Need some 2600k OC Assistance!

[Legio]

Hi everyone!

With my new upgrade, I went the 2600k route, since overclocking seemed so easy. Currently I'm on 40x, VID showed as 1.35v which firstly seems quite high?

On auto the vcore idles at 1.30v, so I manually set it down to 1.25v and it now idles at 1.22v. Turbo disabled. Everything else left to auto, including power saving features. 10 hours of prime95, max temp of 50'c and everything seems peachy.

I then tried to up to 43x, booted into windows but get an instant bsod if I start prime95.

Do I really need more than 1.25v for 4.3Ghz? Is my current overclock safe? I wouldn't be able to replace my chip if something happens!

2600k
Noctua NH-D14
Gigabyte Z68X-UD7-B3

I'd really appreciate any pointers or help, it seemed like a 2600k would be so simple to overclock, but in practice it's been rather confusing and worrisome! All I really want is a decent overclock on safe, stock settings.

 

[BonzaiDuck]

Hi everyone!

With my new upgrade, I went the 2600k route, since overclocking seemed so easy. Currently I'm on 40x, VID showed as 1.35v which firstly seems quite high?

On auto the vcore idles at 1.30v, so I manually set it down to 1.25v and it now idles at 1.22v. Turbo disabled. Everything else left to auto, including power saving features. 10 hours of prime95, max temp of 50'c and everything seems peachy.

I then tried to up to 43x, booted into windows but get an instant bsod if I start prime95.

Do I really need more than 1.25v for 4.3Ghz? Is my current overclock safe? I wouldn't be able to replace my chip if something happens!

2600k
Noctua NH-D14
Gigabyte Z68X-UD7-B3

I'd really appreciate any pointers or help, it seemed like a 2600k would be so simple to overclock, but in practice it's been rather confusing and worrisome! All I really want is a decent overclock on safe, stock settings.

Not new to Overclocking, I was hibernating as socket-1366 came and went (although it's "still there.") Just built my own socket-1155 Z68 rig with i7-2600K. Been doing some catchup reading. And also keep in mind I'm "preparing" to overclock, haven't got started yet. I'm waiting to make the final OS installation on a Raptor drive expected to arrive Friday.

There had been some forum posts and reprints of web-pages dealing with a "controversy" about "maximum" voltage settings for the i7-2600K. someone had asserted that it could go as high as 1.5V. It may actually run for a while at that voltage, but there are two ranges -- the "operable range" and the "safe range." You should be able to verify and confirm these voltage specs on the i7-2600K by finding the spec summary on the INtel web-site. However, the upper limit of the "safe" range is apparently no different than earlier processor models -- including LGA-775 processors: 1.37V approximately.

If you look at reviews for another Z68 motherboard (P8z68-V, V-Pro, and -Deluxe by ASUS), the demonstration of the BIOS-based OC-Tuner resulted in an overclock of 4.4 Ghz with a VCORE of about 1.28V. In fact, it was this review-demonstration that made me "break budget discipline" and buy the parts for my Z68 build.

That would give you around 0.1V headroom of "safety" in further OC'ing the processor.

Now . . . . not all motherboards, even of the same chipset, are the same. Gigabyte makes a good board -- I have three. I've now got two ASUS boards in running configurations.

And these "auto-overclocking" features are a whole new ballgame. But if one board's "OC-Tuner" settles at 1.28V for 4.4 Ghz, I'd think it indeed possible that you need more than 1.25 @ 4.3. Just possible -- maybe . . . likely . . .

At that point, if it boots stably into Windows, and CPU-Z shows the actual voltage in the range of 1.25 to 1.28 or 1.29 -- I don't see a problem about your VCORE.

On attempts to get up to 4.7 and then 5.0 on the ASUS board and i7-2600K, it looked as though you'd have to set the voltage close to 1.4V. In fact, the auto-tuner selected 1.404V (or that was what CPU-Z showed as the result) to achieve 4.7 Ghz. The reviewer was able to do additional tuning to get to 5.0 with 1.400V.

Personally, I wouldn't push the voltage under load beyond 1.34V.

But I don't see how you would've destroyed your processor by setting the voltage at 1.25V, and I don't see how you COULD destroy it by bumping it up another 0.03 to 0.05. And the way I see it -- there's wiggle room beyond that.

Otherwise, you're advised to keep the RAM voltage below 1.65. And at this point, I'd advise looking at the sticky on overclocking Nehalem and later. We both need to get more familiar with the new architecture of the processor and chipset, even though you can only OC the K processor by changing the multiplier.

 

[Legio]

Hi,

Thanks for the friendly, and rather refreshing reply! Haha, nice to see someone else who hibernated during 1366. Grats on your new rig as well!

Voltage wise I think you're pretty much spot on, and I'm happy leaving the high end overclocking to the guys that really know what they're doing and/or are willing to debate safe voltage ranges.

For me, I'm really just looking to get a a bit extra "bang for buck" out of my 2600k, after seeing how seemingly easy and safe guys are getting their 2600k's to 4.4-4.6, it pretty much sold me on getting a K series over non K.

From what I've seen, the auto overclocking applications and even just leaving voltage to Auto in the bios results in the voltage being way too high. For 40x on auto, bios was reporting vcore as 1.30v, whereas I have it stable at 1.25v.

Ram wise I came across a somewhat controversial debate about what is safe for Sandy Bridge, so opted for 1.5v ram.

While looking for guides, I've come across some information about vdroop, which seems to be related to why I'm unstable at 43x. You can set LLC which decrease the gap between idle and load voltages, the asus guide has levels like "high", "Extream", whereas my board has levels 1-10.

I'd love to know exactly what those levels are before I start tinkering with them. I also wonder what would be safer between, running eg: 1.25v constant even when idle, vs 1.20v idle and 1.28v load.

 

[Kenmitch]

Heres somewhat of a guide for your MB/CPU combo. The setting bios in windows is kinda cool.

http://forums.overclockers.co.uk/showthread.php?t=18279474

Of course this guy is going for 5ghz but does give some examples of lower clocks.

 

[bunnyfubbles]

You have a very good CPU cooler, 1.35v will be more than safe as long as your temps don't go out of control (and they shouldn't, not on a Noctua NH-D14)

Ideally you want to stay low 70s (if not below 70) Celsius when under full load, but as long as you don't go over 80C @ 1.35v you really don't have anything to worry about in terms of destroying your CPU or even noticeably shortening its lifespan. I know plenty of crazy people who are confident when pushing more than 1.4v and even close to 1.5v just so they can hit that 5GHz epeen mark, or even a mere 4.8GHz.

Personally I drew the line at 1.35v and saw how far it was able to take me (4.7GHz in my case)

 

[BonzaiDuck]

Hi,

Thanks for the friendly, and rather refreshing reply! Haha, nice to see someone else who hibernated during 1366. Grats on your new rig as well!

Voltage wise I think you're pretty much spot on, and I'm happy leaving the high end overclocking to the guys that really know what they're doing and/or are willing to debate safe voltage ranges.

For me, I'm really just looking to get a a bit extra "bang for buck" out of my 2600k, after seeing how seemingly easy and safe guys are getting their 2600k's to 4.4-4.6, it pretty much sold me on getting a K series over non K.

From what I've seen, the auto overclocking applications and even just leaving voltage to Auto in the bios results in the voltage being way too high. For 40x on auto, bios was reporting vcore as 1.30v, whereas I have it stable at 1.25v.

Ram wise I came across a somewhat controversial debate about what is safe for Sandy Bridge, so opted for 1.5v ram.

While looking for guides, I've come across some information about vdroop, which seems to be related to why I'm unstable at 43x. You can set LLC which decrease the gap between idle and load voltages, the asus guide has levels like "high", "Extream", whereas my board has levels 1-10.

I'd love to know exactly what those levels are before I start tinkering with them. I also wonder what would be safer between, running eg: 1.25v constant even when idle, vs 1.20v idle and 1.28v load.

Your point about "auto" versus "auto-tuned" is a source of some concern, because you're quite right. That's why I feel it necessary to do a lot of reading before I "jump" into this. The auto-tuning features have merit, and someone mentioned that they provide a starting point for finer tuning. But even the review demo I mentioned for my ASUS board noted that you could get an extra few hundred MHz at a slightly lower voltage (although 0.004V difference doesn't seem a lot.)

I also just went back to the Intel site to confirm further what I thought about the 1.37V limit. They used to post it in the spec! I thought I'd even seen it for this CPU, but can't find it.

Some of our friends here also seem to think that's in the ball-park. And I DID find an article or post coming directly from an Intel rep confirming it.

Also, I'm not sure what a "good" VID for the processor is. As said -- still waiting for a hard drive before OS reinstall and drivers, but this prelim install without the chipset drivers has CPU-Z showing my VID at between 0.98V and 1.00+V. I suppose that's good, but have no point of reference.

I'm also stunned by the responsiveness of this core and the system running at its 3.4Ghz stock speed. I need to toy with it, but must wait for the V-Raptor and the final OS install on Friday . . .

On the vDroop issue, this bears again on maximum safe limit. It isn't so much that load causes a disparity, but that there is an "overshoot" when the load ceases and the CPu goes to an idle state. The overshoot supposedly can stress the processor, and if it occurs with some significance above the save range, it will slowly degrade the chip.

EDIT: INFORMATIVE: Forgot to mention. Visiting the Intel site and looking at the spec sheet on the 2600K, it's confirmed that the thermal throttling point is 72+C. I'd just run a prelim CoreTemp on my barely-installed "prelim-OS" before the Raptor arrives Friday. At 78F room ambient, the idle temps on all four cores fluctuate between 25 and 26C. That only suggests that my sensors are pretty much spot-on, and the NH-D14 cooler is "doing its job." And as I said elsewhere -- it's the load temperature that's proof of the pudding.

My best guess at this point though: I expect to see some encouraging results on the load temps, even for pushing the VCore up to maybe 1.30. And I can see this "process" (as VirtualLarry calls it) drawn out a bit more as I tinker with this thing.

 

[Binky]

I just turned my multiplier up to 45 and left everything else alone. Someday when I'm really bored maybe I'll tweak/tune it more.

 

[Legio]

Thanks for all the replies, all insight is much appreciated!

Heres somewhat of a guide for your MB/CPU combo. The setting bios in windows is kinda cool.

Thanks for the link, I had see this guys guide and that he used LLC Level 6 with his 4.4 overclock, I'd love to know more about the LLC levels on my board, and exactly what they're doing.

I'm a bit pedantic, as I don't want to try anything before I know exactly what it's doing.

You have a very good CPU cooler, 1.35v will be more than safe as long as your temps don't go out of control (and they shouldn't, not on a Noctua NH-D14)

Ideally you want to stay low 70s (if not below 70) Celsius when under full load, but as long as you don't go over 80C @ 1.35v you really don't have anything to worry about in terms of destroying your CPU or even noticeably shortening its lifespan. I know plenty of crazy people who are confident when pushing more than 1.4v and even close to 1.5v just so they can hit that 5GHz epeen mark, or even a mere 4.8GHz.

Personally I drew the line at 1.35v and saw how far it was able to take me (4.7GHz in my case)

The NH-D14 is absolutely amazing, well worth all the praise its received. Ideally I think I wouldn't like to go over 60, at 4Ghz I've peaked 52'c during a prime95 run, so hopefully 44x (or whatever final speed I end up with) won't be too much hotter.

Voltage wise, I'm still pretty nervous. I'm cool and stable with 40x @ 1.25v, and debating if a voltage bump is worth the extra 0.4 Ghz. I want to find out more about my boards 10 LLC levels, and how that can change things. I have a feeling I could get 43x stable with LLC, but I'm not happy to try it until I know what it's doing behind the scenes.

I guess I'm a terrible overclocker, as I'm overly cautious with everything ()

Your point about "auto" versus "auto-tuned" is a source of some concern, because you're quite right. That's why I feel it necessary to do a lot of reading before I "jump" into this. The auto-tuning features have merit, and someone mentioned that they provide a starting point for finer tuning. But even the review demo I mentioned for my ASUS board noted that you could get an extra few hundred MHz at a slightly lower voltage (although 0.004V difference doesn't seem a lot.)

I also just went back to the Intel site to confirm further what I thought about the 1.37V limit. They used to post it in the spec! I thought I'd even seen it for this CPU, but can't find it.

Some of our friends here also seem to think that's in the ball-park. And I DID find an article or post coming directly from an Intel rep confirming it.

Also, I'm not sure what a "good" VID for the processor is. As said -- still waiting for a hard drive before OS reinstall and drivers, but this prelim install without the chipset drivers has CPU-Z showing my VID at between 0.98V and 1.00+V. I suppose that's good, but have no point of reference.

I'm also stunned by the responsiveness of this core and the system running at its 3.4Ghz stock speed. I need to toy with it, but must wait for the V-Raptor and the final OS install on Friday . . .

On the vDroop issue, this bears again on maximum safe limit. It isn't so much that load causes a disparity, but that there is an "overshoot" when the load ceases and the CPu goes to an idle state. The overshoot supposedly can stress the processor, and if it occurs with some significance above the save range, it will slowly degrade the chip.

EDIT: INFORMATIVE: Forgot to mention. Visiting the Intel site and looking at the spec sheet on the 2600K, it's confirmed that the thermal throttling point is 72+C. I'd just run a prelim CoreTemp on my barely-installed "prelim-OS" before the Raptor arrives Friday. At 78F room ambient, the idle temps on all four cores fluctuate between 25 and 26C. That only suggests that my sensors are pretty much spot-on, and the NH-D14 cooler is "doing its job." And as I said elsewhere -- it's the load temperature that's proof of the pudding.

My best guess at this point though: I expect to see some encouraging results on the load temps, even for pushing the VCore up to maybe 1.30. And I can see this "process" (as VirtualLarry calls it) drawn out a bit more as I tinker with this thing.

I also couldn't find anything on the intel site, if I do try to bump voltage a bit, I'll try stay within 1.30 and see how things look.

Wow, my VID for stock was 1.25v, hence me using it for 40x.

I've read that LLC can sometimes invert vdroop and your vcore will spike higher than what you've set it to. Gigabyte should have published an explaination of their 10 levels if you ask me.

Ah, you've got a Noctua as well :thumbsup: Those temps are pretty much inline with mine. I've set mine to voltage, so it spins up/down depending on the tempreture, it usually sits between 400 and 500 rpm, basically completely silent.

Stock idle I was also in the 25'c region, at 4Ghz I idle at 30'c and during a prime95 run the hottest core peaked at 52'c. All the while being completely silent, I couldn't love this cooler more if I tried :wub:

 

[BonzaiDuck]

. . .

Voltage wise, I'm still pretty nervous. I'm cool and stable with 40x @ 1.25v, and debating if a voltage bump is worth the extra 0.4 Ghz. I want to find out more about my boards 10 LLC levels, and how that can change things. I have a feeling I could get 43x stable with LLC, but I'm not happy to try it until I know what it's doing behind the scenes.

I guess I'm a terrible overclocker, as I'm overly cautious with everything ()

Quite the contrary. Especially for air-cooling -- caution is wisdom. For water-cooling, lower temperatures can lull the OC'er into a false sense of security with voltage settings that could be excessive.

I also couldn't find anything on the intel site, if I do try to bump voltage a bit, I'll try stay within 1.30 and see how things look.

Wow, my VID for stock was 1.25v, hence me using it for 40x.

It would be useful to share all this information. Was I mistaken, or was the 2600K released around January? During the "big years" for the Conroe, Kenstfield, Wolfdale and Yorkfield cores, there were popular threads at this site where people were encouraged to post their OC'ing results and other useful information. Maybe I missed something, and certainly those type of threads would have dropped into the archival zone of the heap for the Socket-1366 cores. Maybe the unlocked "K's" multiplier makes it too easy; the OC-Tuner and other features make it too easy; and this sort of information seems low-priority. Can't tell.

I've read that LLC can sometimes invert vdroop and your vcore will spike higher than what you've set it to. Gigabyte should have published an explaination of their 10 levels if you ask me.

Ah, you've got a Noctua as well :thumbsup: Those temps are pretty much inline with mine. I've set mine to voltage, so it spins up/down depending on the tempreture, it usually sits between 400 and 500 rpm, basically completely silent.

Stock idle I was also in the 25'c region, at 4Ghz I idle at 30'c and during a prime95 run the hottest core peaked at 52'c. All the while being completely silent, I couldn't love this cooler more if I tried :wub:

That was what I meant about the LLC. Per the Noctua, not too different in design to the old TR IFX-14. With previous coolers like this -- I have a TRUE and a Megahalem currently deployed -- I was designing ducts for enhanced cooling with foam-board and lexan, and contriving ways to mount fans near -- but not on -- those coolers (and with some success.)

After reading Noctua's optimistic appraisal of how its backplate and mounting gear are more than sufficient to relieve worry, I went ahead and mounted the two Noctua fans as shipped. But -- frankly -- if it weren't for the limits on the CPU_FAN and OPT_CPU_FAN headers on my mobo (must use <= 0.10A fans) and the fact that I'd want to have a relevant fan connected to one of those headers, it would be quite easy to craft a Lexan duct around both of those towers so that a sturdy case fan like a 120x38mm 0.40A San-Ace or a 0.60A Panaflo NMB-MAT would be an improvement over the Noctuas, still possible to keep reasonably quiet at some CPU-usage levels, and putting no weight on the cooler.

Given your results, it will be interesting what I also find after Friday . . .