Sign in with
Sign up | Sign in
Your question

Total OC noob w/ i5-2500k questions

Last response: in Overclocking
Share
May 31, 2011 2:28:43 AM

Hey guys,

I'm sure everyone is sick of these threads by now, but I didn't want to necro one of the older ones that had already run its course. Anyway, never OC's a CPU before, and was hoping to get a stable 4.2 or 4.4 on my i5-2500k. I'm running an ASRock Extreme 4 mother board, and I've tried a few different voltage settings but haven't managed to get anything stable yet. Any input is welcome.

Thanks
a b à CPUs
a b K Overclocking
May 31, 2011 2:55:29 AM

Have the same board & an i7 2600K. 4.4 ghz should be doable at around 1.3v, maybe a bit more, maybe a bit less. Change the voltage to manual and use the Offset setting. Increase that up to 0.015+ which will be about 1.3v. Mine runs stable @ 4.5 ghz with that offset

Short duration power limit - 210w
Long duration power limit - 200w
PLL voltage - 1.75v
Offset vcore - 0.040+ (1.32v)
Multiplier 46 x 100

Those are my settings at least
May 31, 2011 3:09:32 AM

Thanks for the input. I'll give those a try and see how it goes.
Related resources
May 31, 2011 8:15:21 PM

One quick note I'll make since you're a OC noob is that every 2500k is different. Some people's chips can do 4.8GHz at 1.32v, some can only get to 4.5 with 1.32v. It just depends.

I found that the best way to get a stable system is to use the offset setting as bearclaw99 notes above.

As an added step, you also want to download CPU-Z and a temperature monitoring software (I use speedfan) to keep an eye on temperatures. You also want to download a stresser utility -- Prime 95 is popular in the OC community. You can find these utilities with a quick google search.

You want to watch the core voltage section of CPU-Z and make sure that the voltage it's reading is not too far out of whack. Intel advises core voltage of no higher than 1.52.

You also want to watch your temps, which brings me to another question -- what cooler/heat sink are you using? I would recommend against overclocking at all with the stock cooler. You really want to keep these chips under 70C. The stock cooler will not cool a chip at load at 4.5GHz below 70C. You will have a few cores over 70C for sure.
May 31, 2011 10:40:28 PM

I grabbed CPU-Z and am using HWMonitor to track temps. I'm running a Scythe CPU cooler and after about 45 mins of Prime95 I my peak temp was 53 degrees at stock settings, so I think I have a decent amount of headroom. Again, I'd like to see 4.4GHz or so, nothing drastic.
a b à CPUs
a b K Overclocking
June 1, 2011 3:03:39 PM

Have Asrock Z68 Extreme4, is yours the Z, or P?
I've tried the 4.2 using the ASRock utility and was stable running prime95 (only ran for 2 Hrs). I backed off to 4.0 untill I complete the build (No GPU yet, will probably get the 6850. Once I complete build I plan on checking out the ASRock utility to see how high I can go and look at their settings, then back off vcore as it tends to be overly aggressive.
My temps @ 4.2 were around 52C using Zalman 9900Max HSF
PS, I know "software" OC is not the way to go.

How is HWMonitoring working on your system.
On monitoring temps and voltages, I love HWMonitor - but with the exceptions of core temps it is hosed. Most voltages are off in left field as is CPU temp. The 4 core temps look rigth (Plan on tring speed fan), But even the core with the core temps the reset min/max does not work. Have the latest version
June 1, 2011 4:29:37 PM

I have the P board. I played with the software utility a bit and really wasn't having much luck so decided to start making changes directly in the BIOS. I guess my biggest problem is just trying to isolate variables when making changes. I kind of feel like I'm trying to manage a half-dozen different settings without having a firm understanding of the impact each has individually, which is an education problem on my part.

I haven't used HWMonitor for voltage checks so far, just checking temps under load, so I can't give any feedback on that front.
June 1, 2011 4:45:54 PM

Penguinopapa said:
I kind of feel like I'm trying to manage a half-dozen different settings without having a firm understanding of the impact each has individually, which is an education problem on my part.


Great point, man. I'm right there with you. I understand the multiplier, I understand the core voltage, I understand the DRAM voltage and those are the only things I touch. Sure, I can find the definitions of other stuff:

Quote:
Load-Line Calibration: As the CPU accepts more work, its power draw increases. This in turn causes a voltage drop (vdrop), which is automatically compensated by the motherboard by increasing the applied voltage slightly. Some motherboards might therefore call this option Vdrop, or even more inaccurately, vdroop. Normally Load-Line Calibration is an enable or disable setting. When overclocking, you always want to enable Load-Line Calibration. Some motherboards provide more advanced options, which we'll detail in turn.

CPU PLL Voltage: Generally this doesn't need changing at all, although if increasing the CPU Voltage or the two Vcc options doesn't work, it's worth a try. Using anything over 2V isn't recommended, but 1.9V can help to stabilise an overclock.

PCH Voltage: Translated, this means the chipset, as PCH standd for Platform Control Hub. Again, this shouldn't need a voltage change, but if all else fails on the other settings above try increasing it just a touch. Don't go crazy though, as most P67 chipset heatsinks are largely for show because the chip doesn't need much cooling at default speeds and voltages.

Spread Spectrum : Spread Spectrum is useful in offices full of PCs, as it fluctuates the Base Clock frequency to prevent the PCs producing a sympathetic EMI field that might interfere with other devices. As such. you should disable Spread Spectrum when overclocking.


The advice of this bit-tech overclocking guide quoted above and that of many of the folks here and on places like overclock.net is real "fuzzy." As in, "Well, why don't you bump up the PCH by .1v?" or "Why don't you go ahead and set load-line calibration to "medium" and tell us what happens?"

I don't think there are many (any???) amateurs here or elsewhere that understand what we are actually doing to the reliability of the processor and chipset or whether bumping up those voltages have a reason to improve stability. After all, without specialized equipment, the only way to assess whether upping the PLL voltage or enabling internal PLL overvoltage or spread spectrum or whatever is to try it and watch the temperatures and see how long it lasts under prime95 -- but we never know whether we're degrading the processor...

I read a post from an enthusiast over at overclock.net that obviously has money to burn -- or a real love for the hobby -- that's buying a few 2600ks and was asking for the best way to degrade the 2600k. That kind of data collection is actually very helpful in helping us determine whether we should be messing with PLL voltage and the rest of these esoteric settings.

So my advice going forward would be to set the multiplier, leave core voltage on auto, and let the system boot and stress test it. If it works, knock down the voltage a little at a time and re-test. At some point you'll find the minimum core voltage needed for the speed you're looking for. That seems to be a better approach than messing around with 5 different variables, the consequences of which no one outside Taipei or Santa Clara knows...
June 1, 2011 6:18:35 PM

fullofzen said:
Great point, man. I'm right there with you. I understand the multiplier, I understand the core voltage, I understand the DRAM voltage and those are the only things I touch. Sure, I can find the definitions of other stuff:

Quote:
Load-Line Calibration: As the CPU accepts more work, its power draw increases. This in turn causes a voltage drop (vdrop), which is automatically compensated by the motherboard by increasing the applied voltage slightly. Some motherboards might therefore call this option Vdrop, or even more inaccurately, vdroop. Normally Load-Line Calibration is an enable or disable setting. When overclocking, you always want to enable Load-Line Calibration. Some motherboards provide more advanced options, which we'll detail in turn.

CPU PLL Voltage: Generally this doesn't need changing at all, although if increasing the CPU Voltage or the two Vcc options doesn't work, it's worth a try. Using anything over 2V isn't recommended, but 1.9V can help to stabilise an overclock.

PCH Voltage: Translated, this means the chipset, as PCH standd for Platform Control Hub. Again, this shouldn't need a voltage change, but if all else fails on the other settings above try increasing it just a touch. Don't go crazy though, as most P67 chipset heatsinks are largely for show because the chip doesn't need much cooling at default speeds and voltages.

Spread Spectrum : Spread Spectrum is useful in offices full of PCs, as it fluctuates the Base Clock frequency to prevent the PCs producing a sympathetic EMI field that might interfere with other devices. As such. you should disable Spread Spectrum when overclocking.


The advice of this bit-tech overclocking guide quoted above and that of many of the folks here and on places like overclock.net is real "fuzzy." As in, "Well, why don't you bump up the PCH by .1v?" or "Why don't you go ahead and set load-line calibration to "medium" and tell us what happens?"

I don't think there are many (any???) amateurs here or elsewhere that understand what we are actually doing to the reliability of the processor and chipset or whether bumping up those voltages have a reason to improve stability. After all, without specialized equipment, the only way to assess whether upping the PLL voltage or enabling internal PLL overvoltage or spread spectrum or whatever is to try it and watch the temperatures and see how long it lasts under prime95 -- but we never know whether we're degrading the processor...

I read a post from an enthusiast over at overclock.net that obviously has money to burn -- or a real love for the hobby -- that's buying a few 2600ks and was asking for the best way to degrade the 2600k. That kind of data collection is actually very helpful in helping us determine whether we should be messing with PLL voltage and the rest of these esoteric settings.

So my advice going forward would be to set the multiplier, leave core voltage on auto, and let the system boot and stress test it. If it works, knock down the voltage a little at a time and re-test. At some point you'll find the minimum core voltage needed for the speed you're looking for. That seems to be a better approach than messing around with 5 different variables, the consequences of which no one outside Taipei or Santa Clara knows...



I can really echo a lot of your sentiments, especially since I value a modest, stable OC over super impressive clock speeds. Just for anyone that's curious I ran a test using the built-in "optimized OC" presets in the BIOS and set the speed at 4.0 GHz. I ran IntelBurnTest for 3 loops and came back with a peak core voltage of 1.312V and a max temp of 55C. Would love to get some feedback based on what others are seeing in their systems.
June 1, 2011 6:32:40 PM

When I ran 4.5GHz, I came up with a voltage of 1.32 or so to be stable. You could probably knock your voltage down a bit from 1.312 -- change the offset and see if you're stable.
a b à CPUs
a b K Overclocking
June 2, 2011 6:06:14 PM

I had mine set to 4.2ghz vcore 1.255v and C1E enable so that the clock speed would go down on idle. Run prime95 for 8hrs max temp 72c (ambient 33c pretty hot in our country) on Noctua cpu cooler. Had it once at 4.7ghz vcore 1.355 but the temps were too high 78c and I don't like it. I just settle at 4.2ghz. My mobo is MSI P67A-GD65.
June 10, 2011 8:08:34 AM

Just wanted to check back in with an update and a few new questions. I got the CPU to run at 4.4 Ghz completely stable with no issues. Now, a couple of questions. Firstly, monitoring CPU-Z I noticed that the Core Voltage figure peaked at 1.352V During a maximum setting stress test using IntelBurn. That seemed a bit high to me, considering the frequency, and I'd like to bump it back but if it's actually drawing that much voltage at that frequesncy I doubt it would be stable if I simply cut back the voltage and left the frequency in place, correct? I know that figure is below the "max" that Intel recommends, but is it a figure that is going to really shorten the useful lifetime of my CPU?

Secondly, my CPU temps. Using the maximum setting in IntelBurn the highest temp I've recorded was a brief spike to 72C, though it generally hovered at about 69C, and the ambient in the room was low 80's F. During actual use the highest I've ever recorded after an extended gaming session was 61C, with it generally hovering in the mid 50's under load. These temperatures didn't seem like a real cause for concern to me, but I thought I'd run by you guys before I do any lasting damage.

Edit: Also wondering what sort of GFlops folks were seeing at their different OC's.
June 10, 2011 3:08:33 PM

Penguinopapa said:
Just wanted to check back in with an update and a few new questions. I got the CPU to run at 4.4 Ghz completely stable with no issues. Now, a couple of questions. Firstly, monitoring CPU-Z I noticed that the Core Voltage figure peaked at 1.352V During a maximum setting stress test using IntelBurn. That seemed a bit high to me, considering the frequency, and I'd like to bump it back but if it's actually drawing that much voltage at that frequesncy I doubt it would be stable if I simply cut back the voltage and left the frequency in place, correct? I know that figure is below the "max" that Intel recommends, but is it a figure that is going to really shorten the useful lifetime of my CPU?

Secondly, my CPU temps. Using the maximum setting in IntelBurn the highest temp I've recorded was a brief spike to 72C, though it generally hovered at about 69C, and the ambient in the room was low 80's F. During actual use the highest I've ever recorded after an extended gaming session was 61C, with it generally hovering in the mid 50's under load. These temperatures didn't seem like a real cause for concern to me, but I thought I'd run by you guys before I do any lasting damage.

Edit: Also wondering what sort of GFlops folks were seeing at their different OC's.


Yeah -- 1.352v is really high for 4.4GHz. What is your vcore offset setting? You could probably put it at least to (-).15 and be stable unless you have a terrible chip.
June 10, 2011 3:29:58 PM

I just tried a manual offset of -.15 and blue screen'd almost immediately upon opening Windows. Restored my old settings, ran IntelBurn again while watching the core voltage on CPU-Z and, sure enough, it spent most of it's time at 1.352. Am I missing something here, or did I seriously draw a dud chip from Intel?
June 10, 2011 3:57:55 PM

Penguinopapa said:
I just tried a manual offset of -.15 and blue screen'd almost immediately upon opening Windows. Restored my old settings, ran IntelBurn again while watching the core voltage on CPU-Z and, sure enough, it spent most of it's time at 1.352. Am I missing something here, or did I seriously draw a dud chip from Intel?


Did you enable PLL overvoltage? What are your load line calibration settings?
June 10, 2011 4:14:13 PM

I was actually just investigating my PLL Voltage. Here's what I'm looking at in the BIOS as far as voltages go:
Core: 1.344
Load-Line: Level 5
DRAM V: 1.5
PCH: 1.059
PLL: 1.832
VTT: 1.051
VCCSA: .925

PLL is set to "Auto", but that 1.832 seems awfully high to me. Think that might be the culprit? Thanks so much for your patience and input, I really appreciate it.
June 10, 2011 4:22:00 PM

Penguinopapa said:
I was actually just investigating my PLL Voltage. Here's what I'm looking at in the BIOS as far as voltages go:
Core: 1.344
Load-Line: Level 5
DRAM V: 1.5
PCH: 1.059
PLL: 1.832
VTT: 1.051
VCCSA: .925

PLL is set to "Auto", but that 1.832 seems awfully high to me. Think that might be the culprit? Thanks so much for your patience and input, I really appreciate it.


Question 1: Let me get this straight -- do you have "PLL voltage" on auto, and that is setting it automatically to 1.832?

Question 2: What does your "internal PLL overvoltage" setting?
June 10, 2011 4:30:13 PM

1. Yessir. From what I read the PLL voltage doesn't generally need adjusting for mild overclocking, and I didn't pay much attention to it, much to my current embarrassment.
2. PLL Overvoltage is currently disabled.
June 10, 2011 4:59:40 PM

1) Just leave PLL voltage as auto -- I wouldn't touch that one. The processor's not going to invite a PLL voltage higher than it's bounds, so if it's 1.9 auto, don't sweat it.

2) Enable internal PLL overvoltage and back-down your core voltage a little back to 1.32 or 1.33. Would you see if this is stable? I don't know the long-term impact of internal PLL overvoltage, but it would be interesting to see if it allows your chip to work at 4.4GHz and 1.32v.
June 10, 2011 5:18:55 PM

Well, I enabled PLL overvoltage and tried setting a fixed core voltage of 1.325 and that crashed immediately. Same story with a -.15 offset. Starting to get a bit concerned.
June 10, 2011 6:29:05 PM

So, I tried restoring the defaults for everything; all voltages and clock speed. The PLL stayed at 1.832. The only voltage change was the core now runs at 1.232 under load. Could this be a BIOS bug, or did I get a lemon?
June 10, 2011 6:34:12 PM

Okay -- I want to go back to load line calibration. I'm confused by the setting that you're using.

You said the following:

Quote:
Load-Line: Level 5


Two things:

1) From what I'm reading in the Extreme 4 manual, the setting for load line calibration (on page 67) is expressed in terms of percentages, not levels. Can you explain what you mean by "level 5"?

2) I also want to suggest another thing -- it is likely that you have made a ton of modifications to the BIOS in the last couple of weeks that you've been working this. May I recommend that you clear your CMOS and do just one thing to it -- push multiplier up to 45. Then see what happens when you try to boot.
a b à CPUs
a b K Overclocking
June 10, 2011 6:34:22 PM

Use a higher offset. Make sure you are using a positive offset instead of negative. You shouldn't need to go much above stock voltage to get 4.4 stable. Some of those chips can even do it on stock
June 10, 2011 6:37:12 PM

bearclaw99 said:
Use a higher offset. Make sure you are using a positive offset instead of negative. You shouldn't need to go much above stock voltage to get 4.4 stable. Some of those chips can even do it on stock


Don't do this -- your voltage is already at 1.35 volts at 4.4GHz. You don't want to do positive offset at all.
June 10, 2011 6:49:51 PM

Default settings and 45x won't even boot to Windows; it BSOD's almost instantly. I have since returned everything to the stock settings as they were the first time I booted the machine. As for the Load Line, my UEFI offers Level 5-Level 1. There's no percentages visible. Also, within the last few days my computer has developed the habit of powering on when I press the case switch and then powering off 3 secs later only to spontaneously power back up a few seconds later and then boot normally. I'm much more apt to believe that I've done something stupid here rather than blame faulty components. I'd like to update to the 1.7 BIOS, but I'm a little hesitant to do anything at the moment, as I'm trying to isolate issues, not potentially create more.
June 10, 2011 8:19:29 PM

Penguinopapa said:
Default settings and 45x won't even boot to Windows; it BSOD's almost instantly. I have since returned everything to the stock settings as they were the first time I booted the machine. As for the Load Line, my UEFI offers Level 5-Level 1. There's no percentages visible. Also, within the last few days my computer has developed the habit of powering on when I press the case switch and then powering off 3 secs later only to spontaneously power back up a few seconds later and then boot normally. I'm much more apt to believe that I've done something stupid here rather than blame faulty components. I'd like to update to the 1.7 BIOS, but I'm a little hesitant to do anything at the moment, as I'm trying to isolate issues, not potentially create more.


Y'know...I am downright flummoxed. There is no reason that your chip and motherboard can't do 4.5GHz at 1.32v with load line calibration at the mid-range and internal PLL overvoltage enabled.

Two questions:

1) What PSU do you have?

2) Are you within 30-day return period(s) for the 2500k and board? I think you need to RMA both -- there must be something wrong hardware-wise. Only explanation, really...
June 10, 2011 9:55:37 PM

Hmm -- yeah, that seaconic is a really great PSU. Seasonic is a hell of a brand. Good luck with the reshipment -- if you would, please come back to this thread and update us on whether the new board and processor work better than the current set.

Sandy Bridge overclocking is not supposed to be this hard.

Hey -- one more quick question. What is your level of confidence in your memory? Did you try running memtest86 on your memory?
June 10, 2011 10:26:56 PM

Haha, yeah I kept hearing stories about how painless OC'ing a Sandy Bridge was and felt really left out. New egg was spectacular, though, as they're advance RMA'ing replacements pieces to me so that I can have them before I move next week. Love those guys.

The only issues I've had with memory was the mobo setting the timing as 9-9-9-28 when the manufacturer specs are 9-9-9-24. I set it manually and haven't had any issues. CPU-Z shows it running at the proper voltage and timings, so I think I'm ok there. Just for reference: http://www.newegg.com/Product/Product.aspx?Item=N82E168...

And I'll absolutely check back in once I get replacement parts and see where I stand with voltage and clock speed. Thanks so much for your help. Hopefully I'll have some good news in the next 4 or 5 days.
June 17, 2011 5:12:59 AM

Got my replacement components and will be installing them tomorrow. Quick question for those that have RMA'd CPU's before: Should I remove the thermal paste and clean up the chip as well as I can before returning it, or will that not impact the return? I'm not trying to be underhanded, but I also don't want to get stuck paying for a defective CPU because I applied aftermarket thermal paste. Thanks.
June 17, 2011 4:55:00 PM

Penguinopapa said:
Got my replacement components and will be installing them tomorrow. Quick question for those that have RMA'd CPU's before: Should I remove the thermal paste and clean up the chip as well as I can before returning it, or will that not impact the return? I'm not trying to be underhanded, but I also don't want to get stuck paying for a defective CPU because I applied aftermarket thermal paste. Thanks.


Not underhanded at all to remove the thermal paste. From my perspective, it came to you clean -- you should send it back clean.

I hope your new components work out alright. Please come back and report on progress OC'ing.
June 18, 2011 4:56:54 AM

Well, so far I'm having a whole new set of issues. I replaced the CPU first, hoping that that was my problem and not the motherboard. Now, I cannot get it to boot to Windows at all. I totally reset the BIOS and the voltage numbers at 3.3 seem promising, but it freezes at the "Starting Windows" screen every time. If I restart and let Windows try to repair the issue, I get a message saying that it cannot be automatically repaired. So, guesses? I'm starting to think I have a short somewhere and should remove the motherboard and re-check all of the stand-offs. Getting a tad discouraged as my efforts seem to be making things worse here.
June 18, 2011 8:04:05 AM

Progress! I took the mobo out and checked all of the connections and stand-offs. After putting it back together things seem to be running much better. I managed to get a stable 4.4 GHz by just manually altering the multiplier and my peak voltage during stress testing came in at 1.302V. The temperature was also considerably lower than with the previous install.

These numbers seem to be much more inline with what others are seeing, so hopefully this is all sorted out now. Not sure if it was a bad CPU or just a sloppy install, but it's definitely a nicer machine now.
!