[SOLVED] Can't get a stable OC on 3770K

[Jumpingmonkey2]

Hey everyone.
So I recently got a Gigabyte GA-Z77X-D3H and a 3770K, which I delided and applied Coolermaster Mastergel Pro V2 both under and on top of IHS. I know it's not optimal, but it was what they had at a local store, so it will have to do while I'm waiting for NT-H1 and Conductonaut to arrive.
Anyway, that's besides the point. My problem is, this chip WILL NOT overclock! No matter what I do, what settings I change, it will always crash while loading into windows or shortly there after.
Furthest I've gotten was 5-10 min Prime95 run small FTT with 1.3V and 4.4GHz. Thing is, even if I reduce the frequency to 4.2GHz at that same voltage it would crash when loading into Windows, which shouldn't be the case, since I was running Prime95 with the same voltage at 4.4GHz??

What gives?
At stock settings it boosts to 3.7 on all cores and it's using 1.2V according to CPU-Z, so why wouldn't 1.3V be enough for 4.2GHz?

Am I missing something? The Vdroop isn't bad, at 1.3V it doesn't drop like at all, temps are fine, 78C on hottest core, although the lowest core is at 69C, which is probably not the greatest mount ever, but it shouldn't affect the OC should it?

Should I mess with any other settings? How do I stabilise this thing? Is the CPU degraded or just bad silicon?
I've never had a CPU be more stable at higher frequency with the same voltage.

Something weird with this chip or is it the motherboard that is to blame? I have no idea what I'm doing wrong.

 

[tennis2]

Faulty RAM?

 

[dorsai]

The motherboard you're using isn't the best as far as VRM goes...it's rather weak so it may not be able to keep the cpu at a stable voltage when over clocking.

Try pushing the core volts to 1.375v, PCH voltage to 1.07-1.08v, and raise the CPU VTT to .5v below your set DDR ram voltage.

Also you may just have a CPU that won't go very far above stock...the silicon lottery doesn't always award a good chip. I had a lousy 3770k that needed 1.4v to go past 4.2ghz so some were definitely not great at OCing.

 

[Jumpingmonkey2]

Faulty RAM?

I'm at work right now do I can't test, but RAM was working fine on asus p8z68-v with 2500K. That thing was stable af at 4,6 GHz.

 

[Jumpingmonkey2]

The motherboard you're using isn't the best as far as VRM goes...it's rather weak so it may not be able to keep the cpu at a stable voltage when over clocking.

Try pushing the core volts to 1.375v, PCH voltage to 1.07-1.08v, and raise the CPU VTT to .5v below your set DDR ram voltage.

Also you may just have a CPU that won't go very far above stock...the silicon lottery doesn't always award a good chip. I had a lousy 3770k that needed 1.4v to go past 4.2ghz so some were definitely not great at OCing.

Sucks to hear that, since iy was the only Z77 board I was able to purchase. Maybe if I can revive my asus p8z68-v with a bios chip I can try on there.
I'm at work rn, I will test your settings when I get home in about 6 hrs and report back. Thank you.

 

[tennis2]

Maybe if I can revive my asus p8z68-v with a bios chip I can try on there.

The Asus P8Z68-V supports Ivy Bridge CPUs if it's been updated to BIOS version 3203 or newer (latest is 3603, likely with Spectre/Meltdown performance limiters patches)

 

[tennis2]

Sucks to hear that, since it was the only Z77 board I was able to purchase.

I hope you didn't pay much for 9 year old hardware.....

 

[Jumpingmonkey2]

I hope you didn't pay much for 9 year old hardware.....

80€ for CPU and 90€ for mobo. Pretty much could have doubled down on that and gotten a 10600KF and a Z590 mobo...
Not my smartest pourchase, my intention was to get a 60€ 2600K for a cheap boost in performance, ended up getting the 3770K the guy had for 20 more, then my mobo bricked when flashing bios and had to get a new... well, old mobo...
Turns out it's pretty hard to get 9 year old Z mobos and definitely not worth it

 

[Karadjgne]

The delid/repaste is a fail from the start.
If hottest core is 78, coldest should be closer to 75/76, not 69. That's too much of a spread. I had only 7° difference hot/cold with the crappy stock paste. When you do get the Conductonaut, don't forget the nail polish and for certain do not use too much, be exacting about its application. MasterGel Maker is the good stuff, the Pro is pretty lackluster in comparison.

Try this. Return to optimal default values, get windows installed, then lock the cores at 4.3GHz. Touch nothing else. Default voltages on those cpus are set high enough that I've not yet seen any 3570k or 3770k that wouldn't run that. From there, change only one thing at a time, don't go changing multiple settings simultaneously without reboots.

3570k @ 4.3GHz 1.108v (wasn't stable at 4.4GHz at any settings or voltages.)
3770k @ 5.0GHz 1.42v, @ 4.9GHz 1.32v

 

[Jumpingmonkey2]

The delid/repaste is a fail from the start.
If hottest core is 78, coldest should be closer to 75/76, not 69. That's too much of a spread. I had only 7° difference hot/cold with the crappy stock paste. When you do get the Conductonaut, don't forget the nail polish and for certain do not use too much, be exacting about its application. MasterGel Maker is the good stuff, the Pro is pretty lackluster in comparison.

Try this. Return to optimal default values, get windows installed, then lock the cores at 4.3GHz. Touch nothing else. Default voltages on those cpus are set high enough that I've not yet seen any 3570k or 3770k that wouldn't run that. From there, change only one thing at a time, don't go changing multiple settings simultaneously without reboots.

3570k @ 4.3GHz 1.108v (wasn't stable at 4.4GHz at any settings or voltages.)
3770k @ 5.0GHz 1.42v, @ 4.9GHz 1.32v

I know the core differences are due to a bad mount, but I won't remount until I get the conductonaut, since the way I see it, as long as the hottest cores aren't too hot there shouldn't be a problem right?

The nail polish thing, where and how much do I apply it? Do I just cover the whole chip around the die? Is there any specific nail polish I should use or will any do?

I have tried putting in 4.2GHz and leaving voltage on auto and it crashes as well, 124 error according to BlueScreenView.
EDIT: The auto voltage put it to ~1.3V when I was testing 4.2 with auto voltage.

Funny thing. I don't actually get a BSOD, the screen goes all fuzzy like on old TV's when there's no channel, then reboots. I have to check the BSOD codes with BlueScreenView. Majority of my codes are mostly vcore related from what I can tell.
Well, something new happened. I just got a crash on default settings...... d1 error code
Here's my BSW screenshot

https://imgur.com/a/7SIAbIK

View: https://imgur.com/a/7SIAbIK


I guess it could be my RAM going bad, since I had occasional crashes on my old 2500K setup, but I bumped vcore and it stopped. I guess it might have been a placeebo?

 

[Karadjgne]

Conductonaut is liquid metal, therefor conductive electrically to a certain extent. If you look at the pcb, surrounding the die itself you'll see tiny gold spots. So you'd nail polish the entire part of the pcb surrounding the die so that if any of the Tim flows off the die, there's a protective barrier between it and the pcb.

View: https://youtu.be/8niTGcJNK_U
0x7F error:
The above STOP error means a trap occurred in kernel mode and the trap is either one the kernel is not allowed to have or is always fatal. The most common causes of a STOP 0x7F are:

3. Low-level hardware corruption, such as corrupt memory (RAM)

4. Mismatched memory modules


5. A malfunctioning motherboard

 

[Jumpingmonkey2]

Conductonaut is liquid metal, therefor conductive electrically to a certain extent. If you look at the pcb, surrounding the die itself you'll see tiny gold spots. So you'd nail polish the entire part of the pcb surrounding the die so that if any of the Tim flows off the die, there's a protective barrier between it and the pcb.

View: https://youtu.be/8niTGcJNK_U
0x7F error:
The above STOP error means a trap occurred in kernel mode and the trap is either one the kernel is not allowed to have or is always fatal. The most common causes of a STOP 0x7F are:

3. Low-level hardware corruption, such as corrupt memory (RAM)

4. Mismatched memory modules


5. A malfunctioning motherboard

I suspect memory, i have hyperx fury ddr3 1866mhz 1.5V. Trying to oc yesterday it was stable for longer when I tried manually setting the ram variables. I will run a memtest when I have time, then wait for conductonaut, update bios and then try to oc, possibly even reinstall bios. Last d1 bsod on stock was probably brcause I had network drivers for revision 1.0 installed, I checked and have rev 1.1 board.
I hope it's not memory, these sticks cost more than ddr4 3600 sticks