Blue Screen even with Slight OC i5 3570k

[BonzaiDuck]

On the topic of LLC (and as a threadjack) I found that although I had less vdroop with LLC (and thus could use less additional turbo voltage) my actual lowest stable full load voltage as reported by CPU-Z was lower with LLC disabled.

My observations: my crashes seemed to happen when switching frequencies or changing loads, ie times when voltage and frequencies were changing, not when voltage was at its absolute lowest (full load). With LLC disabled, because I was using a higher offset, the voltage was higher in periods where frequency/voltage were changing and it prevented crashes, and when the processor settled down at full load the voltage would droop and overall power consumption was lower than if I had LLC enabled.

Your mileage my vary, of course.

[Should be obvious but ] I'm running i7-2600K. In my case, second priority is power-consumption, first priority is minimize low/no load or idle voltage in "Turbo."

I give myself this less-green luxury, because I reduced the power-consumption for "Mah Main Workstation" migrating from Q6600-B3 @ 3.0Ghz and ~1TB across four HDDs in RAID 5. I'm pretty sure it was a reduction of 100W for draw at idle without sleep or hibernation.

 

[Yuriman]

Well you generally wont be at turbo frequencies unless you have some kind of load, in which case having LLC disabled helps (in my case). If you use +turbo voltage (ie with Inte's utility) instead of using +offset, you don't increase your idle voltages either, only those voltages used at turbo frequencies.

For me it's a win-win, as having LLC off raises your voltages when changing frequencies and lowers them at load, without hurting them at idle.

 

[jidery]

Turn Internal PLL Overvoltage on. Set your BCLK to 100. Set your RAM speed to automatic. Work from there.

Followed this exactly as stated, managed to get the system stable up to a 43 multiplier. Crashes about 3 min into the 44 multiplier stress test. Still getting the x0124 error on blue screen.

 

[Yuriman]

Add 100mv (0.1v) and see how much more you can get.

 

[Puppies04]

Also if you want OC higher then you can try taking it off manual voltage and using offset mode with a little +offset (start at +0.05 and work your way up in 0.05v increments) and add a little LLC if your voltage is dropping too low under stress (medium should be fine)

Too everybody jumping on their high horse about the fact the OP won't need LLC at 4.2 after I mentioned it, can I direct you to the bolded part of my previous post.

 

[BonzaiDuck]

Add 100mv (0.1v) and see how much more you can get.

If he got it to boot and then run a stress-test for 3 minutes, a couple millivolts should extend the next stress-run before it crashes. I'd prefer a more incremental approach, but maybe the brief application of 0.1V won't hurt it -- he needs to find his highest multiplier. I just tend to approach that sort of thing in small steps. . . . And also -- he may just want to run the fixed VCORE or "offset/turboV" up just enough to reach 1.30V -- perhaps a sane limit for 24/7.

Puppies is correct. Better to get started sooner with "Offset" mode, which you can call it even if he keeps the Offset constant and changes the "Extra Turbo Voltage."

I never bothered with "Manual" Vcore when tuning my Sandy. I'd read enough "guides" before I began so I could see what had happened to our old-school approach, and I quickly discovered that the same precision awaits you when adjusting the Offset and turboV.

 

[jidery]

Looks like we are going backwards, I adjusted my voltage up slightly, (100mv) and now I got an blue screen immediately on the stress test at the 44 multiplier. I changed the multiplier down to 40 and the voltage back to normal and left it for the night, figured I would work on over-clocking later today. So with the multiplier only at 40, in which I had thought to be stable, I got a blue screen playing portal 2 (not even a stress test!) This is at the 40 multiplier, only 2 above the base of 38.

 

[Yuriman]

Very strange.

What voltage does CPU-Z report when you put your CPU under full load (looking for ballpark numbers)? What power supply do you have? I suspect something else is going on beyond just having a bum chip.

 

[jidery]

Very strange.

What voltage does CPU-Z report when you put your CPU under full load (looking for ballpark numbers)? What power supply do you have? I suspect something else is going on beyond just having a bum chip.

I'll report the voltage readings once I get home later tonight, but the PSU is a CoolerMaster Silent Pro 800w Gold.

 

[jidery]

V core idle 1.074
V core with Prime 95, 1.118

 

[daveybrat]

The one thing i haven't seen you say is what heatsink/fan are you using? The stock one or something better?

 

[jidery]

Cooler master Hyper 620

 

[Termie]

V core idle 1.074
V core with Prime 95, 1.118

Cooler master Hyper 620

Followed this exactly as stated, managed to get the system stable up to a 43 multiplier. Crashes about 3 min into the 44 multiplier stress test. Still getting the x0124 error on blue screen.

So, I'm not entirely sure what the issue is that we're trying to solve at this point. It appears you were able to get 43x stable, which is really all you can expect to achieve on stock voltage, which is what you are running at, given the vcore stated above.

Your cooler is probably capable of more, but your bluescreens have absolutely nothing to do with heat at this point. You are just trying to set an overclock that's too high for the voltage provided. That's a pretty straightforward problem when it comes to overclocking.

So keep it at 43x and be happy, or prepare to do a lot more trial-and-error, including numerous BSODs, on your way to a higher multiplier. If you are indeed unstable at 43x, as you suggest above, please clear the CMOS for your BIOS (pull the lithium battery out of the MB), uninstall the ASUS utilities, and start over. Your first step will be to leave EVERYTHING alone except for your multiplier, and go up ONE step at a time until you reach the limit of stability at stock voltage.

 

[BonzaiDuck]

So, I'm not entirely sure what the issue is that we're trying to solve at this point. It appears you were able to get 43x stable, which is really all you can expect to achieve on stock voltage, which is what you are running at, given the vcore stated above.

Your cooler is probably capable of more, but your bluescreens have absolutely nothing to do with heat at this point. You are just trying to set an overclock that's too high for the voltage provided. That's a pretty straightforward problem when it comes to overclocking.

So keep it at 43x and be happy, or prepare to do a lot more trial-and-error, including numerous BSODs, on your way to a higher multiplier.

That about says it in a nutshell. Even if I'd wonder why his idle V is lower than the load V (and he may have EIST enabled, but I've already said I'm not familiar with the IB CPU to get an idea what the idle would be under that circumstance) -- he should have no problem twisting it up to maybe 1.29V just to see how far it will go within a reasonable voltage objective.

The OC Tuner feature in BIOS would have got him to the 103 bCLK setting, but it would have also volted the processor to reach the "auto-OC" speed, and he might have then dropped the bCLK back to 100. Then, his VCORE would've been excessive, but he might then have started tweaking the voltage down while pushing the mutiplier up in increments.

When I did that as a preliminary action with my Sandy CPU, it actually pushed the voltage up to 1.44V for a speed around 4.4 Ghz. My value for the most severe load @ 4.6 Ghz now is around 1.31 to 1.32V. Of course, that's the Sandy with Sandy voltages.

Also, rechecking OP's BIOS screenies, he still has Spread Spectrum set on Auto. That's always been the rule of thumb -- turn it off or disable it. Regardless of the other energy-saving features -- EIST, C1E, etc. -- I always leave it disabled. Having always followed the rule-of-thumb, I wouldn't know how it might affect his stability, but . . . it's another one of "those things" as I called them earlier in my big-print bold-face post.

The only reservation I might have on what I've said here is his choice of a cooler, but you'd think it would be perfectly adequate . . .

 

[Termie]

That about says it in a nutshell. Even if I'd wonder why his idle V is lower than the load V (and he may have EIST enabled, but I've already said I'm not familiar with the IB CPU to get an idea what the idle would be under that circumstance) -- he should have no problem twisting it up to maybe 1.29V just to see how far it will go within a reasonable voltage objective.

...

Actually, his idle voltage of 1.07 suggests that he is not running at stock settings. It should be closer to 0.85. I'm not sure how it ended up at 1.07, but it could be that an energy-saving option is turned off, or that he's using a manual voltage setting.

Either way, the only way to get to the bottom of this is to start over, and that's what he'll be doing if he clears his CMOS.

 

[BonzaiDuck]

Actually, his idle voltage of 1.07 suggests that he is not running at stock settings. It should be closer to 0.85. I'm not sure how it ended up at 1.07, but it could be that an energy-saving option is turned off, or that he's using a manual voltage setting.

Either way, the only way to get to the bottom of this is to start over, and that's what he'll be doing if he clears his CMOS.

No disagreement there . . . .