Problems? See this thread at archive.org.
Can someone tell me how bad this is? I ran HW Monitor just while I was playing.
http://postimg.org/image/n0rl00cbr/
My core i7-6700K hits 99C and throttles when running the division. My GPU maxes out at 79C (with an average of 74C). Isn't this TOO hot?
-
Man, that's a 91w CPU in a 15.6" frame, you're gonna see thermal maximums hit in that small area.
Out of the box the Clevo 6700k BIOS settings have the voltage set too high. Try undervolting the CPU / CPU cache by -100mV to -150mV. Find the highest stable undervolt, -100mV is good, -150mV is better - less heat generated.
If you are using more CPU than GPU - the GPU is running cooler, try limiting the CPU performance for those apps as the GPU only needs to be kept busy, and even 45w 6700HQ's can do that.
You could reduce the clock rate, fix it less than what you are using now - are you already OC'ing? If you are OC'ing that will exacerbate the high CPU voltage problem, undervolting will fix that.
Another trick to use is to disable Hyperthreading, since most games only need 4 cores. This would be a good solution overall as it will keep the heat down all the time. CS:GO without a FPS limit set is a good example, it will run the CPU to 100%.
Some people with this issue make hotkeys to switch speeds depending on the app.
You can disable Hyperthreading in the BIOS with Prema BIOS, not sure about stock BIOS.
If you can't do it in the BIOS, then in Windows use the TaskManager => Details Tab => right click on initial Steam processes and use Processor Affinity to disable "odd cores" like 1,3,5,7 and then when you start a game from Steam, it will inherit the Affinity setting and only use 4 cores.
You could also get a cooling pad with large fans, or do what the top benchmark score holders do, run their laptops sitting right on top of a portable air-conditioner vent
The full performance desktop CPU's running at 91w need a bit of tuning to get the best performance within the cooling available in a laptop.
If you have the Prema BIOS you can drop the voltage a bit with a -150mV offset, if you don't have that option in the BIOS then install and use Intel Extreme Tuning Utility - XTU - and that will allow you to set the voltage offset for CPU and Cache, as well as multipliers for CPU and Cache.Last edited: Mar 12, 2016Starlight5 and Spartan@HIDevolution like this. -
Actually I don't think at stock a new laptop is suppose to thermal throttle....desktop cpu or mobile cpu that is not suppose to happen at stock speeds, I suggest just first trying to undervolt don't change clock speeds just undervolt with xtu if you don't know how to do this you can look on youtube, google, or ask here You have a new laptop if I got a new (expensive) laptop and it overheated when I played a game I would be PISSED and asking for a fix from who you bought it from (maybe demanding a fix) I would first elimate any software/hardware/paste problems BEFORE you do any undervolting or changing any settings at all because that wouldn't be a fix it would be a bandaid....good luck
hmscott likes this.
-
Well, you are suggesting making changes to reduce thermal output, not running it unaltered out of the box, so at least you realize that when reality meets expectations, you need to adjust to fit, so to speak
The 91w 6700k is generating as much heat as 2x+ the 45w 6700hq, so it's going to take 2x+ the cooling capacity to keep that 91w 6700k cool, and it's hard to double the cooling capacity in a 15.6" frame.
Hopefully, after tuning, some specifically for CPU intensive apps / games, you can optimize the throughput.
For example, I can run my 5950HQ at 4.2ghz stable. But, at that power draw it's going to thermal throttle under load, and power/current throttle before that. It's only got 47w to play with.
If I turn down the cores from 4 x 42x to 4 x 35x and cache from 34x to 32x, and undervolt by -100mv I can keep a constant 35x throughput.
If I leave it at stock or higher clocks, the average throughput is 33x, so by tuning I can reduce heat and actually increase work done over extended time periods.
@Jakamo5 could also delid his CPU and change the IHS paste to CLU to get the best single heat reduction, 10c-20c.
Last edited: Mar 12, 2016 -
The 17" P870DM has 3 fans, 3 separate cooling systems, independent for each of the 1 or 2 GPU's and the 91w 6700k.
It doesn't matter what GPU complement the P870DM has, the stock BIOS comes with a too high CPU voltage setting, and all of them will hit 99c and thermal throttle.
Read through those posts in that thread to learn how to manage it better.
Once you master the voltage, and other settings to reduce temps, you can tune for performance:
Clevo Overclocker's Lounge
http://forum.notebookreview.com/threads/clevo-overclockers-lounge.788975/
It's the way of the 6700k in the new Clevo's, both 17" and 15" frames, at stock BIOS CPU voltage, they will hit 99c and thermal throttle out of the box.
Undervolt by -100mv to -150mV, more or less, whatever your example of 6700k will run stable. That should get rid of the peak heat you are seeing.
That's much easier and quicker than pulling apart the laptop to re-paste
Last edited: Mar 12, 2016Papusan and i_pk_pjers_i like this. -
dude, did you actually look at any of these links you posted? only 1 out of the 6 say anything about the temp of the CPU and the one that does say something about the CPU temp is in reference to a benchmark which is totally different than a real world situation, but seriously please go look at the links you just posted lol I dunno how you came up with them but maybe it was an error or I dunno lol.....okay I am done sorry all, I have strong opinions (because im always right at least in my mind )
-
Undervolting by -100mV - -150mV should be enough.
Disabling Turbo would be a waste of performance, but that would definitely stop the high temps, because it wouldn't clock higher than base clock
-
You are going into it rejecting the concept - you aren't seeing what is right in front of you.
If you went in trying to accept the premise, the stock CPU BIOS voltage is too high in the Clevo's with 6700k, you would work to discover it better to your satisfaction.
Your premise that a laptop shouldn't thermal throttle out of the box is mistaken. It's an optimal idea, but doesn't always happen, even with a great laptop like these are.
Searching on 99c isn't enough, but it gets you into the thread areas for reading more.
Here are some better posts/quotes, some from early delivery when they were just discovering the problem, to more recent posts:
http://forum.notebookreview.com/thr...enix-has-arisen.781814/page-100#post-10138194
"Wow, this Skylake CPU can undervolt great! I was hitting 98-99C with throttling on wPrime at at stock. I set voltage to 1.000V Static (although reading 1.1-1.15V) and it's peaking at no more than 80C at stock 4GHz. I was a bit concerned at first, but now it seems these things can run cool and undervolt great"
http://forum.notebookreview.com/thr...enix-has-arisen.781814/page-152#post-10150299
"Ever thought of putting a peltier in a laptop, fox?
Edit: Also, Mr.Fox is 100% right on the undervolting. I can hold a 4.2ghz easily with a -150mv undervolt. The stock volt is stupidly high. As someone coming from sandy bridge, I am just not use to 1.3v-1.4v being a high voltage."
http://forum.notebookreview.com/thr...enix-has-arisen.781814/page-441#post-10211018
"That means it's drawing more power; it needs more power to run stable. Set that back to 200,000 and then also undervolt as @Mr. Fox and @Phoenix suggested.
You need to have *all* the recommended settings, it's a self-supporting group of system settings, to get the desired result."
http://forum.notebookreview.com/thr...enix-has-arisen.781814/page-254#post-10167330
"1) You should undervolt whether or not you are overclocking. The default voltage is way too high."
http://forum.notebookreview.com/threads/clevo-p870dm-g-review.784082/page-2#post-10137803
Clevo P870DM-G Review
http://forum.notebookreview.com/threads/clevo-p870dm-g-review.784082/page-2#post-10137803
"Yes, all the reports I've seen of people using the 6700K have them getting undervolts of at least -100, some up to -200Mv.
Which means Intel set the default voltages much higher than needed."
There are lots and lots of these mentions in the Clevo 15.6" and 17" threads about the 6700k. The reason I am mostly quoting the P870 thread is because that is where I spend my time. The 15.6" version isn't of interest to me, but I know it has the same basic issues.
The solutions / tuning are similar, but for some reason the PM870DM thread and Clevo OC'ing spend most of the time tuning the 17" variant.Last edited: Mar 11, 2016 -
-
-
-
You've got XTU installed. It's a 5 sec trip through a couple of popup lists of negative voltage offset's for Core and Cache.
You then have the changed elements in Yellow. Click Apply. The Yellow's turn to White as the settings are applied.
Now those settings are currently running your CPU at reduced voltage, and hopefully that setting you are testing is stable
Click Save, name the new Profile. For example you might name it:
Stock Clock -100mV Core -100mV Cache
Now XTU will have a stored Profile, which if you have it set correctly, will Apply that profile every time you boot.
If those settings aren't stable, and your laptop freezes, powers off, and restarts, XTU will notice that, and it won't Apply that Profile at the next boot - or ever again unless you go in to XTU to set it again.
XTU may start at the next boot to complain that you crashed, and you can then try new settings - save to a Profile so you know where you have been and what does / doesn't work. It helps to write this all down too.
So far, all the posts I have seen on this by 6700k users say that -100mV works fine, and some push further to -150mV, and even -200mV.
When tuning the negative voltage offset, you need to try several modes of testing to verify the setting will work in all modes of use.
Load testing is obvious, use XTU Benchmark - for me this is enough to rule out the outlying settings quickly - if it passes that use the XTU Stress test.
If that works, then use the settings in real usage, games, apps, browsing, watching videos, etc.
Remember, you can always go in to XTU and reset to Defaults - I actually have a Stock "Default" Profile that I save so I can quickly switch to it when I need to do work I can't lose due to a crash.
During the time you are testing to find stable settings it's not fun to be OC testing while doing real work and lose said work to a crash.
After all the stress testing, game testing, app testing, let the laptop sit idle for as long as you can - in between the things you need to do - as idle mode crashes are the last you will see when focusing in on the best settings.
If you get an idle crash, drop negative voltage offset by 5mV, like from -150mV to -145mV, and go back to idle.
I have seen CPU's trigger on 2mv differences.
Have fun
Here is an example from my own tuning, with negative offsets for Core / Cache while running long batch jobs. I drop the core multipliers and cache multipliers to reduce power usage and heat, and at those power draws I can also do a nice negative voltage offset.
Notice how the new values have been loaded by selecting a Profile, but need to be Applied.
Note: If you are nervous about crashing, don't be, it's not going to cause any problems. If you want to, after reboot do a right click on your C disk icon, Properties, Tools, and do a disk check every time you crash, and that avoids any disk corruption problems.
It's a good idea not to be doing important work while the OC tuning process is ongoing.
You can start at -50mV and work your way toward -100mV if that makes you feel better. It takes longer, and draws out the process, but maybe I am just jaded after doing this for so long, I just want to get it narrowed down and done.
When I do it, I pick a high value to establish boundaries quickly, and then work within those boundaries for fine tuning. I would likely try -200mV first, then -175mV, then -150mV, etc.
After finding a good starting point, I would then continue to work -10mV at a time till it got unstable, then fall back 5mV, etc. Getting a stable number after testing, I usually drop back 5mV to make sure it's stable.Last edited: Mar 11, 2016Spartan@HIDevolution likes this. -
You might try asking for help in the active threads, pointing back here for help. For whatever reason this board seems to run off 1 thread for each model/make, thousands of posts related will fill them up. Unlike other boards that start a separate thread like you did.
Just a heads up, if you get tired of waiting for a response.
-
-
That's a good start
You still have a lot of room "under" -100mV. Based on your results, I would drop down 10mV to -90mV, and see how that works.
The starting values were just that, the first ones to try based on other reports, you will find an undervolt that works.
I have had CPU's undervolt at -125mV, and one that would only do -20mV, it doesn't take long to figure it out, and it is worth knowing / using undervolting with high power CPU's.
If you go into those threads and pose your question with the undervolt data you have, that would be a good start. There are other tuning parameters people have been using to reduce temps.
You have had the laptop for a while now, last time we talked was in December. Do other games do this as well? Is it only The Division?
Some games do use more CPU than others, and will drive utilization higher resulting in high temps.
Speaking of high temps. Are you only going by the single Maximum value showing in hwinfo64 after a run in The Division?
The undervolt should be removing some heat, temps should drop on average, but peaks can still be hit.
What I have found is that it's necessary to turn on the hwinfo64 logging, and have it log a complete test run, then use Excel to load the csv log results, and then you can see how *often* the peaks hit.
People find the Maximum temperature peaks only happen a couple of times, and the normal temps are in the 70's/80's for most of the time during a game, benchmark, test.
A high Maximum temp can mask what is really going on typically.
The Average temp is supposed to indicate this, but its more reassuring to see the temperature samples every couple of seconds throughout the run to reinforce that the peaks are actually rare.
For example, when I benchmark, I need to usually throw out the first part of the run from hwinfo64 - the fans don't ramp up quick enough to cool immediately when the CPU hits 100%.
If I am interactively watching the hwinfo64 Window I will see Thermal Throttling for 10-30 seconds while the automatic fan curve kick in and cools down the CPU enough to not throttle. Then I can clear the peak/Thermal Throttle readings by clicking the Clock - reset - button to start taking readings fresh again.
You can also stop this behavior by turning on the Max Fan button and start the fans running at Maximum before starting the test.
Are you also looking at the CPU utilization? Do the cores all spin up to 100% and stay there, or do they only peak there once in a while - same idea - the game will vary the load throughout the session - and may only hit the peaks occasionally.
If the fans aren't kicking in quick enough, try running with Max Fans a few minutes before starting to play and see what the peak is then.
This could eventually be a re-pasting, but its a good idea to learn these techniques anyway, as they have the same effect on reducing temps from a maximum that you find acceptable.
Maybe peaks after re-pasting will be in the low 90's, these techniques can reduce further under 90's.
Good luck, and please keep us in the loop how this all works out
Spartan@HIDevolution likes this. -
-
Are you applying the undervolt to both CPU and cache?
The testing you are doing is helping LPC digital decide how to handle the RMA, after this testing you have done your best to mitigate the problem, now it's up to them to solve.
But, knowing these parameters and results before sending it will give you a basis to measure the results as well. Remember to save the logs from the hwinfo runs.
Does The Division have a benchmark? Maybe use that as a test to get readings. If you have a game that does have a benchmark built in, like the Batman games, Tomb Raider, Mordor, etc, use that to get log files with readings you can come after the RMA.
It's fun, right?
-
The fan thing is to get more data points. As well as to recognize that the auto fan profiles on most laptops start too slowly, letting a benchmark run away and heat up the CPU well past normal temp before the auto fan spins the fans fast enough to cool the CPU.
Full fans just gets the fans in motion so your Maximum temp reading isn't polluted with the slow auto fan startup.
You can turn off Full fans after a few minutes, after starting the game, and the temp of the CPU should allow the auto fan profile to keep the fan speed high enough to cool the CPU, and not need to run at 100%.
Doing the logging and looking at the results will give you a good feel as to what is going on throughout the game / test. You can see all the related values - CPU speed, temp, power drawn, etc.
You might consider asking them to paste with Coollaboratory Liquid Ultra (CLU), it's the best cooling compound for high temp applications like a 91w CPU in a laptop.
Coollaboratory Liquid Ultra (CLU)
http://www.coollaboratory.com/en/products/liquid-ultra/
Does LPC do RMA work themselves, or are they sending it off to Sagar for the work? If so you might not have a choice in paste, but it doesn't hurt to ask.
As also mentioned, Delidding - pulling the can off the CPU - and re-pasting the CPU under the can with CLU helps too.
If they can't do this, than get them to swap in another CPU - as it could be the CPU's integrated head spreader paste job that is the real problem.
If they won't re-paste under the CPU can, then odd's are swapping the CPU for another one gives you another roll of the Silicon Lottery dice, and most CPU's IHS is better than what you are seeing now.
At least your GPU runs cool
Last edited: Mar 12, 2016 -
-
-
-
CPU hits 99C and throttles when running division
Discussion in 'Sager and Clevo' started by Jakamo5, Mar 11, 2016.
