6850K CPU Throttling Cores

[Krorghar]

Greetings Guys.

I Have Increase Core Voltage From 1.25 to 1.30v to get more Stability, but When passing Prime95 test to the system, I have noticed that now the CPU is throttling his Cores.

First, as some of you suggested, we thought it may be Thermal Throttling, but after monitorizing the CPU with CoreTemp, none of the cores passed 71 Cº so that possibility is very unlikely.


So now I think that due to the power demand increase, something in the OS or the Bios is limiting the power comsumption and force the cores to throttle when the CPU demands more power.

As It's my first Overclock I'm a bit lost here, any clue about what may be causing this throttling?

Many Thanks.

 

[Burpo]

Without knowing what motherboard you're using, it is impossible to even guess.

 

[Hail The Brain Slug]

1) Can you explain what kind of throttling behavior you are seeing?

2) BW-E has what's called an AVX offset multiplier where it lowers the multiplier by the offset when AVX instructions are detected to improve stability. If you're using a newer P95 it has AVX and could trigger the offset to take effect.

3) If you install Intel XTU and enable all the monitoring, you can see if you have thermal, power limit, current limit, or other throttling occuring.

 

[Krorghar]

Thanks for Answering Guys.

Yup Burpo, sorry forgot to add that data, Motherboard is Asus Rog Rampage V Edition 10, bios is flashed to the last version.

Will try to explain Xabana , I use Coretemp 1.4 to monitor Temp and Load on all the six cores. Cores are acting really weird, when Prime Starts, All cores seems to work properly, but after a few seconds , Load on some cores Drop to 2-3% as they were not working, while others work at allmost max load, then the effect randomly reverts and some of the cores that were working go idle, while some of the other that were not working start to work.

About the AVX Offset multiplier, I remember having set it to 3 because I was following some Internet OC Guide, this may be the problem, should I set it to 0? Anyway, If this is the problem, Why there was no throttling when core Voltage was 1.25v and now that is 1.30 I'm experiencing this issue?

Didn't knew that program, will download and Install it ASAP and see whats the problem.

Many thanks Guys, will keep you Informed

 

[Krorghar]

Ok I have Changed Power Profile In W10 to a custom one with max CPU usage. Then entered the BIOS and set the AVX Offset Multiplier to min I Could that was 1, I mean that I tried to input 0, but the Bios just changed it to 1.

Downloaded Intel XTU and ran Prime95 v289 . Al Cores Stressed properly, but Intel XTU said that the number of Active Cores Was 3. On the other hand, CoreTemp was showing that the 6 cores were Stressed, please check PIC:

Also, sometimes Prime95 says "Unable to Detect some of the Hyperthreaded logical CPUs" And mess with the number of cores my system has. This is begining to give me a Headache, any clues pls Guys?

Many thanks

 

[DAPUNISHER]

What CPU cooling are you using? AIO liquid, custom liquid, air, and which one? More curiosity than anything else, your temps look good.

This is your first overclock and you chose a challenging platform to tackle the challenge on. If you chose a guide like overclock. net provides, they tend to assume a familiarity with the process. Even if not the exact platform, mainboard, or CPU in question.

The Asus thermal control tool may be just what you need. I suggest you read the following links if you have not already done so.

- http://edgeup.asus.com/2016/05/31/g...dwell-e-processors-asus-thermal-control-tool/

Next - http://edgeup.asus.com/2016/06/17/broadwell-e-overclocking-guide/

 

[Hail The Brain Slug]

At this point all I can think of is that it's a reporting error.

Sometimes Prime95 doesn't correctly detect your CPU or cores. In that case, it may not spawn enough worker threads or distribute them among your logical CPUs to fully load your CPU. If the issue only occurs when Prime95 indicates it improperly detected your CPU, I wouldn't worry about it.

Just make sure you get a proper detection like this and try it one more time.

 

[Krorghar]

@DAPUNISHER

Cooling Is Achieved with a Corsair H115i waterp pump. I have modified it a little from stock. First I passed on the preaplied Thermal paste, and used Thermal Grizzly one, after reading some comparatives that placed it on first possition above other well know brands. Then I replaced Stock fans with 4xNoctua P3000 in push pull configuration, the result is quite decent as you can see from the temps.

I have read many guides about overclocking but I'm far from being an expert thats true. Have read the two edgeup.asus guides, in fact were the first I looked into.

@Xabanak

the test from the pics was allready made when P95 was detecting my cores and threads properly, Anyway I have noticed something that could not be in order. If you look at the CPU-Z report it says a max TDP of 140w . It seems that something in the Bios may be capping my system to 140w?, An 4.4Ghz Overclocked 6850K should eat about 200w according with the reviews.

 

[DAPUNISHER]

@DAPUNISHER

Cooling Is Achieved with a Corsair H115i waterp pump. I have modified it a little from stock. First I passed on the preaplied Thermal paste, and used Thermal Grizzly one, after reading some comparatives that placed it on first possition above other well know brands. Then I replaced Stock fans with 4xNoctua P3000 in push pull configuration, the result is quite decent as you can see from the temps.

I have read many guides about overclocking but I'm far from being an expert thats true. Have read the two edgeup.asus guides, in fact were the first I looked into.

I hope I did not offend you, that was not my intent. And happy to read you did your research. Did you give the thermal control tool a go with the advice and suggestions they offered? The trial and error process can be tedious, but that is all part of the OC hobby.

 

[Hail The Brain Slug]

the test from the pics was allready made when P95 was detecting my cores and threads properly, Anyway I have noticed something that could not be in order. If you look at the CPU-Z report it says a max TDP of 140w . It seems that something in the Bios may be capping my system to 140w?, An 4.4Ghz Overclocked 6850K should eat about 200w according with the reviews.

If that were the case, XTU would report it as Power Limit Throttling - enable all types of throttling in the monitor and you will quickly see if that is the case.

 

[Krorghar]

I hope I did not offend you, that was not my intent. And happy to read you did your research. Did you give the thermal control tool a go with the advice and suggestions they offered? The trial and error process can be tedious, but that is all part of the OC hobby.

Sorry If my post sounded as I was offended. Nothing further from reality. I admit I'm a newbie, and I'm here to learn, your help, guys Is Highly appreciated. Still havent test the thermal utility, will do when I reach home and could work a little with the Comp

@Xabanak Awesome, Didnt knew that program could do that, now I only need to find where is that option , will keep you informed guys, many thanks.

 

[Hail The Brain Slug]

Awesome, Didnt knew that program could do that, now I only need to find where is that option , will keep you informed guys, many thanks.

Click the tiny wrench in the upper right corner of the graph to select the graph monitoring options, same for the other monitoring section.

 

[wingman04]

Is your case cooling good enough for the VRM? You can test with the case cover off and a fan blowing in the case while srtess testing.

 

[LTC8K6]

Is this possibly the Turbo Boost 3.0 causing odd readings?

Or do you have this installed and working?

 

[Krorghar]

Ok ppl I have been out from home and couldn't work on the PC

Is your case cooling good enough for the VRM? You can test with the case cover off and a fan blowing in the case while srtess testing.

My case is a corsair Graphite 780T, all test are done with the case oppened, and with the 5 fans I have installed running, anyway there is nothing pointing directly at the VRM, so can't check if problem it's an overheating of this module.

Is this possibly the Turbo Boost 3.0 causing odd readings?

Or do you have this installed and working?

Honestly I have no clue, I'm newbie and don't know what this thing do, will look into it.

OK heres a little update on the situation.

First I have lowered a litte the CPU volt to 2.850v And set Cache volt to Auto. and Raised the current limit in the Digi+ section to 120% just because if the problems were caused from a current shortage. Also enabled All the monitors on Intel XTU. And ran prime95 v289, results were better but still weird.

At first all seemed to run properly, and I thought I finally hitted the JacKpot with the settings:

At this point all were running properly, XTU still was showing that active cores were only 3, but It was reporting 100% load on all cores? This is weird, is any kind of reporting error? or 3 of my cores are really not working? Coretemp also shows 6 cores working at 100%, I can't understand this guys, any help pls?

Then 10 mins into the test, one of the cores started to drop performance from 100 to 60% with no apparent cause:

Then, 10 more mins into the test and another core lowered it's performance to 60%:

At this point I decided to Stop the test and surrender, lowered the Overclock to 4.3Ghz and run test Again, with no problem this time. This pic is taken with the test runnning for more than a Hour there I had to stop the test:

So I must admit I have no freaking clue on what´s going on with my rig. Anyone have any idea pls?

As allways many thanks for your interest guys.

 

[wingman04]

Looks like the utilization is the only thing down, the clock speed stays the same, that would point to a software problem. Do you have any errors?

 

[Krorghar]

Looks like the utilization is the only thing down, the clock speed stays the same, that would point to a software problem. Do you have any errors?

No error that I have noticed, W10 is stable, what kind of error are we looking for? . Anyway If it was a Software error, why is not reproduced when I lower OC from 4.4 to 4.3?

 

[Abwx]

Cores 2 and 4 seems not as efficient as the other cores and are 5-7 degrees hotter, the CPU reference its limitations to theses two cores and limit frequencies accordingly to keep the temp gradients in the silicon within safe operating areas.

Keep in mind that a core temp is not a junction temp, a core can be at 70°C as a whole but with some of its parts being at higher temp, let s say that a FPU that is 1mm2 could well be at 90°C with the rest of the core producing heat at a local rate of 60°C and the whole being at 70°C.

 

[wingman04]

Cores 2 and 4 seems not as efficient as the other cores and are 5-7 degrees hotter, the CPU reference its limitations to theses two cores and limit frequencies accordingly to keep the temp gradients in the silicon within safe operating areas.

Keep in mind that a core temp is not a junction temp, a core can be at 70°C as a whole but with some of its parts being at higher temp, let s say that a FPU that is 1mm2 could well be at 90°C with the rest of the core producing heat at a local rate of 60°C and the whole being at 70°C.

The clock speed is staying the same, so it does not have anything to do with temperature.

No error that I have noticed, W10 is stable, what kind of error are we looking for? . Anyway If it was a Software error, why is not reproduced when I lower OC from 4.4 to 4.3?

Well if the stress tests like prime95 are passing a few hours test at 4.4GHz? The problem would be something else. It's just odd that the CPU utilization drops down to 60%, what that means is the software is not feeding the CPU data as fast a as it should, maybe something inside the CPU like the memory controller. What is the Vccio Vccsa in HWiNFO64

 

[Abwx]

The clock speed is staying the same, so it does not have anything to do with temperature.

At constant frequency and 60% load the core will dissipate about 40% less, so this has to do with temp since a core will skip 40% of the cycles, here it seems to throttle either the culprit when it reach 70°C or a nearby core whose area is then used to cool the core that has reached this threshold.

 

[wingman04]

At constant frequency and 60% load the core will dissipate about 40% less, so this has to do with temp since a core will skip 40% of the cycles, here it seems to throttle either the culprit when it reach 70°C or a nearby core whose area is then used to cool the core that has reached this threshold.

Intel CPUs throttles at 100c TJ Max and you can see it in the clock cycle speed, the multiplier will drop down.

 

[Abwx]

Intel CPUs throttles at 100c TJ Max and you can see it in the clock cycle speed, the multiplier will drop down.

Keep in mind that a core temp is not a junction temp, a core can be at 70°C as a whole but with some of its parts being at higher temp, let s say that a FPU that is 1mm2 could well be at 90°C with the rest of the core producing heat at a local rate of 60°C and the whole being at 70°C.

 

[wingman04]

Keep in mind that a core temp is not a junction temp, a core can be at 70°C as a whole but with some of its parts being at higher temp, let s say that a FPU that is 1mm2 could well be at 90°C with the rest of the core producing heat at a local rate of 60°C and the whole being at 70°C.

Does not work like that the Integer and FPU are on the same core. The temperature on the core or cores designed by Intel does the throttling, I have seen it at 100c the clock speed bounces down and up.

 

[Krorghar]

Hi guys, I won't be able to be at home for 2 days, when I get back will do some testing Again.

Wingman Have downloaded and Instaled Hwinfo64, but can't find Vccio and Vccsa readings, what do these mean and where are these located pls?

Lets recapitulate:

The cores are Acting really weird, only we can Know for sure is that at 4.4 Ghz we got that core problem and at 4.3 Ghz all cores seems to stress properly, with the same program. and settings . Also according with XTU, when a core drops its Load, Frecuency and voltage, and thus the multipplier, remains the same. So there is no apparent cause.


Maybe I should use another test program or another monitoring one? What are the suggested setting for Prime95 btw?

BTW, do you guys want me to take some pic at Bios values, so you can check them maybe?

 

[wingman04]

With Hwinfo64 you will find the Vccsa and Vccio on the sensor status menu when you scroll down to the the motherboard.