Total beginner looking for help taking first OC steps for AMD FX4300

[Orangehat44]

Alright i did 15 min of P95 test on lunch break and took screen shots. temps didnt get as high do we want to test again for longer to make sure the 54C i saw last night was real?

Heres shots:
HW monitor and overdrive thermal margin loging:


hw monitor lower half and overdrive voltage loggin:




CPUZ tabs:


let me know what ya think. THanks

 

[caddi daddi]

I just want you to run it for 30 minutes to be sure its stable.

 

[Orangehat44]

just did 35 min p95 with same settings as the one last night, no errors no warnings. amd overdrive logging thermal margin hit a min of 18 (so 52c) then seemed to settle in around 20. hwmonitor had the weird spike in it again where its max reading was saying 161 which seems to happen ever 3rd test or so.

 

[caddi daddi]

save this clock to a profile in the bios, under the tool tab, asus oc profile just name it 4300, you have to hit enter to make the name stick.

so now you reduce the core clock .5 to 21, then raise the fsb to get the clock back up to 4.3, this should be around 210.

 

[Orangehat44]

and then do another 30 min p95 test?

 

[Orangehat44]

so if i saved profile correctly it should look like this right?:



and i decrease clock to 21 and increase fsb to 210 it actually shows 4410mhz as target:



so i got cautious and set it to 205 and that made a target of 4305mhz:



so i restarted it and went back in to check make sure it all stuck and the updates had taken affect, ram now shows 1912mhz is that okay?:

 

[caddi daddi]

that is all correct, open the cpu-z tabs, cpu and memory, look at the ht, core clock, memory and cpu/nb and see how those have changed.
the 30 minutes of p95, then raise the fsb another 5 and do it all again.
let the core clock rise also, adding vcore, reducing the other busses as needed to get it to boot, you already know the cpu/nb will not make 2600 so as it reaches 2500, back it down one or two notches, back the ht down and try to keep it between 2000 and 2400.
raise the dram voltage up to 1.6, specs say 1.65 is safe, most ram will do +150 with no problems, time to learn to juggle!!!!!

 

[Orangehat44]

haha sounds complicated so i'll prob go slow. but my first question is i dont think i ever tried cpu/nb at 2600, you had told me to try the ramm at 1866 then try cpu/nb 2600, but thats when the ram didnt work so we went into the ram clock thing but never went back and tried at 2600 (was i suppose to?)

also is it useful to start saving more profiles after every or every other step thats stable? like we did with the 4300 stable profile.

do i raise the dram voltage now with the other stuff or raise it after i've bumped up fsb another 5.

also there was a target heat to stay below right? was it 62 on the cpu pacakge temp from hwmonitor (and/or thermal margin of 8 on overdrive) or are we still trying to stay below 55.

also do i need to mind other temps like mother board's cpu temp from hwmontior?

 

[caddi daddi]

save the profiles you wish, saving the 4300 gives you a safe, stable base clock to work from, when it crashes you will reset the bios to default, get it to boot, then boot to bios and load the clock profile and you're back at it.
just set the dram voltage to 1.6 now.

62c is the max temp you want to see, if it drifts a little over, that's fine, 62c on the cores still gives us a safe margin at this point.
go ahead and set the cpu/nb back a notch and set the ht back to 2000.
as you increase the fsb you will have a good idea the ram is the issue if you have a no boot, start with the third digit, the 9, increase it to 10, so it would look like this, 9 10 10, if it still will not boot try making the 10 an 11.

remember, cpu clock is king
memory is queen
cpu/nb is prince
ht is just a court jester.

 

[Orangehat44]

so the 205fsb passed 48 min of p95 fine. bout same temperatures.

so next move is set dram v to 1.6 and move fsb from 205 to 210 and retest?

are we done with cpu ratio and multiplier stays at 21? and we move the clock forward with the fsb freq now?

if it passes test with 210 fsb and dram1.6v do i keep going up by 05 on fsb, till it fails p95, then raise vcore? when you say reducing the other busses as needed to boot, is that reference to cpu/nb freq and ht link speed and set them down by to the next lower value notch (but then that notched value will climb back up as fsb increases?)

Lastly I'm legitimately getting more performance now than we started with mem freq faster than it was and the cpu back up around 4300, correct? (since it was around there when we started and fsb was at 216 with asus's ezmode optimal setting) and as long as it passes a good length p95 test its considered stable and should run everything normal if want to use it or run other programs between OC tweaks?

again thanks for everything this has all been hugely helpful and educational

 

[caddi daddi]

try to keep the cpu speed close to 4300, you know that's stable.
you are now working the ram up to find that wall.

you are getting more cpu throughput due to the core clock.
the ram adds a very little bit.
the cpu/nb adds even less
the ht has never been to help anything.
in gaming or 24/7 use the only thing that you will ever notice is the cpu core clocking.
my fx clocks are all cpu core clocks as high as I dare, ram at default and cpu/nb 2500 or above and I never have to touch it again.

 

[Orangehat44]

okay some what confused me there. if the memory doesnt add much why are we now pushing fsb while trying to keep the speed round 4300 and as you said 'work fram up to find the wall' is this just as you say to find a limit then we do something else?

i might be getting confused on wording. and alittle confused also cause last night you said let core clocks rise with fsb increases, but today sayd try to keep close to 4300, or am i misinterpreting something

 

[caddi daddi]

naaa, on second thought let's just let all the busses rise and use up the rest of our temp overhead up to 62c on the cores.

 

[Orangehat44]

cool, i didnt do much last night wasnt home much. but let me lay out a plan based on what we've said, and assuming i didnt miss type anything here i'll follow this plan and hopefully be able to do a few steps before reporting back(sound good?):

1) set dram v to 1.6 and fsb to 210 then retest p95
2) Assuming that passes raise fsb to 215 and test again? and if one of those tests fail the remedy is to bump cpu voltage like we did before till tests pass again. all the while monitoring heat for that 62C?
3) if this passes do we continue bumping fsb by 5 and cpu voltage till stable test, and repeat till we hit that 62c? (do i still want to knock down nb freq and ht link freq a few notches as they climb with the fsb increase? i think previously you said keep below 2600 and 2400 respectively? or let those rise as well)

 

[Bluefalcon13]

If it fails, then you start the balancing act of juggling cpu/nb, memory dividers, and HTT link multipliers. What it seemed caddi was doing before was finding your max of each of those. It keeps the guess work to a minimum.

 

[Orangehat44]

cool i think i get it sort of haha. do you mean if it fails to boot or fails the 30 min of p95. (now im confusing myself with the word fail lol)

 

[Bluefalcon13]

Ultimate test is the p95 failure. A failure to boot counts as a p95 failure cause you couldn't run the test

 

[Orangehat44]

grr somehow i managed to delete my post... stupid phone.

So i bumped dram voltage to 1.6 and fsb to 210. got 16min into p95 before it failed on the 100 'warnings of illegal summout'. temps didnt go above 54C

is my next move to bump vcore like we did before to get it stable. or juggle the other things to get it stable.

 

[Orangehat44]

so i tried bumping vcore 1 bump and it got to 17 min and different worker failed with error. but then next reboot got weird and gave the memory error on boot (but got to the screen where you got hit f1 to get back into bios so didnt require a cmos reset or the memok button). so i set it back to 205 on fsb, droped the vcore back down one, but left dram voltage at 1.6