A BIOS question about Z68 motherboards

[BonzaiDuck]

Lemme try and make it simple and short -- not my strongest skill.

Built my i7-2600K in July 2011. Did great with the over-clocking -- had 20 11"x17" pages of handwritten notes. Started with EIST and C1E enabled.

Somehow, I never explored how the board handled a CPU with the Turbo feature. And somehow, I had this assumption that with no EIST, the board would simply default to the base multiplier (34 in my case) and then ramp up automatically to 38 -- the chip's spec "turbo" speed.

Now I've finally got it into my head that the "turbo" is either enabled or disabled in BIOS -- except that you could use the software (in my case, the ASUS AI-Suite and Turbo-EVO) to toggle it on or off. It APPARENTLY doesn't just "ramp up" when the CPU is loaded.

It appears that the "ramping-up" only occurs between EIST idle speed and the "turbo" speed or stock speed depending on whether turbo is enabled in BIOS.

This motherboard in my sig is the May, 2011 release of the ASUS P8Z68-V-Pro board.

1) Do BIOS versions after #606 for this board offer more options -- like running the system at 3.4 without EIST so that it "ramps up" to 3.8?

2) Do other motherboards of other manufactures offer those types of options?

Or -- is this "just the way it is?"

 

[Ketchup]

As I understand it, if you remove the IA suite, and leave it up to the Intel chipset software, the Turbo feature will work as Intel intended (assuming BIOS options are enabled).

If you use the AI suite, you may have to manually enter the desired behavior, as it takes over control of this.

 

[BonzaiDuck]

As I understand it, if you remove the IA suite, and leave it up to the Intel chipset software, the Turbo feature will work as Intel intended (assuming BIOS options are enabled).

If you use the AI suite, you may have to manually enter the desired behavior, as it takes over control of this.

Well, let's have a discussion about that -- at everyone's convenience.

The P8Z68-V-Pro User Guide is footnoted extensively (against the deficient ASUS standard or Pacific Rim manufactures in general) -- to clarify that you could (1) use the TPU motherboard switch, (2) the BIOS AI OC menu, (3) manual settings in BIOS, or (4) the AI-Suite-II Turbo-EVO Windows software to over-clock. The last of those "extensive" footnotes states a "Last-In" OC precedence: whichever method you used last determines the ongoing OC settings.

My last and only approach, because I long grew comfortable with it, is "change BIOS manually."

Now . . . . there's still a possibility favoring your view, that merely having the Turbo-EVO AI Suite-component installed -- maybe even the whole enchilada -- could have an effect on these matters. But to argue my side, I unchecked the checkbox in AI-Suite's "Settings" menu to either hide Turbo-EVO, or disable it until the "Settings" checkbox was clicked again.

So . . . . what do you say to that? . . . . Or would you argue that the Intel SW features provide something that neither the BIOS nor ASUS software provides?

 

[Ketchup]

Hi BonzaiDuck, I don't think the Asus software doesn't provide this per se, I just think that it may take a different method to provide the same behavior.

I guess the first thing to do would be to make sure this is actually the case.

What I recommend: Download CPU-Z and make note of the CPU clock speed at idle. What speed is it running?
Start something CPU-intensive and make not of what your CPU speed gets up to.
What is this number?

 

[BonzaiDuck]

Hi BonzaiDuck, I don't think the Asus software doesn't provide this per se, I just think that it may take a different method to provide the same behavior.

I guess the first thing to do would be to make sure this is actually the case.

What I recommend: Download CPU-Z and make note of the CPU clock speed at idle. What speed is it running?
Start something CPU-intensive and make not of what your CPU speed gets up to.
What is this number?

Well, EIST is enabled. At idle, it's at 1.6Ghz. under load, it goes to 4.6. If I disable EIST, it will idle at 4.6. If I turn off turbo in BIOS, it stays at 3.4. If I set back to stock and EIST disabled, turbo-enabled gives 3.8, and disabled -- again -- is 3.4. I'm pretty sure I didn't miss anything there -- always the chance that I didn't cover all the bases, but I think that's about it . . .

 

[Ketchup]

Well, EIST is enabled. At idle, it's at 1.6Ghz. under load, it goes to 4.6. If I disable EIST, it will idle at 4.6. If I turn off turbo in BIOS, it stays at 3.4. If I set back to stock and EIST disabled, turbo-enabled gives 3.8, and disabled -- again -- is 3.4. I'm pretty sure I didn't miss anything there -- always the chance that I didn't cover all the bases, but I think that's about it . . .

Sounds good to me.

 

[BonzaiDuck]

Sounds good to me.

AND! It IS! I was simply living in the misconception that I had more control. Somehow -- I thought I could run the processor with a base IDLE of 3.4 Ghz and a turbo-overclock of 4.6. I have to depend on EIST to keep it at 1.6 idle and 1.008V with the 4.6 setting -- otherwise it stays at 4.6 with the maximum unloaded voltage (~ 1.36V).

And if I change everything back to default or stock, I just get turbo at 3.8, or non-turbo at 3.4. I've got to choose in BIOS, or I have to choose in Turbo-EVO. You'd think . . . if they'd done the engineering to give you EIST speed and power saving, they might have made it possible to swing from non-EIST base clock of 3.4 to whatever you'd given it with turbo-overclock.

But that's just it! It's "Turbo Over-Clocking." It has to be in "turbo" at 3.8 to start. The OC just extended it to 4.6 . . .

 

[coffeejunkee]

C-states and turbo are linked.

Enable either C1E, C3 or C6 and the multiplier will drop to 16x during idle. You need to disable them all (and EIST) to run at constant full speed.

But if you do so you can no longer have the 38x and 37x turbo multipliers for 1/2/3 threaded loads because those depend on C3/C6 states. These states completely turn off cpu cores which is the whole idea behind turbo boost: disable cores to get tdp headroom and use that to clock the remaining ones higher.

EIST is different from the C-states in that it chooses the optimal multiplier for a workload, not necessarily the highest.

Running at 3.4 and turboing to 4.6 is impossible. Either you enable turbo and you will idle at 3.6 or you disable turbo and you no longer boost to 4.6.

I do see a slight difference between EIST enabled and disabled. Enabled my cpu runs at 3.8 turbo constantly, disabled i see very fast fluctuation between 3.8 and 3.7.

 

[BonzaiDuck]

C-states and turbo are linked.

Enable either C1E, C3 or C6 and the multiplier will drop to 16x during idle. You need to disable them all (and EIST) to run at constant full speed.

But if you do so you can no longer have the 38x and 37x turbo multipliers for 1/2/3 threaded loads because those depend on C3/C6 states. These states completely turn off cpu cores which is the whole idea behind turbo boost: disable cores to get tdp headroom and use that to clock the remaining ones higher.

EIST is different from the C-states in that it chooses the optimal multiplier for a workload, not necessarily the highest.

Running at 3.4 and turboing to 4.6 is impossible. Either you enable turbo and you will idle at 3.6 or you disable turbo and you no longer boost to 4.6.

I do see a slight difference between EIST enabled and disabled. Enabled my cpu runs at 3.8 turbo constantly, disabled i see very fast fluctuation between 3.8 and 3.7.

Your explanation clarifies. It's clear as day now: Turbo is an all or nothing proposition. Funny I hadn't explored this angle for my own better understanding when I first built the system. But as I left EIST and C1E enabled, my overclocking procedure was very methodical and patient. Once I got the satisfying level and fine-tuned it, I didn't bother getting clear on these issues.

The other thing I did early in the game was to enable PLL-Overvoltage, simply because others suggested that "it assisted stability at clocks higher than 4.5Ghz." It also caused problems with some folks with waking from sleep-states. Oddly, although I don't use sleep much on this machine, I never had that problem with PLL_OverV_____ enabled.

I finally just chose to disable it, and my OC stability seems fine.