Manual or Adaptive Voltage - i7 4770k

Associate
Joined
14 Jun 2009
Posts
432
Hi,

I recently purchased an i7 4770K and manage to get it to overclock it to 4.6Ghz at 1.25v by manually setting the voltage.

Would it be better if I set it to adaptive voltage and cap it at 1.25v? So at least when the PC is idling it doesn't need to be constantly running 1.25v?
 
Associate
Joined
11 Jun 2013
Posts
1,087
Location
Nottingham
As far as I know it doesn't ever lower the voltage on a manual ("override") voltage, so it'll be warmer and consume more power at idle. Yeah it'll lower the clock but not the voltage. Adaptive voltage is better in my experience. Though there may be differences between boards too I guess.
 
Soldato
Joined
29 Aug 2010
Posts
7,839
Location
Cornwall
Yeah, I can only go from my experience, but I'm using manual voltage (1.175v) and while typing this my voltage is 0.768v and my clock is 800MHz. I can't see what adaptive voltage would do that this isn't?
 
Soldato
Joined
29 Aug 2010
Posts
7,839
Location
Cornwall
its probably different on different boards,i know the asus haswell downclocks with a fixed voltage too

but anything below haswell wouldn't hence the need to use offset

I'm using a Gigabyte, so that's two we know of.

EDIT:
I don't want to argue with Shiari, but this OC guide (admittedly for the Gigabyte Z87X-OC) suggests it's down to the chip, not the motherboard.

If you want to use an offset for the VCore you should type in or select “Normal” instead of “Auto” and then you will be able to set offset. However now that Haswell has a built in VRM there is almost little reason to use offset mode b/c if you enable C3 power state and EIST the CPU multiplier and voltage drop together even if you have set a manual voltage without offset. Leaving the VCore on “Auto” will result in pretty high auto VCore as the CPU usually over compensates.

Maybe some manufacturers have chosen to be awkward and prevent it working like this?
 
Last edited:
Caporegime
Joined
11 Jul 2009
Posts
27,049
Location
BenefitStreetBirmingham
it depends on the board/cpu

auto is the worst you can use
manual tends to use less than auto
offset tends to use slighty more than a manual fixed voltage,but if you tweak it right you can get it to use less if you allow for vdroop aslong as you remain stable ect

and the added benefit on chips lower than haswell (ivy/sandy/Nehalem ect) is the cpu isn't running full volts 24/7

but haswell seems to have it built in,so fixed/offset isn't as important but id still like to have a play see what's the best option
 
Soldato
Joined
25 Jun 2013
Posts
5,046
Location
Warks
My Asus board would lower the voltage with the clock when set to manual, my Gigabyte board doesn't.

Adaptive tends to use very high voltages when pushed, I think.
 
Associate
Joined
11 Jun 2013
Posts
1,087
Location
Nottingham
Adaptive tends to use very high voltages when pushed, I think.

Not really. It's Haswell's internal VRM logic that may decide to provide more voltage than was requested while under heavy load, but that's not specific to adaptive voltage, it happens with manual voltage override as well. The 3 step guide to Haswell overclocking says as much that the cores getting more voltage than you configured may happen with benchmarking in particular.

I'm running adaptive voltage with 1.096V and 150.4mV offset, and that adds up to ~1.25V ... which is what I get under regular load: just testing Handbrake encoding which pushes the voltage to 1.259V, and temps at 69-78c. But running benchmarks like Intel XTU's bechmark pushes it to about 1.32V (with corresponding higher temperatures to boot: 93c max, but no throttling yet). My idle voltage is 0.852V btw, with cores idling at 30-31c.
 
Associate
Joined
11 Jun 2013
Posts
1,087
Location
Nottingham
I don't want to argue with Shiari, but this OC guide (admittedly for the Gigabyte Z87X-OC) suggests it's down to the chip, not the motherboard.

Well, you may be right. It was quite a while ago now that I've experienced my system not lowering the voltage, but I may have been using a power plan that doesn't allow for it, and I may not have been using EIST and C3 power state either .. there are a fair few variables involved in getting it all working perfectly, and at the time my focus was on figuring out how much voltage I needed to get it stable and what clocks I could get away with.
 
Back
Top Bottom