i5 750 temps too high

[BonzaiDuck]

Not to waste anyone's time on this . . .

First, ShintaiDK is correct about Tcase and Tjunction specs.

He knows that I "shoot for" overclock settings which, under load, only exceed TCASE by a few degrees.

However, you'll never very likely reach that point with most real-world usage. I think I agree with him: You could overclock closer to TJunction under load, because real-world use may still likely show temperatures peaked in the mid to high 70s C.

Now, on the fan ports of that motherboard, while I know the OP and possibly this or that poster may lose patience with me here, I would scrutinize the BIOS very carefully to see if some feature or function needs to be set and tweaked to meet the promise of the manual: thermal control of the fan speed.

Moreover, I'm only familiar with ASUS, Gigabyte and EVGA boards these days. ASUS will thermally control a 3-pin fan properly connected to a 4-pin PWM header. I do not know what provision Intel made (if any) for this dual capability. And -- I'm sorry, but I haven't looked at the detail about the fans in previous posts. Without having done so, I'm wondering if Intel actually requires PWM fans on the CPU and chassis 4-pin ports.

I have motherboards that have been running 24/7 since ~2008: ASUS, EVGA and Gigabyte boards (LGA-775). We've never had the fan-control features suddenly die. They just "keep on tickin', and takin' a lickin'."

It's always possible that the gurus who tower over me in these forums may have missed some feature of the Intel boards. I don't know of any (of whom I remember, anyway) who used Intel boards for those Lynnfield chips or later.

Also, check your chipset drivers, as they would include the "ACPI fan" features which should also show up as nodes in your Device Manager under the System Devices node.

 

[dtbecker]

Not to waste anyone's time on this . . .

First, ShintaiDK is correct about Tcase and Tjunction specs.

He knows that I "shoot for" overclock settings which, under load, only exceed TCASE by a few degrees.

However, you'll never very likely reach that point with most real-world usage. I think I agree with him: You could overclock closer to TJunction under load, because real-world use may still likely show temperatures peaked in the mid to high 70s C.

Now, on the fan ports of that motherboard, while I know the OP and possibly this or that poster may lose patience with me here, I would scrutinize the BIOS very carefully to see if some feature or function needs to be set and tweaked to meet the promise of the manual: thermal control of the fan speed.

Moreover, I'm only familiar with ASUS, Gigabyte and EVGA boards these days. ASUS will thermally control a 3-pin fan properly connected to a 4-pin PWM header. I do not know what provision Intel made (if any) for this dual capability. And -- I'm sorry, but I haven't looked at the detail about the fans in previous posts. Without having done so, I'm wondering if Intel actually requires PWM fans on the CPU and chassis 4-pin ports.

I have motherboards that have been running 24/7 since ~2008: ASUS, EVGA and Gigabyte boards (LGA-775). We've never had the fan-control features suddenly die. They just "keep on tickin', and takin' a lickin'."

It's always possible that the gurus who tower over me in these forums may have missed some feature of the Intel boards. I don't know of any (of whom I remember, anyway) who used Intel boards for those Lynnfield chips or later.

Also, check your chipset drivers, as they would include the "ACPI fan" features which should also show up as nodes in your Device Manager under the System Devices node.

Is it possible to calculate Tcase without a thermal probe? (Since Tjunction temperature is software reading)

 

[BonzaiDuck]

Is it possible to calculate Tcase without a thermal probe? (Since Tjunction temperature is software reading)

Axiomatic, my dear Watson! Axiomatic!

TCASE <= SUM(Ti)/n for i = core 1 to k for k cores.

That is, the temperature which could actually be measured for all the trouble ShintaiDK mentioned would always be less than or equal to the average temperature among the four core sensors.

Some time ago, when Conroe and Kentsfield emerged for socket-LGA-775, someone at Tom's Hardware concluded that the TCASE measured temperature was ~ 10C less than the average of the cores. But these days, it's more likely the two values will be closer together or almost equal.

As ShintaiDK will argue (and "no contest here!") it is a still-published spec intended as a guideline for computer case design and airflow, and I will argue that it is intended as a boundary beyond which the chip may very slowly and gradually undergo thermal degradation over continued, constant use at that temperature. As temperature increases, more power consumption will occur (and see threads by IDontCare), more voltage would be necessary to maintain stability, which would in turn create higher temperature.

But stress-testing an OC'd chip for several hours @ 90C would be nothing compared to what that spec is designed to avoid.

I don't want to get into a dispute about this or the usefulness or meaning of the TCASE spec. [ShintaiDK and I already did that.] I argue that it has some meaning or Intel wouldn't still publish it in CPU specs, while they discontinued printing a spec for "maximum safe voltage" and "maximum operable voltage" after Nehalem -- which would be "Gen 1" more or less.

 

[Deders]

A decent cooler will make a huge difference, as will good case airflow. I'm running my i5-750 at 3800MHz, it never goes above 60c when fully stressed, maybe 64c on a hot day.

My case allows me relocate some of my drive bays in a way that unblocks the top front fan, and some of the bottom one so the air can flow unheeded towards the GPU and CPU. It makes a huge difference for both.

I'm using a mid range cooler (Thermaltake Contac 29), with a low RPM but high static pressure Noctua fan. The heatpipes directly touch the IHS which helps a lot. I use MX4 (better temps than Arctic Silver 5) and spread it with a card so it covers the entire IHS. Best method for Direct-touch heatpipes.

 

[ShintaiDK]

Is it possible to calculate Tcase without a thermal probe? (Since Tjunction temperature is software reading)

No. Because Tcase depends on what type of heatsink, the bond with the heatsink and so on.

Tcase only function is a guideline for heatsink/OEM manufactors. And it got no relevance as such to the CPU temperature limits.

http://www.intel.com/support/processors/sb/CS-033342.htm

Tcase calculations based on Tdiode and Tjunction can be off with 20-30C. Or even as much as 50C if you use water.

 

[BonzaiDuck]

No. Because Tcase depends on what type of heatsink, the bond with the heatsink and so on.

Tcase only function is a guideline for heatsink/OEM manufactors. And it got no relevance as such to the CPU temperature limits.

http://www.intel.com/support/processors/sb/CS-033342.htm

Tcase calculations based on Tdiode and Tjunction can be off with 20-30C. Or even as much as 50C if you use water.

I'd have to go back and rummage through the lengthy Intel spec sheets. As I mentioned, we'd had this discussion before, and I came away thinking I'd been mistaken as to how Intel expected the TCASE to be measured. All along, I thought it was measured at the upper surface of the IHS, but in our discussions you convinced me that it would be measured with a probe at the point where the die meets the copper IHS.

Yet your thoughts about this would suggest my original understanding (or misunderstanding) in the matter, since a water-block or HSF-base and the thermal material would figure more in it.

But it's really a moot and trivial point, since it could never be more than what the Tjunction sensors report.

It's merely a guideline one could choose to observe or ignore. Since the spec itself -- ~73C for my i7-2x00K processors or ~67C for an i7-5820K and considerably less than the Tjunction throttle point, choosing to apply it against readings from the core sensors does no harm, but would only limit the extent of an extreme overclock.

The throttle-point is an immediate limit of protection to the processor. If TCASE has any relevance, it's a guideline for maximum continuous long-term usage.

 

[ShintaiDK]

I'd have to go back and rummage through the lengthy Intel spec sheets. As I mentioned, we'd had this discussion before, and I came away thinking I'd been mistaken as to how Intel expected the TCASE to be measured. All along, I thought it was measured at the upper surface of the IHS, but in our discussions you convinced me that it would be measured with a probe at the point where the die meets the copper IHS.

Then you didnt argue with me. And that is wrong. Its the top of the IHS.

Yet your thoughts about this would suggest my original understanding (or misunderstanding) in the matter, since a water-block or HSF-base and the thermal material would figure more in it.

But it's really a moot and trivial point, since it could never be more than what the Tjunction sensors report.

It's merely a guideline one could choose to observe or ignore. Since the spec itself -- ~73C for my i7-2x00K processors or ~67C for an i7-5820K and considerably less than the Tjunction throttle point, choosing to apply it against readings from the core sensors does no harm, but would only limit the extent of an extreme overclock.

The throttle-point is an immediate limit of protection to the processor. If TCASE has any relevance, it's a guideline for maximum continuous long-term usage.

Tcase is only useful for designing cooling solutions. Nothing else.

 

[BonzaiDuck]

Then you didnt argue with me. And that is wrong. Its the top of the IHS.

And that's what I'd originally assumed.

Tcase is only useful for designing cooling solutions. Nothing else.

Sure -- cooling solutions, to cool "what?" Certainly, the CPU.

I don't think we really have a conflict or serious difference of opinion about any of this. I also think I properly articulated your view of the TJunction throttle-point.

Of course, as IDontCare had pointed out, you can run up the temperatures to these processors into the 90's C, but then -- more power consumption and the vicious circle of more voltage needed.