Overclocking - A few questions

Soldato
Joined
6 Aug 2010
Posts
5,775
Location
Birmingham
Hi

I am looking at overclocking my pc which is spec as per my sig. Now I have read the thread above about overclocking but can't see in there about cpu multipliers. Now my questions relates to this bit

Finding Max CPU Clockspeed

1. Increase the FSB by 5MHz.
2. Decrease the RAM divider if required to keep it at stock or less

So according to that step 1 I increase fsb to 205 x 16 = 3280 Mhz and step 2 decrease ram to keep it at (or below) the 1333Mhz it runs at now then stability test it. Right so far?

Now my question is, I have looked AMD Overdrive which allows me to increase cpu multiplier from 16 to 18 giving a max of 3.6Ghz. Now as the multiplier is unlocked instead of increasing the fsb could I just increase the multiplier in bios to 19? or is the 200mhz step that it means too high?

If 200mhz is too big, at which point can the CPU Multiplier be increased? If I increase fsb to 212.5 at 16 multiplier = 3.4 Ghz, could I drop fsb back to 200 and 17 multiplier which gives 3.4Ghz. Though if I do this if I try to increase fsb some more the higher multiplier means a bigger step in overclock. So logic would tell me that increase fsb as far as it will remain stable then decrease the fsb and increase the multiplier.

Using simple math (I know some of these figures can't actually be used). If we say 3.8 Ghz is max I can go with increasing fsb it means it would be
237.5 x 16 = 3800 Mhz, would I then decrease fsb to keep the 3800Mhz (200 x 19)

Now this leaves it open that I have got it wrong and can increase the multiplier to overclock cpu higher than fsb alone allows. Using example above of 237.5 x 16 = 3800Mhz, could I then say decrease fsb to say 220 but increase multiplier to 17 = 3740 Mhz and if stable go for 220 x 18 = 3960Mhz?

Edit: Just thought could also - Increase multiplier +1 each time until it becomes unstable, decrease by 1 then use small increases in fsb as far as it will go. e.g. 200 x 19 = 3.8Ghz ok, 200 x 20 = 4Ghz but unstable, so try 205 x 19 = 3895Mhz?

I feel I may have gone on a bit but hopefully it makes enough sense for experienced guys to give ansnwer.

Thank so much in advance

Greboth.
 
Last edited:
It really is a good idea to find your max FSB before you move on to the multi.

But that is just how I do it.
 
I prefer finding the chips max stable frequency using the multiplyer, in my case 4ghz.

Once I've found that I drop the multiplyer to 10x and keep upping the FSB to find the limit of the HT Link and NB. I settled for a round 2.5ghz for both giving me 250 FSB. This gives the sweet spot for performance as your NB HT and CPU are overclocked.

This resulted in me running 250x16
 
Ah I see so the fsb limit isnt the limit of the over clock when the cpu has an unlocked multiplier? So i can increase fsb as far as possible keeping it stable then increase multiplier as far as its stable (voltages and temps allowing)

It just seems to make more sense to just increase multiplier as then ram speed is left alone but if this isnt the ideal way then fsb first it shall be.

Edit: So making sure I got this right frozennova,you found highest you go with multiplier was 200x20=4GHz,then found max fsb you could use being 250, then using the relevant multiplier again to go back to the 4ghz you found from the multiplier increase? or is it just coincidence that both equal 4GHz?
 
Last edited:
I just did 4000/250 to give me the multiplyer for 4ghz. 250fsb wasnt actually my wall I never bothered to test it fully yet, from my experience I knew that most crap out at 2.5-2.6ghz HT link

edit: 250fsb also left me with a nice round 1332mhz memory clock allowing me to run it at stock
 
Ah ok I see, makes sense if 250 gives you a round number for memory sounds like an ideal number for me to aim for. Would mean a multiplier of 15 - 16 would give me a nice overclock. Thats fsb and cpu multipliers out the way.

In regards to your HT and NB, I don't know if it is the same on all boards but for me auto is a multiplier of 10, so as standard running at 2000Mhz. This is on auto though, so I assume I set both to manual and set manual multiplier of 10x which as standard gives me my 2000Mhz then increase fsb to increase both HT and NB? So for you frozennova your HT and NB are at 2500Mhz?

Is there any reason why you would increase fsb but then run a lower multiplier for HT and NB? Say use 9x250 to give 2250Mhz? I am thinking possibly for stability but while allowing cpu overclock to remain the same.

Sorry for all the questions but I can happily build PC's but overclocking has always been a mystery before now.
 
Last edited:
Back
Top Bottom