Can't get above 4.4 stably with i7-2600k

icebox

Distinguished
May 9, 2008
18
0
18,510
I've had a i7-2600k setup for about a year now and after a burst of tweaking that got me to a stable 4.3 ghz on the processor I let things stay. This is my first OCed machine. Now I've started tweaking again, and find as before that I can't get above 4.4 ghz. I've got the following system specs:

CPU i7-2600k, air cooled with Noctua NH14
Mobo MSI Z68-GD80 (B3)
PS OCZ Z1000M
Mem: 2x4gb Corsair Dominator CMP8GX3M2A1600C9
GPUs are absent on RMA at the moment, using integrated graphics
running win 7 64

I have voltage set to auto, and CPU-Z reports it at 1.248 at load. Right now I have multiplier at 43 and Bclock at 101. This is stable. I had the memory running at 1066 until now, when I enabled the XMP mode and now have it running at 1600. Every attempt at either 1066 or 1600 on the memory to exceed 4.4 ghz on the CPU results in blue screens. I've tried multipliers of 44, which at 1066 will run stably at load for no longer than 24 hours before bluescreening, and I've tried increasing Bclock to 103, which resulted in a bluescreen quickly at load. 102, which is just below 4.4ghz, may be stable. It appears that auto voltage will work at least to 1.4 v on this motherboard, so I don't think there's a voltage problem.

What I'm wondering is whether something with the memory is the problem. The processor should be capable of a couple more ticks in the multiplier, as voltages are not problematic and temps are under 60 C. I'm no pro at overclocking, but I saw elsewhere that the DDR Vref voltages, of which I'm unsure of the function, should be half the Dram voltage. When I entered XMP, the motherboard automatically raised the DRAM voltage from 1.5 to 1.65, but didn't change the DDR Vrefs. Might that be an issue?

 
Solution
upping the BLCK also overclocks your RAM. so its just another point of failure for very little gain.

on non K series processors this is the only way to overclock and makes it much harder. hence the release of the K series chips (extra price just to have an unlocked multiplier).

be very carefull of heat as you will be reaching the limit of air cooling (even aftermarket air)

as was said by amuffin. start with BLCK 100. multi 45x and 1.31 vcore. if the computer is stable and not crashing try dropping the vcore down bit by bit until its at its lowest spot but the system is still stable.

same goes if its crashing @ 1.31. try upping vcore little by little until its stable

icebox

Distinguished
May 9, 2008
18
0
18,510

I've been told 101 or 102 are safe on Bclock, and seen on intel's website that the setting can be changed +/- 5%. Why shouldn't it be touched? I'll try the voltage change in the AM.
 


"Safe" is usually considered to be 103-105 max (even though I've seen some people go as far as 110 without problems), but the fact is, you just don't need to. There's seriously no reason at all with an unlocked CPU. Just put it back to 100 and never think about it again.
 

HugoStiglitz

Distinguished
upping the BLCK also overclocks your RAM. so its just another point of failure for very little gain.

on non K series processors this is the only way to overclock and makes it much harder. hence the release of the K series chips (extra price just to have an unlocked multiplier).

be very carefull of heat as you will be reaching the limit of air cooling (even aftermarket air)

as was said by amuffin. start with BLCK 100. multi 45x and 1.31 vcore. if the computer is stable and not crashing try dropping the vcore down bit by bit until its at its lowest spot but the system is still stable.

same goes if its crashing @ 1.31. try upping vcore little by little until its stable
 
Solution

icebox

Distinguished
May 9, 2008
18
0
18,510
One thing I noticed since I've been running CPU-Z a lot is that the voltage it reports at load is lower than what it reports at idle. E.g., Voltage at load, 1.248 volts, at idle, 1.304. Why would this be the case?
 


A phenomenon known as Vdroop. Not necessarily a bad thing, but LLC can fix it. My idle voltage at 4.5 with SpeedStep disabled is 1.328V and @ 100% load it's 1.277V. It's perfectly stable though, and runs cooler than it would at 1.328V, so I don't complain.
 

icebox

Distinguished
May 9, 2008
18
0
18,510
I hadn't previously disabled speed step as I'd seen mixed advice about that on the web, but I did this time when I went into the bios to make changes. First looks at 45x100 with voltage hard set to 1.31 and speed step disabled shows almost no vdroop at all. 1.304 idle/1.298 load. (though I don't understand why at idle I'm not seeing 1.310 from CPU-Z). Amazing difference. 2 and a half hours in, this is the longest I've gotten 4.5 to last. Here's hoping it stays stable. Before I had 4.4 last up to 18 hours at load before bluescreening.
 

Haserath

Distinguished
Apr 13, 2010
1,377
0
19,360
Vdroop is an Intel specification specifically designed to lower voltage at load. It's to make sure the voltage in the BIOS is the highest voltage the processor ever receives from the VRMs.

I think 1.35V Vcore should easily get you to 4.4 even with Vdroop in effect. You can tweak it down after you know it's stable.

Intel also specifies 1.575V max for the ram. I would try to keep them around 1.5V if you can; no need to push them, they don't affect performance much.
 

icebox

Distinguished
May 9, 2008
18
0
18,510

Temp is 61 C right now running at 4.5, about 4 C above the temps I had running at 4.3. I believe intel recommends keeping the processor below 72 C. So it appears I have a bit of heat headroom, no? Though once I get the video cards back, we'll see. Those threw off unbeliveable amounts of heat at load folding and will likely raise the CPU temp by a few degrees as well.