SB Overclocking results?

dnottis said:
Eventually he will crash and probably come here posting about it blaming something other than his "stable for me overclock". If you cant pass 8- 10 hours of prime and linx then you just aren't 100% stable regardless of what you do with it.
Click to expand...

I disagree. If you can't pass 24 hours of Prime95, with Intel Burn Test and Furmark both running at the same time, then you aren't stable. Can't let any spare cycles get away, or let your video card and PSU get a free ride. And it helps if it is a full moon, during high tide, so the gravitational effects are maximized.

See, anyone can make up arbitrary "stable requirements". Trying to define stablity by using some test, or combination of tests is just a waste of time. Unless you want to spend Intel-level dollars on qualification testing, you'll never know if your system is really stable at any given speed. Which is why chips have "rated" speeds. Overclocking, by definition, is unstable.
 
Last edited:
I'm running my 2500k at 4.3 with 1.3v. Was running 4.4 at same voltage but system would occasionally fail to post or crash while idle at the desktop. Haven't tried playing with anything beyond core voltage and load calibration yet.

I run some stress tests but don't go crazy. If I encounter a crash I'll usually either tweak voltages or drop the clock a multiplier. Earlier today I chose to drop the multiplier and continue working.
 
some more results... (system spec in sig)

4.5ghz with the Auto-OC app (105x43), sets vcore at 1.2v
4.6ghz (100x46) at 1.22v, mid 50c
4.8ghz (100x48) at 1.35v, 60-70c
5.0ghz (100x50) at 1.42v, 80c+

each good for a 10 min on P95 blended. i'm not concerned about longer testing.

Note, actual voltage reported in CPU-Z and Asus' own monitoring app is about +0.1v higher than my settings listed above, under load. Think i'm going to use 4.6ghz for day to day use.
 
Well, looks like I got prime stable at decent settings. Will try 4.5ghz overnight.

LLC is off, cpu vcore is set at +0.060. Prime running for more than 6h.

44ghz-1.jpg
 
My 2500k capped out at 4.4ghz @ 1.368v.

Can't do anything to get it to boot into windows at 4.5ghz.
 
4.6 @ 1.275 stable
Had it 4.8 but temps hit 80+ under 100% load. Need to push/pull the H50 w/ intake vs exhaust spec in sig
 
Forceman said:
I disagree. If you can't pass 24 hours of Prime95, with Intel Burn Test and Furmark both running at the same time, then you aren't stable. Can't let any spare cycles get away, or let your video card and PSU get a free ride. And it helps if it is a full moon, during high tide, so the gravitational effects are maximized.

See, anyone can make up arbitrary "stable requirements". Trying to define stablity by using some test, or combination of tests is just a waste of time. Unless you want to spend Intel-level dollars on qualification testing, you'll never know if your system is really stable at any given speed. Which is why chips have "rated" speeds. Overclocking, by definition, is unstable.
Click to expand...

So according to your theory no one should over clock. Definitely not [H] man.
 
dnottis said:
So according to your theory no one should over clock. Definitely not [H] man.
Click to expand...

No, overclock all you want, just don't kid yourself that 8 or 10 or 24 hours of Prime95 proves you have a stable system. Stable enough, maybe, but not stable. And if all you want is stable enough, than maybe 10 passes of Linpack is also good enough.
 
Another thing guys, that I just discovered .... your memory timings are absolutely CRUCIAL to a successful stable overclock.

My memory is rated at 2000mhz @ 9 9 9 24 T2

I've been getting this random / instant BSOD's when stress testing with Prime 95 and Super PI.

I loosened my timings out of spec to 9 10 10 26 T2 and now, as far as I can tell that fixed that problem.

So if you have some of the issues I mentioned you might want to look at your memory settings.

So right now I am working on vcore and heat. I don't think my heat is an issue tho it gets into the mid 70's with a touch of 80 - 81. A bit warm for me. I am ok with anything in the 70's under load.
 
That's from the Intel white paper on the sandy bridge.
Click to expand...

Where did you happen to get that white paper from? I've been itching to read it but can't find it anywhere.
 
How have you guys been dealing with the multiplier throttling issues when testing? I dont have speedstep enabled, and I increased the watt limit for that specific line by a significant margin. Also, temps are not an issue.

edit: I also have all of the throttling options disabled.
 
Last edited:
Sorry if this has been posted before (I can't go through 15 pages to check), but this i how I got my 2600K stable at 4.8Ghz

Flashed beta BIOS
Internal CPU PLL Overvoltage = Enabled
CPU Vcore = 1.38
LLC = Ultra high
RAM = 9-9-9-24 @ 1600Mhz 1.60V (Corsair XMS3 rated at 9-9-9-24 @ 1600Mhz 1.65V)
VRM Frequency = 350

IIRC, these are all the changes I made.

Its rock solid stable at this config. Its my dedicated folding rig and has been folding at 100% load 24/7 for almost 2 weeks now without any incidents. Temps do not cross 62 degrees for most of the day except between 2-5AM, when they increase by 3-5 degrees.
I'm running it watercooled though.

I could OC higher, but I don't wanna push over 1.38V just yet (I know Intel Spec Sheets say 1.52 is max, but it doesn't seem right to me for a 32nm chip - might be just me being apranoid, but I'm happy at 4.8).
 
dnottis said:
So according to your theory no one should over clock. Definitely not [H] man.
Click to expand...

Wow....people are COMPLETELY missing the point.

I can run Deep Rybka 4 chess engine all day, temps don't exceed 75C, 8 threads.NONE of my games get close to what Rybka can do. If Rybka passes, nothing else crashes.
I think CPU test in vantage reached 68C. Black Ops didn't reach that high.

But if I run prime, temps reach 90C.
So you're telling me I should run Prime to see if I'm TRULY stable? BULLOCKS. Vcore is too high for prime (1.4 bios, 1.456 under load, despite my Ven-X and Panaflo fan). UNLESS you want to pay for my next chip---care to toss me $350 for a new 2600K, then sure I'll run prime all day at 90C, since it's not my money being wasted....

The instant someone releases some new stress tester that heats temps 10C higher than Prime and LinX with AVX(?) extensions (SP1), everyone will start flocking to that....

Stability has to do with temps, current (voltage) and flip flops (mhz). Not running prime to avoid having your temps reach 90C to remain stable is the same thing basically as buying a bigger /more extreme cooler to do the same thing. But it sounds like people are equating not running Prime to be as sinful as switchiing from a Ven-X to a Hyper212 (which will raise your temps, and thus LOWER your chances for stability also). Ok,so I'm unstable because I can't keep my temps below 90C with prime, when nothing else has exceeded 75C...whatever....

If you can keep the temps low, below a target point for your mhz and vcore, then you're stable. And if nothing you use willi cause the chip to heat up past that point, you're stable. If you know you can run your rig all day and night without exceeding 80C, and nothing will crash if you don't exceed 80C-but Prime goes to 90C, guess what, youre STABLE. I know nothing about folding. I don't fold. I game. So maybe this applies to folders or something....

Anything else is for people with OCD or something (Notice: *I* have OCD, so I don't need to get pulled into doing REPEATED testing unless something I use normally starts crashing...)

We'll see what happens with the new extensions....

--------

On previous chips, people used Prime as a stress tester, since even though no game or other application ran as hot as prime, if you could pass prime, you were going to be stable in virtually anything, and Prime's maximum temps were not that much hotter than what a multicore enabled game could pump. I think on my Core 2 Quad, prime was about 6C hotter than Rybka.

But when Prime is 15C (!!) hotter than a multicore chess engine, THEN you really need to sit back and ask yourself if it's really necessary...
 
Last edited:
xbonez said:
Sorry if this has been posted before (I can't go through 15 pages to check), but this i how I got my 2600K stable at 4.8Ghz

Flashed beta BIOS
Internal CPU PLL Overvoltage = Enabled
CPU Vcore = 1.38
LLC = Ultra high
RAM = 9-9-9-24 @ 1600Mhz 1.60V (Corsair XMS3 rated at 9-9-9-24 @ 1600Mhz 1.65V)
VRM Frequency = 350

IIRC, these are all the changes I made.

Its rock solid stable at this config. Its my dedicated folding rig and has been folding at 100% load 24/7 for almost 2 weeks now without any incidents. Temps do not cross 62 degrees for most of the day except between 2-5AM, when they increase by 3-5 degrees.
I'm running it watercooled though.

I could OC higher, but I don't wanna push over 1.38V just yet (I know Intel Spec Sheets say 1.52 is max, but it doesn't seem right to me for a 32nm chip - might be just me being apranoid, but I'm happy at 4.8).
Click to expand...

That 1.52 thing you're reading is the same as what USED to be called "Absolute maximum" back during the P4 and Core 2 days, when Absolute maximum and "Functional Limits (this may be the same as Max VID)" were clearly shown on the intel spec sheets. Absolute max for core i7 45nm versions was 1.55v. I don't remember what the max VID was, but Intel didn't list "functional" maximums anymore.
 
So far, running 4.8 ghz @ 1.325v. Looks to be Prime95 stable after a couple hours. Heats up to the low 60's-70' depending on core. Tried pushing to 5.0 ghz just to have a nice, round number, but it meant pushing the vcore up to 1.34. Not a comfy level considering I also have LLC on Ultra high.

Have speedstep and the various energy saving features on, they don't seem to hurt.

I have the build in my sig.
 
Last edited:
Jayemji said:
So far, running 4.8 ghz @ 3.25v. Looks to be Prime95 stable after a couple hours. Heats up to the low 60's-70' depending on core. Tried pushing to 5.0 ghz just to have a nice, round number, but it meant pushing the vcore up to 3.4. Not a comfy level considering I also have LLC on Ultra high.

Have speedstep and the various energy saving features on, they don't seem to hurt.

I have the build in my sig.
Click to expand...

I hope thats 1.325 not 3 volts lol.
 
Jayemji said:
So far, running 4.8 ghz @ 3.25v. Looks to be Prime95 stable after a couple hours. Heats up to the low 60's-70' depending on core. Tried pushing to 5.0 ghz just to have a nice, round number, but it meant pushing the vcore up to 3.4. Not a comfy level considering I also have LLC on Ultra high.

Have speedstep and the various energy saving features on, they don't seem to hurt.

I have the build in my sig.
Click to expand...

LOL....
Are you drunk, or just sleepy?
3.25v? 3.4v????
 
Falkentyne said:
Wow....people are COMPLETELY missing the point.

I can run Deep Rybka 4 chess engine all day, temps don't exceed 75C, 8 threads.NONE of my games get close to what Rybka can do. If Rybka passes, nothing else crashes.
I think CPU test in vantage reached 68C. Black Ops didn't reach that high.

But if I run prime, temps reach 90C.
So you're telling me I should run Prime to see if I'm TRULY stable? BULLOCKS. Vcore is too high for prime (1.4 bios, 1.456 under load, despite my Ven-X and Panaflo fan). UNLESS you want to pay for my next chip---care to toss me $350 for a new 2600K, then sure I'll run prime all day at 90C, since it's not my money being wasted....

The instant someone releases some new stress tester that heats temps 10C higher than Prime and LinX with AVX(?) extensions (SP1), everyone will start flocking to that....

Stability has to do with temps, current (voltage) and flip flops (mhz). Not running prime to avoid having your temps reach 90C to remain stable is the same thing basically as buying a bigger /more extreme cooler to do the same thing. But it sounds like people are equating not running Prime to be as sinful as switchiing from a Ven-X to a Hyper212 (which will raise your temps, and thus LOWER your chances for stability also). Ok,so I'm unstable because I can't keep my temps below 90C with prime, when nothing else has exceeded 75C...whatever....

If you can keep the temps low, below a target point for your mhz and vcore, then you're stable. And if nothing you use willi cause the chip to heat up past that point, you're stable. If you know you can run your rig all day and night without exceeding 80C, and nothing will crash if you don't exceed 80C-but Prime goes to 90C, guess what, youre STABLE. I know nothing about folding. I don't fold. I game. So maybe this applies to folders or something....

Anything else is for people with OCD or something (Notice: *I* have OCD, so I don't need to get pulled into doing REPEATED testing unless something I use normally starts crashing...)

We'll see what happens with the new extensions....

--------

On previous chips, people used Prime as a stress tester, since even though no game or other application ran as hot as prime, if you could pass prime, you were going to be stable in virtually anything, and Prime's maximum temps were not that much hotter than what a multicore enabled game could pump. I think on my Core 2 Quad, prime was about 6C hotter than Rybka.

But when Prime is 15C (!!) hotter than a multicore chess engine, THEN you really need to sit back and ask yourself if it's really necessary...
Click to expand...

Folding is a real world application, and it will heat your chip as much as Prime. I've had cases where my PC has been Prime stable for hours upon hours, but failed a Folding Work Unit within the hour.
 
Falkentyne said:
But when Prime is 15C (!!) hotter than a multicore chess engine, THEN you really need to sit back and ask yourself if it's really necessary...
Click to expand...

You are putting too much emphasis on temperature. Unstable over clocks will often crash before they reach a high temperature. Prime is great in these situations.
 
xqz2 said:
You are putting too much emphasis on temperature. Unstable over clocks will often crash before they reach a high temperature. Prime is great in these situations.
Click to expand...

You are right in a way, but you can also be fully stable, and then start crashing if you get too hot. Otherwise, no one would buy any high end air cooling if temps weren't important, now would they? You basically just implied "air cooling quality isn't that important for stability." But the whole reason people try to keep temps down is precisely BECAUSE temps getting too hot causes instability if you're on the edge.

90C is just too hot for prime. MANY People have problems if prime goes over 80C when doing high overclocks. I'm happy with my Rybka test.
 
xbonez said:
Folding is a real world application, and it will heat your chip as much as Prime. I've had cases where my PC has been Prime stable for hours upon hours, but failed a Folding Work Unit within the hour.
Click to expand...

I mentioned folding as being real world (and also extremely important), but I don't fold. So it doesn't apply to me XD
 
Jayemji said:
So far, running 4.8 ghz @ 1.325v. Looks to be Prime95 stable after a couple hours. Heats up to the low 60's-70' depending on core. Tried pushing to 5.0 ghz just to have a nice, round number, but it meant pushing the vcore up to 1.34. Not a comfy level considering I also have LLC on Ultra high.

Have speedstep and the various energy saving features on, they don't seem to hurt.

I have the build in my sig.
Click to expand...

Can you post your settings, please? What is your batch?
 
Falkentyne said:
You are right in a way, but you can also be fully stable, and then start crashing if you get too hot. Otherwise, no one would buy any high end air cooling if temps weren't important, now would they? You basically just implied "air cooling quality isn't that important for stability." But the whole reason people try to keep temps down is precisely BECAUSE temps getting too hot causes instability if you're on the edge.

90C is just too hot for prime. MANY People have problems if prime goes over 80C when doing high overclocks. I'm happy with my Rybka test.
Click to expand...

Your incorrectly associating temps with stability. While they are related, voltage has much more to do with it. Higher end air coolers enable users to pump more voltage to maintain stability, not to keep temps down to maintain stability.
 
So when you guys are reporting 4+ghz on the sandy bridge, thats when they are using turbo boost correct? Well doesn't that mean they park a certain amount of cores? So at that 4+ghz you guys are reaching you are actually only running 1 or 2 cores instead of all 4 because of the turbo boost feature?
 
Last couple of nights ive started prime 95 blend test, checked it in the morning after 12 or so hours and prime always seemed to have crashed with a "program is not responding" message.

Ive 1.525 going to the ram (ram at stock speeds) and slightly over 1.3 going to the cpu. I done a 13 hour small ftt run a while back and it was fine. More memory volts needed maybe?

2600k @ 4.5GHZ\Asus P8P67 Deluxe\4 Gigs geil 1600 ddr3
 
I have a Asus P8P67 Deluxe paired with a i7 2600k.

Some things I've discovered through hours and hours of tweeking.

vcore. The higher your overclock, the more juice you need. If you start crashing, gently raise this up. Yes, your temps will go up. I can prime 95 for hours with LLC set to extreme with my vcore at 1.415 @ 4.8GHz, anything else and it crashes out or gets too hot for comfort.

Memory timings. Just cause your memory is rated x x x 24 whatever whatever by the manufacturer, don't think you don't have to adjust those values. I only found stability when I loosened my timings a hair. From the rated 9 9 9 24 to 9 10 10 26 T2. This caused my Prime95 and Super PI benches to become stable. I'm currently @ 1952 memory.

I know it's more than obvious to all of us that voltage needs to be raised the higher your overclock is. The reason I bring it up is it is probably the main reason you do not have stability. So look here FIRST.

Also, in my experience, LLC needs to be set at extreme. CPU vcore under load is crucial to stability. Very high causes my system to randomly crash as were extreme holds vcore to a more precise level preventing crashes. Most notably in Prime 95.

I know many of you out there have LLC set to very high and are over volting to give you stability. Changing to extreme will allow you to zero in on your actual needed vcore, thus lowering your temps a few degrees.
 
I had the LLC on extreme, but caused me some issues though, so I disabled it.

I found that the offset > auto mode works the best so far. The vcore idles around 1.0, and 1.34 under load. I have it OCed to 4.5Ghz though.
 
AliceCooper said:
So when you guys are reporting 4+ghz on the sandy bridge, thats when they are using turbo boost correct? Well doesn't that mean they park a certain amount of cores? So at that 4+ghz you guys are reaching you are actually only running 1 or 2 cores instead of all 4 because of the turbo boost feature?
Click to expand...

You are correct if you don't touch the bios - only 1 core will reach 3.8.

With the K series, you can freely unlock the turbo an all cores. So, my 2500K does 4.7 on all four cores. If I wanted to, I could set it to do 4.7/4.0/3.8/1.6 or any arbitrary set of multipliers.
 
AliceCooper said:
So when you guys are reporting 4+ghz on the sandy bridge, thats when they are using turbo boost correct? Well doesn't that mean they park a certain amount of cores? So at that 4+ghz you guys are reaching you are actually only running 1 or 2 cores instead of all 4 because of the turbo boost feature?
Click to expand...

No, these motherboards let you set a multiplier that applies when all cores are active. So mine is running at 46x even with 4 cores active. You can also set (on most/all boards?) the multipliers based on the number of cores active (like the standard turbo boost) but I don't think anybody actually uses that.
 
Forceman said:
No, these motherboards let you set a multiplier that applies when all cores are active. So mine is running at 46x even with 4 cores active. You can also set (on most/all boards?) the multipliers based on the number of cores active (like the standard turbo boost) but I don't think anybody actually uses that.
Click to expand...

Ah ok thanks, I was just thinking like wow ok cool 4.6ghz, but only 1 core, but nevermind thanks for the info :cool:. Hows the stock cooling for OCing on these processors?
 
AliceCooper said:
Ah ok thanks, I was just thinking like wow ok cool 4.6ghz, but only 1 core, but nevermind thanks for the info :cool:. Hows the stock cooling for OCing on these processors?
Click to expand...

Just like every other processor: don't bother.
 
AliceCooper said:
Ah ok thanks, I was just thinking like wow ok cool 4.6ghz, but only 1 core, but nevermind thanks for the info :cool:. Hows the stock cooling for OCing on these processors?
Click to expand...

It'll get you till about 4ish Ghz, after which you'll need something better
 
You must log in or register to reply here.
Back
Top