Overclock.net › Forums › Intel › Intel CPUs › Huge temperature difference on cores (4690K 4,5 GHz)
New Posts  All Forums:Forum Nav:

Huge temperature difference on cores (4690K 4,5 GHz)

post #1 of 12
Thread Starter 
Hi,

I ran into strange problem recently and I want to ask if this is something that normally happens or there is something wrong.

Thing is that temperatures on different cores of my i5-4690K (4.5GHz 1.276V) are very uneven. Here are some screenshots to show how much:



As you can see core #0 and #1 are way hotter than #2 and #3 and while those last two are sitting at comfortable temperatures for such demanding test like OCCT, first two cores are much hotter, especially core #0 that is very close to overheating at times.

Here are some more screens:



Direct link for better visibility: http://i.imgur.com/qxAkaAJ.jpg?1

Differences are reaching 20C sometimes and first two cores are almost always living completely different life than remaining two. I am loosing frequency stability because of those first cores, last two are sitting at comfortable temps and first two are, especially first one, are way hotter.

My CPU cooler is SPC Fortis 3 Malik Customs: https://www.techpowerup.com/reviews/SilentiumPC/Fortis_3_HE1425/
My thermal paste is Noctua NT-H1

Both newly bought few days ago. I don't know how those temps were at my previous cooler (Fortis 2) and if they were uneven too or not.

So far I tried:

-Loosening or tightening screws of CPU cooler, with no result.
-Applying thermal paste again and making sure that it is evenly spread, also no result.

Is it normal? What may be causing it?

I am not desperate for solving this problem because CPU won't reach even 80C in games anyway, but still, 20C difference on cores temps, and two cores running much hotter than other two is a bit strange.

Thanks for any help,
Edited by Krzych04650 - 6/18/16 at 2:35pm
post #2 of 12
I think it is physically impossible to have 15-20C difference within a 3x3cm processor, CPU's mechanism of providing temperature data is at fault here, I believe.
My Rig
(14 items)
 
Ex-wife's Rig
(15 items)
 
 
CPUMotherboardGraphicsRAM
Core i5 4460 AsRock H81M-DG4 Sapphire Rx470 Platinum KVR 1600 16Gb 
Hard DriveHard DriveCoolingOS
2x Seagate 3Tb Samsung 850 EVO 120 Scythe Ninja 3 Rev.B Windows 10 Pro 
MonitorKeyboardPowerCase
Fujitsu Siemens A17-2A Logitech K280e SuperFlower SF-550K12XP Thermaltake Versa H25 
MouseAudio
Logitech G402 Sony MDR XD150 
CPUMotherboardGraphicsRAM
Athlon 750K 4.0Ghz AsRock FM2A75 Pro4+ Sapphire R9 270X Dual-X Kingston 2x4Gb 1600 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 EVO 120  Western Digital 320Gb LiteON DVD-RW CoolerMaster Hyper Z600 
OSMonitorKeyboardPower
Windows 7 Pro x64 Toshiba 32" FullHD TV Logitech FSP Hexa 550 
CaseMouse
DeLUX Logitech 
  hide details  
Reply
My Rig
(14 items)
 
Ex-wife's Rig
(15 items)
 
 
CPUMotherboardGraphicsRAM
Core i5 4460 AsRock H81M-DG4 Sapphire Rx470 Platinum KVR 1600 16Gb 
Hard DriveHard DriveCoolingOS
2x Seagate 3Tb Samsung 850 EVO 120 Scythe Ninja 3 Rev.B Windows 10 Pro 
MonitorKeyboardPowerCase
Fujitsu Siemens A17-2A Logitech K280e SuperFlower SF-550K12XP Thermaltake Versa H25 
MouseAudio
Logitech G402 Sony MDR XD150 
CPUMotherboardGraphicsRAM
Athlon 750K 4.0Ghz AsRock FM2A75 Pro4+ Sapphire R9 270X Dual-X Kingston 2x4Gb 1600 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 EVO 120  Western Digital 320Gb LiteON DVD-RW CoolerMaster Hyper Z600 
OSMonitorKeyboardPower
Windows 7 Pro x64 Toshiba 32" FullHD TV Logitech FSP Hexa 550 
CaseMouse
DeLUX Logitech 
  hide details  
Reply
post #3 of 12
Thread Starter 
Quote:
Originally Posted by ronnin426850 View Post

I think it is physically impossible to have 15-20C difference within a 3x3cm processor, CPU's mechanism of providing temperature data is at fault here, I believe.

This make sense, last two cores are sitting at around 80C even with 4,7GHz 1,42V, this doesn't seem to be possible on mid-end air cooler in such demanding test like OCCT. AIDA 64 reports some throttling on 4,7 GHz 1,42V, so probably all cores are as hot as first two and temperature values for last two cores are incorrect. Thanks for answer.
post #4 of 12
There is a reason people delid CPUs. 10-20c difference between the hottest and coolest cores isn't unheard of. You probably just got a chip with a poor thermal paste job in between the IHS and Core + Intel went cheap with TIM that generation.
post #5 of 12
Quote:
Originally Posted by ronnin426850 View Post

I think it is physically impossible to have 15-20C difference within a 3x3cm processor, CPU's mechanism of providing temperature data is at fault here, I believe.

As a matter of fact it is possible, even within tens of micrometers, let alone centimeters depending on how well the structures are thermally coupled/isolated. Here's a thermal image of a transistor I'd taken (just a random device, not even a silicon cpu but it points out the physical possibility) where every pixel represents the integrated temperature of a 5x5 micrometer area. A temperature gradient of 20 degrees can clearly occur even within 20-25 micrometers.



Back to the topic, I also think this is a poorly applied adhesive issue between the chip surface and the cpu lid. I do have similar temperature differences between my first and fourth cores up to 14 degrees as well which is quite well known and apparently such drastic core temperature imbalance is what Intel thinks worth trading off to eliminate those solder cracking risks they've had in the past.
post #6 of 12
Quote:
Originally Posted by audiotest View Post

As a matter of fact it is possible, even within tens of micrometers, let alone centimeters depending on how well the structures are thermally coupled/isolated. Here's a thermal image of a transistor I'd taken (just a random device, not even a silicon cpu but it points out the physical possibility) where every pixel represents the integrated temperature of a 5x5 micrometer area. A temperature gradient of 20 degrees can clearly occur even within 20-25 micrometers.



Back to the topic, I also think this is a poorly applied adhesive issue between the chip surface and the cpu lid. I do have similar temperature differences between my first and fourth cores up to 14 degrees as well which is quite well known and apparently such drastic core temperature imbalance is what Intel thinks worth trading off to eliminate those solder cracking risks they've had in the past.

Oh, nice! Rep+ thumb.gif
My Rig
(14 items)
 
Ex-wife's Rig
(15 items)
 
 
CPUMotherboardGraphicsRAM
Core i5 4460 AsRock H81M-DG4 Sapphire Rx470 Platinum KVR 1600 16Gb 
Hard DriveHard DriveCoolingOS
2x Seagate 3Tb Samsung 850 EVO 120 Scythe Ninja 3 Rev.B Windows 10 Pro 
MonitorKeyboardPowerCase
Fujitsu Siemens A17-2A Logitech K280e SuperFlower SF-550K12XP Thermaltake Versa H25 
MouseAudio
Logitech G402 Sony MDR XD150 
CPUMotherboardGraphicsRAM
Athlon 750K 4.0Ghz AsRock FM2A75 Pro4+ Sapphire R9 270X Dual-X Kingston 2x4Gb 1600 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 EVO 120  Western Digital 320Gb LiteON DVD-RW CoolerMaster Hyper Z600 
OSMonitorKeyboardPower
Windows 7 Pro x64 Toshiba 32" FullHD TV Logitech FSP Hexa 550 
CaseMouse
DeLUX Logitech 
  hide details  
Reply
My Rig
(14 items)
 
Ex-wife's Rig
(15 items)
 
 
CPUMotherboardGraphicsRAM
Core i5 4460 AsRock H81M-DG4 Sapphire Rx470 Platinum KVR 1600 16Gb 
Hard DriveHard DriveCoolingOS
2x Seagate 3Tb Samsung 850 EVO 120 Scythe Ninja 3 Rev.B Windows 10 Pro 
MonitorKeyboardPowerCase
Fujitsu Siemens A17-2A Logitech K280e SuperFlower SF-550K12XP Thermaltake Versa H25 
MouseAudio
Logitech G402 Sony MDR XD150 
CPUMotherboardGraphicsRAM
Athlon 750K 4.0Ghz AsRock FM2A75 Pro4+ Sapphire R9 270X Dual-X Kingston 2x4Gb 1600 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 EVO 120  Western Digital 320Gb LiteON DVD-RW CoolerMaster Hyper Z600 
OSMonitorKeyboardPower
Windows 7 Pro x64 Toshiba 32" FullHD TV Logitech FSP Hexa 550 
CaseMouse
DeLUX Logitech 
  hide details  
Reply
post #7 of 12
Quote:
Originally Posted by boot318 View Post

There is a reason people delid CPUs. 10-20c difference between the hottest and coolest cores isn't unheard of. You probably just got a chip with a poor thermal paste job in between the IHS and Core + Intel went cheap with TIM that generation.

And this is why people believe the misinformation that floats around online and on the forums lol.

1. The Intel TIM is probably better then all TIMs on the market period.

2. A 15°C difference isn't unheard of, just to show you the difference between delided and not delided. Mine has a variance of 4°C right now just browsing around and what not, under load its probably around the same if not lower.

3. The issue isn't the TIM at all, any TIM, no matter how cheap, will not perform at its best if the pressure applied between surfaces is poor. And thats the issue here. Heres a picture of a freshly delided 2 years 3350p ivy bridge. See how the paste on the TIM isnt even pushed/squished on the IHS? Thats because the silicone holds the IHS slightly off the die. Even a .04mm (which if you have a digital caliper is ridiculously tiny) is enough to raise your temps.

4. The reason people use liquid metal is, it does not push out or turn to liquid compared to TIM, and you can use a bit more then you would a TIM and it wont affect thermal efficiency.

    
CPUMotherboardGraphicsRAM
R7 1700X asus crosshair 6 Sapphire R9 390 Nitro CMK16GX4M2B3200C16 
Hard DriveHard DriveCoolingCooling
Kingston HyperX 3K Seagate ST2000DM006 EKWB Supremacy Evo EK-XRES 100 Revo D5 - Acetal 
CoolingCoolingCoolingCooling
Alphacool GPX r9 390 m01 alphacool xt45 240 Alphacool xt45 360 Alphacool D5 PWM 
OSMonitorKeyboardPower
Windows 10 LG 29UM68 Corsair Strafe Corsair RM1000 
CaseMouseMouse Pad
Thermaltake X5 Corsair M65 Pro RGB Razer Goliathus Speed 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
R7 1700X asus crosshair 6 Sapphire R9 390 Nitro CMK16GX4M2B3200C16 
Hard DriveHard DriveCoolingCooling
Kingston HyperX 3K Seagate ST2000DM006 EKWB Supremacy Evo EK-XRES 100 Revo D5 - Acetal 
CoolingCoolingCoolingCooling
Alphacool GPX r9 390 m01 alphacool xt45 240 Alphacool xt45 360 Alphacool D5 PWM 
OSMonitorKeyboardPower
Windows 10 LG 29UM68 Corsair Strafe Corsair RM1000 
CaseMouseMouse Pad
Thermaltake X5 Corsair M65 Pro RGB Razer Goliathus Speed 
  hide details  
Reply
post #8 of 12
Intels polymer TIM is still a paste. Intel white papers stated reason for solder is 80 w/mk versus intels polymer TIM at 5 w/mk heat conductance. Heat conductance is governed by 3 factors, bulk conductance of TIM, surface contact resistance, and bondline thickness (how thin it is spread). So not just the bondline thickness.

Intel tim at 5 w/mk is equivalent with the best pastes on the market that is true. But it pales in comparison to liquid metal at 40 w/mk. It is partly thinner bondline of liquid metal, partly lower contact resistance, and partly the higher bulk conductance. Most that tested after delidding (myself included) got about 1/2 the effect of using paste after delidding vs liquid metal.

After delidding, applying mx 2, NTH 1, and couple others, best was 10-12C decrease at testing parameters I was using. But 22C drop with CLP. There is no way to determine exactly how much is contact resistance vs thinner bondline vs higher bulk conductance.
My System
(15 items)
 
  
CPUMotherboardGraphicsRAM
6900K 4.4ghz  MSI X99A Krait Titan XP Gskill 32GB 3200 C14 
Hard DriveHard DriveCoolingOS
Samsung 960 Pro 1TB OS Samsung 850 pro 1TB games D5, XT45 280 rad, Nemesis 280 rad,  Windows 10 64bit 
MonitorPowerCase
Acer Predator XB271HU 144hz Seasonic Prime Titanium 850 Fractal Design R5 
  hide details  
Reply
My System
(15 items)
 
  
CPUMotherboardGraphicsRAM
6900K 4.4ghz  MSI X99A Krait Titan XP Gskill 32GB 3200 C14 
Hard DriveHard DriveCoolingOS
Samsung 960 Pro 1TB OS Samsung 850 pro 1TB games D5, XT45 280 rad, Nemesis 280 rad,  Windows 10 64bit 
MonitorPowerCase
Acer Predator XB271HU 144hz Seasonic Prime Titanium 850 Fractal Design R5 
  hide details  
Reply
post #9 of 12
Quote:
Originally Posted by opt33 View Post

Intels polymer TIM is still a paste. Intel white papers stated reason for solder is 80 w/mk versus intels polymer TIM at 5 w/mk heat conductance. Heat conductance is governed by 3 factors, bulk conductance of TIM, surface contact resistance, and bondline thickness (how thin it is spread). So not just the bondline thickness.

Intel tim at 5 w/mk is equivalent with the best pastes on the market that is true. But it pales in comparison to liquid metal at 40 w/mk. It is partly thinner bondline of liquid metal, partly lower contact resistance, and partly the higher bulk conductance. Most that tested after delidding (myself included) got about 1/2 the effect of using paste after delidding vs liquid metal.

After delidding, applying mx 2, NTH 1, and couple others, best was 10-12C decrease at testing parameters I was using. But 22C drop with CLP. There is no way to determine exactly how much is contact resistance vs thinner bondline vs higher bulk conductance.

The difference between a tim and liquid metal is that liquid metal has better thermal conductivity not just a higher w/mK (which btw is meaningless because nobody measures temperatures on their PC in Kelvin lol.)

I got absolutely IDENTICAL temps, within a margin of error, when using a TIM and CLU on my bare die. This throws your argument out the window. Oh and i used 2 different TIMs as well.

That w/mK is meaningless, if you convert it to Celcius like all computer parts are measure in you get .05w/cmC compared to .36w/cmC. Sounds like a big difference but its not, its a difference of .31w (watts), for every cm. Heat on cpus doesnt have to travel a meter let alone a cm.

The thing is no matter what you do youre going to get a temp drop, and the reason why you get more in CLU isnt the rating but the fact that it makes WAY better contact under an IHS then the paste does. Again ive tested this and went from 53°C with clu/ihs to 46°C with bare die. Running NH-T1 on bare die got about 48°C, running Hydronaut got about 48°C, running CLU got about 46°C. But clu is far more reliable in the long run. Bare die gets way too hot for MOST TIMs.

The fact that deliding works is it gets rid of the gap caused by the silicone, pretty much has nothing to do with the TIM.

P.S. Someone tested conductonaut which is rated at 86 w/mK!!!!!!!!!! And saw no difference between CLU/CLP.

For any metal under pressure, you want the LEAST amount of TIM possible, you want the metals to make as much contact as possible as they while dissipate far more heat then 5-36w/mK.
    
CPUMotherboardGraphicsRAM
R7 1700X asus crosshair 6 Sapphire R9 390 Nitro CMK16GX4M2B3200C16 
Hard DriveHard DriveCoolingCooling
Kingston HyperX 3K Seagate ST2000DM006 EKWB Supremacy Evo EK-XRES 100 Revo D5 - Acetal 
CoolingCoolingCoolingCooling
Alphacool GPX r9 390 m01 alphacool xt45 240 Alphacool xt45 360 Alphacool D5 PWM 
OSMonitorKeyboardPower
Windows 10 LG 29UM68 Corsair Strafe Corsair RM1000 
CaseMouseMouse Pad
Thermaltake X5 Corsair M65 Pro RGB Razer Goliathus Speed 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
R7 1700X asus crosshair 6 Sapphire R9 390 Nitro CMK16GX4M2B3200C16 
Hard DriveHard DriveCoolingCooling
Kingston HyperX 3K Seagate ST2000DM006 EKWB Supremacy Evo EK-XRES 100 Revo D5 - Acetal 
CoolingCoolingCoolingCooling
Alphacool GPX r9 390 m01 alphacool xt45 240 Alphacool xt45 360 Alphacool D5 PWM 
OSMonitorKeyboardPower
Windows 10 LG 29UM68 Corsair Strafe Corsair RM1000 
CaseMouseMouse Pad
Thermaltake X5 Corsair M65 Pro RGB Razer Goliathus Speed 
  hide details  
Reply
post #10 of 12
Quote:
Originally Posted by bluej511 View Post

The difference between a tim and liquid metal is that liquid metal has better thermal conductivity not just a higher w/mK (which btw is meaningless because nobody measures temperatures on their PC in Kelvin lol.)

I got absolutely IDENTICAL temps, within a margin of error, when using a TIM and CLU on my bare die. This throws your argument out the window. Oh and i used 2 different TIMs as well.

That w/mK is meaningless, if you convert it to Celcius like all computer parts are measure in you get .05w/cmC compared to .36w/cmC. Sounds like a big difference but its not, its a difference of .31w (watts), for every cm. Heat on cpus doesnt have to travel a meter let alone a cm.

The thing is no matter what you do youre going to get a temp drop, and the reason why you get more in CLU isnt the rating but the fact that it makes WAY better contact under an IHS then the paste does. Again ive tested this and went from 53°C with clu/ihs to 46°C with bare die. Running NH-T1 on bare die got about 48°C, running Hydronaut got about 48°C, running CLU got about 46°C. But clu is far more reliable in the long run. Bare die gets way too hot for MOST TIMs.

The fact that deliding works is it gets rid of the gap caused by the silicone, pretty much has nothing to do with the TIM.

P.S. Someone tested conductonaut which is rated at 86 w/mK!!!!!!!!!! And saw no difference between CLU/CLP.

For any metal under pressure, you want the LEAST amount of TIM possible, you want the metals to make as much contact as possible as they while dissipate far more heat then 5-36w/mK.

thermal conductivity is w/mk. You just said better thermal conductivity not just higher thermal conductivity, since you didnt understand units. you need to learn how to convert units, if you did it correctly would be the same.

About 50 people found exactly what I did in the delid thread, you get about 50% improvement with paste, the other 50% with liquid metal. There is no liquid metal or other that is 86 w/mk, as all based on 40 w/mk gallium. Only intel mostly indium solder is 80 w/mk.

Yes part of it is the gap, which is correctly called bondline thickness, obviously the thinner the better. Solder however, per intel, has roughly the same bondline thickness as its paste, has the same silicon causing the same "gap", yet it is 80 w/mk thermal conductance and it performs very much like liquid metal.

Having a thick bondline or "gap' at 5 w/mk is a problem. Having a thick bondline or "gap" at 80 w/mk of solder is less of a problem.
My System
(15 items)
 
  
CPUMotherboardGraphicsRAM
6900K 4.4ghz  MSI X99A Krait Titan XP Gskill 32GB 3200 C14 
Hard DriveHard DriveCoolingOS
Samsung 960 Pro 1TB OS Samsung 850 pro 1TB games D5, XT45 280 rad, Nemesis 280 rad,  Windows 10 64bit 
MonitorPowerCase
Acer Predator XB271HU 144hz Seasonic Prime Titanium 850 Fractal Design R5 
  hide details  
Reply
My System
(15 items)
 
  
CPUMotherboardGraphicsRAM
6900K 4.4ghz  MSI X99A Krait Titan XP Gskill 32GB 3200 C14 
Hard DriveHard DriveCoolingOS
Samsung 960 Pro 1TB OS Samsung 850 pro 1TB games D5, XT45 280 rad, Nemesis 280 rad,  Windows 10 64bit 
MonitorPowerCase
Acer Predator XB271HU 144hz Seasonic Prime Titanium 850 Fractal Design R5 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel CPUs
Overclock.net › Forums › Intel › Intel CPUs › Huge temperature difference on cores (4690K 4,5 GHz)