OC using Turbo Boost Multipliers

[NickelPlate]

Hi Folks,

I'm curious if anyone is using the turbo boost multipliers to OC their processor? My i5-3570k has default turbo boost multipliers of:

38 - (1 active cores)
38 - (2 active cores)
37 - (3 active cores)
36 - (4 active cores)

I first noticed these using Intel XTU and these can all be changed within XTU as well as within my MB BIOS. Although I'm uncertain changes to them within XTU will persist if it's not running. I can already OC my processor up to 4.5Ghz stable and have been running at 4.3Ghz by just changing the master multiplier in BIOS. But it seems to me a possible benefit of OC'ing using the turbo multipliers is I could set them all to 43 and the CPU would only OC as much as needed, thereby saving power and potentially increasing it's life. Or does it already do that automagically anyway just using the master multiplier? Thoughts?

Thanks,

NP

 

[Borealis7]

i'm not sure what you mean, but there are power saving options (called C-states) that can be active even while your CPU is overclocked to save power - the CPU will downclock.

BTW..."automagically" is my new favorite word

 

[NickelPlate]

i'm not sure what you mean, but there are power saving options (called C-states) that can be active even while your CPU is overclocked to save power - the CPU will downclock.

BTW..."automagically" is my new favorite word

Haha! Glad you like the word. I guess it's the C-states then that allow the CPU to downclock itself. I was just curious if overclocking the CPU using the turbo boost multipliers is effectively the same as just changing the master multiplier (at least that's what I call it and it's the setting most people refer to when talking about OC'ing). In other words one could just set the master at 43 or set all turbos to 43. What's the difference? I dunno.

I would insert a screenshot of the Intel XTU showing these settings, but I don't know how do it on this board.

NP

 

[IntelEnthusiast]

I know some people that is the way they overclock their "K" for the main reason is that you have the thermal coupling (downclock based on temperature) with the turbo boost.

 

[Saylick]

Yeah, I'm am running a "dynamic overclock" on my i5 750. It currently runs at a base overclock of 3.33 Ghz but OCs to 3.83 Ghz when only 2 cores are loaded and to 4 Ghz when 1 core is loaded. There were no "tricks" involved to make this work; it was simply running the processor at a higher base clock and a higher CPU multiplier than the default and leaving Turbo Mode ON. Voltages were adjusted such that the CPU would still be fed with the proper amount of voltage across all possible loading scenarios. Everything runs smoothly, even with all the power saving features ON as well.

As for you, my experience with SNB and IVB processors is limited (as I built my rig back when Nehalem/Lynnfield were mainstream) but I figure it should be possible for you to achieve a dynamic overclock with SNB/IVB CPUs. Give the master CPU multiplier a kick to 43 and leave Turbo Mode ON. Go run P95 or another program that lets you vary the number of threads and see if the chip is maintaining the use of its turbo bins.

 

[Yuriman]

I use the Intel utility for overclocking and I've had great success with it. I'm not certain, but I think that changing the multiplier in bios changes the max turbo multiplier too?

 

[Ed1]

Yeah, I'm am running a "dynamic overclock" on my i5 750. It currently runs at a base overclock of 3.33 Ghz but OCs to 3.83 Ghz when only 2 cores are loaded and to 4 Ghz when 1 core is loaded. There were no "tricks" involved to make this work; it was simply running the processor at a higher base clock and a higher CPU multiplier than the default and leaving Turbo Mode ON. Voltages were adjusted such that the CPU would still be fed with the proper amount of voltage across all possible loading scenarios. Everything runs smoothly, even with all the power saving features ON as well.

As for you, my experience with SNB and IVB processors is limited (as I built my rig back when Nehalem/Lynnfield were mainstream) but I figure it should be possible for you to achieve a dynamic overclock with SNB/IVB CPUs. Give the master CPU multiplier a kick to 43 and leave Turbo Mode ON. Go run P95 or another program that lets you vary the number of threads and see if the chip is maintaining the use of its turbo bins.

thats how I have mine to right now , I don't want to much temps or above base vcore .
So right now I have turbo 40, 40 (single and 2 core), 39 (3 core) and 38 (4 core) .

This gives boost w/o much added heat , I can probably go another 100 or two (41-42) and still have it stock voltage with little added heat.
Seems once you go above 43-44 you add a lot more heat .

Also note I disable the "asus multi-core enhancement" which setts all cores at max turbo boost setting .

I find keeping the turbo staggered keeps temps down and most apps don't use all cores at once so you still get good boost .
thermally , 2 cores running 4.0ghz are still cooler than 4 cores at 3.8 , so gives higher clocks while not going over what you get with all cores 100% load .

 

[AaronGolliver]

Haha! Glad you like the word. I guess it's the C-states then that allow the CPU to downclock itself.

I'll clarify.

P-states are what downclock the CPU. C-states are a bit different.

Basically, if the CPU is in C0, the core is active, instructions are retiring, etc. But if the CPU goes into any other C state (1,2...6) the core is idling. Depending on how deep the C-state is, the deeper in idle the chip is, the lower power it's using, and the longer it takes to wake up.

P-states are a (sometimes configurable, depending on motherboard, bios, and chip firmware) table of multipliers for the chip the use. P0 being max speed/possibly turbo, and P(>0) being consecutively slower frequencies/voltages.

I know from my work with Intel, the operating system is able to request specific P-states, and while the chip sometimes honors the OS, it usually just ignores it and runs at whatever it sees is the best (and the CPU can run at a much finer granularity than the OS, so this usually benefits performance unless the OS has a very good idea about the performance characteristics of the running thread)

And don't even get me started on t-states, those still confuse me. From wiki:

"T"-states (throttling states) which will further throttle down a CPU (but not the actual clock rate) by inserting STPCLK (stop clock) signals and thus omitting duty cycles.