View Full Version : i7 3770K can't get stable past 4.4 Ghz

04-17-2014, 01:55 AM
So I have overclocked my system but can't get totally stable at 4.5 or beyond. I want to get some advice on whether there is anything else I can try or if I just got a bad overclocker. So first off, here are my system specs. Also, you can see some pics and video of my custom build here.


Case: Custom built desk mod.
CPU: Intel Ivy Bridge Core i7-3770K @ 4.5 Ghz (still tinkering with the overclock)
Motherboard: ASUS Maximus V Formula Z77 ATX board
Memory: 16 GB of DDR3 from GSkill
Hard Drive: RAID 0 array using 2x 128 GB OCZ Vertex 4 solid state drives for the operating system. I also have 2 Western Digital 500 GB black drives in a RAID 1 array for data storage and backup.
GPU: EVGA Nvidia GeForce 670 (x2) in SLI
Power Supply: OCZ Fatality 1 Series 1000 Watt PSU with red LEDs
Optical Drive: LG DVD/CD Burner Combo and a LG BluRay Burner drive
Operating System: Windows 7 Pro
Monitors: ASUS 23" 120Hz 3D Monitor (x3)
Cooling System: EK D5 140mm reservoir with D5 vario pump, EK acrylic CPU water block, EK acrylic GPU water blocks, EK acrylic SLI bridge, 2 240mm EK radiators, and EK blood red coolant. I have 3 140mm fans by Aerocool up front as intake fans and 4 120mm fans on the radiators as outlet fans.

Here are some pics of the settings I have for running 4.4 and 1866 on the memory.

35851 35852 35853 35854

35855 35856 35857

When I tried 4.5 I was still at 1866 on the memory and tried increasing the LLC to extreme, the fixed frequency to 500, power phase control to extreme, and cpu voltage up to 1.32. What happens is I keep getting "WHEA Logger" errors in event viewer when trying to stress test or play BF3. BF3 will sometimes crash and sometimes I can play without crashing but the WHEA Logger errors will still show up. Also, Prime95 will crash after an hour or less. It won't crash Windows but the app will stop. I also have tried using AIDA64 and that I can run longer without the app crashing (have only tried for about 2 to 3 hours), but still get WHEA Logger errors. Also temps using Prime95 will get in the low to mid 80's. I don't think it's good to leave it that hot for extensive times like a 24 hours stress test. So, I know that the usual knob to turn is to keep uping the voltage to get stable, but I went up to 1.33 setting and with the other setting maxed out I get about 1.366 under full load. Temps getting close to 90. So I got scared to go any higher. Also I'm referencing the highest core temperature using reading from either Core Temp or AIDA64. The "CPU" temp is lower but I'm guessing the Tjmax temp of 105 for the 3770K is off of the "core temps" not the "CPU temp"?

I have read plenty of other posts that say people are getting even higher overclocks with even less voltage, so I'm wondering if I can do some more tweaking to get rid of the WHEA Loggers and still run at 4.5 or higher. So, any advice on other settings to tweak or things to try would be helpful. Thanks!

04-17-2014, 02:54 AM
Your temp will dictate your clock speed and that is your limit. what is your temp @ idle?. I would think you would have better temps with your current cooling. What TIM are you using. You might consider checking your cpu water block make sure your getting good contact. also make sure you dont have any air trapped in your loop. Use a 5 mm pea in the center of the cpu and make sure you properly tighten the water block. (DO NOT OVER TIGHTEN) Use the XMP setting on your memory. After you check your cooling problem. (Temps are a bit high for that speed and voltage) check out this guide for a 4.7 OC


Melting Point
04-17-2014, 07:11 AM
I agree with meankeys, but also...

You should most definitely be using offset CPU voltage rather than manual. Also, most people don't go past medium or high LLC. I believe setting it to extreme will cause the VRM to generate more heat (Someone correct me if I'm wrong), and that's not what you want. Changing either or both of these will mean you'll need to play around again to get the optimum CPU voltage.

However, I'm not so sure it's your CPU that's the issue. It could well be memory related. The reason I say this is, every time I push my CPU too far, it BSODs. It doesn't nicely log errors in the windows event log. You could try bumping up your VCCSA a little, and/or lower the speed of the RAM to 1600, just to test that theory. If it runs stable after that, you might find you can drop your CPU voltage somewhat, and reduce your TDP.

04-19-2014, 09:53 PM
Right now I'm back running 4.4 Ghz with memory at 1866 and my idle temps are 28-30 on the core temperature.

I'm using Artic silver 5 and applied it like they suggest on their website. That is using a vertical line instead of a dot in the center. I may try to redo it but if you saw my setup you'd see that it's not that easy to get to. Moving that triple display is a pain and takes awhile to get all three monitors lined back up after moving.

I did try 4.5 again with the memory at 1333 (stock settings) to see if that would make a difference. I went all the way up to 1.33 core voltage which would put me close to 1.37 under load (using extreme LLC). Still would get the errors. Also if I went down to high LLC and had any vdroop it would BSOD. So I went back to 4.4 and 1866 using core voltage of 1.31, high LLC, power phase control at extreme, but other settings in Auto. Right now everything is stable and I get no WHEA logger errors. I may even be able to reduce the voltage some more.

It just seems that it doesn't want to run at 4.5+. What's the most voltage anyone else has run on this chip? I had heard not to go past 1.35 but I think I would rather just run at 4.4 with low voltage for a 24/7 overclock.

I do plan to experiment with the offset mode vs manual. This is the first board I've had that used the offset mode. I will post back with some more results after I have a chance to tinker with it. Thanks for the advice!

04-20-2014, 02:10 AM
Your idle temps seem about right. the only way your going to get a higher clock is more voltage. use high LLC - T-probe - 4.4 is good for 24/7 @ that voltage your temps are good. Are you using the XMP setting on your memory? you should be. As melting point said 24/7 OC offset voltage may be a better option. use like a +.005 and adjust it from there. use CPUZ and watch the voltage the CPU will draw under load that will give you an idea of were you need to be. I killed a 3770K with 1.85v on SS @ 5798 MHz on ice water I have another 3770K doing 5400 @ 1.585v Temps are 7-15c idle 55-68c loaded. Any voltage will eventually degrade your cpu. Higher voltages for prolong periods of time will shorten it's life span.


04-20-2014, 04:44 AM
You may just have to reside to the fact that your chip can't do any higher than 4.4GHz.

04-20-2014, 11:20 AM
The answer is simple: your CPU needs more Vcore for higher clocks... and you can not do it because of high temps... - 4.4GHz/1.3V should not cause the core temps to get close to 80C...

Your ambient temp may be high... but if I see correctly that you're using 2x240mm rads for a CPU + 2x graphics cards, then you may have a cooling setup trouble too... specially if those things are on a single loop...

04-20-2014, 11:30 PM
So my loop is setup as follows: Pump discharge >Rad1>CPU>Rad2>VRMs>GPUs>Res. Yes both rads are 240mm. It seems as I have enough cooling capacity, due to my GPU temps only get to 42-44 degrees after several hours of gaming and they are overclocked also. It does seem like I'm just not getting enough heat transfer from the cpu to the water. I am planning on reseating the cpu block and reapplying the TIM. How about the whole delidding thing? Anyone done that and how did it go?

Oh and ambient temp in my room stays about 72 to 76 F.

04-22-2014, 12:49 AM
The thing is that when you start OC'ing the CPU, then for a while the Vcore goes linearly with the frequency... then you get to a point where you need a "jump" in Vcore... - your "jump" seemingly is after 4.4GHz... you need more Vcore to be stable at 4.5GHz...

Then with higher Vcore comes the higher core temps... and then your cooling won't be enough...

To be absolutely on the safe side, I would stick to 4.4GHz... but if you want to experiment, changing TIM and reseating the cooler may be your first step... - if those help, it means that you had a bad mounting before... based on your experience, let's say you have some, I don't think that will be issue... (but certainly worth a try)

Then yes, you can try de-lidding the CPU... in some cases it really helps, specially with air- and water-cooling (when you go sub-zero, I really not recommend it)... - you have 2 choices: 1.) run the naked die (risky, better to get the specific mounting kit from EK along with their block), 2.) just change the TIM between the die and IHS and remount the IHS...

I did both ways, I did killed some CPUs during the process, but also still have one which is working (it is re-lidded)... I have used them on water-cooling, but with chilled water - went up on the Vcore over 1.5V, but just for short runs... I'm not gaming, just benchmarking...

Melting Point
04-22-2014, 11:08 AM
1.) run the naked die (risky, better to get the specific mounting kit from EK along with their block)
Hey Zka17, are you saying EK make a CPU block specifically for cooling a CPU with the IHS removed? I'd buy that for a dollar.

04-22-2014, 11:14 AM
Not a block, but a mounting screw set for their Supremacy block...

Melting Point
04-22-2014, 12:03 PM
Cool. I'm thinking of delidding and moving to WC soon. This might be an option I'll consider.


04-25-2014, 01:15 AM
It's just a mounting kit to go with a Supremacy Block. Check out this article on it.


Melting Point
04-25-2014, 08:18 AM
Thanx jcrane929, but I've since found out that the IHS of my 3930K is soldered on, so delidding is not an option.