RealTemp and I7 4970K

[davidst99]

Hi,

I have a I7 4970K and a Gigabyte GT Z97 motherboard. When I run a Prime95 or the IntelBurnTest RealTemp says the temperatures are in the mid 90s. When I look at the Gigabyte Hardware monitor it says the temperature is in the high 70s low 80s. I'm running at 4.4 ghz. Could someone tell me what temperature is more accurate? Thanks.

David

 

[BonzaiDuck]

Hi,

I have a I7 4970K and a Gigabyte GT Z97 motherboard. When I run a Prime95 or the IntelBurnTest RealTemp says the temperatures are in the mid 90s. When I look at the Gigabyte Hardware monitor it says the temperature is in the high 70s low 80s. I'm running at 4.4 ghz. Could someone tell me what temperature is more accurate? Thanks.

David

The only thought I can offer at the moment involves "specification" history during the year or two after the Kentsfield Q6600 was introduced. Somebody miss-stated the T-junction spec for the processor: the temperature where the processor throttles to protect itself.

Software reporting temperatures of the internal processor sensors uses Tj to compute the actual temperature. I've forgotten the specifics, but an incorrect Tj will lead to incorrect temperature readings.

That's just one thing that can happen. Use of different softwares will also give different readings: for instance, the ASUS Ai-Suite Monitor is notorious for the temperature shown lower than what HWMonitor or AIDA-64 would show. There might be a "method to their madness," since the temperature reported is usually about 10C lower than the average-of-cores, or what one might expect for a "TCASE"
[*] spec temperature, but this latter is a spec so archaic it would be "mythical."

The ASUS example parallels my guess about your Gigabyte software. RealTemp is more "realistic," as would be "CoreTemp," CPUID's HWMonitor or the AIDA-64 readings. But you should assure that any freeware or paid license involves a version produced after your processor release. The difference between the i7-4790K and the 4770K wouldn't be great enough -- they're essentially Haswell and "same-socket." So just get the latest extant version of the temperature-monitoring software, and go from there.

* TCASE was a thermal guideline for computer-case design. It supposedly specified a temperature taken at the heatspreader (IHS) dead-center, usually found to be about 10C or so lower than the core-average. If your processor is within this spec, you could figure the processor is "reasonably safe," but then -- you can't easily measure such a temperature, can you?

 

[davidst99]

Thanks for the info. It's very helpful. I'm going to test again with CoreTemp.

 

[monkeydelmagico]

I'd trust your mobo maufacturer's gigabyte hardware as more likely to be correct. I've found some wonky reporting for HWMonitor and Argus monitor for my mobo but the Asus AI suite seems to get all the values right.

 

[BonzaiDuck]

I'd trust your mobo maufacturer's gigabyte hardware as more likely to be correct. I've found some wonky reporting for HWMonitor and Argus monitor for my mobo but the Asus AI suite seems to get all the values right.

I could only speak for those of us with the "ASUS" experience and their temperature monitor. Their voltage monitor seems spot-on. Temperature -- always ~10C below the core-average reported by alternative software. Oddly, this had born out for load temperatures, but the difference appears to attenuate somewhat at the idle state. Even so -- it is still lower than what's reported for the cores, and ASUS doesn't report core temperatures -- just a single CPU value.

But can't speak for the Gigabyte hardware and proprietary software. Maybe you're using a later version of ASUS Suite: my mobo is three years old.

Best to look at several software monitoring options and cross-verify.

 

[digitalbuda]

Hey davidst99

Did you try updating the BIOS? I have the exact same processor with a z97 board also from Gigabyte. At first my temps were very high at idle, I'm talking 50+ Celsius. The voltage was also very high, almost 1.4v. After updating the BIOS, the voltage dropped down to 1.18 with temps @39 Celsius idle.

The Z97 boards were released before Devil's Canyon processors were released so updating the BIOS will improve support for that processor.

I don't know if you already did update the BIOS, if you haven't, give it a try. If you already have, then maybe you need better cooling? Are you running stock heatsink/fan? Stock heatsink/fan performance are abysmal at best

 

[coffeejunkee]

The only thought I can offer at the moment involves "specification" history during the year or two after the Kentsfield Q6600 was introduced. Somebody miss-stated the T-junction spec for the processor: the temperature where the processor throttles to protect itself.

Software reporting temperatures of the internal processor sensors uses Tj to compute the actual temperature. I've forgotten the specifics, but an incorrect Tj will lead to incorrect temperature readings.

That's history now. Modern Intel cpu's have their Tjunction value stored in a register which can be read by programs like Realtemp.

So, it really is the real temp that Realtemp displays. Gigabyte software might display something like the temp for the cpu as a whole. But for temp discussions it's assumed you are talking about core temps as displayed by Realtemp (or any of the many other temp monitor programs).

 

[Dufus]

I could only speak for those of us with the "ASUS" experience and their temperature monitor. Their voltage monitor seems spot-on. Temperature -- always ~10C below the core-average reported by alternative software.

Yes I remember that, with P67 boards ASUS took the PECI temperature which reported the hottest core, subtracted 10 from it via the EC and called it "Case Temperature" which is incorrect. Temperature differential between case and cores will be load dependent, the bigger the load the bigger the differential, not fixed at 10C.

@davidst99 Run Intel's XTU if you want to check if other software is correctly reporting temperatures.