we are in dire need of an overclocking guide, which I will work on as time permits......but to make it as simple as possible....isolate each component, find its "sweet spot", relax everything else.....then move on to the next component......for example.....worry about your processor, relax your memory timings, FSB, voltages, etc.......tinker and toy with the system to find a good balance of processor speed, voltage and temperatures (there is always a point of diminishing returns)....once you fibnd that, move on to the memory, etc.

Its really not that difficult, but mnst people don't have the patience to do it correctly....they see someone else claim to have spectacular results....which can be misleading (either not stable, or not safe), and they think they can just plug in the same settings and achieve the same results.

[[Dr. V]]

Yes it does make sense.

The funny thing is I've spend 2-3 hours overclocking others' systems and got amazing results (I do know how to O/C), but I never seem to get luck with MY hardware.

BUT, before overclocking, temps are my main concern. Nothing seems to be able to explain why CoreTemp gives others sub-30 temps with a GeminII but I'm getting high-30s - mid-40s.

[Yellowhello]

Maybe you should consider water cooling.

[Reloadron]

Originally Posted by ST!X

Yes it does make sense.

The funny thing is I've spend 2-3 hours overclocking others' systems and got amazing results (I do know how to O/C), but I never seem to get luck with MY hardware.

BUT, before overclocking, temps are my main concern. Nothing seems to be able to explain why CoreTemp gives others sub-30 temps with a GeminII but I'm getting high-30s - mid-40s.

First, and before all else, if I were to trust something it would be the motherboard software written and designed specifically for the motherboard.

One reason people see deviations is how many of the third party software packages read the information. Looking at things in a nutshell.

Back in 1995 Intel developed a bus for motherboard design. This is called the SMBus (System Managment Bus). This explains it pretty well:

What is SMBus?

The SMBus reads from devices on the bus using a standard protocol. Enter the devices on the motherboard. The devices are small Integrated Circuit chips that perform a variety of managment functions, among them temperature monitoring and fan speed monitoring. In the case of a laptop even battery power level.

Speed Fan was likely one of the first user programs developed to read from these devices and provide users with information they could see about the system. Typically that data was reserved for system use. If I recall correctly the home page for Speed Fan mentioned a few of the chips it supported.

Chips like the LM85 (National Instruments) were popular hardware monitoring chips:

LM85 - Hardware Monitor with Integrated Fan Control [Obsolete]

However, note in the link that chip is discontinued! Lately many of these chips evolve and better ones are developed daily. This presents a problem and part of the problem is what you see. The software needs to be able to read the monitoring device and to do that it needs to correctly ID the device. It's all on the 2 wire SMBus but the software needs to know what the data breakdown actually is.

The monitoring software can't give accurate numbers unless it knows what math functions to apply. Each new device needs to be loaded into the software and devices keep popping up. The software is reading bits and not the numbers we like to see or can relate to.

Overall, that is why we see differences.

Regarding Ambient temperatures. Back in October 07 I did a post about an experiment I did.

A Little About Temperatures (Just A Tad)

Though not intended as a guide I tried to make it informational and looked at heat in general in and exiting a case.

Anyway, without beating the inner workings of the SMBus and the chips it monitors I hope this sort of explains the delta from program to program you see in temperatures.

Ron

[[Dr. V]]

Great info Ron, thanks!

So, I got my new CPU, E6550 (333x7 = 2.331GHz stock), in my new motherboard, GA-P35-DS3L, with a new cooler, ZALMAN CNPS9500A LED. This is odd. I'm actually ecstatic and confused at the same time.

Both CoreTemp and TAT are telling me that idle temps are between 18 and 21. Everest tells me it's between 35 and 39. Load temps for CoreTemp and TAT are around 33-36. Everest shows abour 49-51. (Note: this is a rough average of both cores - by the way they're never more than 2 degrees apart now).

In either case, whichever is right, I'm very happy. If CT and TAT are right, I'm unbelievably happy. Let's just say I booted straight to Windows as I am now with 429x7 = 3GHz. Voltage from 1.35 to 1.36875 (before droop).

So...this is still confusing, but in either case it's good news!