Overclock.net banner

i7 2600K 4.7 Ghz overclock BSOD: 0x0000001e. What causes this?

5K views 34 replies 8 participants last post by  TwoCables 
#1 ·
I'm trying to get a stable 4.7 Ghz overclock on this 2600K CPU I've just bought. I previously was running a 2500K @ 4.6 Ghz in this same PC. The motherboard I'm using is an Asus P8P67 Pro Rev 3.1.

This isn't one of the typical BSODs for an overclock, like 0x101, or 0x124. It sounds like it's a memory or driver error. But I haven't changed any of my software, and the only hardware I've changed is going from a 2500K to a 2600K CPU. Anyone know what could be causing this?

This is my overclock configuration:

Voltage offset: -0.010v
PLL Overvoltage: Disabled
LLC: 100%
VRM Spread Spectrum: Disabled
VRM frequency: 350
Phase control: Asus optimized
Duty control: Extreme

Speed-step, C1e: Enabled
C3, C6: Auto

RAM profile: XMP

Passes Intel Burn Test, Prime 95, run for ~10 minutes, max settings.

min core voltage - 0.992v
max peak voltage - 1.416v
min core temps - 30, 28, 26, 30
max core temps - 68, 76, 75, 72

Things seem fine in general, but every so often, and while my PC isn't under any serious load, I get BSOD 0x0000001e, KMODE_EXCEPTION_NOT_HANDLED, crash address: ntoskrnl.exe+6f3d0, 6.1.7601.23418 (win7sp1_ldr.160408-2045):

Problem signature:
Problem Event Name: BlueScreen
OS Version: 6.1.7601.2.1.0.256.1
Locale ID: 1033

Additional information about the problem:
BCCode: 1e
BCP1: 0000000000000000
BCP2: 0000000000000000
BCP3: 0000000000000000
BCP4: 0000000000000000
OS Version: 6_1_7601
Service Pack: 1_0
Product: 256_1

Files that help describe the problem:
C:\Windows\Minidump\101916-15802-01.dmp
C:\Windows\Temp\WER-17316-0.sysdata.xml
 
See less See more
#2 ·
I think you need to disable C3 and C6. When using an offset, these should be disabled. With them set to Auto or Enabled (either one), you can have instability while the system is idling or under light loads. This will raise the lowest core voltage slightly.

Also, is that all the testing you're going to do? 10 minutes?
 
#4 ·
I just loaded my 4.7 Ghz overclock profile in the BIOS, and it looks like I already had C3 and C6 disabled. C1e and Speed-Step are enabled, though.

I'm trying reducing the voltage offset from -0.010v to -0.005v.
 
#6 ·
Quote:
Originally Posted by TwoCables View Post

Also, is that all the testing you're going to do? 10 minutes?
For now. My i5 2500K @ 4.6 Ghz would crash after 30 seconds of Prime 95, but it never gave me a BSOD during my normal PC usage, while gaming, audio editing, watching movies, etc. I figured 10 minutes would be comparatively better.
 
#7 ·
Quote:
Originally Posted by Delicieuxz View Post

For now. My i5 2500K @ 4.6 Ghz would crash after 30 seconds of Prime 95, but it never gave me a BSOD during my normal PC usage, while gaming, audio editing, watching movies, etc. I figured 10 minutes would be comparatively better.
Time is not how Prime95's effectiveness as a stability test can be judged, unless you know precisely how many FFTs your version of Prime95 has and how long it will take you to get through all of them (there are 82 in version 27.9). If you're not getting through all of them, then you're not testing the stability and thus you're wasting your time. 10 minutes is pretty much the same as nothing. So is 30 minutes and even a few short hours.

It's not a constant, unchanging load (and it would be far less effective if it were). Depending on your setting for how many minutes it spends on each FFT (less is not always better), you are looking at running Prime95 for several hours at minimum to do a proper test of your computer's stability with the CPU overclocked that high.
 
#8 ·
I have my i7 2600k @ 5ghz 1.42 24/7 with 68c/70c across the cores using cpuz stress test. in all fairness I wasn't stable below 1.45 until I realised that I had my northbridge voltage undervolted. fixed it by going to 1.050v from 1.042v. silly me. stable in prime95 for 2 hours @ 76c/77c across all cores
smile.gif
 
#9 ·
Quote:
Originally Posted by TwoCables View Post

Time is not how Prime95's effectiveness as a stability test can be judged, unless you know precisely how many FFTs your version of Prime95 has and how long it will take you to get through all of them (there are 82 in version 27.9). If you're not getting through all of them, then you're not testing the stability and thus you're wasting your time. 10 minutes is pretty much the same as nothing. So is 30 minutes and even a few short hours.

It's not a constant, unchanging load (and it would be far less effective if it were). Depending on your setting for how many minutes it spends on each FFT (less is not always better), you are looking at running Prime95 for several hours at minimum to do a proper test of your computer's stability with the CPU overclocked that high.
Thanks for that information, TwoCables.

Quote:
Originally Posted by Mr Mcgoo View Post

I have my i7 2600k @ 5ghz 1.42 24/7 with 68c/70c across the cores using cpuz stress test. in all fairness I wasn't stable below 1.45 until I realised that I had my northbridge voltage undervolted. fixed it by going to 1.050v from 1.042v. silly me. stable in prime95 for 2 hours @ 76c/77c across all cores
smile.gif
Well that's interesting,to me. I can't seem to get my CPU stable at 4.7 Ghz, no matter how much vcore it receives. Though, I am using a voltage offset, and not a constant voltage. But, even when allowing it to reach peaks of 1.440v, it still trips at unpredictable times, and crashes the PC - sometimes at basically no load, sometimes while playing a games...

So, I've set it to 4.6 Ghz for now, and it's stable at that Ghz, so far, with a voltage offset that gives it voltage peaks of 1.424v, or maybe even 1.432v. The voltage peaks are nuts, and I thought it was bad when my i5 2500k needed 1.336v to handle 4.6 Ghz.

Anyway, either my i7 2600K is the worst overclocking 2600K I've ever heard of, or something is messing up the OC, regardless of all the settings I've tried - and I've tried pretty much every non-destructible configuration I could think of, and a lot of other people's own configurations, or parts of them.

So, maybe that NB voltage is something I have to look into. I didn't notice it in the Asus P8P67 Pro BIOS, but I'll look for it now.
 
#10 ·
There are actually some things that I haven't adjusted, which are the voltages in this image that are set to Auto.



And I haven't changed my RAM rimings, or downclocked it. I have it set the the 1866 Mhz XMP profile, and I don't want to be running RAM @ less than 1866 Mhz.

I didn't find a NB voltage setting in the BIOS, BTW.

Edit: Actually, the NB voltage setting is apparently the setting labeled "VCCSA Voltage."

http://www.hardocp.com/article/2011/08/16/asus_p8p67_pro_new_b3_revision_motherboard_review/2#.WAiln48rI-U

The CPU integrated Northbridge controller voltage is settable to a 1.70V maximum using the VCCSA Voltage setting.
The VCCIO Voltage setting configures the CPU's internal memory controller voltage, also with a 1.70V ceiling.
The CPU PLL Voltage setting configures the CPU's power regulation circuitry voltage, with up to 2.20V allowed.
The Intel P67 chipset voltage can be configured using the PCH Voltage option, for operation at up to 1.70V.
 
#12 ·
I think you just need to stop using Offset. I hate using Offset because I get slightly inferior performance and overall slightly worse stability no matter what settings I use to help increase its stability. With Manual Voltage, it just works.

Also, try to resist comparing the way your CPU overclocks to the way other CPUs overclock. Some are great overclockers, some are more average, and some just suck. Mine sucks. It's The Silicon Lottery. You also have to consider the motherboard and memory a person has when they tell you how great their CPU overclocks. However, even with the exact same motherboard and memory as yours, it's still The Silicon Lottery, even for motherboards and memory.

By the way, you don't have to take a photo of your UEFI BIOS. You can take screenshots. You just insert a USB flash drive and press F12. It only captures what you can see, so scroll to show more in each screenshot. Each screenshot is saved as a BMP, so you will need to do a Save As in like Microsoft Paint to change them over to PNGs.
 
#13 ·
Quote:
Originally Posted by TwoCables View Post

I think you need to disable C3 and C6. When using an offset, these should be disabled. With them set to Auto or Enabled (either one), you can have instability while the system is idling or under light loads. This will raise the lowest core voltage slightly.

Also, is that all the testing you're going to do? 10 minutes?
Quote:
Originally Posted by TwoCables View Post

Time is not how Prime95's effectiveness as a stability test can be judged, unless you know precisely how many FFTs your version of Prime95 has and how long it will take you to get through all of them (there are 82 in version 27.9). If you're not getting through all of them, then you're not testing the stability and thus you're wasting your time. 10 minutes is pretty much the same as nothing. So is 30 minutes and even a few short hours.

It's not a constant, unchanging load (and it would be far less effective if it were). Depending on your setting for how many minutes it spends on each FFT (less is not always better), you are looking at running Prime95 for several hours at minimum to do a proper test of your computer's stability with the CPU overclocked that high.
I disagree. A proper custom prime95 test only needs to run for 1-2 hours in order to establish 99.999 percent stability. Running prime 95 for half a day or longer proves nothing in my opinion and all you are doing is degrading your CPU.

Also, using offset is the way to go unless you like wasting electricity. Manual doesnt improve stability nowadays. In the Core2 days, yeah, it was better but not nowadays unless you are running some kind of ludicrous 5ghz 1.5 volts setup.
 
#14 ·
Quote:
Originally Posted by Lex Luger View Post

I disagree. A proper custom prime95 test only needs to run for 1-2 hours in order to establish 99.999 percent stability. Running prime 95 for half a day or longer proves nothing in my opinion and all you are doing is degrading your CPU.
That's wrong because you're not getting through enough of the FFTs, especially the ones toward the middle and end of the test that are much more effective at testing the system's stability. You want to get through all of the FFTs, not just the first like 15% of them, which is exactly what you're recommending even if you don't even realize it.

Every once in a while, I see people coming to OCN complaining like, "I'm having problems. I don't know why because my CPU is super stable! I tested it for 2 hours in Prime95!! It should be 100% stable! Is my CPU dying?!". No. It's just not stable because 2 hours is almost as good as nothing. So when I see this, I always try to get them to test for long enough to get through all 82. At 15 minutes per FFT, that's about 24 hours. At the new default of 3 minutes per FFT, that can be 7-10 hours. Every single time I have convinced the person to get their system stable enough to run through all of the FFTs in Prime95, all of their problems go away. Yeah, every single time. Without fail.

So, you can keep doing incomplete tests if you want, but I will never recommend it! Prime95 does NOT put a constant, unchanging load on the CPU and it's NOT harmful unless you're doing it 24/7 for several months in a row with a dangerously-high core voltage! With each new FFT that it starts up during the test, you get a much different test of your system's stability. There were 82 FFTs back in Prime95 version 27.9, and with how many they seem to add every couple of versions or so, I wouldn't be a bit surprised if they have added 5-15 more FFTs by now. I just haven't had enough time to test it to find out (it's not exactly easy to find out). I suppose I could just contact them and ask, but I like to avoid bothering someone with a question if I can find out the answer myself.

Anyway, yeah, it's not harmful. The only way you're going to have a chance of degrading the CPU is if you have like say 1.45V or higher and you're putting a constant load on the CPU for months and months in a row. Just ask @$ilent. He knows firsthand. He can tell you about his adventures with his 2500K. He had the clock at around 4.7 or 4.8 GHz with his voltage set to over be over 1.4V in Windows (it was necessary for his CPU just like it is for mine) and was doing like Folding or Mining on the CPU 24/7 for several months (over 6 if I remember correctly), and he finally began seeing some mild degradation toward like the 7th or 8th month. It wasn't a big deal, he just had to lower the clock speed to 4.5 GHz and it was fine. It's not like he killed his CPU or anything. Y'know? We need to quit worrying so damn much.

So, a few days of testing with Prime95 isn't going to hurt the CPU in the LEAST bit. When you are increasing the clock speed that high, you need to make sure your system is stable. You need to properly test it. The best way we know of to do that is by using Prime95's Blend test. Or better yet, you select "Custom" after making sure Blend is selected, and this creates what's usually called the Custom Blend test. With this, you enter about 80-85% of your memory where it says 'memory to use', and you get a MUCH more effective test of your system's stability. If your system can successfully run ALL of the FFTs with this test, then you can pretty much be guaranteed it's extremely stable. Again, it can take 24 hours to get through 82 FFTs at 15 minutes per FFT (there's a pause in between each FFT, and that adds up very quickly), and it can take 7-10 with the new default of 3 minutes per FFT.
 
#15 ·
Quote:
Originally Posted by Lex Luger View Post

Also, using offset is the way to go unless you like wasting electricity. Manual doesnt improve stability nowadays. In the Core2 days, yeah, it was better but not nowadays unless you are running some kind of ludicrous 5ghz 1.5 volts setup.
What you're seeing in CPU-Z (or other similar programs) while you're not doing anything is not true idle because CPU-Z is forcing the CPU to stay "awake" so that it can poll it for the information you see in CPU-Z. If your computer is truly not using the CPU at all (that is, if you don't have any monitoring software open and nothing else is possibly using the CPU), then it could be using pretty much no power practically speaking with pretty much no voltage at all, even with the Manual Voltage, and yes this depends of course on your power-saving settings. You have to have that set up correctly, of course. So, an Offset Voltage doesn't mean your true idle is less harmful on the CPU. It can be exactly the same as Manual, if you have your power-saving settings set up for it. Offset Voltage is just more "green", but I don't care for it due to my experiences with it.

As I said, I have compared using Offset and Manual for my 2500K, and over time I have discovered slightly superior stability and slightly superior performance with Manual vs Offset, even after spending days trying to stabilize Offset more than I had it. During my time on OCN ever since I discovered this, I have found that I'm not the only one who experiences this.

Did you really think that I was preferring Manual just because of overclocking Core2 CPUs? It has nothing to do with that! This is an extremely different CPU, and thus everything I learned and experienced with Core2 CPU overclocking goes right out the window. You have to start over and learn how these CPUs work and how these CPUs overclock. You can't say, "Well, we did this with Core2 CPUs, and so we do this with Sandy Bridge". I know it doesn't work that way. Give me some credit!
 
#16 ·
You're right that the middle and end FFT are the ones to test. Thats why I run a custom test for around 2 hours with middle ones and another test for an hour with the last ones. I have never had a system blue screen after passing both the middle and end FFT's in last 5+ years and were talking around a dozen or so overclocked systems. My main rig especially gets heavily tested with lots of gaming and transcoding in handbrake. 10 years ago, yeah I ran 24 hour prime 95 tests. Nowadays I find Im wasting electricity and time.

My 2500k degraded after running 1.47+ for long hours with prime 95. It was kept around 70C and I used a good motherboard, asus p67 deluxe.
 
#18 ·
Quote:
Originally Posted by Lex Luger View Post

You're right that the middle and end FFT are the ones to test. Thats why I run a custom test for around 2 hours with middle ones and another test for an hour with the last ones. I have never had a system blue screen after passing both the middle and end FFT's in last 5+ years and were talking around a dozen or so overclocked systems. My main rig especially gets heavily tested with lots of gaming and transcoding in handbrake. 10 years ago, yeah I ran 24 hour prime 95 tests. Nowadays I find Im wasting electricity and time.

My 2500k degraded after running 1.47+ for long hours with prime 95. It was kept around 70C and I used a good motherboard, asus p67 deluxe.
Yeah, well not everyone's experience will be the same. You can't say that just because that type of stability testing works well for you (or seems to) that it will work well for me or anyone else. It's The Silicon Lottery, y'know? The first test that should be done after overclocking should be the same for everyone. You need to make sure you test it as thoroughly as possible. To do that, you want to get your system to run all of the FFTs. They are there for a reason.

How do you know your 2500K degraded? All Sandy Bridge CPUs eventually reach a point after a year or two where they suddenly need more voltage (or you just reduce the clock speed, or do both a little bit). It's a break-in period that they go through, as many of us here on OCN have discovered. It's normal (it's not degradation). After this, the CPU seems to be good to go from that point forward at the new settings. It's like it snaps or something, but again, after that it's fine at the new settings.

Degradation is gradual and you have to keep reducing the clock speed and/or the voltage more and more. Sometimes it's so bad that the CPU can't even overclock anymore but instead can only be stable at stock. That to me still isn't a big deal.
 
#19 ·
Quote:
Originally Posted by Delicieuxz View Post

There are actually some things that I haven't adjusted, which are the voltages in this image that are set to Auto.



And I haven't changed my RAM rimings, or downclocked it. I have it set the the 1866 Mhz XMP profile, and I don't want to be running RAM @ less than 1866 Mhz.

I didn't find a NB voltage setting in the BIOS, BTW.

Edit: Actually, the NB voltage setting is apparently the setting labeled "VCCSA Voltage."

http://www.hardocp.com/article/2011/08/16/asus_p8p67_pro_new_b3_revision_motherboard_review/2#.WAiln48rI-U

The CPU integrated Northbridge controller voltage is settable to a 1.70V maximum using the VCCSA Voltage setting.
The VCCIO Voltage setting configures the CPU's internal memory controller voltage, also with a 1.70V ceiling.
The CPU PLL Voltage setting configures the CPU's power regulation circuitry voltage, with up to 2.20V allowed.
The Intel P67 chipset voltage can be configured using the PCH Voltage option, for operation at up to 1.70V.
sorry I meant vccsa as you stated. that helped stabilize my cpu to 5ghz 1.42v. my zalman lq315 keeps it under 70c @ 100% load on cpuz. vccio voltage increase 1 notch got me to 2133mhz 10-10-10-30 from 1600mhz 9-9-9-24
 
#20 ·
I feel that something pretty particular is causing the BSODs, because no matter how much I increase the voltage, it still runs generally fine, but then randomly BSODs. Something has to be going on in those moments, and I wish I knew what to change to stop it.

On another forum, I was advised to not use 100% LLC and negative offset, but to use 50% or 75% LLC, and then apply a positive voltage offset, to raise the lowest voltage. I tried that with the CPU set at 4.6 Ghz, and it runs generally fine, though it crashed about an hour or so into Prime 95 blend test. I used LLC 75%, and a voltage offset of +0.45v.

Also, voltage offset is weird, in that if I leave it at auto, the voltage goes very high, like 1.52v, and even a bit higher, with the multiplier at 47. However, setting the voltage offset to just -0.005v results in the voltage staying below 1.432. That's obviously not anything close to -0.005v of difference. And that is very inconvenient, because I don't have a means to adjust the voltage between way too high, and significantly lower than way too high, to find something inbetween. My BIOS only accepts VO intervals of 0.005v.

I might try giving a little bit extra to the VCCIO.
 
#21 ·
Try switching to Manual Mode for the voltage control instead of Offset Mode. Seriously.

Auto Offset should be avoided when overclocking.

The way Offset works is, it adds or subtracts that exact amount either to or from the VID of the current multiplier and clock. Every multiplier and clock speed has its own VID. The resulting voltage you get depends on how much vDroop or vRise there is, and that's the voltage you get. If you can eliminate the vDroop and vRise, then you can just do a simple calculation to know what your voltage in Windows under full load should be if you choose X amount of Offset either negative or positive. If you have it set to Auto, then there's no way to know how much it's adding or subtracting, and it's usually going to result in voltages that are WAAAYYY too high. If you want to use Auto for Offset, then everything else needs to be set to Auto as well. That means, no overclocking.

Just try Manual Mode for now though. I bet you will find it to be superior just as I do.
 
#22 ·
Yeah, I know I could do manual, though I really want to keep the speed-stepping. I really like that at 95%+ of the time, the voltage is beneath 1.3v, usually close to 1.1v.

I tried a manual voltage of 1.45v, with the CPU multiplier at 48, and was able to run Prime 95 and IBT for a few minutes, until I stopped the testing. I don't like leaving the voltage that high, though. Is heat, or voltage more responsible for wear on a CPU?

What do you think about what this Prime 95 guide says, about using specific FFTs for overclock-testing?

http://overclocking.guide/stability-testing-with-prime-95/

The correct torture test settings are important to test the right things of your system. These are the most important FFT sizes:

1344K: High load on the core and the core voltage
8K: Maximum heat production to test your cooling solution
800K: Test RAM stability
864K: Test the whole system

Quote:
The way Offset works is, it adds or subtracts that exact amount either to or from the VID of the current multiplier and clock. Every multiplier and clock speed has its own VID. The resulting voltage you get depends on how much vDroop or vRise there is, and that's the voltage you get. If you can eliminate the vDroop and vRise, then you can just do a simple calculation to know what your voltage in Windows under full load should be if you choose X amount of Offset either negative or positive. If you have it set to Auto, then there's no way to know how much it's adding or subtracting, and it's usually going to result in voltages that are WAAAYYY too high. If you want to use Auto for Offset, then everything else needs to be set to Auto as well. That means, no overclocking.
That explains it, thanks.
 
#23 ·
Quote:
Originally Posted by Delicieuxz View Post

Yeah, I know I could do manual, though I really want to keep the speed-stepping. I really like that at 95%+ of the time, the voltage is beneath 1.3v, usually close to 1.1v.

I tried a manual voltage of 1.45v, with the CPU multiplier at 48, and was able to run Prime 95 and IBT for a few minutes, until I stopped the testing. I don't like leaving the voltage that high, though. Is heat, or voltage more responsible for wear on a CPU?

What do you think about what this Prime 95 guide says, about using specific FFTs for overclock-testing?

http://overclocking.guide/stability-testing-with-prime-95/

The correct torture test settings are important to test the right things of your system. These are the most important FFT sizes:

1344K: High load on the core and the core voltage
8K: Maximum heat production to test your cooling solution
800K: Test RAM stability
864K: Test the whole system
That explains it, thanks.
What you are seeing in CPU-Z isn't what's actually happening when the CPU is truly doing nothing at all. When you have a monitoring program open (such as CPU-Z), the CPU is constantly being polled. It's being kept "awake". So, what you're seeing in the monitoring software can lead you to believe that when you close the monitoring software, the voltage is still that high. In reality, it's not - but only if you have all power-saving settings fully enabled in the BIOS such as C1E, SpeedStep, C3 and C6 (because it's harmless to leave C3 and C6 enabled in Manual Mode vs. Offset Mode). These settings enable the deepest sleep mode possible so that when the CPU isn't being asked to do anything, it can go into a nearly zero-power mode type of sleep. It doesn't matter what CPU-Z says because when you have CPU-Z open (or any other similar program), it's preventing the CPU from truly idling and going to sleep.

So again, I recommend avoiding Offset Mode because it's clearly not going to work no matter what you do - which is exactly the same problem my CPU has. Manual Mode is fine and very safe, provided you have your power-saving settings enabled. Even if you don't, it's still not going to hurt your CPU. As I said, just ask $ilent. Here's the post I made to talk about his experience: http://www.overclock.net/t/1614022/i7-2600k-4-7-ghz-overclock-bsod-0x0000001e-what-causes-this/10#post_25595646

As you can see, these CPUs are tough. They're like the toughest tanks on the battlefield.

Regarding testing with Prime95, it's harmless, just as I explained using $ilent's experience. It's not like you're running Prime95 24/7 for a year straight. It's just a day or two. Y'know? It will not hurt the CPU in the least bit. If this were a video game and your CPU had a "hull integrity" readout, it would still be at 100.00% after all Prime95 testing. It can take it. It's built to be used. Y'know? Intel wouldn't succeed if they built wimpy CPUs that are fast but can't really be used much due to falling apart too easily under load. Think about it. ;)

So just use Manual Mode and run a proper test of Prime95. I recommend opening Prime95, selecting 'Custom', and then typing in about 80-85% of your installed memory, whatever that may be. If the default time to run each FFT is 3 minutes, then just go with that and try to get your system stable enough to last for around 8-9 hours of this. That should be plenty of time to get through ALL of the FFTs. If it's defaulting to 15 minutes, then shoot for around 24 hours. Yes, that's how long it would take at 15 minutes per FFT.

You don't want to focus on just a few FFTs because that's only for quick tests while you're in the middle of setting up your desired overclock. You've already done that. You're at the final testing stage right now, so you want to get your system stable enough to pass all of the FFTs, not just a few specific ones.
 
#24 ·
So, I gave my 2600K a manual voltage of 1.425v, and 100% LLC, and got these readings after successful, though short, Prime 95 and IBT runs (10 minutes of FFT 1344, 15 minutes of IBT maximum).



That gave me the idea that the CPU handles 4.7 Ghz around the 1.45v area (though it also lasted short testing at 4.8 Ghz around that same voltage).

After that, I did some more testing with voltage offset, this time using a larger positive voltage offset, and using a positive voltage offset, rather than a negative one, seems to have helped with the random BSODs, as I've been running 4.7 Ghz with 75% LLC, and an offset of +0.060v, since yesterday, gaming, watching videos, running benchmarks, idling overnight, and haven't had any crashes (other than one mid Prime 95 test).

The max voltage peak with this configuration is 1.440v, in Prime 95. In general use, and gaming, max voltage peak is 1.424v, though I have seen it touch 1.432v. Minimum voltage that I'm seeing during idle is 1.048v.

The minimum idle core temperatures, so far, in Celsius, are 27, 27, 26, 29. And the max core temps have been 62, 73, 72, 70.

I did get one BSOD, while testing the CPU in Prime 95 with FFT 1344, after about 10 minutes of successful test passes. However, I have had no issues outside of that test, so I might leave the voltage offset as it is, right now, and if I ever get a BSOD during regular usage, then maybe I'll bump the VO up to +0.065v.

But like I said about my 2500K 4.6 Ghz overclock, that overclock (negative voltage offset, max voltage peak of something like 1.336v) didn't last 30 seconds in Prime 95, but I ran that OC for a year or maybe longer, and it never gave me a single BSOD in regular PC usage, which included gaming, heavy audio processing, playing videos, and everything else. I thought 1.336v was a little high for comfort when I did that OC, so I didn't try to stabilize it with further testing. However, I've learned a lot about proper overclocking this time around, and now I wonder if I could get a stable 4.8 Ghz from the 2500K.
 
#25 ·
When you say "100%" LLC, do you mean that you're using the 'Extreme' setting? The 'Extrme' setting should really be avoided because it makes the Voltage Regulator Module hotter which can kill stability unless you are using special cooling for it. At the highest with the stock cooler on the VRM, I'd avoid going above Ultra High. With Offset Mode, the 'High' setting is the highest you should really set it to. Offset Mode isn't any better than Manual, and I am certain I explained why. I think I even explained that Manual is better.

You're not properly testing your system's stability. Doing 10 minutes of ANY Prime95 run is nothing, as I'm sure I explained already. It's the same as not testing your stability. Just waiting for a BSOD one day could result in corrupt files, and yes there's a risk of very critical files becoming corrupt or even irreplaceable files becoming corrupt. It can happen while it's being accessed. When you're overclocking the CPU this high, it's absolutely critical to properly test your system's stability, and if memory serves me correctly, I already discussed at length and in detail how to properly test stability and why.

So, I hate to tell you this, but you're not really learning proper overclocking, you're trying to take shortcuts and with overclocking, shortcuts don't exist. If you take shortcuts, then you do so at the risk of corrupting your data or your entire OS. Also, the less stable your system is, the harder it is in your CPU and your PSU and your motherboard.

Look at the voltage your CPU requires just for 4.7 GHz. Is it worth it? Consider the reason why you think it's better to use Offset, and then test to see what voltage 4.5 GHz would need. It's a difference of 200Mhz, but I'll bet you that the voltage requirement would drop quite considerably.

Again, you need to do a proper test of stability. These little 10-minute runs aren't showing you anything. Yes, I know the 2500K couldn't last 30 seconds and all that, but that doesn't mean that 10 minutes is a very good test or anything. We can't judge the effectiveness of Prime95 or IBT by the length of time it runs. I am pretty sure I explained why too.

Speaking of IBT, I recommend to you the same thing that was recommended to me when I was trained in on overclocking Sandy Bridge: don't use it. Even if you have Linpack that uses AVX, it can be harmful due to the high temps. Just stay with Prime95's Blend and Custom Blend tests, as I believe I described before.
 
#26 ·
I had to run 100% LLC with the manual voltage, in order for the overclock to be stable in testing. I tried 75% first, and it didn't work. Mind you, I also might have bumped the manual voltage up a tiny bit when I also set LLC to 100%, so maybe the 100% LLC wouldn't have been needed at the higher voltage.

With my current voltage offset configuration, I have LLC at 75%, which, again, seems to be where it needs to be (though I will try a higher VO, with 50% LLC, in some further testing).

I did read all your explanations on the various points you raise, in the posts you made them in. I am aware that 10 minutes of testing is not thorough or complete testing. However, an OC passing 10 - 20 minutes is better than it not passing those 10 - 20 minutes. I'll be doing further testing, for sure, though not necessarily by running a full Prime 95 FFT test (though I might also do that). Other testing that I do, is letting my PC run idle for long periods of time, like overnight, for some days, to find out if it will randomly BSOD. Another form of testing is playing my most OC-sensitive games, over a period of weeks, to see if they ever encounter BSODs.

Quote:
Look at the voltage your CPU requires just for 4.7 GHz. Is it worth it? Consider the reason why you think it's better to use Offset, and then test to see what voltage 4.5 GHz would need. It's a difference of 200Mhz, but I'll bet you that the voltage requirement would drop quite considerably.
Yes, I notice that the voltage needed for the overclock I want is a bit high. And the reason I want to use VO is to not have the CPU running at a high voltage constantly. With VO, the CPU voltage is probably above 1.4v for less than 5% of its total operational time, and probably is above 1.440v for less than 0.5% of its total operational time.

I read $ilent's account of his CPU degradation, here:

http://www.overclock.net/t/1100100/info-intel-2500k-2600k-overclocking-tips
Quote:
Update: As of August 2012, after 9 months of overclock at 4.9ghz my cpu initially required only 1.41v. It now needs 1.45v to remain stable. So thats an extra 0.04v my cpu needs. Please bear in mind I used to fold 24/7, so I estimate my cpu has been at 100% full load for about 7 months at temperatures between early 70s to just under 80C.
After minutes under 100% Prime 95 load, my core temperatures are still under 70C for two of the cores, and the other two max out around 73C. In games, the cores don't reach those temperatures. However, this is why I asked in an earlier post, "Is heat, or voltage more responsible for wear on a CPU?"

If the voltage is more responsible for degradation, regardless of heat, then I might be more concerned. If heat, because of voltage, is primarily responsible for degradation, then I'm less concerned about having the maximum peak voltages that I do, right now.

If $ilent's 2600K CPU withstood 7 months of 72 - 79C temperature and at 1.41v, then I think my 2600K, at 30 - 65C, and under 1.35v for 80+% of the time, and over 1.4v for less than 5% of the time, can withstand many times the time-period that $ilent's CPU did, before notable degradation was a thing.

I bought this 2600K for $180 CAD, from someone who said it never saw serious stress, and that they didn't really overclock it (they told me they put a 200 Mhz OC on it, just to try it out). And my goal for this CPU is to not have to upgrade my PC for another 1 - 3 years, and so I'm really only concerned about its ability to survive that period, while delivering good performance. Each addition 100 Mhz on the core speed brings my Unigine Valley minimum FPS up around 2 FPS:

From 4.6 - 4.7 Ghz, 38 to 40 FPS.

I want to run this 2600K at 4.7 Ghz, because my 2500K was running at 4.6 Ghz, and want to know that I'm at least getting the same performance from this CPU as I was from the previous - and since, in some particular games, the 2600K slightly under-performs the 2500K at equal clock-speeds, I want to add a little Mhz to the 2600K OC, so that it's at least performing equally to the 2500K in everything, and out-performing it in many things.

For $180 CAD, if I get 3 years out of it (which is likely more than I'll need to), then I'm satisfied. But, with the voltage offset that I'm currently running, and with the loads the CPU is being subjected to, the CPU will likely last 10+ years.

I'm not done tweaking and testing just yet, though.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top