need help with a 2500k overclock on gigabyte Z68X UD4 B3

[poohbear]

Hey all,

so i'm overclocking this 2500k on default vcore, and its going lovely all the way up to 4.3ghz, then i hit a wall.... hard! default voltage is 1.245v as read in CPU-Z @ 4.3ghz stable in P95 (blended test), but to get to 4.5ghz (a mere 200mhz increase), i need a whopping 1.485v just to get it stable in Prime95! that's a .24 vcore increase for those 200mhz! I get 67c under load (using P95) @ that setting (4.5ghz). This is all on a Noctua D14.

Now, there are alot of settings in the voltage area that i'm not familiar with, what are these terms as neither my manual nor the bios explain what they are:

Multi-Steps Load -Line (disabled by default, but has levels 1-9)
QPI/Vtt voltage 1.050v (default)
System Agent Voltage 0.920 (default)
CPU PLL 1.800v (default)


are any of you familiar with those settings? they're found in the voltage section of the cpu and i'm wondering if adjusting them would effect my overclock? is the 1.24 to 1.48 jump normal to stabilize the 2500k @ 4.5ghz?

Now, the strange thing is when i run the Dawn of War 2 performance test @ 4.5ghz (its CPU intensive strategy game) my minimum FPS drops to 9-14 fps!!!! When its at stock it's @ 33fps minimum fps! why this strange disparity? i ran the test numerous times and the same thing: overclocked @ 4.5ghz the minimum FPS is 9-14, but @ stock its 33-36 minimum FPS. Can anyone explain that?
EDIT: Nevermind, that was because Nvidia's CP had adaptive vsync on, screws up with performance test. With Vsync off its 40 minimum FPS.

thanks in advance for any clarification.

 

[BonzaiDuck]

This is page 2.

http://www.bit-tech.net/hardware/cpus/2011/01/07/how-to-overclock-the-intel-core-i5-2500k/2

Perhaps you need to read the whole article starting with page 1.

This is but one article on OC'ing the Sandy Bridge processors. It seems this article refers to the P67 chipset, which was precursor to the Z68.

There should also be some threads in this forum for last year, particularly for months following April.

 

[poohbear]

thanks, that lead to a treasure trove of info! seems i need to manipulate more than just the vcore to get it stable....

 

[poohbear]

hey all, something strange is happening with my overclock: I'm @ 4.6ghz and im using 1.475vcore in bios (1.440 in CPUz), but when my cpu is under load the vcore drops to 1.39 and when its idle it goes back up to 1.44 (CPUZ settings). Which setting is allowing it to drop like that? is there any way i can just keep it at 1.39 all the time instead of having it drop when its under load?

i'm using "Multi-Steps Load -Line" @ Level 3 (default its off), and i read that's the setting that combats vdroop. It's also called load line calibration in other bioses & there's only an enable/disable feature, but on this Gigabyte one its called Multi step load line. Is that the one i want?

thanks in advance for any clarification.

 

[BonzaiDuck]

hey all, something strange is happening with my overclock: I'm @ 4.6ghz and im using 1.475vcore in bios (1.440 in CPUz), but when my cpu is under load the vcore drops to 1.39 and when its idle it goes back up to 1.44 (CPUZ settings). Which setting is allowing it to drop like that? is there any way i can just keep it at 1.39 all the time instead of having it drop when its under load?

i'm using "Multi-Steps Load -Line" @ Level 3 (default its off), and i read that's the setting that combats vdroop. It's also called load line calibration in other bioses & there's only an enable/disable feature, but on this Gigabyte one its called Multi step load line. Is that the one i want?

thanks in advance for any clarification.

I'm guessing that your LLC or load-line calibration has "multiple steps" just like my ASUS mobo and others.

I also think you're using excessive voltage at that speed. You should be able to get to 4.6 with a voltage less than 1.35V under load. The difference you see is "vDroop," which is reduced progressively by higher and higher LLC settings. The problem with using "extreme" LLC: if you think of harmonic motion, when the processor unloads, the load voltage (with vDroop) snaps back to the idle level and actually overshoots it before equilibrating to that 1.44 level. And you can't measure or see the overshoot.

If I can get my i7-2600K to 4.6+ with a maximum "idle" or unloaded turbo voltage of 1.36 to 1.38 and showing full-load voltage at around 1.33V, you shouldn't need to have settings too different from that with an i5-2500K.

Set your VCCIO to 1.1V, leave the PCH and PLL voltages on "auto," and use "Offset mode" over-clocking. Also, look somewhere in your BIOS screens for a voltage item that suggests "Extra voltage in Turbo." Find out what the reported values are under the "auto" setting while in BIOS.

You can either adjust the "offset" (with sign "+" or "-") to increase or decrease the load voltage total, or use both settings.

IF your system can run Prime95 small-FFT test for 10 to 20 minutes without fail or BSOD, you would have reached a point where single step adjustments to Offset will extend that running time until it will run all day without error. You would also want to test with Large-FFT after that. At the point where you think the voltages are getting you stability at whatever speed you're at or hoping for, you'd run IntelBurnTest or LinX for maybe 10 iterations.

The validating LinX or IBT would probably need 20 or 30 iterations.

Suggest before you really start doing this you use a bootable "memory test" program like HCI Memtest or Memtest86+ to assure your memory hardware is "good."

There are a lot of guides, review articles that function as guides, forum posts, and so on which are "out there" on the web. Somehow, I just get the impression from your posts that you need to do some reading and approach this effort with a little more patience and attention to detail.

Keep notes -- write down everything -- settings, values, reported values, etc.

 

[poohbear]

yes i always keep notes. This chip seems to need alot of vcore to 4.5ghz and beyond, it was on default auto voltage all the way up to 4.3ghz, then after that is when i had to start tweaking.

The Multi steps load line is what's effecting vdroop? so, if i have a higher level it'll show less of a drop? ie now its level 3, and it drops from 1.44v to 1.392 under load, so if i raise the level to 4-6 i'll see less of a drop, and therefore i can set the vcore to 1.3x in the Bios?

EDIT: ok i tried it and i see that is indeed the case. Why would they do this though? i have to change a setting to reduce vdroop? why can't the mobo automatically readjust the vcore to match what we're seeing in the Bios & what we're seeing the desktop? Am3+ platforms do that..... if i set 1.4vcore in the Bios, it does what it has to do to ensure 1.4vcore is displayed in system & that when the system is underload it continues to recieve 1.4v. The user having to go and meddle with another setting to get the correct output seems so redundant. I read some of the higher end z68 mobos automatically do this for you, too bad they could'nt make it a standard feature.

 

[pantsaregood]

Your VCore is really high. Maybe back your clock speed down to 4.4 GHz. 4.4 is still 33% faster than stock, and will usually run at or under 1.3v.

 

[BonzaiDuck]

OK . . . you'd been an AMD/AM3 junkie. Now it makes more sense.

I'm just rather surprised that you can't get the 2500K stable at the 4.4 or 4.5 clocks without using that much extra voltage. there are "not so good," "average," "better" and "phenomenally stellar" units coming off the INtel assembly line, but yours can't be that bad.

Even so, it's true -- some motherboards handle power better than others, and maybe that's a factor. I don't go for the flagship board models, but I choose them for "industrial-strength" phase-power design. That feature varies from the low-end to higher-end of the price spectrum.

 

[poohbear]

when you say voltage are you referring to vcore or the other voltages like QPI/VTT voltage, System agent voltage, & CPU PLL voltage? I'm just used to seeing vcore, vdimm (RAM voltage), & NB voltage when doing AM3+ overclocking, but with this 2500k & Z68 there are these 4 different voltages all for the cpu that we hafta tinker with to get the most out of our overclock.

 

[pantsaregood]

There really shouldn't be much you have to mess with other than DVID. Some other things can help if you've significantly passed the point of diminishing returns, but I really don't see a reason to bother with those.

Hit 4.4 GHz and be happy with 33.3% extra performance.

 

[poohbear]

man i've been reading so many guides on SB overclocking and many of them contradict other guides on what to do, ie you should turn off power saving features while another says you should leave them on, VTT/VCIO is important for stability while another says just leave it alone and don't touch it etc etc

Why did intel make 4 different types of vcore for the cpu & on top of that not resolve vdroop issues?!? its so annoying and makes overclocking past 4.5ghz a complete headache. It's supposed to be an unlocked cpu like AMD's Black Editions, but atleast with AMD there's ONE vcore and that's it, no vdroop, no VTT/VCIO, system agent voltage, CPU PLL voltage and all this other nonsense! The user has to "offset" the vcore under load and idle to avoid vdropp? kinda shoddy engineering. its just such a mess of options in the bios and coming from an AMD setup its looks really shoddy.

 

[2is]

vdroop isn't an issue, thus does not need to be resolved. It's there for a reason. Remember, Intel designs their processors to work as reliably as possible at their default speeds. vdroop may be an issue to extreme overclocks but is beneficial to the long term reliability/stability of the CPU when used as was intended. Catering to the extreme overclockers is where the motherboard manufacturers come in with things like LLC to combat vdroop.

As was stated, overclocking SB or IB is very easy for light to moderate clocks. When you go above that, you need to do more tweaking. If you can show me a modern CPU that can get a 50% OC with one or two simple adjustments, which seems like what you're kind of expecting here, I'd love to know what processor that is.

I also think you're killing yourself with information overload. The power savings deal is simple really. Disabling power savings certainly will NOT hurt your overclocking, the debate is weather or not it will help. Some say yes, some say no. In your case, you can go ahead and disable it as it's one less troubleshooting factor you need to worry about when you're troubleshooitng/testing. Once you've hit the speeds you want to hit, you can go back an enable it and see if it remains stable. If it does, leave the power savings enabled since at that point, there's no benefit to disabling it.

 

[pantsaregood]

Core i7-920s with the D0 stepping would manage 50% overclocks pretty regularly by just bumping VCore and BCLK. Bumping a 2500K or 2600K 50% would land you over 5 GHz, which is just beyond what Sandy Bridge can scale to with reasonable voltage.

I think a 33% overclock is a nice place to aim for most of the time. It doesn't require extreme tweaking or high voltage, and it results in a very significant increase in performance. The extra 300-400 MHz you can get out of SB/IB just aren't worth it due to the diminishing returns.

 

[2is]

I've played with a few D0's. I wouldn't say 50% was regular with just those tweaks, more like an exception.

 

[pantsaregood]

More common than a 50% overclock on a 2500K/2600K, definitely. A Sandy Bridge CPU that will run at 5.0 GHz without breaking 1.4v is pretty unusual. Seeing a D0 920 hit 4.0 GHz may not be "common," but it isn't really impressive either.

 

[2is]

I agree, my point is he seems to be expecting massive overclocks with minimal effort, and that just isn't going to happen. I can run 4.3GHz fairly easily, above that requires a bit more effort. You can just tell the CPU is fighting you to hit that speed and remain stable, and when you do get there, there's always that thought in the back of my head "am I really stable?" considering the additional tweaks that were necessary. Particularly if a program stops responding. So instead I opted for 4.2GHz which is 100MHz less than what the chip can do pretty comfortably which gives me that extra cushion. Knowing my system is running stable is more important than 2-300MHz of performance I likely will not notice.

 

[pantsaregood]

I absolutely agree. The vast majority of SB units will run at a 33% overclock without any significant tweaking. Ivy Bridge is probably the same, but it is more likely to become thermally limited. I would assume your 3770K will clock above 4.2 GHz without too much difficulty, but the voltage increases give higher temperatures than you'd like?

 

[poohbear]

yea i guess i expected too much. People are posting about 4.7ghz overclocks on SB as if its common, its NOT (and comments like "show me a modern CPU that does a 50% OC with minimal tweaks" surely don't help, SB does NOT overclock by 50%, that means would mean it can reach 5.1ghz with "minimal tweaks"). they max out @ 4.3-4.5 before massive tweaking is needed. IMO its simply not worth it to go beyond that. Maybe i'll change my mind in a few months, but the difference in performance in the benchmarks i ran (mostly gaming) is very minimal between 4.4ghz and 4.6ghz (mine won't boot with 4.7, but i can't be bothered doing all the tweaking for that).

What's more frustrating is the different recomendations from different guides that completely contradict each other. Seems like people don't really know what some settings do, they say things like "try it to see if it helps". lol why would i try it on a whim just to see if its effecting my OC??

anyways thanks all for your help.

 

[2is]

I absolutely agree. The vast majority of SB units will run at a 33% overclock without any significant tweaking. Ivy Bridge is probably the same, but it is more likely to become thermally limited. I would assume your 3770K will clock above 4.2 GHz without too much difficulty, but the voltage increases give higher temperatures than you'd like?

Yeah exactly, I can do 4.3 by simply leaving everything the way it is and enabling the lowest level of LLC. I can do 4.4 by then enabling Internal PLL Overvoltage and bumping up the offset by another couple notches, but as you said, temps can get in the way.

yea i guess i expected too much. People are posting about 4.7ghz overclocks on SB as if its common, its NOT (and comments like "show me a modern CPU that does a 50% OC with minimal tweaks" surely don't help, SB does NOT overclock by 50%, that means would mean it can reach 5.1ghz with "minimal tweaks"). they max out @ 4.3-4.5 before massive tweaking is needed. IMO its simply not worth it to go beyond that. Maybe i'll change my mind in a few months, but the difference in performance in the benchmarks i ran (mostly gaming) is very minimal between 4.4ghz and 4.6ghz (mine won't boot with 4.7, but i can't be bothered doing all the tweaking for that).

What's more frustrating is the different recommendations from different guides that completely contradict each other. Seems like people don't really know what some settings do, they say things like "try it to see if it helps". lol why would i try it on a whim just to see if its effecting my OC??

anyways thanks all for your help.

I can do 4.7 on my 2500k but I'm only running a Hyper 212 evo on there and temps can get a bit toasty at that speed. I'm running it at 4.4GHz for the same reasons i'm running my 3770k at 4.2. SB can do 4.7 and quite commonly, but you're expecting to achieve that by simply bumping the multi and vcore and viola, a stable 4.7GHz. 4.7GHz is certainly very possible, just not as easy as you may have liked it to be.

as far as "why would I try it on a whim just to see if its affecting my OC??" question... You answered in in the same sentence. You would try it to see if it affects your OC. Not all chips are created equal and not all motherboards are the same. In this very thread we have people who were affected by power savings while others were not. There is no magic bullet here. Sometimes you have to put some work in when trying to achieve higher clock speeds. Even if it means taking a few minutes out of your life to change a setting on a whim to see if it affects your OC. Expecting 4.7GHz isn't necessarily unreasonable (not gauranteed but not unreasonable) expecting to hit it with ease combined with yoru unwillingness to change bios settings without knowing exactly the outcome before hand, is.

 

[poohbear]

its not unwillingness, it just seems u're being arbitrary about it with no explanation or reasoning. There has to be some method to it, just changing a few settings at random and then being even more confused isnt gonna get me any nearer to stability. From what i read the VCCIO/VTT/QPI (name depending on mobo) is only to be tinkered with if u're overclocking your RAM, which im not, so why touch it?

 

[Red Hawk]

Just a note about overvolting on a Gigabyte motherboard -- when I tried it on my Gigabyte motherboard and 2500k, CPU-Z did not report the correct voltage. So I wouldn't put too much stock in it.