Why won't my 2500K go past 4.5GHz?

[iiLooNiE]

I just got my new Phantek PH-TC14PE and I am wanting to overclock my 2500K past 4.5GHz. But everytime I try to go further my PC won't post. I have tried 4.8GHz and 5GHz and given it more voltage, probably much more than it needs, and it still won't post. The only way for me to fix it to get it to post again is by clearing the CMOS. Here are the setting that I have, and also the motherboard.

ASRock Extreme3 Gen3 Z68

All power saving features are off, 45x multiplier, 1.270 Vcore (set in offset mode with the voltage offset +10), Spread Spectrum off, C3 and C6 is on, Short Duration Maintained 230, Long Duration Maintained 200, and something else is 230. I forgot exactly what it's called.

EDIT: The V droop compensation is set to Level 5 (the Highest)

Does anyone know why my computer is doing this? I really only want to go to 4.8GHz or 4.9GHz.

 

[Janus67]

You may need to turn on CPU pull overvoltage

So a 46x multi freezes? What's the max multi you can hit with 1.35v. Try setting the vcore manually instead of using offsets

 

[iiLooNiE]

Anything past 4.5GHz doesn't allow me to post. And when I was going further with it, I wasn't using offset mode. I was using Fixed.

 

[gregmacknass]

Try raising PLL voltage some. On my 2500k, I had to raise several of the voltages to get it to work. Don't use offset voltages, set the voltage manually. Turn of EIST and C3 C6. Disable turbo mode. These steps will get you past 4.5GHz. When shooting for max clocks, most of the time you will have to disable any power saving options.

 

[RJARRRPCGP]

I just got my new Phantek PH-TC14PE and I am wanting to overclock my 2500K past 4.5GHz. But everytime I try to go further my PC won't post. I have tried 4.8GHz and 5GHz and given it more voltage, probably much more than it needs, and it still won't post. The only way for me to fix it to get it to post again is by clearing the CMOS.

Sounds like a bad OC'er (bad batch) or DRAM clock too high.

And with a bad batch, it's possible that it will get to the desktop once then Windows gives error messages and never makes it to the desktop again at that clock.

My dreaded T'bred 2000+ AIUCB did that kind of bull. I got 2.1 once, but never again!

 

[iiLooNiE]

Try raising PLL voltage some. On my 2500k, I had to raise several of the voltages to get it to work. Don't use offset voltages, set the voltage manually. Turn of EIST and C3 C6. Disable turbo mode. These steps will get you past 4.5GHz. When shooting for max clocks, most of the time you will have to disable any power saving options.

That actually got me to post and get into Windows! Thanks! I have it set to 1.380 Vcore right now at 4.8GHz. I am about to do some stress testing. Thanks again. I disabled EC1E, C3, and C6. Upped the PLL voltage from 1.832 to 1.849 and it worked.

 

[Janus67]

I'd still try enabling CPU pll overvoltage to see if it makes a difference

 

[iiLooNiE]

When I started to stress test it instantly froze my computer. So I tryed a couple other things, got a couple BSOD's, but now I think I have it. I went back and re-enabled EC1E, and I have the voltage set in the BIOS to 1.450 and the PLL voltage set to 1.857, but the Vdroop setting that I use keeps it around 1.336 when stress testing. That is the max that I have seen with just a couple minutes of Prime95. What doesn't make sense to me is that the idle Vcore is INSANE. When running at 1.6GHz just idling the Vcore is at 1.400. And it drops when I stress test.

Also, I was stress testing while typing this, and core 4 has always been the douche for me. Even when I was overclocking to 4.5. I forgot exactly what I did to fix it. Lol. I am going to try some other things.

 

[GeneO]

When I started to stress test it instantly froze my computer. So I tryed a couple other things, got a couple BSOD's, but now I think I have it. I went back and re-enabled EC1E, and I have the voltage set in the BIOS to 1.450 and the PLL voltage set to 1.857, but the Vdroop setting that I use keeps it around 1.336 when stress testing. That is the max that I have seen with just a couple minutes of Prime95. What doesn't make sense to me is that the idle Vcore is INSANE. When running at 1.6GHz just idling the Vcore is at 1.400. And it drops when I stress test.

Also, I was stress testing while typing this, and core 4 has always been the douche for me. Even when I was overclocking to 4.5. I forgot exactly what I did to fix it. Lol. I am going to try some other things.

You are seeing the droop in voltage under load - the voltage will increase when not under load with a fixed voltage setting. Put it back to offset and the voltage will be lowered with the frequency when the load is low. You need offset for the voltage to drop with the frequency.

 

[iiLooNiE]

You are seeing the droop in voltage under load - the voltage will increase when not under load with a fixed voltage setting. Put it back to offset and the voltage will be lowered with the frequency when the load is low. You need offset for the voltage to drop with the frequency.

Is it better to have it in Fixed mode, or is it better to have it in Offset mode?

 

[djscrew]

Is it better to have it in Fixed mode, or is it better to have it in Offset mode?

seems to depend on a variety of factors but you just need to find whichever way works for you... generally offset is required for higher OCs but the advise you are given was good... your chip might not be the best OCer

are you OCing your ram too?

 

[Janus67]

that is a HUGE amount of vdroop. Are you sure that level 5 is the highest? I think with ASRock boards with LLC 5 is actually the lowest and 1 is the highest (I haven't messed with one but others have said similar things).

 

[manu2b]

It's the other way around with Asrock boards: Level 5 is the lowest level.

If you set it to level 2, you will likely have no vdroop.

EDIT: as Janus stated.

 

[iiLooNiE]

I just figured that I would do a little update. I go it stable at 4.9GHz. Stress tested with Prime for 3 hours, no errors. Some people say that you need to do it for 12, but to me, if nothing happens within the 3 hour period, then nothing is going to happen later on.

Max temperatures are 73C. Which isn't that bad because I am on air. Thank lord for the Phantek PH-TC14PE.

Here are the settings that I am using.

C1E, C3 and C6 enabled. Offset mode is being used with it set to +145. It hits 1.4 Vcore when stress testing. At one point for a split second it jumped to 1.41, but no biggie. Long duration maintained is 300, short duration maintained is 270, and something else is set to 300 (Forgot the name of it, again). PLL is set to 1.849. Vdroop is still at Level 5. Solid as a rock.

I know this isn't the place for RAM questions. But after I got this stable I was messing around with my RAM timings. I have G.SKILL RipJaws X Series 2133MHz RAM, and I was wondering what is better. It running at 2133MHz with 9-11-10-28 2T timings, or it running at 1866MHz with 8-10-9-26 1T. Both of those settings are with the stock voltage of 1.642.

 

[djscrew]

I just figured that I would do a little update. I go it stable at 4.9GHz. Stress tested with Prime for 3 hours, no errors. Some people say that you need to do it for 12, but to me, if nothing happens within the 3 hour period, then nothing is going to happen later on.

Max temperatures are 73C. Which isn't that bad because I am on air. Thank lord for the Phantek PH-TC14PE.

Here are the settings that I am using.

C1E, C3 and C6 enabled. Offset mode is being used with it set to +145. It hits 1.4 Vcore when stress testing. At one point for a split second it jumped to 1.41, but no biggie. Long duration maintained is 300, short duration maintained is 270, and something else is set to 300 (Forgot the name of it, again). PLL is set to 1.849. Vdroop is still at Level 5. Solid as a rock.

I know this isn't the place for RAM questions. But after I got this stable I was messing around with my RAM timings. I have G.SKILL RipJaws X Series 2133MHz RAM, and I was wondering what is better. It running at 2133MHz with 9-11-10-28 2T timings, or it running at 1866MHz with 8-10-9-26 1T. Both of those settings are with the stock voltage of 1.642.

Unless this is a mission critical server, you probably wouldn't ever need to stress test more then an hour or two really.

How about 2133 9-11-10-28-1T instead ;P

I would go with the higher clock. You might be able to OC the ram too. Try tightening the timings up at 2133 and see what happens. Start with 1T.

That being said, that is some extremly high performance ram. You are not likely to notice any difference with better timings or overclock.

 

[iiLooNiE]

Unless this is a mission critical server, you probably wouldn't ever need to stress test more then an hour or two really.

How about 2133 9-11-10-28-1T instead ;P

I would go with the higher clock. You might be able to OC the ram too. Try tightening the timings up at 2133 and see what happens. Start with 1T.

That being said, that is some extremly high performance ram. You are not likely to notice any difference with better timings or overclock.

Well I am running it a the 8-10-9-26 1T timings right now, and it does seem a tad bit snappier when opening programs. Reason I am asking, is because I have heard that the difference between 2133Mhz and 1866MHz is about 1%. So I was wondering if would be faster at 1866 with tighter timings. Also, when I set it to 1T at 2133Mhz is doesn't want to seem to get stable.

 

[gregmacknass]

Well I am running it a the 8-10-9-26 1T timings right now, and it does seem a tad bit snappier when opening programs. Reason I am asking, is because I have heard that the difference between 2133Mhz and 1866MHz is about 1%. So I was wondering if would be faster at 1866 with tighter timings. Also, when I set it to 1T at 2133Mhz is doesn't want to seem to get stable.

Outside of benchmarks you really won't see any difference in performance with higher speeds or tighter timings. Just run it at it's best 100% stable speed and timings, and be done with it.

 

[wingman99]

+1 you won't notice any faster speeds with faster ram and tighter timings.

 

[Janus67]

I just figured that I would do a little update. I go it stable at 4.9GHz. Stress tested with Prime for 3 hours, no errors. Some people say that you need to do it for 12, but to me, if nothing happens within the 3 hour period, then nothing is going to happen later on.

Max temperatures are 73C. Which isn't that bad because I am on air. Thank lord for the Phantek PH-TC14PE.

Here are the settings that I am using.

C1E, C3 and C6 enabled. Offset mode is being used with it set to +145. It hits 1.4 Vcore when stress testing. At one point for a split second it jumped to 1.41, but no biggie. Long duration maintained is 300, short duration maintained is 270, and something else is set to 300 (Forgot the name of it, again). PLL is set to 1.849. Vdroop is still at Level 5. Solid as a rock.

I know this isn't the place for RAM questions. But after I got this stable I was messing around with my RAM timings. I have G.SKILL RipJaws X Series 2133MHz RAM, and I was wondering what is better. It running at 2133MHz with 9-11-10-28 2T timings, or it running at 1866MHz with 8-10-9-26 1T. Both of those settings are with the stock voltage of 1.642.

I'm glad to see that you got it up and stable. Where is the voltage when you are at idle?

As I stated previously LLC at level 5 actually will cause more vdroop than level 1 (no idea why ASRock would number it that way). So if you run LLC at 1 your voltage under load will go up, as opposed to go down (which I prefer as I figure under load it will need more voltage than at idle)

 

[GunzRX]

Outside of benchmarks you really won't see any difference in performance with higher speeds or tighter timings. Just run it at it's best 100% stable speed and timings, and be done with it.

I havn't tried with i5 so remember that, but when I had a e6600 at 3.6 ghz speeding up ram would give 1-3%, not much but we're the kind who re-route a loop for 2deg off a gpu.

+1 you won't notice any faster speeds with faster ram and tighter timings.

Above, and for benches btw my exp. has been that mhz esp. with ddr3 gives more bandwidth in sandra than timings. latency matters I guess, you don't want it too high but you'll notice if it is.

Command rate is what mattes mostly I guess, but an ssd will make everything a-ok anyway.