Novice (trying) to overclock E6420 @(stock) 2.13 GHz

[courteous]

The relevant pieces (all stock):

• C2D E6420 @ 2.13 GHz
• Asus P5K with latest BIOS (supports FSB @ 1333 MHz and RAM @ 1066 MHz)
• Patriot DDR2 800

In BIOS I've raised FSB from 266 to 400, which should give 3.2 GHz (which is what it says in POST message). But ... when I get into OS, I see that the multiplier was lowered (from 8x to 6x), giving only 6x 400 = 2,4 GHz. What can be done about this multiplier lowering?

What I'm more worried about, is that the (generic) Patriot DDR2-800 RAM was (automatically) overclocked from 400 MHz to 500 MHz! I know, because of FSB, but is this too much?

BTW, the CPU temperature was raised from 38° to about 40°C (using HWiNFO software). Now, which software would allow you to also see the RAM temperatures?

Any other advice much welcome ... I'm not looking for an extreme overclock, just about say 3 GHz ... and with not melting the RAM.


EDIT: Would it be best to raise just the CPU clock frequency, while leaving FSB alone (or at least at max. 333 MHz) for the sake of RAM?

 

[Dangman]

That multiplier lowering should be Intel's Speedstep feature working. Intel CPUs will downclock when not heavily loaded. Use Prime95 or LinX or Intel Burn Tool to load the CPU which will therefore load the multi again. Also note that the above three programs should have been run to begin with to make sure that the OC is stable.

As for the RAM, an extra 100Mhz might hurt certain low-end RAM. There should be an option that sets the RAM frequency as the same as the FSB frequency. That option may be called System Memory Multiplier or something similar to it. So if you raise the FSB to 400MHZ, the RAM will be kept at 400Mhz.

You cannot raise the CPU clock while leaving the FSB alone unless you have an extreme edition C2D based CPU. So the only way to OC your CPU is by FSB increasing.

 

[courteous]

As for the RAM, an extra 100Mhz might hurt certain low-end RAM. There should be an option that sets the RAM frequency as the same as the FSB frequency. That option may be called System Memory Multiplier or something similar to it. So if you raise the FSB to 400MHZ, the RAM will be kept at 400Mhz.

Danny, thank you! I'll try to find this option. Also, should anyone else have the P5K motherboard, please tell me.

EDIT: Got it ... RAM now fixed at @ 400 MHz. Next, I ran the Prime95 (simplest) test for 10 minutes and the CPU core temperature rose to 66°C (as reported by HWiNFO software). Now, is this considered "stable" overclock? (Idle temperatures are back at 40°C.)

BTW, the where does the extra energy come from? Namely, this was @ 2.13 GHz

8 x 266.7 MHz @ 1.2875 V

and after @ 3.2 GHz

8 x 400.4 MHz @ 1.2875 V

The voltage stayed the same ... so, where is an extra load?

 

[Talonz]

Danny, thank you! I'll try to find this option. Also, should anyone else have the P5K motherboard, please tell me.

EDIT: Got it ... RAM now fixed at @ 400 MHz. Next, I ran the Prime95 (simplest) test for 10 minutes and the CPU core temperature rose to 66°C (as reported by HWiNFO software). Now, is this considered "stable" overclock? (Idle temperatures are back at 40°C.)

BTW, the where does the extra energy come from? Namely, this was @ 2.13 GHz and after @ 3.2 GHz The voltage stayed the same ... so, where is an extra load?

If I remember correctly, heat dissipated is a linear function of frequency and an exponential function of voltage (assuming the same workload).

 

[courteous]

Talonz, I was asking about power consumption. Is it increased through larger overall current (because of higher frequency), as the voltage seemed to stay the same?

 

[Dangman]

EDIT: Got it ... RAM now fixed at @ 400 MHz. Next, I ran the Prime95 (simplest) test for 10 minutes and the CPU core temperature rose to 66°C (as reported by HWiNFO software). Now, is this considered "stable" overclock? (Idle temperatures are back at 40°C.)

Not really. Ten minutes is far far too short of a testing period with Prime 95. A several hour run (I personally aim for 18 to 24 when I can) of Prime 95 is usually recommended. With that said, running Intel Burn Tool as the load tool does take significantly less time but puts an extremely high level of stress on the CPU.

However the load temp of 66C is a tad high. You should be thinking about getting a better 3rd party HSF or increase the cooling to the CPU.

BTW, the where does the extra energy come from? Namely, this was @ 2.13 GHz and after @ 3.2 GHz The voltage stayed the same ... so, where is an extra load?

Talonz, I was asking about power consumption. Is it increased through larger overall current (because of higher frequency), as the voltage seemed to stay the same?[/QUOTE]

The power consumption increase as the result of both voltage increases and higher frequency. So yes, even if you keep the same voltage, you'll still see an increase in power consumption as well as heat and vice versa.

 

[Silent Knight]

Next, I ran the Prime95 (simplest) test for 10 minutes and the CPU core temperature rose to 66°C (as reported by HWiNFO software). Now, is this considered "stable" overclock? (Idle temperatures are back at 40°C.)

No.
For quick test,you should use Linx or IBT instead of Prime.They'd detect errors and unstability quicker.

 

[courteous]

However the load temp of 66C is a tad high. You should be thinking about getting a better 3rd party HSF or increase the cooling to the CPU.

I have Scythe NINJA ... but I also had Antec's fans on LOW mode. Should I put the fans on MEDIUM at least (or can they stay on LOW)? EDIT: For reference, the temperature in the room is ~24°C ... and there's an 8800 GTS inside.

No.
For quick test,you should use Linx or IBT instead of Prime.They'd detect errors and unstability quicker.

I'll go ahead with IBT ... hope it won't fry the CPU. What's the (approximately) max. temperature for C2D? Around 80°C?

 

[courteous]

Just ran the IBT (log below), the max. temperature reached was 74°C ... is this too high and should I go lower @ 2.6 GHz? On the other hand, there aren't many such processes like IBT (not even games) that put such stress on a CPU ... right?

BTW, is this CPU temperature (with room temp. of 24°C) high enough to imply a problem with the thermal paste* (beneath the Ninja heatsink)?

I'll also replace (1) the 8800 GTS (idle temp. of 50°C ) with a passive HD 5450 ... and (2) the mechanical disk (temp. of 40°C) by an SSD. Will this make an appreciable difference?

----------------------------
IntelBurnTest v2.51
Created by AgentGOD
----------------------------

Processor: Intel(R) Core(TM)2 CPU 6420 @ 2.13GHz
Clock Speed: 0,67 GHz
Active Physical Cores: 2
Total System Memory: 2047 MB

Stress Level: Standard (1024 MB)
Testing started on 12.6.2011 12:56:06
Time (s) Speed (GFlops) Result
[12:57:06] 50.336 17.7584 3.677164e-002
[12:58:06] 50.277 17.7790 3.677164e-002
[12:59:05] 50.357 17.7508 3.677164e-002
[13:00:04] 50.223 17.7983 3.677164e-002
[13:01:04] 51.007 17.5248 3.677164e-002
Testing ended on 12.6.2011 13:01:05
Test Result: Success.
----------------------------

* This article only talks about the "burn-in period" of increasing the paste performance.

 

[Silent Knight]

74 degrees is still in the acceptable range.You should try to keep it under 80 degrees when running IBT and under 70 degrees for daily use.

And IBT/Linx pretty much push the CPU to the point where no daily apps can.

 

[courteous]

Previously, the BIOS had "Auto" on the core voltage ("V_core") and the CPU-Z showed it to be ~1.36 V.

Now, I set it manually in BIOS to 1.325 V and got 3° to 4° lower temperatures ... but the CPU-Z now shows core voltage as 1.256 V. Why the difference?

BTW, what else would you do to lower the temperatures (and retain the 3.2 GHz frequency)? Also, this V_core of 1.325 V is not stable for, say, IBT stress-testing ... but for "normal" usage it seems good enough. Is this dangerous thinking?

EDIT: Artoxian at whirlpool forums reported of achieving the same 3.2 GHz @ 1.1 V?! Mine E6420 won't even boot @ 1.25 V, much less @ 1.1 V! What's the secret?

 

[Silent Knight]

Now, I set it manually in BIOS to 1.325 V and got 3° to 4° lower temperatures ... but the CPU-Z now shows core voltage as 1.256 V. Why the difference?

It's vdroop.You can reduce the amount of dropping by enabling Load-line Calibration.

BTW, what else would you do to lower the temperatures (and retain the 3.2 GHz frequency)? Also, this V_core of 1.325 V is not stable for, say, IBT stress-testing ... but for "normal" usage it seems good enough. Is this dangerous thinking?

Get a better HSF or lower the vcore.

I think you should get it stable on IBT.Hard to explain,but stable on IBT is better than unstable on it and "might" stable on normal use.

 

[courteous]

Get a better HSF or lower the vcore.

Scythe Ninja should be OK? So, how do I lower Vcore, if even at 1.325 V (set in BIOS) the system isn't stable (at least regarding the stress-testing, as in IBT, Prime95, etc.)?

On the other, some make it stable @ 1.1V!

 

[Silent Knight]

That guy got a E6600 which has higher multiplier than your E6420.Also,he's lucky to have a CPU with low vcore like that.

I'm not a big fan of Scythe HSF,i'd look into Thermalright or Prolimatech if i was you.

So, how do I lower Vcore, if even at 1.325 V (set in BIOS) the system isn't stable (at least regarding the stress-testing, as in IBT, Prime95, etc.)?

Seems like you can't get it stable with any lower vcore.You should bump it a little more to get it stable.
Lowering vcore only works if you feel it has too much vcore for a certain frequency.

 

[courteous]

You're right, I misread (twice).
So the best option would be to set it back to "Auto" (in BIOS) which set it @~1.36 V?

 

[Silent Knight]

You should set it manually,so the vcore wouldn't jump up all over the place.

Try 1.38 to 1.4 and see if it's stable or not,then either bump up the FSB or lower the vcore.

 

[courteous]

The very first increment to 1.35 V got me a stable (under IBT's "Standard" stress level) CPU. The temperature went to 66°C only for a moment, otherwise the max. was at 65°C ... much better than 74°C. And idle temperature is now 35°C with 25°C room temp.

This went better than I thought for my first time ... of course, only with your help: thank you Mr. Silent Knight, Danny and Talonz!