[Total Noob] Sufficient Voltage?

[extremepilot]

I am overclocking an E3110 and I don't know how much voltage to give. I upped the FSB from 333 to 400 and the multiplier is on 9.5 so that the frequency is 3.8Ghz. The ram is running at native 400mhz @ 5-5-5-18 so a 1:1 ratio is there. It would not boot or sometimes restart so I upped it from 1.2250 to 1.23750. This is an increase of only 0.0120 volts.... which seems way too little. I haven't tested for stability yet but I am going to test this overclock tonight with prime95. Anywho...
1) is this too little of an increase? I just want to get a general idea.
2) Also... in what case will I need to overclock the MCH and FSB?

 

[Evilonigiri]

1.) Nope, it's possible, but make sure it's stable though, 12hours of small FFTs.

2.) It's OCed already isn't it?

 

[dagger]

If it runs stable at low vcore, it's a good thing. Don't complain. You'll probably need to increase it to be stable under prime95 though.

And wouldn't large FFTs be a better test for long term stability than small FFTs? More energy consumption, heat production, and... vdroop.

 

[Evilonigiri]

I thought small FFTs were more intense than large FFTs? So then small FFTs would ensure longer stability?

 

[dagger]

I thought small FFTs were more intense than large FFTs? So then small FFTs would ensure longer stability?

It is. The conventional wisdom is small FFTs for cpu stability. Also it don't stress the ram so you can rule that out for unstability. I was just wondering if, under realistic conditions where energy consumption and heat matters, larger ffts might actually be better.

 

[Evilonigiri]

It is. The conventional wisdom is small FFTs for cpu stability. Also it don't stress the ram so you can rule that out for unstability.

I wouldn't be too sure about that. It doesn't stress the ram much, but according to my experience, if the ram is very unstable, it will show up as an error in small FFTs, especially if you're using the pc while stress testing.

 

[dagger]

I wouldn't be too sure about that. It doesn't stress the ram much, but according to my experience, if the ram is very unstable, it will show up as an error in small FFTs, especially if you're using the pc while stress testing.

Meh, who would oc their ram at the same time?

 

[Evilonigiri]

Meh, who would oc their ram at the same time?

Sometimes you can't help it if the ram is linked to the fsb right? You know, 1:1 ratio and all.

While the conventional way is to get the cpu stable first, I think a more advance way is to get the ram first. But it's probably better for newbies to follow the conventional way, it's much simpler.

 

[dagger]

Sometimes you can't help it if the ram is linked to the fsb right? You know, 1:1 ratio and all.

While the conventional way is to get the cpu stable first, I think a more advance way is to get the ram first. But it's probably better for newbies to follow the conventional way, it's much simpler.

Wait, what are the benefits of the "advanced way?" Thou shall pass thy 1337ness onto me.

 

[Evilonigiri]

Wait, what are the benefits of the "advanced way?" Thou shall pass thy 1337ness onto me.

Well, the basic idea is to get the ram stable first, while having the cpu clocked really low. This so called "advanced" way is meant for more extreme OCing where the conventional method creates massive headaches (ie, >3.4GHz OC for Q6600).

Simply stated, you set the fsb to a frequency you want, let's take 500MHz as an example. Using 1:1 fsb to ram ratio, this will leave you 1000MHz for ram, which might need massive tweaking to get stable if you have DDR2 800MHz. So, dropping the cpu multiplier to the lowest possible, which is say x6, to get 500MHz x 6 = 3GHz (let's assume this is the Q6600). From previous OCing, you know the Q6600 at 3GHz is stable at 1.32500V, so that's where you set it at. Then you concentrate on the ram, laxing timings and increasing voltage to get it 24hours stable in blend test in Prime or Orthos. Of course you'll need to get the mobo stable too by increasing the voltage in such, and instabilities in blend test could indicate insufficient voltage for mobo or unstable ram.

Getting the ram and mobo all stable, you move on to the cpu, bringing the multiplier up to whatever, say x9. As you know, 4.5GHz on a Q6600 is rather hard, so with all the ram and mobo variables gone, this task becomes much simpler.

This is the gist of the 'advance' way. Of course there's an even more advance way to do this, but this should solve many problems. The other advance way separates the ram and mobo variables, making it an accurate yet rather long process of OCing.

 

[dagger]

24 hours stable... ouch. This is too 1337 for me.

 

[Evilonigiri]

24 hours stable... ouch. This is too 1337 for me.

Reason why I said so was because I had mine fail at 12hours and 7mins. Imagine that, I woke up, turned on my monitor to see if it failed or not. At that instant, it failed. I saw for a split second the green "GO" that orthos displays turning to "STOP". Well, that solved the issue of my instability problem.

 

[dagger]

Reason why I said so was because I had mine fail at 12hours and 7mins. Imagine that, I woke up, turned on my monitor to see if it failed or not. At that instant, it failed. I saw for a split second the green "GO" that orthos displays turning to "STOP". Well, that solved the issue of my instability problem.

Are you sure you didn't just imagine seeing the icon go red? Getting up in the middle of the night checking on computer oc might be stressing you mind's stability.

 

[Dunkel]

Why did this conversation go from the difference between large and small prime runs to running Orthos?

 

[Evilonigiri]

Yeah, why did it? Before that it was about voltage?

Sometimes I feel the urge to derail threads, and it happens naturally, unexpectedly.

 

[Evilonigiri]

Are you sure you didn't just imagine seeing the icon go red? Getting up in the middle of the night checking on computer oc might be stressing you mind's stability.

Who said anything about middle of night? It was morning. And I'm very sure it turned red the instant I turned it on, since I was running 2 instances of Orthos at the same time, so the timer of both of the were the same until the instant I turned on the monitor. One of them stopped at 12hours 7mins 15seconds and the other one kept on going.