EIST and VCore on auto: implications for an i7 920 & Asus P6T V2

[Mr McC]

Any time I have tried a mild overclock, I have simply disabled all of the energy saving features in the Bios. I would greatly appreciate it if someone could take the time to answer a couple of questions, as I prefer not to experiment without prior knowledge. I have scourged various forums but I have failed to find straight answers to ther following:

a) To enable EIST, I must have Vcore on "auto" in the Bios. Is this correct?

b) Vcore on auto will dynamically adjust voltage according to the demands placed on the system, in other words, it is not a fixed quantity. Is this correct?

c) I cannot enable the 21 multiplier if EIST is disabled. Is this correct?

In the event that the statements above are correct:

d) Overclocked to 3.20, should I worry about powers saving, would it have much of an impact?

e) How high can I expect the Vcore on auto to jump when the system is stressed?

f) With a BCLK of 160, if I enabled Turbo, the multiplier of 21 would raise an overclock of 3.20 to 3.36 on a single core, is this correct?

Thank you in advance.

 

[{JNT}Raptor]

a. No
b. Not sure....I'v never left Vcore on auto on any rig I've run.
c. 21 multi can be enabled anytime.....I'm at 21x200 with EIST on or off.
d. No...I'm at 4200 with power saving on with no issues.
e. Dunno....refer to b.
f. I remember reading something on this but cannot recall where.

I'm on the same board..same CPU....same memory.

Hope it helps.

 

[Mr McC]

It does indeed help.

So i can have EIST and manually set the Vcore, that was one of my concerns.

Cheers

 

[{JNT}Raptor]

Hmmm...looking around the web some it seems you have to run the default MULTI for EIST to work properly...along with power saving features turned on in Windows......I'm running with it off now but will play with this over the next few days......gives me something to do.

But...as I said earlier though.....I'm at 4200 21x200 with all bios power saving features turned on......stability isn't an issue here.

 

[Mr McC]

There's the thing. Have you seen EIST make a noticeable difference at your clocks? That's another concern: even if you can enable it, is it doing its job?

 

[somebody]

b. Vcore should change with respect to multi.

d. Up to you and your personal definition of impact.

d. Depends on the CPU, but should be well within Intels spec.

e. I think with the 920 you should be able to run 21 on all cores and 22 on a single core if the BIOS is set for it. With BCLK at 160 that would be 21x160 = 3360MHz for 4 cores and 3520MHz for a single core when other cores are non-active.

Power saving while idle might be something like below where 1 is best and 5 is worst*

EIST Vcore C1E C3/C6 Rating
on Auto X on 1
on Auto on off 2
on Auto off off 2
on Fixed X on 3
off Fixed X on 3
on Fixed X off 4
off Fixed on off 4
off Fixed off off 5

*5 can be made worse still by disabling all c-state transitions.

 

[Mussels]

Any time I have tried a mild overclock, I have simply disabled all of the energy saving features in the Bios. I would greatly appreciate it if someone could take the time to answer a couple of questions, as I prefer not to experiment without prior knowledge. I have scourged various forums but I have failed to find straight answers to ther following:

a) To enable EIST, I must have Vcore on "auto" in the Bios. Is this correct?

b) Vcore on auto will dynamically adjust voltage according to the demands placed on the system, in other words, it is not a fixed quantity. Is this correct?

c) I cannot enable the 21 multiplier if EIST is disabled. Is this correct?

In the event that the statements above are correct:

d) Overclocked to 3.20, should I worry about powers saving, would it have much of an impact?

d) How high can I expect the Vcore on auto to jump when the system is stressed?

e) With a BCLK of 160, if I enabled Turbo, the multiplier of 21 would raise an overclock of 3.20 to 3.36 on a single core, is this correct?

Thank you in advance.

a. half yes, the multi will change but the voltage wont - thus giving no power savings. i HAVE tested this with a wall meter. its possible this varies between motherboards, but i find setting a manual voltage breaks idle volts on just about every system - AMD, intel, asus, gigabyte, and DFI. some boards have two voltage options to offset this (the voltage one, and a % boost - the % boost one raises the values without breaking idle states)

b. its fixed to two set amounts - stock load, and stock idle. not sure what they are for your chip.

c. dunno

d. yes. my housemates system went up 50W at idle disabling these features, and thats at stock.

d. it wont jump. it will stay at stock.

e. not a clue.

 

[Mr McC]

a. half yes, the multi will change but the voltage wont - thus giving no power savings. i HAVE tested this with a wall meter. its possible this varies between motherboards, but i find setting a manual voltage breaks idle volts on just about every system - AMD, intel, asus, gigabyte, and DFI. some boards have two voltage options to offset this (the voltage one, and a % boost - the % boost one raises the values without breaking idle states)

In other words, I need to leave this on auto for EIST to work properrly?

d. it wont jump. it will stay at stock.

Are you entirely sure about this? I have read that auto on Asus boards, rather than fixing the parameter at a certain level will actually allow it to do its thing once I start overclocking? This is a concern and it is what led me to manually set the Vcore any time I have been fiddling around.

Basically what we are saying is that I can take it up to 3.20 without EIST and with Vcore on auto, check stability, enable EIST and Bob's your uncle?

e. I think with the 920 you should be able to run 21 on all cores and 22 on a single core if the BIOS is set for it. With BCLK at 160 that would be 21x160 = 3360MHz for 4 cores and 3520MHz for a single core when other cores are non-active.

Really? So i can manually tell all the cores that the 21 multiplier is the new"stock" and have the system apply Turbo to that? Interesting, to say the least.

Thanks you both for the responses.

 

[Mussels]

In other words, I need to leave this on auto for EIST to work properrly?



Are you entirely sure about this? I have read that auto on Asus boards, rather than fixing the parameter at a certain level will actually allow it to do its thing once I start overclocking? This is a concern and it is what led me to manually set the Vcore any time I have been fiddling around.

Basically what we are saying is that I can take it up to 3.20 without EIST and with Vcore on auto, check stability, enable EIST and Bob's your uncle?



Really? So i can manually tell all the cores that the 21 multiplier is the new"stock" and have the system apply Turbo to that? Interesting, to say the least.

Thanks you both for the responses.

actually, i DONT know if auto will raise volts on your board - i too HAVE heard of some that do, but i've never owned one. You have no control on how much it raises or when, so its very unreliable to trust it.

 

[Mr McC]

actually, i DONT know if auto will raise volts on your board - i too HAVE heard of some that do, but i've never owned one. You have no control on how much it raises or when, so its very unreliable to trust it.

There's the rub, I don't want to put Vcore on auto, but I am concerned that if I don't, I won't be in a position to enable EIST, or I will impair EIST's funtion.

 

[Mussels]

There's the rub, I don't want to put Vcore on auto, but I am concerned that if I don't, I won't be in a position to enable EIST, or I will impair EIST's funtion.

mmm thats why i went AMD this time around, their idle voltage stays the same (1.10v) - and since my FSB hasnt changed (black edition chip), it doesnt need more volts at idle, so no stability issues.


grab yourself a wall meter and go from there, imo. you can tell what kind of impact various settings are having on power consumption, as you test the various combinations.

 

[somebody]

Really? So i can manually tell all the cores that the 21 multiplier is the new"stock" and have the system apply Turbo to that? Interesting, to say the least.

21 is already a turbo. The 920 has 2 turbo bins, 21 & 22.

I can't tell you what your 920 VID's are for auto but for my 860 it's roughly the following for auto/normal Vcore which is basically the CPU controlling the VRM through the VID signals. (the 860 has 5 turbo bins, 22-26).

Multi Vcore
9 0.864
10 0.896
11 0.912
12 0.944
13 0.960
14 0.976
15 1.008
16 1.024
17 1.056
18 1.072
19 1.104
20 1.120
21 1.152
26 1.248

Since Vcore is tied to the multiplier then if I increase BCLK there comes a point where Vcore is not enough anymore to ensure stability. This can be got around by using a fixed Vcore and ignoring the VID requests from the CPU. My BIOS also has a thing called dynamic Vcore but AFAIK this just adds a static offset to the standard VID request from the CPU rather than dynamically changing each VID for each multi. In other words each VID is shifted the same amount.

IMHO if you need to run with a fixed Vcore then EIST doesn't really matter if C3/C6 are enabled. C6 is the big power saver.

 

[overclocking101]

about 90% on the intel results (i7) on hwbot are all done with Turbo on which means eist and speedstep must be enabled as well, you are thinking way to much into it man just enable eist/speedstep/turbo set your vcore and go to town. but imo you will see better performance with it off due to higher bclck/QPI/UNCORE frequency

 

[somebody]

about 90% on the intel results (i7) on hwbot are all done with Turbo on which means eist and speedstep must be enabled as well

Just to clarify, Turbo Boost does not require EIST. EIST is speedstep BTW. Perhaps it's a problem with the way your BIOS handles Turbo / EIST that leads you and many other people to think this. If you are interested in what is actually required for Turbo Boost the white paper can be downloaded from here.

http://download.intel.com/design/processor/applnots/320354.pdf?iid=tech_tb+paper

 

[Mr McC]

I just set the vcore manually to 1.2 and changed nothing else, using the latest Bios, and OCCT has shown the same underclock and frequency reduction, so the multipliers and/or voltages are going down as normal: the energy saving features still work, at least at stock clocks.

The plan is to take the chip up to 3.2 with EIST and hyperthreading enabled, I'll let you know, might write a short guide.