by congo » Mon Mar 05, 2007 10:22 pm
My CPU is 2.2ghz stock with an 11x multi.
If orthos is hanging at all @ 2.7ghz, then don't go higher until you can get stable at the present speed.
To get cpu max, you set your memory divider to DDR200 or DDR266 so that the ram is running so slow that memory will not top out or cause any error to confuse matters.
Lock the memory timings to your normal timings you use or to the SPD timings you normally use at the memory speed you will be using. This will ensure that the machine does not throw up some incompatible ram timings when you set the new low divider. This is just a precaution to prevent a poor bios from setting bad timings through changing the memory divider to such a low level, so we can avoid any possible memory conflicts. We want to purely test the CPU speed and it alone should throw errors.
Set the HT multiplier to 3x so the HTT speed is relaxed so that won't throw any errors as well.
Set Vcore to 1.4v (true) at first. My mainboard bios skews the Vcore upwards for example. If I set 1.375v, I get 1.4v true. The higher my Vcore goes, the higher it gets skewed, so be careful and monitor temps ALWAYS while Priming orthos.
Set all fan control to maximum CPU fan RPM, we are testing max CPU, so we want max cooling available.
Set yourself a max temp limit and end the test or any further Vcore increase if that limit is reached or exceeded, I set a personal limit of 58*C and I will reduce my overclock settings so my CPU never goes over 55*C at any time in usage. But that's just my preference so I don't overcook the poor little thing.
Now you can run the FSB up a bit at a time until the CPU fails ten minutes of small FFT testing in Orthos. Do that at 1.4Vcore (make sure Vcore is read correctly by cross checking bios readings with two third party utilities so you are sure of the true Vcore.
Ten minutes is long enough for the CPU to reach 95% or so of it's max temp due to case heating etc. This is why I only prime for tem minutes during initial testing, to get rough idea of stability.
Once you have a Orthos failure in under ten minutes, increase Vcore and see if it will pass ten more minutes of Orthos.... if it passes, raise the FSB again so the cpu runs 50mhz faster.... ie, go from 2700 to 2750mhz.
Scenario:
Let's say you top out (Orthos throws a small FFT test error within ten minutes) at 2794mhz with 1.42Vcore and the CPU temp reached 52*C during the test.
You then add Vcore to 1.44v and the test still fails but your temps got to 54*C.
You then add Vcore to 1.5v and you pass at 2794mhz, but your temps reached 58*C !
You then try 2820mhz at 1.5v and your temps hit 58*C but you get an error in Orthos.
You then take Vcore to 1.52v and you still get errors in Orthos, but temps hit 60*C
So, you try 1.55v and you still get Orthos errors but the CPU quickly exceeds 60*C
In this scenario, the CPU is at it's limit......at least with the present cooling system and most likely, even if it was more effectively cooled, you won't get a lot more out of it.
If the above was true, I would back off to 2.75ghz and try getting stable at 1.4Vcore, and only raise it to 1.42v if it was required to get stable. If 1.4Vcore turned out to be stable at 2.75ghz, I would try reducing Vcore until I got an orthos error, then go back up one notch.
The above is pretty much what happens on my San Diego 3700+ with the stock cooler. I ended up with a Vcore of 1.39v and a cpu speed of 2.75ghz. This is good, because there is no excess voltage to slowly kill off my Sandy and I still have a good overclock without stressing my cpu.
My overclock philosophy is to find max clocks, then back off to relax voltages and reduce heat. It's a happy system then with a good performance boost. Most voltage increases are required near the hardware's limit, so running just under max with reduced voltage is a good thing.
Less Volts, less heat, less stress on components = better long term overclock stability.
Now you can go back to bios and set your memory to it's max settings with the tightest timings you can get stable, knowing full well what the CPU will max out at. Be careful when adding ram voltage to tighten timings. It may be stable until the ram warms up due to case heating and heat radiation from other components. If the ram feels hot, it is! Better to have slightly higher timings than dead ram modules from overvolting/overheating. My Kingmax hardcore got hot fast so I kept the voltage down at 2.65v. The new stuff I have runs nice and cool.
Your HT multiplier can go back up to 4x AFTER your memory is stable with the preferred CPU overclock. If Orthos then throws an error, you know your HTT bus needs to come back down to a 3x multi, but many boards should run HTT well in excess of 1000mhz
I noticed your HT link speed is 900mhz so the multi is at 4x. With a 225mhz FSB the HT link will likely be stable at a 5x multi, ie 1125mhz. You just need to try it is all. Do that last though, when everything else is stable.
Summary:
With any cpu, there will be a point where adding Vcore does little to improve stability but the temps start to rise sharply, this is close to max CPU speed. If adding Vcore makes the cpu stable, then your only concern is temps. When Vcore creates stability with excess temperature, you need better cooling to push on ahead. Don't keep overclocking if the temps keep going up, you might frag your CPU, get a cooling mod, then go ahead for more speed.
Last edited by
congo on Tue Mar 06, 2007 7:54 pm, edited 1 time in total.
Mainboard: Asus P5K-Premium, CPU=Intel E6850 @ x8x450fsb 3.6ghz, RAM: 4gb PC8500 Team Dark, Video: NV8800GT, HDD: 2x1Tb Samsung F3 RAID-0 + 1Tb F3, PSU: Antec 550 Basiq, OS: Win7x64, Display: 24&
