Calculate your A64 overclocking potential

[tzitzos]

Make sure you can ignore your ram when maxing your FSB. Just choose a safe setting

yeap thx dude

[XiN]

Works fine for my (crap) Winchester too.

[tzitzos]

Works fine for my (crap) Winchester too.

[RTB]

Just finished testing, 2700 stable at 1.525 volts (CPU-Z says 1.472). According to the charts, it should be stable up to 2750 at the same voltage.

[s7e9h3n]

Ok, I decided to give this formula another shot at predicting my OC potential, but I think I've run into a little flaw. There has to be a range of working speeds that this forumla applies to, otherwise, it will give you an odd result. Once again, I'll use my Opty 154 as an example - 32m stable @ 3753, 1.648Vcore. I wanted to predict what my OC could be @ 1.7Vcore. Sooooo, the forumla works out as such:

Freq = 3753 + (1.700-1.648)*620 =

3785.24Mhz -> an increase of a bit over 30Mhz It may just be me, but that seems like a pretty small increase for that amount of Vcore.....

[vanrossum]

I'll use my Opty 154 as an example - 32m stable @ 3753, 1.648Vcore. I wanted to predict what my OC could be @ 1.7Vcore.

Freq = 3753 + (1.700-1.648)*620 =

3785.24Mhz -> an increase of a bit over 30Mhz It may just be me, but that seems like a pretty small increase for that amount of core.....

Try it out, I don't believe you will be getting out much more, I give you 40Mhz at most.
That's the gain of the formula! You don't need to start from scratch, with too high expectations, then decrease... decrease your FSB untill it gets stable again.

[s7e9h3n]

Try it out, I don't believe you will be getting out much more, I give you 40Mhz at most.

You mean 40mhz at this Vcore, or @ 1.70Vcore?

[chew*]

I think this would work IF all chips were equal.....But they arent......I had 2 manchesters x2 3800's one was a dud no matter what and one wasn't No math no magic just a plain dud. No amount of cooling, amount of voltage or amount of rubbing a magic lamp will get more out of it. And I would think I know what im doing......

[vanrossum]

You mean 40mhz at this Vcore, or @ 1.70Vcore?

Well if 3753Mhz is your best overclock @1.648V, I give you 40MHz at most @1.7Vcore. Try it out!...

To chew*: Therefore you insert your own (x,y), the dud line is parallel to the others, just a bit lower. Or is your IHS making bad contact? Than you have the cooling problem.

[chew*]

Well if 3753Mhz is your best overclock @1.648V, I give you 40MHz at most @1.7Vcore. Try it out!...

To chew*: Therefore you insert your own (x,y), the dud line is parallel to the others, just a bit lower. Or is your IHS making bad contact? Than you have the cooling problem.

One chip does 2.4 gig one does 3.0. There are way to many variables for this to work IMHO. Oh and the dud does 2.4 stock volts all the way up to 1.7.......Does not scale at all whatsoever no matter what board is used.....Contact is not an issue...

[cdelatte]

Pretty close to target for me with my current setup

[vanrossum]

One chip does 2.4 gig one does 3.0. There are way to many variables for this to work IMHO. Oh and the dud does 2.4 stock volts all the way up to 1.7.......Does not scale at all whatsoever no matter what board is used.....Contact is not an issue...

Never seen that before, but that might be an exceptional case. Is the second one behaving linearly?

Can you guys with Venices report a few oc results from a single chip. Than I can update the z_ven value as soon as I have enough data.

[Trice]

So tell me if I got this correct.

If I run 2.7 stable at 1.475 and want to run 1.6 to see what I could get it would be.

2700 + (1.600 – 1.475)*870

2700 + 108.75
which = 2808.75 at 1.6v?

[Marios]

I'm not a matematical wonder, but if I understand correctly there is a linear connection between raising Vcore and overclock frequency?

That's simply not true I think. Many preocessors (including mine) gain more relatively in the low Vcore segments. For example: from 1.4v to 1.5v you get 600Mhz more and from 1.5 to 1.6 yon get only 300 more. That's only half.

Correct me if I'm wrong please.

With my Toloedo I can O/C up to 2650MHz stable with only 1.35v.
Then I need to give 1.5v to get from 2.700 up to 2.800MHz stable.
This is strange and surely not a linear vcore - CPU frequency behavior.
Though for a small range of 100MHz the equation might work
The equation:
Vcore needed at A1MHz CPU frequency =(A1-2596+1.35*870)/870
gives me the following table.

CPU vcore
2400 1,12
2450 1,18
2500 1,24
2550 1,3
2600 1,35
2650 1,41
2700 1,47
2750 1,53
2800 1,58
2850 1,64
2900 1,7
2950 1,76
3000 1,81

These values might be good for the original CPU frequency starting point (2600 at 1.35) +100MHz. This is the bold part.

[vanrossum]

@ v0dka:
take
(x,y)=(1.35,2650)
then, for 1.5V you should be able to achieve 2770MHz, according to the formula. That corresponds to the range you determined (2700-2800@1.5V)! It is not possible to exclude the possibility of a linear relation with your two single points.

Did you try out the table that you posted? Do the values match with your findings?

[biohead]

So first we're all naming our steppings, asking ourselves how far it might go... and next thing you know, all you need is a calculator?

If frequency/vcore would run linear we would've found out years ago.

�t might sound sceptical, but it's just plain logic. I'm with chew*.

[vanrossum]

If frequency/vcore would run linear we would've found out years ago.

It has been indeed, http://www.ocforums.com/showthread.p...06#post2588006. In practice, however, your oc potential always will be limited by your cooling and voltage instability. So your curve will gradually stop growing from a certain Vcore onward.

Nevertheless it is interesting to calculate the potential of your bare A64. The formula also gives a way to compare oc results that are achieved at different Vcores.
For example, who is better Yoe:2630@1.425 or Mary:2700@1.575? Yoe is, he could do 2750@1.575.

[Dragoran]

If you insert the x=1.51 bios, then take the 1.6V bios. You should be able to get 2778MHz with that. But remeber that you get 34% more heat output. Your cooler should be able to hadle that, otherwise you won't make it to 2778.

it won't work...
any ideas why?
the most stable overclock seems to be 2700 vcore does not help anything...
it can boot up to 2950 but won't run stable ram dividers does not help too
I have 2GB GSKILL ZX.

[afireinside]

This won't work. Far to many variables. PSU, mobo, skill, cooling contact, quality of memory controller, memory controller, cold bug, cooling, not all CPUs scale well with high volts, drooping, ect.

According to your formula going from 1.65v to 1.7v on my 165 should take me from 2800mhz to 2843.5mhz. That much voltage takes me to 2925.

[vanrossum]

it won't work... any ideas why?
the most stable overclock seems to be 2700 vcore does not help anything...
it can boot up to 2950 but won't run stable ram dividers does not help too
I have 2GB GSKILL ZX.

Your curve stops at some point and then gets flat, then voltage increase doesn't lead to higher clock speeds anymore. This limit can be related to your cpu, but more probably has to do with all other factors afireinside listed above. IMO you can blame the cooling in most cases.

As I am on air, my curve stops at 299*9, no matter how much I increase Vcore, I don't make it to 300*9...

So what could you try to reach out further? Maybe better cooling helps???