Thermal sensor issue i7-7700k?

[CK7]

I have a brand new build; EVERYTHING NEW. i7-7700k is running at stock speeds. I have the RAM set to XMP for DDR4-2666. Motherboard is Asus Maximus IX Hero Z270.

I have found that the i7-7700k reports a momentary (a second or less) temperature spike +25 > 35 degrees Celsius anytime a program is opened, a webpage is opened, a background app runs etc. The temperature blip cascades through the cores in random order; not the same every time. This causes my heatsink fan to constantly cycle up and down. Temperatures otherwise report as steady, normal increases. Peak temperature under Prime95 blend test is 71 degrees Celsius.

Attempted solutions:

I have re-installed my heatsink and thermal paste with no change.

I have tried to manually set my fan speed in the bios. The only setting that avoids this issue is setting the temperature / fan at a constant (and loud) 80-100%. I've tried PWM and DC mode.

I have found a few user reports elsewhere on the web, all reasoning that it's just the way it is. I don't accept that. Opening a folder or browser should not spike temps +30 degrees. Not only is the fan cycling annoying, it puts undue stress on my fan; possibly shortening its lifespan.

What's the answer, if any? RMA?

[TGrab1]

Ok so here is the basics.

To answer your question first the Cache Frequency should be left alone. That is for overclocking something different and it something you do later on once you have a stable core overclock. Even then most people leave the cache alone as the gains are minimal.

So first You should start with leaving everything at auto except for 2 things. You want to see your turbo clocks on your cpu 1-4 to 48x. That will make sure they are all running at 4.8ghz. Next for now I would dial in my voltage manually instead of offset. Offset and other auto settings are bad about setting too high a voltage which will increase your temps.

I would suggest setting your manual voltage to 1.285 to start out.

So here is where it gets time consuming. What you will want to do is set that manual voltage. Then start windows and run a stress test of your choosing for a set amount of time you are comfortable with. If it passes drop the voltage each time by .05. If it fails bump the voltage up .05.

So doing this you will be slowly moving up/down in voltage looking for a point where it fails/passes.. So if it is failing then passes after a voltage bump you would then slowly drop the voltage by .01 steps until it fails again then go back to the previous good voltage (.01 higher). If it was passing and you dropped it .05 and it fails... then you would just add .01 voltage until it passes again.

This is of course a very simple way of explaining it, but it will get you there. I would suggest checking out overclockers.com for further more in detail information. Because once you do get a good manual voltage you can then see what is needed at those speeds and then use adaptive or offset mode to ramp your voltage up and down based on usage etc.. which will make it more energy efficient while also dropping temps at idle etc.

Hope this helps a little.

[idata]

Many thanks! ... I'll give that a try and take this up further on Overclockers if necessary.

Cheers,

Robert

[idata]

Or you can do like I did and go just for 4.7gz (47x) and use 1.2v on Vcore. And your temp problems are gone (check my video couple post back on this thread).

[idata]

You have a good chip - I needed just a little extra - 1.24v before I could get 4.7 and not fatal error in Prime.

[LRibe2]

Hi! Sure.

But first let me run a few more tests and I'll put it all here. probably tomorrow I'll post it!

[idata]

That has made a huge difference ... many thanks!

At 47 and with your settings AIDA64 stress test gave a max of 70C and average of 60C which is a good 15C lower than the max I was getting.

At 48 and with your settings AIDA64 stress test blue-screened.

I increased the VCore to 1.22 and the system was stable at pretty much the same temps. I also ran the Intel Extreme Tunining Utility stress test and this gives a lower temperature of about 66C max which is really great.

I take it that the VCore is now always at 1.22? A couple of questions:

- How much can I safely increase VCore to? I would like to see how far I can push the CPU without smoke ... not that I need it or would run at anything higher than 4.8GHz.

- Is there a setting that would throttle back VCore automatically from 1.22 at lower frequencies? I've had a look at apart from SVID Support (which you say to disable) none jump out at me.

Cheers

Robert

[TGrab1]

Yes, You can use either offset or Adaptive. Basically what you do is put it back on auto and then boot into windows and check what your ACTUAL VID is during a stress test ... This can be done with HWINFO. After you have found out what your auto usage us at those volts you can figure out what settings need to be for offset and adaptive.

And example with offset would be... When running on auto you are using 1.32Volts, but to be stable at 4.8 you only needed 1.22 volts. So in offset mod you would set and offset of a - and the value being .10. Which would make it run are 1.22 at the top end and lower at lower clocks etc.. basically still an auto config just you are setting how much to increase/decrease it by.

Adaptive would be much the same. You would set your maximum boost amount to = your stable volt amount etc.

[idata]

Thank you!

Well I've partially succeeded, but not to a running system.

To start off with I clocked up to 4.9GHz with a vCore of 1.26 (it was stable at 1.24 but I gave it a little extra juice for comfort ... whether that is a good idea or not I do not know). Temperatures are Max 72C, Average 65C with AIDA64 Stress Test (stree CPU, FPU, Cache and System Memory), which is pretty extreme, so I'm happy with that setting. I also ran a bunch of applications (Photoshop with some heavy filters and resizing plug-ins running at pretty much 100%, web etc) and it seems fine.

I then reset to Auto with a clock of 4.9GHz and checked the voltage using AIDA64 again. The max value was 1.45v (with temperatures going above 90C, so I didn't run the test for long!).

I then went back and set:

- CPU Core/Cache to Offset Mode

- CPU Mode Sign to '-'

- CPU Core Voltage Offset to 0.19 (so 1.45-0.19 = 1.26)

- CPU VCCIO to 1.1 (as before)

- CPU System I/O Voltage to 1.1 (as before)

On exit the BIOS just sat there with a blank screen. I powered off and tried again, same thing.

Should I be setting the VCCIO and System I/O Voltage to auto?

Thanks

Robert

[idata]

I've tried various combinations to set the CPU Core/Cache Voltage to Offset but none of them work.

I would really appreciate it if you could point me in the right direction, starting from:

- SVID: Disabled

- CPU Core Cache Voltage: Manual

- CPU Core Voltage Override: 1.26

- CPU VCCIO: 1.10

- CPU System I/O Voltage: 1.10

I've measured the offset that I need to apply as -0.1V (the max under load is 1.36V).

Many thanks!

Robert

[LRibe2]

Hi again guys.

So after a few days of testing I got to this conclusion, there is definitly a overheating issue with this CPUs...

I've tried a bunch of setups, manually switching CPU voltage values. It's impossible to find a balance between voltage / temperature and base clock running stable.

At this time I have my CPU running at 4.6Ghz | Vcore Voltage set to 1.23V | PLL Voltage 0.9V | and other voltages set to auto and this is WHY....

I can really start and run the PC with lower Voltage and higher base clocks but... if I really push it to the limit of workload it will fail... ignoring the temperatures of prime95, the pc will allways fail the stability tests here or eventually crash randomly when I really thought it was stable.

The minimum stable OC for me is 4.6Ghz 1.23V and values set to auto and pll 0.9V and this are the temperatures:

Intel burn test 79º 2 min test...

Prime95 85º - 15 minute test....

Aida 64... dont even bother... this test isnt reliable since is the only stress test that never gave me stability or temperature problems so I assume it's not a full load test.

Asus Realbench - 79º 10min

Really high right? minimum OC....

So... as you all can see, I can live in the ilusion that my cpu is able to OC preety well cuz I can set it up to 4.8 Ghz with 1.23 or less Volts but.. the truth is... it will fail.. it's not working properly.

I can play BF1 or use other workload programs and I may not have a problem but the thing is... I dare anyone on air cooling to OC or even without OC run an intel burn test or prime95 for 1 Hour or even more.. Then check the temperatures and be amazed.

My brother uses an Intel i7 2600 And all this tests dont even touch 70º on an 5 year old cooler and thermal paste and a lousy case that dosent have a good airflow.

Just a few last words... I was really able to down this temperatures about 10º but never got the system stable.. even if it seems to be stable if I ran An exaustive stress test the PC will eventually Crash / Blue screen.. even if it takes an hour or two.

With higher temperatures I'm not sure because I dont even dare to toast the CPU...

Starting to get really tired of this subject.

Thank you all!!

PS: By the way, I believe that this is a TIM problem because if you compare the temperatures rise with older generation CPUs you will find that the 7700k dosent have the ability to keep it self cool,. A few seconds after you start a stress test, temperatures escalate instantly. In older generations like Sandy bridge, the maximum temperatures rise slowly and only after minutes of full load...like it's starting to heat up naturally and not instantly.

[idata]

Yes, well that's not good. There is clearly a huge variation between these chips, which certainly indicates that the TIM has been applied in a very inconsistent manner. With my CPU I can get 4.9GHz at 1.24 volts with the temperature not going over 73C. But I may be fooling myself because I've only run the AIDA64 stress test for any length of time (couple hours).

I take it that the Prime 95 is not doing AVX instructions? Because if so you would want to set an offset of maybe 2 to make sure that the AVX test isn't what is causing the crashing.

[idata]

Uhm..somehow I dont think you have your OC settings on..(Turbo mode..)

Just to be sure check from your Cpu-Z program that you are running right core speed ==>

I was doing Prime test and wondering why I can get so low core voltages but I had this on..lol ==>

My test numbers (these are with my settings and with my CPU so might not work on onthers):

- 4.7Ghz need 1.20 Vcore (100% working)

- 4.8Ghz need 1.26 Vcore (100% working)

- 4.9Ghz need 1.30 Vcore (100% working)

- 5.0Ghz need 1.35 Vcore (100% working)

So not much point for me to go 4.9..only 200 mhz more and have to push 0.10 Vcore..

[LRibe2]

In that case I advise you to try an extensive stress test with intel burn test and prime95.

Aida64 shows temperatures under 70 in my build with no problems. So, I think that you may find temperature issues or even test failures in this two.

If you try it, let me know because a few days ago I was really happy to have been able to reduce temperatures and OC pretty well, and later found out that the CPU fails a few demanding tests or if it dosent fail, temperatures reach insane values.

Thanks!

[idata]

Are you running in Manual or Adaptive?

Also what is the Advance Calibration Settings from? It doesn't look like the Prime95 I have.

[idata]

Well I've run Prime95 in various modes and here are the results - with the BIOS set to Adaptive Mode with a boost voltage of 1.30V and multipler of 47 (just as an initial rough setting).

and this is the package temperature graph:

as you can see the temperature is very stable with a peak of 77C and average of 75C. I've run the tests for well over an hour and everything seems stable ... CPU temperature, coolant temperature, PSU etc.

I would expect to be able to improve on this by lowering the boost voltage (in previous testing I had a stable 49 multipler at 1.27V. I'll try that after I've left the system running as is for a while.

I HAVE disable AVX testing in Prime95 though. If I was to enable that I would most likely use the ASUS Thermal Control Tool. But as I have no progams that use AVX instructions I'm not too worried at this point.

One thing that hasn't been mentioned is the PSU. I have a very good power supply (Corsair HX750i) and it holds the rails absolutely steady under load changes. It's also well over-spec for the loads I have. A lesser power supply might cause problems ... it would be interesting to get some feedback on this from you guys.

Robert

[LRibe2]

Hi again..

You are right, I wasnt paying attention.

Changing the negative offset on the avx instructions (option that I've never cared while setting voltages), makes all the diference and I'm able to OC much better and maintain stability.

Even so, to be able to OC to 4.7Ghz at 1.23V and run prime95 under 75 with no crashs I have to set negative offset on the AVX instructions to 8.

And, of course, any benchmark will show a much lower score than usual..

Dosent seem like a good solution.

Any way, thank you!

[TGrab1]

OverDriver_PT83 wrote:

Hi again guys.

So after a few days of testing I got to this conclusion, there is definitly a overheating issue with this CPUs...

I've tried a bunch of setups, manually switching CPU voltage values. It's impossible to find a balance between voltage / temperature and base clock running stable.

At this time I have my CPU running at 4.6Ghz | Vcore Voltage set to 1.23V | PLL Voltage 0.9V | and other voltages set to auto and this is WHY....

I can really start and run the PC with lower Voltage and higher base clocks but... if I really push it to the limit of workload it will fail... ignoring the temperatures of prime95, the pc will allways fail the stability tests here or eventually crash randomly when I really thought it was stable.

The minimum stable OC for me is 4.6Ghz 1.23V and values set to auto and pll 0.9V and this are the temperatures:

Intel burn test 79º 2 min test...

Prime95 85º - 15 minute test....

Aida 64... dont even bother... this test isnt reliable since is the only stress test that never gave me stability or temperature problems so I assume it's not a full load test.

Asus Realbench - 79º 10min

Really high right? minimum OC....

So... as you all can see, I can live in the ilusion that my cpu is able to OC preety well cuz I can set it up to 4.8 Ghz with 1.23 or less Volts but.. the truth is... it will fail.. it's not working properly.

I can play BF1 or use other workload programs and I may not have a problem but the thing is... I dare anyone on air cooling to OC or even without OC run an intel burn test or prime95 for 1 Hour or even more.. Then check the temperatures and be amazed.

My brother uses an Intel i7 2600 And all this tests dont even touch 70º on an 5 year old cooler and thermal paste and a lousy case that dosent have a good airflow.

Just a few last words... I was really able to down this temperatures about 10º but never got the system stable.. even if it seems to be stable if I ran An exaustive stress test the PC will eventually Crash / Blue screen.. even if it takes an hour or two.

With higher temperatures I'm not sure because I dont even dare to toast the CPU...

Starting to get really tired of this subject.

Thank you all!!

PS: By the way, I believe that this is a TIM problem because if you compare the temperatures rise with older generation CPUs you will find that the 7700k dosent have the ability to keep it self cool,. A few seconds after you start a stress test, temperatures escalate instantly. In older generations like Sandy bridge, the maximum temperatures rise slowly and only after minutes of full load...like it's starting to heat up naturally and not instantly.

First off the 2600 you are referring to would have much lower temps. That is because the DIE is soldered onto the IHS which leads to much better thermal transfer than the TIM being used on the 7700k.

Second each CPU is different. A setting that will work for me, might not work for you, might not work for anyone else for that matter. Each CPU requires different voltage for different clocks and can have different thermal levels a long the way.

You unfortunately seem to be suffering from the same problem I had initially. Your TIM/Silicon Seal/ etc is not working properly. I was having major temp issues that pretty much made overclocking for me impossible. I could get a nice overclock at decent voltages, but my temps would instantly jump up into the 90's which on my custom loop made little to no sense.

I corrected my issue by delidding my cpu, but I would not suggest that because of the risk you take in doing so. Depending on where you purchased your CPU you might be able to swap it for another. Microcenter for one will let you return CPU's like crazy within a 15 day period. Which can be good for people playing the silicon lottery.

I complained to intel in another thread about my temps and they offered me an RMA option. So that could also be a possibility for you. In the end some chips are just better than others for various reasons and you might unfortunately have a under-performing chip via temps.

[idata]

Regarding the AVX offset ... you misunderstood me. What I meant was to leave the offset at 0, but to disable AVX testing in Prime95, because AVX instructions are apparently very heavy and very few applications use them.

I found that putting an offset in the AVX field reduces all the frequency by that amount for all loads, with or without AVX. That might just me on my board (Prime Z270-A), so it's useless. That might be because I have Adaptive mode on ... something I need to check (or perhaps someone knows the answer).

I think 1.23V at 4.7GHz sounds way too low ... you may only be running at 3.9GHz

[LRibe2]

Hi TGrable!

I understand that the CPUs vary much from unit to unit but, seriously, this dosent seem like a limitation of my CPU only.

After trying out the AVX offset settings I found that overclocking voltages and Bclock dont really make a huge difference in temperatures between 4.6 to 4.9 Ghz

Meenwhile, AVX instructions, used by prime95 and intel burn test completly frie the CPU.

I no longer have temperature issues but at the same time, I'm loosing all the AVX capabilities. I know that AVX is not often used but.. it dosent make sense to sell a CPU supporting AVX2 that simply cant handle it right.. I've also found out that this issue affects other CPU gens from intel, like skylake, broadwell-e.. and so on...

here's a few quotes:

Broadwell-e

"Unlike on server platforms, the AVX Offset register is open for adjustment on Broadwell-E processors, which makes it possible for us to define the AVX workload ratio. The adjustment option is certainly welcome. However, the mandatory condition is that we have to leave processor voltage in Auto mode to reduce operating temperatures under AVX workloads, and this has implications for overclocking "

Kaby Lake

"

Ordinarily, an overclock would be constrained by the hottest, most stressful application we run on the system. By using AVX offset, we can run non-AVX workloads, which don't consume as much power, at a higher CPU frequency than AVX applications.

As an example, that means we can apply 50X ratio with a BCLK of 100MHz, resulting in a 5GHz overclock for non-AVX applications. We can then set the AVX offset parameter to a value 2 (or lower if required), which will reduce the CPU core ratio to 48X (4.8GHz) when an AVX workload is detected. This ensures the system stays within its thermal envelope, without our overclock being constrained solely by the AVX workload."

Skylake

"If you overclock and run applications which use AVX, you may need to reduce Vcore and clock speed and / or upgrade your cooling so that Core temperatures don't exceed 80C."

All info from Asus.

Ok, I get it. AVX isn't used very often in real life applications, and the offset option gives us the margin to overclock and keep teperatures controled under AVX workload..

This way I am able to overclock up to 5Ghz below 80º Max.. But once again loosing the option to maintain the same performance on AVX.

I think I can live with it.

This makes me a bit sad, because even though Sandy bridge has better thermal performance, it's hard to accept that Sandy bridge can handle AVX and overclock with no limitations while kaby lake simply can't.

Any way. I've learned much with all of you and this thread. Also, got to the conclusion that, there is no solution to this issue or the spikes issue (the last one dosen't bother me that much).

Regards!

[TGrab1]

Except sandybridge doesn't handle avx 2.0. That is where the issue lies. Sandybridge does not support AVX 2.0 and that is also what is causing issues on the 7700k. It is much more intensive on the cpu than 1.0 was.

That being said your problem is heat. So that means it could be one of a few things.

• Your Contact between DIE and IHS is bad. This could be TIM issues or Gap issues.

• Your Contact between IHS and Cooling Solution is bad. Same thing could be mounting pressure, tim, or another issue in that same range.

• Your Cooling solution does not dissipate enough heat. This could be too little surface area, weak fans, bad case air flow, etc..

• Lastly your voltage may be too high which only makes any issue with the previous listed items even worse.

So if you have one of those issues that will cause temps to be higher than they should. If you have multiple issues from above then it will only multiply from there.

So make sure you check all other options before you completely blame the chip. If you are running stock speeds what temps and voltage does it run? What is your cooling solution? What TIM did you use and what Application method. What type of case and what type of fan configuration are being used?

These are all important questions and until you have checked all other possibilities will you know if it is 100% chip related.

[LRibe2]

Hi again.

I find it really hard to accept that this processor has thermal issues besides the TIM and this is why.

My build:

Case: Bitfenix prodigy M + 5 Fans (2 in / 3 Out + CPU Fan all of them are corsair SP120 1500RPM)

MOBO: Asus Prime Z270M

Cooler:Cooler master Evo 212 - Everyone knows this one

PSU: Corsair VS650

Graphics card: Gigabyte GTX 1060 6G WF OC

RAM: Crucial Ballistix LT 32GB DDR4 2400Mhz @ 1.2V

If I have heating problems, then why am I able to have idle temperatures of 21C/22C in all cores, case temperatures of 16/17C, Run non-AVX instructions at full load up to 4.9Ghz below 73C But when using AVX instructions the CPU simply blows up to more than 85C? even if I use offset down to 4.4Ghz on the AVX instructions, it goes up to 79/ 80C in the first 2minutes.. It simply dosent make sense... and if I turn off the AVX instructions, temperatures go down to under 30C immediatly. This CPU dosent show signs of thermal problems, the thermal problem here is that if I leave AVX offset alone, the power consumption goes up to 120W or something like that while running this instructions, causing heat issues or instability if voltage is too low.

This happens even at stock speeds.. reaching 80C or higher

Every single review that made exhaustive testing on this CPU shows the same thing .. temperatures blowing up to 80C+ even 90 or 100C.

I know that deliding the CPU would help but I have an 3 year warranty that I dont want to lose.