What are your Overclocking Methods?

[fAlCoNNiAn]

Hey all,

Just curious on how each of our approach overclocking and what methods you use.

for example, i lower the voltage to see what the cpu is most stable at (my 3.2e works wonderfully at 1.185, stock 1.385) then i start bumping up the fsb from there, increaseing voltage one step at a time when i get a superpi error (stability test by running dual prime (max heat/power) and superpi 32m). if superpi passes and prime is still good, then i use clockgen to increase the fsb one step, then repeat. ofcourse bumping up voltage as i go.

i have found this to give me a decent overclock and im statisfied. for example, im testing right now at 221 fsb with vcore of 1.285 (.1v under stock) so far so good.

so lets hear them, any special methods used?

[EvlUndrWareNome]

Usually I up the fsb, prime it, up fsb prime fails, more volts, then up fsb again and prime it. its a fun process =p

And I always make sure i set the pci/agp lock before any of it.

God, I LOVE Intel. No LDT, HTT, HTT multi...


Intel for me.

[timpanogos]

amd man here ... I get into bios, enable memtest boot .. start tweaking whatever it is I'm working on (fsb/memtimings or cpu meg). I try to only tweak one variable at a time, no more than 3, boot to memtest 1 pass of #5 and #8 - tweak till max. Now boot out to windows, 1 test of prime95 blend (about 4 minutes), back to tweak/memtest for the next set of 1 to 3 variables

After getting throw all settings like this, pushing for the fasted mem bandwidth (as shown in memtest) and the tightest timings .. then I go for the prolonged basic benches - superpi 32m, occt, all the 3dmarkxx dudes etc.

Usually I've tweaked too tight for some of these and now you have a huge guessing game .. but the first step gives you a real good feel for which individuals settings are most sensitive to change.

I also maintain an xsl spreadsheet and note all settings as I go.

I've also found a stock/loose timings config that is totally stable, all of these values are in the database. The final tweaks are often guesses on interpolating between the stock/loose config and your super tight failing config

[Lithan]

Usually I up the fsb, prime it, up fsb prime fails, more volts, then up fsb again and prime it. its a fun process =p

And I always make sure i set the pci/agp lock before any of it.

God, I LOVE Intel. No LDT, HTT, HTT multi...


Intel for me.

Hehe, I never thought someone would consider locks to be good for overclocking.

[Dillusion]

I set the vcore to the default for the chip, then i up the FSB 1:1 with the max multi from there, find the max FSB with stock vcore. then i bump up the vcore until i get more headroom, or until vcore does nothing more- while upping the fsb. (all within good temps of course)

then i lower the multi and look for my max 1:1 fsb ...i dont really mess with dividers at all to tell you the truth.

[charlie]

Here's how I do it:

I immediately eliminate the RAM/MOBO as bottlenecks... by running low fsb and immediately up the Vcore on an A64 to about 1.55v...

So on a new 3500+ Venice I'll leave it at 11x multi, set LDT to 3x, 1:1 and I'll just use clockgen in desktop to up the fsb until it "won't" pass a superPi 1M. I know in 15 minutes if I have a good chip or not... maybe 11 x 255 = 2805.... if 11 x 256 won't pass, I'll up the Vcore, 1.575v, 1/6v, 1.625v... and repeat. Some chips like the volts, some don't. Later on I'll test for long term stability and such nonsense :stick:

C

[fAlCoNNiAn]

I immediately eliminate the RAM/MOBO as bottlenecks... by running low fsb and immediately up the Vcore on an A64 to about 1.55v...


im confused on how you would do that, since you state that you set to 1:1, which would give the memory a chance to fail.
:confused:

[Delirious]

amd man here ... I get into bios, enable memtest boot .. start tweaking whatever it is I'm working on (fsb/memtimings or cpu meg). I try to only tweak one variable at a time, no more than 3, boot to memtest 1 pass of #5 and #8 - tweak till max. Now boot out to windows, 1 test of prime95 blend (about 4 minutes), back to tweak/memtest for the next set of 1 to 3 variables

After getting throw all settings like this, pushing for the fasted mem bandwidth (as shown in memtest) and the tightest timings .. then I go for the prolonged basic benches - superpi 32m, occt, all the 3dmarkxx dudes etc.

You can tweak memory timings in memtest?

[Jayw28]

Usually I up the fsb, prime it, up fsb prime fails, more volts, then up fsb again and prime it. its a fun process =p

And I always make sure i set the pci/agp lock before any of it.

God, I LOVE Intel. No LDT, HTT, HTT multi...


Intel for me.


im with him

:slap:

[fAlCoNNiAn]

You can tweak memory timings in memtest?
as far as i know, you cant.

so nobody here tries to undervolt then raise fsb as high as you can, while increaseing voltage? :(

[timpanogos]

all, there are a few timings that you can tweak in memtest, but that is not what I'm talking about. long slow, likely a better way, but I tweak in bios, boot to memtest run 5/8 once, boot back to bios ... rinse and repeat

[ZX7891]

I do almost the exact same thing as timpanogos. I guess ive just gotten used to it.

[Delirious]

all, there are a few timings that you can tweak in memtest, but that is not what I'm talking about. long slow, likely a better way, but I tweak in bios, boot to memtest run 5/8 once, boot back to bios ... rinse and repeat

yah i like your method for ocing the memory. Use low cpu multi, and ht multi to eliminate both as bottlenecks, crank the fsb, test in memtest, then once u find the max for that start raising to cpu multi to oc the chip.

[fAlCoNNiAn]

great read so far guys, keep them comming, id like to hear what fugger's method is. ;)

[fAlCoNNiAn]

Come on Fugger, spill it ;)