Good multiplier setting for i5-2500k?

[bbhaag]

I was wondering what a good multiplier setting was for an i5-2500k. I want to achieve a stable OC without messing with voltages, ram timings, or bridge settings ect.

For example on my computer running a 965be I set the multi to 18.5 in the bios and achieved a stable 3.7ghz oc. A sweet easy oc.

I want to do the same for my sons 2500k so what is the standard multiplier setting to achieve similar results?

 

[Burpo]

37.. bclk is 100 = 3700

 

[Charlie98]

Only increase your multiplier, don't fool with the backclock.

You should be able to get 4.0 pretty easy, but heat may or may not become an issue (it was for me on my first 2500K.) If you want to reduce heat a bit, you can lower the vCore a little... you didn't say if you have an aftermarket CPU cooler or just the Intel cooler.

 

[Yuriman]

How far you can overclock without touching voltage will vary widely from chip to chip. I bet 42x (4200mhz) is going to be safe/stable for 95% of 2500K's though.

 

[bbhaag]

Hmm I'm not sure I understand. Here is a pic, as you can see the multi is set to 42 so lowering it to 37 like Burpo suggested doesn't seem to make sense.
Here is the mb his pc uses. http://us.msi.com/product/mb/P67AGD55_B3.html#hero-specification

 

[bbhaag]

How far you can overclock without touching voltage will vary widely from chip to chip. I bet 42x (4200mhz) is going to be safe/stable for 95% of 2500K's though.

Thanks for the advice. I'm very new to Intel cpu's. I only bought my son this because I've read that Intel is the best for gaming and he games a lot.

Is his multi already set to 42? That is what it looks like to me but like I said I'm new to Intel boards and oc'ing.

 

[Yuriman]

Yes, it looks like it's already overclocked. Base clock is 100mhz, and CPU multiplier is set at 42x, resulting in 4.2ghz, up from the 3.3ghz / 3.7 turbo stock clocks.

 

[bbhaag]

I guess without me knowing he already oc'ed his cpu. That is very strange though because he didn't tell me.
Anyway, thanks for the advice guys. Much appreciated!

 

[Majcric]

I'm using 40x multiplier on mine. I had my overclock to 4.5 but wasn't seeing much scaling past 4ghz aside from benchmarking.

 

[BonzaiDuck]

If you trust your son to "fix his own stuff," You might just casually ask. If you're "deeply concerned," you might find out how he may have tested it for stability. Otherwise, he "fixes his own stuff." You should know.

42 seems to be close to the limit with "Auto" voltage settings. That's why I mention testing. But it's only more a rule and not an absolute one that 42 to 43 reaches a point where the voltage(s) need tuning.

If it's a P67 motherboard, it's more likely the only voltage settings for Offset mode are the +/- sign and Offset voltage. "+" is better even if minimum near 0. If by chance there's another adjustment for "extra voltage for turbo," it gives you more flexibility while keeping the idle power-saving voltage stable in some situations. There would definitely be an LLC load-line calibration setting, and on a scale of four or five levels, the 2nd or 3rd higher levels are recommended if use of LLC is preferred as needed. That is, LLC should never totally eliminate vdroop.

At those levels -- and probably up to 44 or 45 -- you don't need to use any LLC. So you'd most likely bump up the Offset voltage a notch or two if any LOAD instability arises, beginning at 43 -- maybe 42. If the second "Extra" voltage setting is included in BIOS, then a notch or two on that one is also a possible option.

Otherwise -- he fixes his own stuff?

 

[bbhaag]

Sorry it took so long to reply. My son was at boy scout camp this week and just got home today so I finally had a chance to ask him about this.

He had no idea that his multi was set to 42 and that his cpu was oc'ed. His cousin was in town awhile ago and is more computer savvy than him so maybe he did it. Either way the oc seems to be stable and my son says that he hasn't experienced any stability issues on his pc.

Thanks again for the help guys.

 

[BonzaiDuck]

Sorry it took so long to reply. My son was at boy scout camp this week and just got home today so I finally had a chance to ask him about this.

He had no idea that his multi was set to 42 and that his cpu was oc'ed. His cousin was in town awhile ago and is more computer savvy than him so maybe he did it. Either way the oc seems to be stable and my son says that he hasn't experienced any stability issues on his pc.

Thanks again for the help guys.

Somehow, the Over-Clocker in me wants to tell you to test it to assure stability. Even if it hasn't crashed so far and your kid does some gaming, there's no satisfying yourself unless you talk to the cousin or run, say, OCCT:CPU for 4 hours. If it proves out, fine. If it doesn't . . . well . . . I suppose you can raise the Offset a couple notches . . . Or you could wait and see . . .

 

[stockwiz]

I didn't mess with the voltage on my 2600K... I messed with the load line calibration and voltage offset so that I got the combination that I wanted.. I believe I have my offset at -.1 volt and load line calibration on high which results in a load voltage of 1.32 volts which is what my chip needs to do 4.5 GHZ... with the load line calibration on medium there was enough vdroop to cause a lockup when priming. If you crash while idle your voltage drops too low while idle. I have not had this issue.

With these very minor adjustments you should be able to hit 4.4 to 4.5... I picked 1.32 to 1.33 volts as my max because I've already had this chip since early 2011 and I plan to have it until 6 core 10nm parts are out.... so 2017-2018. Actually for quite some time I ran it at 4.3 with the same load voltage because I was too lazy to mess with any other settings.

 

[BonzaiDuck]

I didn't mess with the voltage on my 2600K... I messed with the load line calibration and voltage offset so that I got the combination that I wanted.. I believe I have my offset at -.1 volt and load line calibration on high which results in a load voltage of 1.32 volts which is what my chip needs to do 4.5 GHZ... with the load line calibration on medium there was enough vdroop to cause a lockup when priming. If you crash while idle your voltage drops too low while idle. I have not had this issue.

With these very minor adjustments you should be able to hit 4.4 to 4.5... I picked 1.32 to 1.33 volts as my max because I've already had this chip since early 2011 and I plan to have it until 6 core 10nm parts are out.... so 2017-2018.

Heh-heh. . . . tweaking LLC and Offset IS messing with voltage. If "High" LLC is third-highest setting, I wouldn't increase it anymore for any reason. I'm willing to bet that such a level still leaves between 10 and 20mV of droop. Also, I'm holding back on comments I might otherwise make because there are differences in board generations for the same CPU-- P67 versus Z68, for instance.

Gotta say it! None of the Sandy users seem to be giving them up too soon! they work for me!

 

[steve wilson]

How far you can overclock without touching voltage will vary widely from chip to chip. I bet 42x (4200mhz) is going to be safe/stable for 95% of 2500K's though.

I would agree with this if you have a decent Heatsink. If you are using the standard intel one I don't know if I'd go over 4ghz.

 

[BonzaiDuck]

I would agree with this if you have a decent Heatsink. If you are using the standard intel one I don't know if I'd go over 4ghz.

Actually, I think you're spot-on with that.

Usually, I don't even bother to install the stock cooler. Last time around with a 2700K, I tried it initially, and attempted to find that threshold around 4.2 Ghz which requires more voltage.

I never got there with the stock cooler. I was testing with OCCT:CPU, and its default settings would stop the test at 85C.

for what I'd call a limp overclock for any Sandy K, it just shows how the stock cooler will limit or defeat anything but a very mild overclock. 4.0, 4.1 -- maybe.

I've watch the Ping-Pong go back and forth speculating about the performance distribution among Intel Sandy Bridgers released for sale.

Personally, while I DO believe there are those "one-in-a-hundred" chips, or that some larger fraction may allow a 4.9 Ghz overclock on air, I also believe that the expectations are fairly predictable from one chip to the next for lower objectives like 4.2, 4.4, 4.5. So I'm pretty confident that 4.1 is more likely without taking voltage off "Auto;" 4.2 is possible but not certain.

 

[PhIlLy ChEeSe]

Make sure you have the latest Bios, the old ones would over volt the PLL. Also if auto voltage, keep an eye on it in windows as it is possible to over auto volt.

 

[steve wilson]

Actually, I think you're spot-on with that.

Usually, I don't even bother to install the stock cooler. Last time around with a 2700K, I tried it initially, and attempted to find that threshold around 4.2 Ghz which requires more voltage.

I never got there with the stock cooler. I was testing with OCCT:CPU, and its default settings would stop the test at 85C.

for what I'd call a limp overclock for any Sandy K, it just shows how the stock cooler will limit or defeat anything but a very mild overclock. 4.0, 4.1 -- maybe.

I've watch the Ping-Pong go back and forth speculating about the performance distribution among Intel Sandy Bridgers released for sale.

Personally, while I DO believe there are those "one-in-a-hundred" chips, or that some larger fraction may allow a 4.9 Ghz overclock on air, I also believe that the expectations are fairly predictable from one chip to the next for lower objectives like 4.2, 4.4, 4.5. So I'm pretty confident that 4.1 is more likely without taking voltage off "Auto;" 4.2 is possible but not certain.

Well everyone has different thresholds and different chips perform differently. I'm a bit of a softy and won't push my chip because I don't want to pay out for a new one if I brick it. I'm at 4.3ghz with a Noctua cooler and temps don't go above 60 Celsius on my day to day gaming (CS GO mostly and a bit of SC2 desert strike) and are usually around low 50's. The reason I don't OC anymore is because I just don't need anything faster atm, so there is no point risking it.

I look forward to Skylake and hope it really is the chip that a lot of us Sandybridge owners have been waiting for.

 

[BonzaiDuck]

Well everyone has different thresholds and different chips perform differently. I'm a bit of a softy and won't push my chip because I don't want to pay out for a new one if I brick it. I'm at 4.3ghz with a Noctua cooler and temps don't go above 60 Celsius on my day to day gaming (CS GO mostly and a bit of SC2 desert strike) and are usually around low 50's. The reason I don't OC anymore is because I just don't need anything faster atm, so there is no point risking it.

I look forward to Skylake and hope it really is the chip that a lot of us Sandybridge owners have been waiting for.

I've found that the SB-K chips approach "voltage-limits" for certain heatpipe cooler solutions just about where a certain still-published temperature spec (dismissed as irrelevant) is "close." In fact, the "ultimate" throttling temperature-limit gets closer with lesser cooling solutions if you do the same thing. More specifically, if the prior-gen 32nm CPUs were spec'd with a safe-range limit around 1.38V, then the consensus among Sandy users also leaned that way. You could water-cool to your heart's content, but to get a few more 100Mhz at lower temperatures, you would likely go over that self-imposed voltage limit, unless you are favored with one of those few "golden chips."

This explains why I favor getting the OC work done before putting the PC into regular usage. But an OC'er like me is still going to fiddle with settings long after that. BSODs in testing are . . . shall we say . . "an unpleasant experience." Suffering through those little traumas is the only way you'll likely go beyond your comfort level of 4.3.

And as you say, if you're happy with it running that way, no reason to bother.

 

[stockwiz]

I would agree with this if you have a decent Heatsink. If you are using the standard intel one I don't know if I'd go over 4ghz.

Good point. I sometimes forget I have a NH-d14 on mine... But even a hyper 212 evo would be plenty.

Heh-heh. . . . tweaking LLC and Offset IS messing with voltage. If "High" LLC is third-highest setting, I wouldn't increase it anymore for any reason. I'm willing to bet that such a level still leaves between 10 and 20mV of droop. Also, I'm holding back on comments I might otherwise make because there are differences in board generations for the same CPU-- P67 versus Z68, for instance.

Gotta say it! None of the Sandy users seem to be giving them up too soon! they work for me!

True... but when I just set the multiplier on 43 and did nothing else the motherboard overvolted it modestly so I basically gained 200 MHZ with no voltage change aside from the increased voltage at high loads from the load level calibration. The idle voltages decreased because of the -.1 offset.

 

[BonzaiDuck]

Good point. I sometimes forget I have a NH-d14 on mine... But even a hyper 212 evo would be plenty.



True... but when I just set the multiplier on 43 and did nothing else the motherboard overvolted it modestly so I basically gained 200 MHZ with no voltage change aside from the increased voltage at high loads from the load level calibration. The idle voltages decreased because of the -.1 offset.

Have you ever had any unexplainable idle instability with your configuration? If not, no problem.

Enough of forum posts and guides suggest to me that the best place to start with the Offset is to choose a "+" sign and put it as close to zero as possible. This would assist to reduce the chances of idle instability when using that level of LLC. But if the processor is "good enough," if you don't have an "Extra Voltage for Turbo" item in the BIOS, the -.1 may work for you. Still, that's a 10th of a volt. On my Z68 board, Offset voltage is adjusted in 0.005V increments. Are you sure you don't really mean "0.010V" for your setting?

 

[steve wilson]

I've found that the SB-K chips approach "voltage-limits" for certain heatpipe cooler solutions just about where a certain still-published temperature spec (dismissed as irrelevant) is "close." In fact, the "ultimate" throttling temperature-limit gets closer with lesser cooling solutions if you do the same thing. More specifically, if the prior-gen 32nm CPUs were spec'd with a safe-range limit around 1.38V, then the consensus among Sandy users also leaned that way. You could water-cool to your heart's content, but to get a few more 100Mhz at lower temperatures, you would likely go over that self-imposed voltage limit, unless you are favored with one of those few "golden chips."

This explains why I favor getting the OC work done before putting the PC into regular usage. But an OC'er like me is still going to fiddle with settings long after that. BSODs in testing are . . . shall we say . . "an unpleasant experience." Suffering through those little traumas is the only way you'll likely go beyond your comfort level of 4.3.

And as you say, if you're happy with it running that way, no reason to bother.

Yeah I'm not that enthusiastic about OCing my CPU. I only did it after about 2 years of having the chip because I needed the extra horse power. What is the highest you have managed to get your chip stable? How long did it take you?