Overclocking 6850k with custom water loop

Toggle sidebar Toggle sidebar
F

fidelity101

Honorable
Mar 8, 2013
54
0
10,640
I can't seem to find good information on comparable setups to see if I'm in the right ball park. My temps are higher than I'd like them to be and I'm getting temp spikes above 80C which seems odd given the water loop is staying cool. Maybe I need to redo the thermal paste but hopefully you guys can offer some pointers.

Using Asus AI 5 way tuner.
System sets voltage to 1.375, 4.4ghz with 100bclk. This is not stable for gaming at all. I drop the settings to 1.35v, 4.2ghz with 102bclk and this causes my wireless signal to drop and I need to restart to get it back...temps must be to high? So I drop back to 1.35v, 4.2ghz overclock, 100bclk, and run prime95. Temps stay under 78C but occasionally core 3 jumps to 85C and then comes back down...which seems odd. Core 3 is usually about 15C hotter than core 6...is this something I can fix with a new thermal paste job?

I spent 4 hours messing with overclocking last night and the results were nothing like I expected. The water temperature never really passed 36C, the VRM was about 55-65C. I want to keep the hottest core of the cpu under 78C to be safe and I'd like to drop the voltage if possible. Asus 5 way optimizer wants to jack the settings way up and that would result in some high temps and crashes.

I'm going to keep researching but I ultimately just want a decent overclocking job for every day use. I want to be able to play games for hours without worrying about temps. I'm also water cooling my Nvidia 1080 which seems to be perfectly happy running 4K at 60fps with gsync on my brand new Asus pg27aq monitor although I did have to overclock the gpu +160 and the memory +200 so it stays at 60fps.

 
Solution
Eximo
Not that familiar with your monitoring tool to say. Package temp could be the socket temperature, which doesn't have direct cooling.

CPU core temps seem to be about right the chip. I have a decent watercooling setup on my CPU and I still see peaks in the 70s.
Sort by date Sort by votes
Eximo

Eximo

Titan
Ambassador
May 27, 2010
25,839
4,186
139,340
That is a pretty big difference between cores. Cooler might be on slightly crooked. Also check the chip itself, it may have high or low spots.

Certainly take a close look at the thermal compound application when you remove the cooler.

My understanding is that these 6th gen HEDT chips don't do so well with overclocking compared to their Haswell-E 5th gen predecessors, so where you are at might be pushing it. Don't be afraid to try odd voltages and frequencies, sometimes the big leaps are unstable, but inbetween points are fine.
 
Upvote 0 Downvote
F

fidelity101

Honorable
Mar 8, 2013
54
0
10,640
After a lot more reading, I decided to set CPU Core voltage to Adaptive and install Asus Thermal Control Tool. What I'm really looking for is "is this normal for a water cooled system" or is this well below what I should be expecting?

After setting adaptive mode, +, Auto, 1.200 in the bios for CPU core voltage. I set 4.2ghz for all cores.
I loaded the thermal control tool and set it to drop to 36x after hitting 75C

Intel XTU scores are 1742 at 4.2mhz hitting 70C max. Is this what I should expect from a custom water loop?
The voltage seems to be sticking to around 1.2970 volts whether it's idle or under load according to the thermal tool.

I then loaded Asus real bench. Running the stress test, I got these numbers:
CPU temp: 62
CPU Package temp: 80
Core 1 temp: 64 (marked as my best core)
Core 2 temp: 61
Core 3 temp: 75
Core 4 temp: 60
Core 5 temp: 54
Core 6 temp: 56
There seems to be a large difference between package temp and core temp. Should package temp be that high? What's the difference?

Then I ran realbench benchmark and got a score of 150248.

Idle temps are 44C and max temp seems to be 80C for the CPU package.



 
Upvote 0 Downvote
Eximo

Eximo

Titan
Ambassador
May 27, 2010
25,839
4,186
139,340
Not that familiar with your monitoring tool to say. Package temp could be the socket temperature, which doesn't have direct cooling.

CPU core temps seem to be about right the chip. I have a decent watercooling setup on my CPU and I still see peaks in the 70s.
 
Upvote 0 Downvote
Solution
F

fidelity101

Honorable
Mar 8, 2013
54
0
10,640
I appreciate your help Eximo! I pulled the water block off and redid the paste job just to be on the safe side. Temperatures aren't much better but they also don't jump as quickly so hopefully that's a good thing.

I also noticed a 5C increase in temperatures between turning XMP on and off for my memory...I hadn't thought about that as an option but things are nice and cool when the memory is running at 1.2v vs 1.35v. I'm experimenting with that now to see if I can get temps lower.

Currently still at around 1742 score for Intel XTU Benchmark at 4.2ghz and 73C max temperature.
I also obtained a score of 155565 with Asus' Benchmark and the max temp was running about 65C.

After the paste job, the temperatures of each core seem to be staying within 10C of each other which is a little better than it was before.
 
Upvote 0 Downvote
F

fidelity101

Honorable
Mar 8, 2013
54
0
10,640
So it appears that my VID is locked to 1.298v when Turbo is active, so I can no force voltage below 1.298v while in Adaptive mode. This turned out to be a good thing because my occasional system restarts and crashes were directly related to Vcore voltage being to low. Now I'm running 4.3ghz at 1.310v and I've been stable for 24 hours. Several hours of Asus RealBench and "The Division" video game. I'm tempted to drop voltages back down to 1.305 voltage or lower to see what the lowest stable voltage can be, but I know it's not stable at 1.298v so 1.310 is only 0.012v off from that...which seems like an extremely small adjustment already.

I've also set my Vcache voltage to Auto because Adaptive doesn't seem to be a good settings from everything I've read so far. I haven't bothered to overclock the cache multiplier more than 32x so it's stable at Auto and I don't see much of a reason to mess with it at this point.

The one thing I still can't seem to figure out is power dissipation for the EK X360 waterblock kit. What is this system capable of dissipating and what voltage should I be safe to run at? Does anyone else have an X360 EKWB kit out there? If so, what voltages are you running and what temperatures?
 
Upvote 0 Downvote
F

fidelity101

Honorable
Mar 8, 2013
54
0
10,640
That's a huge amount considering my cpu is 140 and gpu is 160. Guess I didn't need the other se240 radiator.

My biggest problem isn't total dissipation, but overall rise to cpu temps under load. Voltage at 1.310v is doable on a 6 core 6850k but going over 1.325v causes temps to jump into 80C pretty quickly. So maybe it's more a matter of how much heat I can pull from the cpu rather than total dissipation.
 
Upvote 0 Downvote
Eximo

Eximo

Titan
Ambassador
May 27, 2010
25,839
4,186
139,340
Well one thing you have to understand about water cooling vs air cooling is that your cooling is happening away from the CPU. When you first turn it on you see amazing temps because the water coming into the block is cool. After a while though the cooled water is going to be fairly warm itself once the radiator reaches equilibrium. What I have seen is that you might see higher temperatures, but be able to keep going much farther at that point then if you were air cooling, where you get an immediate effect.

Though I think you are just at the limit of that CPU. If you look at CPU temperatures to voltage on a chart they shoot nearly straight up at a certain point. For Haswell it was around 1.35 volts at roughly 150W. You are starting with a 140W larger CPU with a slightly lower tolerance for voltage, with a better heat spreader, but the cores themselves are much smaller. 14nm vs 22nm. So you are correct in saying it will be more difficult in getting the heat out.
 
Upvote 0 Downvote
F

fidelity101

Honorable
Mar 8, 2013
54
0
10,640
The thing that really surprised me at first is the sudden jump in CPU temperatures under load. I think I have it straightened out now and I'm only running 1.310v with a 4.3ghz overclock, but going into 4.4ghz and 1.350v makes the CPU temperatures jump from 40C at Idle to 80C plus almost instantly and then it drops down a bit. It seems like the system is having trouble pulling heat away from the CPU fast enough vs. the radiators not being able to keep up. The water temperature never gets over 35C and starts at 33C so the radiators seem to be doing a very good job.

My initial assumption was that temperatures would gradually increase over time (or more rapidly over time) but I became concerned when temperatures shot up significantly in very short order. Apparently that's to be expected with the 14nm process as you pointed out. It would be very helpful to us newbies if the tech reviews mentioned what kind of temperature spikes you should expect under load with certain setups. My EK X360 kit seems fairly popular along with the 6850k CPU so doing a review and including temperature ranges should be available as a review by now...or at least it's wishful thinking on my part.

My main question and concern lies more in the amount of heat that can be pulled away from the CPU vs how much the system can dissipate via the radiators. Most reviews focus on the radiators and how much energy they can dissipate over time, but no one really points out that the new Intel CPU's just can't be kept cool enough under heavy loads at certain voltages.
 
Upvote 0 Downvote
Eximo

Eximo

Titan
Ambassador
May 27, 2010
25,839
4,186
139,340
At least from what I have seen Broadwell-E didn't receive much attention because it doesn't perform significantly better than Haswell-E and there were no changes to the chipset/socket/motherboard aside from out of the box support for DDR4-2400. On top of the price increase, just not very many out there.

Haswell-E suffered similar problems in that there weren't a lot of overclocking guides made for them. They are so similar to Haswell and Devil's Canyon that nothing else was really needed. Devil's Canyon got a lot of press because they fixed the temperature problems that plague most early Haswell samples.
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom