I recently tried to overclock my Intel Pentium E2220 2.4 GHz (prethe viously I had no trouble overclocking my Athlon 64) - however when I tried recently to raise it to 2.6 GHz, it crashes pretty quickly when running Orthos. Temps seem okay, voltage is set at 1.325v, memory divider is set to auto (1:1) - RAM is Corsair XMS2 6400 (800). I tried raising the voltage to 1.35v and setting the memory divider to 2.0. However when I tried this, after POSTing nothing happens. I restarted a few times and sometimes it didn't even post. Sometimes I could get into the BIOS menu but when it would freeze. Eventually I got lucky and managed to get into the BIOS long enough to reset the settings and now it's fine. I was wondering if anyone could shed some light on this problem? I'm pretty sure a 200 MHz overclock isn't massive for this CPU but I don't want to risk changing settings again in case something goes wrong again and I can't get it to boot. Thanks for any advice.
Which mobo are you using? Is 1.325V the VID of the chip? How about your memory voltage - have you left that to auto, or not? I'm trying to think of any more questions to ask, but I'm drawing a blank atm. In all honesty, though, the chip should handle 200MHz overclock easily, and I'd expect it to do so at stock speeds. However, your mileage may vary. Overclocking's not an exact science.
Ah sorry I should have posted a bit more info in hindsight. Voltages are all set to manual but I have left them all at default except the Vcore which says it's default is 1.325v. Other than that, PCI frequency is set to 100 MHz, CPU multiplier at x12 (max). Motherboard is Gigabyte GA-P31-DSL (when I was upgrading I used the CPC recommended budget components at the time).
Beaut. So, assuming your power supply's not hideously crippled, we've got a few lines of enquiry here. 1) Cooling. The chip could be overheating under the Intel reference cooler. The one for Pentiums/Celerons is smaller than Core 2 coolers. If you're using after-market cooling, it's possible it may not be mounted properly. Check your temps. 2) The components. You may just have a stinker in the chip department (though I very much doubt it, tbh - chances are Intel wouldn't ship a processor at its limit, in case it failed quickly). What I'd try and do, is get the settings sorted to check the FSB first - drop the multiplier as low as it goes (think it's about 6, normally), then try upping the FSB slowly. I'd guess you'll be able to push it up (take it in steps first - start at 210, then go up in 10s till you hit a wall). If it doesn't want to go up much, something's wrong (though I'd tweak the voltages before abandoning hope). A bizarre thought, here - could there be a problem with your installation of Orthos? Try another stress-test, like Prime95. Not sure how it'd be like that, but always worth checking every factor.
Hmm thanks for the tips. PSU is Tagan 450W Whisper so it's reasonably well made (I assume). The CPU cooler is Arctic Cooler Freezer Pro, and temps idle about high 20's/30. I'll give Prime95 a shot. Do you think it's okay to leave the memory multiplier at 1:1 or best to make sure the RAM stays under 800 MHz (its DDR2 6400 800 MHz). One thing I did notice checking the CoreTemp log was that before it crashed, I noticed the multiplier dropped from x12 to x6 sometimes. Maybe this is just the intel power saving feature though.
Keep the RAM under 800MHz for now. Turn off speedstep etc in the BIOS before you overclock. It gets in the way & causes crashes.
I'm not sure I have speedstep (maybe something similar under a similar name though - I'll just turn off all the weird options like Turbo boost, etc.).
Don't have that option in my BIOS sadly. It's Gigabyte's M.I.T. interface. Will try lowering the multiplier and the memory divider and use Prime95 and see if that makes a difference. Thanks for the advice - stay tuned!
Okay I managed to run P95 small FFTs test for a few hours no problem at 2.5 GHz (10x250 FSB), max temps at about 44 degrees (Vcore at 1.3v). Running a test at 2.6 GHz now to see if this works. One additional question though - with my memory multiplier at 1:1 ratio (so at FSB = 250, DDR2 = 500 MHz), this means my memory is quite underclocked. So say I reach my limit before FSB = 400 (therefore my RAM would be operating at its optimal 800 MHz), how can I fiddle the numbers to get it closer to 800 MHz? Is it simply a matter of lowering the CPU multiplier and increasing the FSB? I have read somewhere that it is best to have the multiplier as high as possible so as not to stress the FSB. p.s. Just checked my RAM info and the "Tested Voltage" is 1.8v - is that the base or maximum?
It's a question of balancing multiplier and FSB. For example, let's say your processor will max out at 3.6GHz. You could run it at 400x9, but that might be too high for your motherboard to take. At 300x12, you'd have your RAM running at 600MHz, so again, not good. At 360x10, you could run your RAM at 720MHz. Might be enough to allow you to tighten the timings, if you're fussy. I still prefer running RAM as fast as possible, even if it means switching from 1:1, but that's my opinion. EDIT: Just a point I should reinforce - your motherboard will have a maxmimum FSB it can handle. If you want to find it, start overclocking with the multiplier set as low as possible.
Though again, it's not an exact science, what with FSB holes and such. Sometimes finding stable settings can be a matter of blind luck.
I Agree, I had to make many test before finding a balance. O/C = Needed more Wattage = Bigger PSU. Higher Frequency = higher temperature = a better Cooling system. I'm tryin Overclock with the Intel Stock Heatsink..... at 3Ghz my CPU was at 80 Degree Rofl. Maybe you can Update your Mobo Bios hummmm for advenced Option in the Bios CTRL + F1
Well I got it to work but I seem to have hit a barrier. I have set it to FSB = 333 and CPU Multiplier to x9 to get 3 GHz. The Vcore is 1.325v, however after running P95 for about 5 minutes, one core fails but the other keeps going. Is this likely to be a CPU limitation or voltage issue?
That'll be a voltage issue. 1.325 is the default, so you'll have to go up from there. Without a really good air cooler, or water, I'd stop before 1.45V. Course, depends on your case, mobo, etc.
Ah okay thanks. I just noticed, I set the Vcore to 1.35v now in the BIOS but CPU-Z is reading it as 1.28v. I have power saving options turned off and the 1.28v doesn't seem to fluctuate when the CPU is running at 100%. Is this something called Vdroop (I have only heard this, don't really have much understanding of it, or if this is even the cause - maybe CPU-Z is just dodgey at reporting voltages?)
That's actually known as vDrop. vDroop is the lowering of voltage when the CPU is under load, as shown here:
What was vDrop intended to do? Running the CPU at a higher frequency could benefit from an increase in voltage, not a decrease. So why was it implemented, besides lowering the power consumption (and thus heat dissipated) when under load? Logic dictates that if at it's maximum frequency it can run stable with a lower voltage, why not run at even lower voltages when not in "full throttle"? I don't have experience in overcloking, but in undervolting I do. The lower the multiplier SpeedStep uses, the lower the voltage my CPU requires in order to be stable.
Vdroop isnt a feature that was added by anyone. its simply a side effect of the CPU and electrical properties that i cant explain (im sure someone will do it for me). A feature that was added to combat this as its a bad thing on the whole is LOAD LINE CALIBRATION which pumps more voltage to the CPU to prevent it dipping below the voltage you set in the bios. so if your CPU is stable at say 1.3V but under load Vdroop causes it to dip down to 1.28V and your computer crashes turning on loadline calibration would boost the initial 1.3V to say +10% or whatever so it acutally sends 1.313V to the CPU and under load your CPU voltage now read 1.3V and remains stable. of course your could simply increase the voltage to compensate for vdroop which is what i do but you just have to be aware of the Vdroop then
Incorrect. It's a native feature used for all Intel processors to increase stability. You're looking at it from the wrong perspective, we're not talking about running the CPU at a higher frequency. Assume the CPU is at stock speeds and not overclocked, vDroop is a way of regulating the voltage spike when a CPU transitions from an idle phase to a load phase, and vice versa. This is what was shown in the image I posted above. By regulating or levelling out the voltage that is supplied to the CPU depending on the demand placed upon the CPU, it ensures the voltage does not exceed CPU's specified limits and helps to increase system stability.
