Intel I7-7700K Temperatures

[zshs2011]

Great and Powerful Forum Members:

I am looking to upgrade my computer for VR awesomeness in the next 2-3 weeks. I understand the I7-7700K is probably more processor then current gen games and hardware require, but this should prevent my CPU from ever bottlenecking my rig and make a good launching point for SkyNet. With that said I am getting tripped up on delidding or not.

The Kaby lake processors are said to run hot. Most enthusiasts recommend de-liding to get the temperatures under control which I am "cool" with.

My concern comes from a different forum that had a gentleman trying to explain that the temperature HAD to run high and throttling was the CPUs way to protect ALL components on the die. This reasoning provided was that the throttle temperature (100C) was based on the core getting to a certain temperature indicating other components on the die which aren't monitored were also reaching their limits. The outer components temperatures were directly proportional to core temperature. By de-lidding and replacing the TIM, you are cooling the core but leaving the other components to potentially reach damaging temperatures without being monitored. Under modified conditions, the core could be cool but the outskirts of the die could be hot.

It doesn't seem to pass the smell test in all honesty, but I also can imagine a scenario where it could be designed that way (where the TIM is NOT protecting all components equally) ... so any truth to this? Does de-liding put my irresponsibly repurposed rent payment in danger as they claim?

Thanks!

 

[CompuTronix]

zshs2011,

The "gentleman" to whom you are referring is seriously misinformed.

Intel's Digital Thermal Sensors (DTS) measure each individual Core independent of one another. Each DTS is an "array" of 8 analog thermal diodes, so a quad core has 32 thermal diodes. Each diode is located on a "hot spot" within a Core, which are the most active and hottest clusters of transistor junctions. The hot spots migrate according to differences in workload, thus the need for 8 diodes per Core.

The transistor junctions are the hottest components anywhere in the Die. Any other components in the Die or on the Package operate at lower temperatures.

The analog levels of the 8 diodes are input to a comparator circuit where the single hottest diode's level is then converted from analog to digital (A to D), which then outputs the DTS value per each individual Core. This is how Core temperature works.

Here's the recommended operating range for Core temperature:

80°C Hot (100% Load)
75°C Warm
70°C Warm (Heavy Load)
60°C Norm
50°C Norm (Medium Load)
40°C Norm
30°C Cool (Idle)
25°C Cool

Core temperatures up to 80°C are safe.

Your highest temperatures will occur when running test utilities. Temperatures are typically lower during real-world everyday workloads such as processor intensive applications or gaming.

There's a Sticky near the top of the CPU's Forum which you might want to read: Intel Temperature Guide - http://www.tomshardware.com/forum/id-1800828/intel-temperature-guide.html

CT

 

[zshs2011]

zshs2011,

The "gentleman" to whom you are referring is seriously misinformed.

Intel's Digital Thermal Sensors (DTS) measure each individual Core independent of one another. Each DTS is an "array" of 8 analog thermal diodes, so a quad core has 32 thermal diodes. Each diode is located on a "hot spot" within a Core, which are the most active and hottest clusters of transistor junctions. The hot spots migrate according to differences in workload, thus the need for 8 diodes per Core.

The transistor junctions are the hottest components anywhere in the Die. Any other components in the Die or on the Package operate at lower temperatures.

The analog levels of the 8 diodes are input to a comparator circuit where the single hottest diode's level is then converted from analog to digital (A to D), which then outputs the DTS value per each individual Core. This is how Core temperature works.

Here's the recommended operating range for Core temperature:

80°C Hot (100% Load)
75°C Warm
70°C Warm (Heavy Load)
60°C Norm
50°C Norm (Medium Load)
40°C Norm
30°C Cool (Idle)
25°C Cool

Core temperatures up to 80°C are safe.

Your highest temperatures will occur when running test utilities. Temperatures are typically lower during real-world everyday workloads such as processor intensive applications or gaming.

There's a Sticky near the top of the CPU's Forum which you might want to read: Intel Temperature Guide - http://www.tomshardware.com/forum/id-1800828/intel-temperature-guide.html

CT

Thanks for the response! I did thumb through the guide you mention before I posted, but for some reason it made more sense when I re-read it after your response. Probably because you concisely and cleanly summarized it!

Just to be thorough, he was talking about the CPU cache producing the most heat on the die (and not thermally monitored) which would be a prime example of a section of the CPU to fail due to this. By extracting more heat from the cores, they stay cool but the CPU cache would overheat.

One of the commenters mentioned he was probably full of it, since primarily these de-lidded CPUs should be failing left and right especially since they have been offered for months now...

 

[mrobscura]

while im not sure about the accuracy of the info in the op, one thing i did read about delidding(it may have been an article on toms actually) is that intels tim, while maligned by enthusiasts, and not offering the best thermals, may be designed to cope with the constant temp fluctuations the 7th gen cpus see, whereas after market stuff used when delidding is not and as a result delidded cpus may see issues in the future.

 

[CompuTronix]

zshs2011,

Everyone tends forget that we're talking about 1.75 billion 3D transistors on the 14 nanometer scale operating in incredibly close proximity to one another in an extremely small 122 sq mm multi-layered Die. Intel doesn't measure the Cache portion of the processor because that's not where the majority of the workload is being processed; the hot spots are the most active clusters of transistor junctions, which are located within the IA Cores, and are the primary heat sources within the Die.

Heat is removed by conduction away from the hottest components and dissipates into the Die, where it dissipates further to the surface of the Die, which is relatively cool. However, the more efficiently we cool the entire Die, such as with Thermal Grizzly Contuctonaut liquid metal TIM, the cooler all nanometer components within the Die will operate.

Liquid Metal Examples:

Thermal Grizzly Conductonaut - 73.0 W/mk
Coollaboratory Liquid Ultra - 38.4 W/mk
Coollaboratory Liquid Pro - 32.6 W/mk

Popular Silicon Example:

Arctic MX-4 - 8.5 W/mk

The reason that TGC is superior is because it contains a significant amount of Indium, which has very good thermal conductivity compared to most other elements. Intel stopped using Indium solder due to cost after 2nd Generation Sandy Bridge, which were the last thermally predictable and well behaved processors. Intel's "High End" processors are still soldered due to being higher TDP processors with more Cores. Intel's change to TIM generally works well enough ... until you overclock, which is an extremely small enthusiast segment of their market.

mrobscura,

Temperature fluctuations are not related to TIM or delidding. Intel’s specification for DTS sensor response time is 256 milliseconds, or about 1/4th of a second. Since Windows has dozens of Processes and Services running in the background, it’s normal to see rapid and random Core temperature fluctuations, especially during the first few minutes after startup.

Core temperatures are directly proportional to changes in workload. Intel's 6th and 7th Generation processors now use "Speed Shift" instead of the older "SpeedStep" technology, which is faster and more responsive to changes in workload. Speed Shift is internally controlled by the CPU, whereas SpeedStep is controlled by the Operating System. Also, Speed Shift has many more clock and voltage levels than SpeedStep, so it produces more temperature spikes than 4th Generation and previous processors.

The biggest culprit responsible for Core temperature spiking is "tray-trash" or too many unnecessary background processes running, which needlessly devour system resources such as memory and CPU cycles, and collectively have a pronounced effect on temperature spiking. Many users neglect to clean up their startups after software installations and program updates, which is just part of routine computer maintenance. It's like the wife who forgets that the car needs oil changes and proper tire pressure.

CT

 

[mrobscura]

compu, im aware of that. what i said is, intels tim is designed with those fluctuations in mind whereas after market tim applied to delidded dies is not, and this might cause issues with delidded cpus in the future. at least thats what the article suggested.

 

[CompuTronix]

"Pump-out" is a phenomenon where silicon based TIM oozes out from between the Die and Integrated Heat Spreader over time due to thermal cycling. As such, silicon based TIM needs to be replaced frequently as it fails relatively quickly over months. Instead, liquid metal TIM is essential for de-lidding because of it's long term resistance to pump-out over years, as well as unmatched thermal conductivity.

Intel's TIM suffers from dry out, particularly at sustained high temperatures, and becomes less thermally conductive over time. Regardless, it's still far inferior to Indium solder. These figures from Silicon Lottery - https://siliconlottery.com/collections/all/products/delid - will give you an idea of just how poorly Intel's TIM really performs:

"Temperature improvements vary depending on processor sample and architecture. Temperatures under an overclocked load (1.3V-1.4V) will decrease anywhere from 5° to 25°C. Typical temperature improvements are listed below:"

Ivy Bridge: 10°C to 25°C
Haswell: 10°C to 25°C
Devil's Canyon: 7°C to 15°C
Broadwell: 8°C to 18°C
Skylake: 8°C to 18°C
Kaby Lake: 12° to 25°C

In any case, Core temperature spiking is not caused by TIM. However, the better the TIM and the lower the Core temperatures, the less pronounced the spiking. In this respect, Intel's TIM is again inferior. In my opinion, Intel's change from Indium solder to TIM just to save pennies per unit has done it's customers a disservice, and is a failing typical of corporate quest for profits.

CT

 

[zshs2011]

CompuTronix,

Thank you for the explanation(s). Very well done and much clearer now.

(its also incredible how much detail and engineering goes into everything from material selection to heat dissipation.)

Thanks again!