post #1 of 1
Thread Starter 
The title says it all. First off though, I'd like to thank everyone here at Overclock.net for all of the great information that I've used as research here over the past few months. This forum has actually been housed in a few tabs in my browser for quite sometime over the past couple months as I learned and tweaked my 2600k to the 4Ghz I wanted.

Now, before I get flamed, I know 4.0Ghz on a 2600k is weak. But for real, what do I need 5.0Ghz for? I don't, not in the least. I have owned this chip for a year and a half and it was time to step up by 600Mhz and see what all I could make her do.

Let me give you a little background on myself. I've always been an overclocking fiend up until my first processor that ran at 3.0Ghz or faster. But that E8400 did get pushed to it's absolute limit a few times as well. I actually had it boot into windows at 4Ghz on a GeminII air cooler! Wasn't stable enough at all, but it was fun getting there. Ever since my first "real" build sporting the Katmai based P3@450Mhz on an Intel 440bx chipset; I've loved seeing what I could get a CPU to do. Heck, that Katmai on air cooling at 633mhz was a TANK! Specially being backed with 128mb of PC133! I've clocked CeleronA's to insane speeds with water cooling, Athlon XP's with 40% overclocks on air (Luck of the stepping gods on that XP 1700+ lol) I even had the pencil trick on those chips down to the T! So, "This SB chip ain't gonna be nuffin` I can't handle." Or will it?

So let me start by explaining my beginning temps and stresses. I do use Prime95 which seems quite popular here. But only for overall stability tests. I wouldn't dare consider it a thermal testing platform when IntelBurnTest(IBT) does such a better job at sending my SB to the thermal slowdown area. Since I bolted this thing together in 2011, this chip ran stock, stock settings, stock multiplier, you name it, it was stock. Never had too much of a heat issue but I never really stressed it with IBT or Prime either.

So in the beginning; you ALWAYS need those stock temps and voltages and yada yada jotted down for reference points.

@3.4Ghz (stock voltage settings)

Stock voltages ranged from 0.89v to 1.34v
Stock temperatures idle was at 27-30C while IBT could actually push it into the 80C range. (IRRITATED ME!)
Of course, the speedstep would range from 1600Mhz to 3400Mhz and on occasion that BS turbo setup would hit 3800Mhz. (I honestly didn't see the point in the processor shutting off cores to reach a higher frequency, but whatever it's a nice market ploy and on a single thread; maybe it will help.)

So, first I did a little research to find what others have reached and the general consensus basically told me that 4.2Ghz was easy; even with the stock cooler. I wanted 4Ghz, so I buckled to it right off rip. Switched the multiplier and went to town.

@4Ghz (stock voltage settings)

Voltages went from 0.9v clear to the 1.4v area
Temperatures; well lets just say with IBT, it certainly hit the thermal throttle point before I got IBT shut off. Didn't hit the max however and didn't have a problem with the CPU shutting clear off.

So how is it EASY to reach 4.2Ghz if I can't even pull a 3.6Ghz overclock with stock settings without reaching 80C? I don't want 70C much less anywhere near 80C.

I know, let's fix that voltage. Now I know sometimes this can be harder than finding that needle in a haystack; and with my ASRock P67 Extreme6, it was. Finally, after tinkering around in the EUFI bios for a bit, I figured out the way to fix the voltage. Now, as most of us know; even if you fix voltages to a certain number, they won't be perfect. You will have droop at load and it won't ever be perfect. So you have to give/take the numbers a bit. After a few different attempts I learned that my chip WILL not boot at 4Ghz under 1.1v. However, at 1.150v the processor is rock solid at 4Ghz. And, at that voltage... well take a look.

@4Ghz fixed voltage of 1.150v

Temperatures went to a max of 68C (Ambient was about 72F which is the normal for my house)

Just so you know, these temperatures are the average of all four of my cores; and my cores are all within a degree or two of each other. So I did do the thermal paste quite well! smile.gif

So I have my overclock and really nice temperatures at a full load with extreme stress mode and maximum stress level on IBT. But when I'm idle, my CPU is at 4000Mhz and still drawing 1.15v? I want my speedstep back and I want those ultra low under a volt vcore's at idle. This saves power on my electric bill (while not much; a penny is a penny in all respect).

So now I go to researching since I disabled speedstep and fixed the voltages. How can I get these features back? Well; come to find out, the only way to drop the voltages is to have either Auto set (which I read comes straight from the chips VID? Please correct me if I'm wrong here) or set on a mode called "Offset Voltage". Time to learn about that!

@4Ghz at -0.05v Voltage offset

Voltages ranged from 0.89v clear to 1.42v
Temperatures, the idle was great but once I saw the 1.42v; I stopped the stress test immediately. CPU only reached about 50C by that time. I could see it heading straight to the thermal max so I ended that quick.

What caused that major fluctuation? I think, correct me if I'm wrong, there's some sort of variable that the motherboard uses against the chips VID with the LLC (Load Level Calibration) and the voltage offset to set voltages.

So, from my experimentation, this is what I've learned.

Load Level Calibration adjusts how much the voltage will vary by different levels. It's like having a sliding scale where level 1 will allow nearly stock changes to the voltage; the level 5 option on my board allows for more fine tuning between the changes in voltages. In easier terms, the LLC seems to narrow the amount the voltage will increase as the processors load increases. This is where I literally spent days testing and testing tweaking and testing. It seems that on my motherboard that the larger level the LLC is set to the more narrow this spectrum is.

A couple of tests revealed these findings.

@3.4Ghz LLC level 1 and voltage offset of -0.050V

Voltage would reach 1.37 at max load
Temperatures were in the high end of the 70C range and didn't feel safe enough for me. However, it was pretty close to having the LLC and voltage set to Auto. Aside from the actual Vcore, it reached a little less on the LLC level 1 than it did with the LLC on auto. It was better than hitting 1.4v and 80C+ temperatures for sure.

@4Ghz LLC level 1 and voltage offset of -0.050v

Voltage maxed out at 1.4V and sent the temperatures up and I did hit 80C but didn't break into 81C. So it was a little better than the automatic control of the vcore. Still not sure on the stability; but I was making progress.

@4Ghz LLC level 3 and voltage offset of -0.050v

Turned the voltage down to hit only 1.25v and one spike to 1.27v which also left my temperatures in the lower 70C ranges. I was getting somewhere. I learned that tweaking with level 3 on the LLC I got really close to what I wanted at a voltage offset of -0.125v.

@4Ghz LLC level 5 with a voltage offset of -0.050v

Voltages would only reach about 1.25 on a max spike and temperatures were a little lower than they were with the LLC at level 3. I hit a 73C maximum.

After a multitude of different settings, I swore I heard my chip say, "THERE IT IS!" Now I knew I was grade A stable at 1.150v at 4Ghz and wanted to stay in that range. My goal was to actually stay under 1.2v and have a stable 4Ghz overclock with grade A temperatures.

@4Ghz LLC level 5 with a voltage offset of -0.1v

The voltages never have broken 1.2v with a max I believe I saw 1.197 volts.
My temperatures really do excite me here. Mind you, I'm running a Zalman 9500A which isn't the best cooler (I have a hyper 212+ coming for more tinkering xD). I idle at 27C to 30C depending on ambient temperatures and I have yet to break into the 70C range even with IBT set on extreme stress level and maximum stress size. I have ran IBT like that with a 20 unit range which lasted about an hour of full on load on the processor giving even my mouse the "slow and hindered" action. Still didn't break past 69C. I ran three sets of Prime95 blend for 24 hours and still haven't had a single problem.

So now I have my 1600mhz idle with under a volt for power consumption and a nice stable 4Ghz overclock that beats even the stock setup temperatures. The entire time of my tweaking, I learned a TON about this chip and these settings. It brought me back to those Athlon XP days where you'd tweak the CPU and let that voltage set in for a few days and notice you could get another 33Mhz out of the chip! Google was my assistant for a majority of my tour in the overclocking of my SB chip. And for the most part; google always brought me here. So again, I want to thank everyone here for their invaluable information that helped me so much along the way. I am happy to say that my chip finally scored a 7.9 on the windows experience index. Now if I could just get my SSD to flip a 7.9, I'd have 7.9 across the board. LOL

What did I learn?

Voltage offset and Load Level Calibration go hand in hand.
Stock voltage offsets actually can push more voltage to these chips than they require to be stable which causes more heat.
Anything is possible when you have a good community and plentiful information backing you.

Now, as for the future. I actually have a Rosewill Thor V2 case coming to replace my beaten and battered Sunbeam Transformer case that's been modded quite few times. I also have a Hyper 212+ coming along with two of my favorite 120mm fans ever. I actually have a pair in this system I bought in 2008. Silverstone fans that push 110CFM! I may just put them onto my Hyper 212 just for S&G's. LOL But I love to tweak out things my way. I know 4Ghz is quite modest, but I don't "need" more. But if you must know, I was pretty stable at 4.6Ghz at 1.35v xD

Tell me what you think; and for anyone who was/is in my situation as I had here. Remember, reading can do wonders. I hope this helps someone one day when they first experience the woes of voltage offsets like I did when I first off just boosted my multiplier and hit the thermal throttling temperatures. It was quite a scare and I even thought to myself; "Is this chip not going to overclock?"

Thanks for reading!