2500k @ 5GHz?

[996GT2]

The OP's friends are claiming 5.0 Ghz -- likely as a matter of bragging rights. I am skeptical about this. First -- air-cooling wouldn't likely be sufficient to handle the temperatures for such an overclock -- whether 2500K, 2600K -- even 2700K.

I have been running for over 2 years @ 5 GHz on air cooling My max load temp after over an hour of LinX is 76.75C averaged across all 4 cores, well under Intel's recommended maximum of 90C. I do have EIST enabled so my chip is mostly idling at 1.6 GHz when I'm not running anything intense on it. I ran 12 hours of Prime95/LinX and it passed with flying colors, so I would consider that a stable overclock. Not to mention it hasn't crashed once in 2 years, haha.

You just need to have a very good chip in order to do 5 GHz stable at under 1.4V. And a good cooler + a case with good airflow characteristics.

 

[Headfoot]

I have been running for over 2 years @ 5 GHz on air cooling My max load temp after over an hour of LinX is 76.75C averaged across all 4 cores. Well under Intel's recommended maximum of 90C. I do have EIST enabled so my chip is mostly idling at 1.6 GHz when I'm not running anything intense on it.

You just need to have a very good chip in order to do 5 GHz stable at under 1.4V. And a good cooler + a case with good airflow characteristics.

Did you buy that one used looking for a golden chip or just a really lucky purchase? I suppose the fact it was a 2700k helps with the chance of it being well binned

 

[996GT2]

Did you buy that one used looking for a golden chip or just a really lucky purchase? I suppose the fact it was a 2700k helps with the chance of it being well binned

I had a "golden" 2500K which I was very reluctant to sell because it did 5 GHz easily on air, but I eventually did sell it to upgrade to i7. Got lucky with my chip, although 2700Ks in general tend to be well-binned.

 

[BonzaiDuck]

I have been running for over 2 years @ 5 GHz on air cooling My max load temp after over an hour of LinX is 76.75C averaged across all 4 cores, well under Intel's recommended maximum of 90C. I do have EIST enabled so my chip is mostly idling at 1.6 GHz when I'm not running anything intense on it. I ran 12 hours of Prime95/LinX and it passed with flying colors, so I would consider that a stable overclock. Not to mention it hasn't crashed once in 2 years, haha.

You just need to have a very good chip in order to do 5 GHz stable at under 1.4V. And a good cooler + a case with good airflow characteristics.

Well, good for you! Even as it's the 2700K, I still think that's a good CPU, there. I assume you selected "time" versus number of iterations with LinX, but you should be able to confirm equivalent of 12hrs PRIME95 with maybe half that time under either IBT "Maximum" or LinX. You didn't say which cooler you're using, and I'm curious. Your thermal average is only about 8C above mine. I'd expect to have your voltage profile just for 4.8.

But if you remember back in 2011 and 2012, statistics had emerged: "X percent would go to 4.5, Y percent to 4.7, Z percent to 4.8 and above."

Makes me jealous!

PS You must have read some of my other posts about my "intermittent problem" [ha-ha -- "ha-ha."] I can explain that, and just the other day discovered a bad piece of hardware with driver conflicts. But -- no problem -- don't need to. It was never the OC, though!

 

[996GT2]

Well, good for you! Even as it's the 2700K, I still think that's a good CPU, there. I assume you selected "time" versus number of iterations with LinX, but you should be able to confirm equivalent of 12hrs PRIME95 with maybe half that time under either IBT "Maximum" or LinX. You didn't say which cooler you're using, and I'm curious. Your thermal average is only about 8C above mine. I'd expect to have your voltage profile just for 4.8.

But if you remember back in 2011 and 2012, statistics had emerged: "X percent would go to 4.5, Y percent to 4.7, Z percent to 4.8 and above."

Makes me jealous!

PS You must have read some of my other posts about my "intermittent problem" [ha-ha -- "ha-ha."] I can explain that, and just the other day discovered a bad piece of hardware with driver conflicts. But -- no problem -- don't need to. It was never the OC, though!

I did run Prime95 as well as LinX.

I'm using a Thermalright HR-02. 2x Gentle Typhoon AP-15 push-pull with an additional 180mm fan providing airflow over the heatsink, RAM, VRMs. The fans run at <800 RPM when idle to keep things quiet. I had a Noctua D14 but swapped it for the HR-02 because of much easier RAM access. There was a few C penalty in cooling performance but the HR02 is still a beast of a cooler.

My GPU isn't the same anymore, but everything else is:

 

[BonzaiDuck]

I did run Prime95 as well as LinX.

I'm using a Thermalright HR-02. 2x Gentle Typhoon AP-15 push-pull with an additional 180mm fan providing airflow over the heatsink, RAM, VRMs. The fans run at <800 RPM when idle to keep things quiet. I had a Noctua D14 but swapped it for the HR-02 because of much easier RAM access. There was a few C penalty in cooling performance but the HR02 is still a beast of a cooler.

My GPU isn't the same anymore, but everything else is: . . .

Can't figure what to say. You have a similar cooling strategy -- generally. It dawn on me, with even a single 180mm fan, you have a very compact case. so . . . more effective pressurization? I thought I'd done good to reduce the number of fans to four! I hadn't looked into newer coolers since I put the D14 in this system. The AP-30 exhaust fan -- with my noise-deadening strategy -- doesn't help much over 3,200 RPM. I've exploited the cooling capability of the D14 to the limit: the average of IBT maximums is 69.5C around 77F room ambient. If I ran it up to 4,200, I'd still get those temperature results.

I WILL post my success -- if I have any -- getting to 4.8. When I first built this rig, I thought the maximum expectation was 4.6.

It also dawns on me that I might want to reevaluate my voltage settings, now that I likely expunged the other hardware and software problems. It might be stable at a few millivolts lower. Not in a hurry to do that, though. I could bet that you get some extra edge with the Maximus IV board, but I remember picking the P8Z68-v Pro because I got approximately the same phase-power design for less.

 

[Yuriman]

If you've been overclocking for 15 years and actually knew what you were doing, you'd know that it's not very hard to see if a Sandy Bridge chip will do 5 GHz. They're about as easy as it gets when it comes to overclocking.

1) Set voltage to 1.5V, highest LLC, all power saving features off. If your motherboard doesn't let you adjust LLC, get a better motherboard.
2) Set multipler to 50 and BCLK to 100
3) If your CPU doesn't boot, then it's most likely not a good enough chip to do 5 GHz stable.

This is pretty solid advice, though you would also want to enable iPLL overvoltage.

I'd probably also suggest you try again with ~1.35-1.375v for a "maximum safe overclock for continuous use". Keep raising the multiplier until it no longer passes stress testing, at which point you can back down a little and begin reenabling power saving features. In my case, I need C3/C6/package C states disabled or I get random BSODs. iPLL overvoltage is not necessary on my chip (YMMV).

Ivy Bridge is not Sandy Bridge, but here are my rough voltage requirements:

4.3GHz - 1.17v @ 56c IBT loaded
4.4GHz - 1.19v @ (didn't record temps)
4.5GHz - 1.24v @ ~75c IBT
4.6GHz - 1.29v @ ~80c IBT
4.7GHz - 1.35v @ 87c IBT
4.8GHz - 1.41v @ 98c IBT

No amount of voltage allows me to even post at 4.9GHz, I've tried in excess of 1.55v. I have an expensive ($100+) waterblock mounted bare die with liquid metal, a powerful pump and 1/2"ID tubing, and a triple radiator cooling my chip.

 

[BonzaiDuck]

This is pretty solid advice, though you would also want to enable iPLL overvoltage.
I'd probably also suggest you try again with ~1.35-1.375v for a "maximum safe overclock for continuous use". Keep raising the multiplier until it no longer passes stress testing, at which point you can back down a little and begin reenabling power saving features. In my case, I need C3/C6/package C states disabled or I get random BSODs. iPLL overvoltage is not necessary on my chip (YMMV).
Ivy Bridge is not Sandy Bridge, but here are my rough voltage requirements:
4.3GHz - 1.17v @ 56c IBT loaded
4.4GHz - 1.19v @ (didn't record temps)
4.5GHz - 1.24v @ ~75c IBT
4.6GHz - 1.29v @ ~80c IBT
4.7GHz - 1.35v @ 87c IBT
4.8GHz - 1.41v @ 98c IBT
No amount of voltage allows me to even post at 4.9GHz, I've tried in excess of 1.55v. I have an expensive ($100+) waterblock mounted bare die with liquid metal, a powerful pump and 1/2"ID tubing, and a triple radiator cooling my chip.

I'd sent 99GT2 a PM asking for more details, but it is clear I overlooked his previous post which you quoted. It is of course one thing to raise the VID to 1.5V, and another as to what voltage you actually get as vCORE.

The OP asked the question, and he has options for choosing these kinds of settings. You (the rhetorical "you") may actually get a safe and stable 24/7 overclock to 5.0 by raising LLC to the "Extreme" limit and VID to 1.5V, but a lot of folks -- whose veteran OC'ing "service" exceeds mine -- had always noted the risks.

Note that we're talking about the SB-K chips. But I found firsthand that if you raise LLC above a certain level, you can actually show a vCORE above VID. One "school of thought" urges that "a little vDROOP is a good thing."

Then there was the issue about the PLL Overvoltage setting, which, on some motherboards, prevented the system from emerging out of "sleep." I and others noted that you could actually increase temperatures and stress on other components by choosing "Extreme" phase-power and VRM settings.

So -- yes -- you may actually get to 5.0 Ghz, and it may be stable for 24/7 use (although we would suggest enabling the EIST power-saving features -- possibly even the C1E and C3/C6.) I seem to have discovered (unless it were mere coincidence, but I think not) that these particular features caused problems themselves with the sleep function if hibernation is enabled in the OS.

With an i7-2700K, we might argue you get an extra edge for the stock turbo spec of 3.9, or a "free" 100 Mhz. It might seem reasonable, but there is no complete assurance that this means you get a free 100 Mhz between -- say -- 4.7 and 4.8. It only seems reasonable.

So comparing 996GT2's voltage settings to mine, it makes more sense that he could reach 5.0Ghz than it does for a 2500K or 2600K to reach 4.8 or 4.9.

The OP has options and choices, so it would be in the "eye of the beholder" as to what constitutes a satisfactory 5.0 setting for 24/7. Will you compromise sleep state? Will you enable LLC so that load vCORE actually exceeds the VID at that speed? Will there be implications for other motherboard components? I can't say. I can only guess.

But I can say that my Cinebench 15 scores for 4.7 match those for a Haswell 4770K @ 4.4, or it's at least "neck-and-neck." It is even pretty close if I run the benchmark at 4.6. What that means, we are also not so sure, since there would be enhancements to the instruction-set and other features on the Haswell. Cinebench is after all, just a benchmark.

 

[john3850]

Most of the new people here that ran 5k ask the same thing how do I get the temps down with 141-145vcore.
In the winter my pc room is at 65f and x48-49 was ok but not in the summer.

 

[Arkaign]

I've installed well over 2,000 SB/IB/Haswells now, with about a third being OC units.

YMMV is a HUGE caveat.

I've hit 5Ghz rock solid with 212+ Evo and near stock volts, and had others that wouldn't pass 4.2Ghz under H100. I even had a couple of defective Ivys and Haswells that were unstable at 100% stock, verified across multiple boards/ram/PSUs.

That said, H60 is a waste. Honestly with water go big or skip it. 212+ Evo, Noctua DH14 are both superior options in terms of ease / effectiveness when used properly. H80 is where you start, and you can go up from there if you really want to. A great chip can ride under air though. And a bad chip can't be helped that much by anything short of angel tears.

 

[996GT2]

Most of the new people here that ran 5k ask the same thing how do I get the temps down with 141-145vcore.
In the winter my pc room is at 65f and x48-49 was ok but not in the summer.

Enabling Speedstep is really useful for keeping the temperatures/noise down. Most of the time the CPU is doing little more than idling, so it doesn't make sense to keep it pegged at an overclocked frequency/voltage all the time.

Currently my 2700K is sitting at 1.6 GHz, 1.016V

 

[BonzaiDuck]

Enabling Speedstep is really useful for keeping the temperatures/noise down. Most of the time the CPU is doing little more than idling, so it doesn't make sense to keep it pegged at an overclocked frequency/voltage all the time.

Currently my 2700K is sitting at 1.6 GHz, 1.016V

Same that my HW monitoring shows here -- to the last millivolt. Like I may have said, I bumped up my offset voltage from a minimum 5 mV to about 20, and compensated with the "Xtra voltage" setting. But this had been in response to various forum trails around the web for people who experienced "idle instability." And -- certainly turns out -- that C1E (at least) will cause the idle voltage to vary around a mean. Later I may back down the offset and bump up the "Extra" and run some stability tests. But the problem there: stability tests are less likely to reveal idle instability.

Arkaign has been around here a long time. If 2,000 SB, IB and Haswells had passed through his hands, he carries with him some statistical distributions even if he didn't record anything on paper.

 

[996GT2]

Same that my HW monitoring shows here -- to the last millivolt. Like I may have said, I bumped up my offset voltage from a minimum 5 mV to about 20, and compensated with the "Xtra voltage" setting. But this had been in response to various forum trails around the web for people who experienced "idle instability." And -- certainly turns out -- that C1E (at least) will cause the idle voltage to vary around a mean. Later I may back down the offset and bump up the "Extra" and run some stability tests. But the problem there: stability tests are less likely to reveal idle instability.

Arkaign has been around here a long time. If 2,000 SB, IB and Haswells had passed through his hands, he carries with him some statistical distributions even if he didn't record anything on paper.

You can test idle stability by locking the CPU to 1.6 GHz (either in BIOS or with software like Throttlestop), then running a benchmark or stability test for a little bit. That's what I did.

 

[BonzaiDuck]

You can test idle stability by locking the CPU to 1.6 GHz (either in BIOS or with software like Throttlestop), then running a benchmark or stability test for a little bit. That's what I did.

I'm going to think about it some more, but I think you have something there -- I'm not sure . . . I'm not entirely sure that loading the CPU at that speed would be the only condition in which idle instability occurs.

 

[BonzaiDuck]

I did run Prime95 as well as LinX.

I'm using a Thermalright HR-02. 2x Gentle Typhoon AP-15 push-pull with an additional 180mm fan providing airflow over the heatsink, RAM, VRMs. The fans run at <800 RPM when idle to keep things quiet. I had a Noctua D14 but swapped it for the HR-02 because of much easier RAM access. There was a few C penalty in cooling performance but the HR02 is still a beast of a cooler.

My GPU isn't the same anymore, but everything else is:

Lemme get this straight . . .

http://www.newegg.com/Product/Produc...82E16811163222

No vents on the bottom panel, but because of the case's size, no need. And because of the case's size, no need for caster wheels. the case comes with a top-panel 180 or 200mm intake fan, but the fan can be mounted instead in the front panel? So you could block off the top-panel with black foam board, or -- you could install another fan -- intake or exhaust according to your wise or untutored strategy, however it may be.

You have enough room for all your SSDs or HDDs, you have a thin or slim optical drive in the top, and my guess -- if the case supported a 1000W Silverstone, it likely supports some other beefy PSUs of alternate manufacture.

So your worst limitation is in your mATX options -- which these days, isn't so limiting. Is that correct?

That's a decent price for a good case, and it looks like a very promising case, too . . .

Maybe you could even block off the top vent, craft some reinforcing removable brackets from aluminum or steel stock for insertion under the case top-panel, to which you secure an aluminum handle probably obtained at MNPC. And away you go to the LAN party, the overnight trip on the boat in Puget Sound, or some other excursion. Thus, an easily transportable gaming rig?

 

[pantsaregood]

You're probably not going to hit 5.0 GHz on a 2500K/2600K/2700K at a reasonable voltage. It isn't unheard of for Sandy Bridge to clock that high (even on air), but it is rare. 4.4 GHz is honestly something you should be quite happy with - that's a 33% increase in frequency, which will yield around 33% more CPU performance.

 

[996GT2]

Lemme get this straight . . .

http://www.newegg.com/Product/Produc...82E16811163222

No vents on the bottom panel, but because of the case's size, no need. And because of the case's size, no need for caster wheels. the case comes with a top-panel 180 or 200mm intake fan, but the fan can be mounted instead in the front panel? So you could block off the top-panel with black foam board, or -- you could install another fan -- intake or exhaust according to your wise or untutored strategy, however it may be.

You have enough room for all your SSDs or HDDs, you have a thin or slim optical drive in the top, and my guess -- if the case supported a 1000W Silverstone, it likely supports some other beefy PSUs of alternate manufacture.

So your worst limitation is in your mATX options -- which these days, isn't so limiting. Is that correct?

That's a decent price for a good case, and it looks like a very promising case, too . . .

Maybe you could even block off the top vent, craft some reinforcing removable brackets from aluminum or steel stock for insertion under the case top-panel, to which you secure an aluminum handle probably obtained at MNPC. And away you go to the LAN party, the overnight trip on the boat in Puget Sound, or some other excursion. Thus, an easily transportable gaming rig?

The intake is the top fan. There is no front fan. The PSU is mounted in such a way that it intakes from the front and exhausts out of the side, thereby making its airflow independent of the other components.

CPU intake is handled by the 180mm fan, exhaust is through the rear of the case.

GPU intake is via the lower side panel vents. There is the option to install 2x80 or 92mm fans and 1x120mm fan in the lower side panel for high-end SLI/XFire setups. GPUs exhaust through the rear of the case, so their airflow is relatively independent as well.

Overall its cooling ability is better than most ATX cases out there. There is nothing in the way of intake airflow, unlike in a standard ATX case, which places the hard drive cages behind the intake fans.

I didn't actually build the system with the intention of transporting it a lot, but it's certainly easier with a rig of this size compared to a full-size ATX rig. I built it because I have a preference for small systems, and because it's easier to swap parts around (you don't need much room, just put the system on a desk and have at it). It's also a bit more of a challenge to build a small rig, which I like.

 

[BonzaiDuck]

. . . . I built it because I have a preference for small systems, and because it's easier to swap parts around (you don't need much room, just put the system on a desk and have at it). It's also a bit more of a challenge to build a small rig, which I like.

Also a good point about less work-space contention with a "smaller piece of work."

As to the challenge, I'd been thinking about that for some time. In other threads ("Cases & Cooling") I'd mentioned a gurr-eat project I made in 2007 out of a '95 Compaq Proliant server case, which utilized the drive rails and plastic latches for the old SCSI's -- twice the height of today's standard desktop HDD. WoodButcher (who may also do carpentry or cabinetmaking in addition to his automobile and computer obsession) had mentioned it in recollection of forum posts for that year. I said I was a bit loath to post about it, having gotten great mileage out of the accomplishment here and there on these forums since then. But -- hey -- here it is:

Shame, really. After building my SB-K with intention of giving the old Q6600 to my bro in a new HAF 922 case, I was so weary of the prospects for disassembly and rebuild that I just gave him the whole enchilada instead.

And of course the disarray of my office doesn't lend itself to any bling-bling promotion. Might have taken some snaps of the fluorescent lighting inside the case, which is switched inside the front panel door.

Smaller is better.

If the moderator says "Hey! This is 'CPUs and Overclocking' -- post that over in 'Cases and Cooling," I'll bring up ADA [Jill Hennessey or Diane Neal] arguments on "Law and Order" that the defendant's lawyer "opened the door" to admit that evidence . . . by posting a picture of the Silverstone Sugo in this forum. :biggrin:

 

[Eric1987]

This is pretty solid advice, though you would also want to enable iPLL overvoltage.

I'd probably also suggest you try again with ~1.35-1.375v for a "maximum safe overclock for continuous use". Keep raising the multiplier until it no longer passes stress testing, at which point you can back down a little and begin reenabling power saving features. In my case, I need C3/C6/package C states disabled or I get random BSODs. iPLL overvoltage is not necessary on my chip (YMMV).

Ivy Bridge is not Sandy Bridge, but here are my rough voltage requirements:

4.3GHz - 1.17v @ 56c IBT loaded
4.4GHz - 1.19v @ (didn't record temps)
4.5GHz - 1.24v @ ~75c IBT
4.6GHz - 1.29v @ ~80c IBT
4.7GHz - 1.35v @ 87c IBT
4.8GHz - 1.41v @ 98c IBT

No amount of voltage allows me to even post at 4.9GHz, I've tried in excess of 1.55v. I have an expensive ($100+) waterblock mounted bare die with liquid metal, a powerful pump and 1/2"ID tubing, and a triple radiator cooling my chip.

What's a good program to make sure I'm setting the volts correctly? My motherboard is weird with the options.

 

[ImpulsE69]

When I first got my 2500k, I could take it to 4.6 easy with little tweaking which was good enough for me, then tweaked it a bit more to get the voltage/temps a bit lower and it ran for months w/o issue, then one day it simply refused to go over 4.2 and will not go to 4.2 from the MB settings at all, I have to do it all in software. I got tired of trying to figure out why and just leave it like that now. Haven't had any issues since (for over a year).

 

[Essence_of_War]

What's a good program to make sure I'm setting the volts correctly? My motherboard is weird with the options.

What is weird about it? Can you find the 'voltage offset', 'turbo voltage', and 'all core multiplier' options?

Or are you looking for a way to measure v_core while you're stress testing?

 

[Ketchup]

Sounds like you got a very nice chip 996GT2. Here is my story, for those interested:

Shortly after getting the system in my sig put together, I decided to go ahead and see what she would do:

1. She made it up to about 4.5 on AUTO voltage. It wouldn't go over 1.31 volts, and Core #2 (always the hottest on mine) would max out at 63 C during stress testing.
2. After that point, it was a battle (seemingly) and had to really ramp up the volts on each step up. I stopped at 4.9, where max load voltage hit 1.48 and Core #2 hit 80 C during stress testing. Putting it up to 5 was an almost instant blue screen when loading Windows, and I believe I did try a tiny voltage increase before calling it done.

4.9 is still a pretty amazing overclock over stock, and I am very happy at 4 flat for now.

 

[Headfoot]

I had a "golden" 2500K which I was very reluctant to sell because it did 5 GHz easily on air, but I eventually did sell it to upgrade to i7. Got lucky with my chip, although 2700Ks in general tend to be well-binned.

Lucky!! I'm super jealous

 

[lehtv]

The H60 is outperformed by several air coolers, and even the best air coolers will struggle with temperatures on a chip that is capable of 5GHz. It's an overclock that only a custom water setup will handle properly, and it still needs a relatively rare chip to achieve.