To Vdroop or not to Vdroop?

[ale91]

Hello guys,
Basically this is the situation.I am overclocking an Intel 2600k on an ASUS P8Z68-V board.
At the moment it's sitting at 4Ghz with a negative offset of -0.070 (my board overvolted it a lot...).
During gaming/emulation(main things I do) the voltage arrives to 1.23/1.24V however during Prime 95 the voltage goes down to 1.2V cause of Vdroop.
If I try to increase the negative offset the system BSODs when testing with Prime.
Now,I was tempted to use LLC to compensate for Vdroop but I read that it's kind of pointless to add voltage when using a negative offset.
So,given the situation,do you guys think that I should keep things like they are now or ignore Prime and drop the voltage even more/turn up the multiplier?
If Prime is stable at 1.2V with a 100% load I think that games will probably be fine with 1.17/1.18V,only problem is that Prime will always be unstable because of Vdroop.
What do you guys think?

 

[chris987]

over 1.2v for 4ghz sandy bridge cpu it way too much. it can easily do better! supposedly you have something better than the intel stock cooler (and a solid psu) you could hit 4.4ghz with such voltage. i would further increase multiplier, one step at a time and see if still produces such voltage outcome.if yes keep increasing multiplier to the point the voltage droping stops.then LLC shoud be increased. however, i would start over by setting a manually fixed voltage, begin with default values, increase multiplier but increase voltage only when needed.! it may be a more time consuming process but it will provide better resultes and GHZ / voltage ratio.

 

[ale91]

over 1.2v for 4ghz sandy bridge cpu it way too much. it can easily do better! supposedly you have something better than the intel stock cooler (and a solid psu) you could hit 4.4ghz with such voltage. i would further increase multiplier, one step at a time and see if still produces such voltage outcome.if yes keep increasing multiplier to the point the voltage droping stops.then LLC shoud be increased. however, i would start over by setting a manually fixed voltage, begin with default values, increase multiplier but increase voltage only when needed.! it may be a more time consuming process but it will provide better resultes and GHZ / voltage ratio.

I tried using a manual voltage,I only changed voltage and multiplier,I tried to use 1.25 and below and the system BSODs at windows logo...

 

[chris987]

I tried using a manual voltage,I only changed voltage and multiplier,I tried to use 1.25 and below and the system BSODs at windows logo...

what multiplier are we talking about? i wouldnt try 44 x 100 for a first try. as i said try with default values first. set everything on auto and try stresstesting while monitoring cpuz or real temp or any other software to examine voltage behavior. lets say is tops at 1.115v , so use that as a manual set and try stress testing again but with 39x100. then try 40x100 and increase voltage only if stresstesting says so. be petient dont stress test for a couple of minutes only. that is my approach.

my stable result : 44x100 = 4400mhz / 1.216v / LLC very high / intel speedstep disabled / also +0.05v to memory voltage for extra stabilty. i cannot get stable 45x100 even with 1.25v or above so there isnt a point to keep pushing. also my corsair H60 is a limiting factor for such high voltage.

it is important to ensure stability with defaults specs first, else there isnt a point searching for a stable overclocking anyway.!

 

[ale91]

over 1.2v for 4ghz sandy bridge cpu it way too much. it can easily do better! supposedly you have something better than the intel stock cooler (and a solid psu) you could hit 4.4ghz with such voltage. i would further increase multiplier, one step at a time and see if still produces such voltage outcome.if yes keep increasing multiplier to the point the voltage droping stops.then LLC shoud be increased. however, i would start over by setting a manually fixed voltage, begin with default values, increase multiplier but increase voltage only when needed.! it may be a more time consuming process but it will provide better resultes and GHZ / voltage ratio.

I tried using a manual voltage,I only changed voltage and multiplier,I tried to use 1.25 and below and the system BSODs at windows logo...

what multiplier are we talking about? i wouldnt try 44 x 100 for a first try. as i said try with default values first. set everything on auto and try stresstesting while monitoring cpuz or real temp or any other software to examine voltage behavior. lets say is tops at 1.115v , so use that as a manual set and try stress testing again but with 39x100. then try 40x100 and increase voltage only if stresstesting says so. be petient dont stress test for a couple of minutes only. that is my approach.

my stable result : 44x100 = 4400mhz / 1.216v / LLC very high / intel speedstep disabled / also +0.05v to memory voltage for extra stabilty. i cannot get stable 45x100 even with 1.25v or above so there isnt a point to keep pushing. also my corsair H60 is a limiting factor for such high voltage.

it is important to ensure stability with defaults specs first, else there isnt a point searching for a stable overclocking anyway.!

Thanks Chris for your answers,I am still learning so every input is welcome!
I changed C states(C1E,C3,C6)from the BIOS from auto to enabled and now things are working much better!
Here's my configuration:
Multi:42
Internal PLL Overvoltage : Disabled
LLC:Regular 0%
Phase Control : Optimized
Offset: -0.030
Everything else on Auto.
Right now Vdroop is gone after enabling Cstates,Under full load i have a Vcore of 1.25 (With everything at stock I got 1.2V @3.5Ghz so I think that the voltage is about right,also if I try to lower it Prime starts to show errors).
Temps are 47°C during Blend test with fans at 45%.
Everything seems stable.Ran Prime for an hour and Battlefield 4 for another hour.
What do you think?

Thanks again for the advices!

 

[Undying89]

chris987, Ok, lets get something clear. Sandy is 32nm chip and it can go even beyond 1.4v. For instance, my 2500k stock voltage is 1.243v.

I'm at 4.8ghz at 1.4v on air for last 2 years. This isnt an Ivy that runs hoter and limit at 1.3v. Stop scaring him, lol.

 

[ale91]

chris987, Ok, lets get something clear. Sandy is 32nm chip and it can go even beyond 1.4v. For instance, my 2500k stock voltage is 1.243v.

I'm at 4.8ghz at 1.4v on air for last 2 years. This isnt an Ivy that runs hoter and limit at 1.3v. Stop scaring him, lol.

Ahahaha no scares here.I checked online some max safe voltages before even attempting to OC and I know that until 1.4 my chip will be relatively fine.
What do you think about the OC I posted before?
I guess I could go higher but at the moment I really don't have a use for more power...

 

[Undying89]

You can easily push it higher. Sandy is indeed safe up to 1.4v if you keep your temps under control and thats not hard with this chip. Ivy and Haswell are much hotter.

I'll say try how high can you go and find that wall. Also you try reducing PLL from 1.8v to 1.7/1.6. It should help with temps and maybe allow you reduce vcore little more.

 

[ale91]

You can easily push it higher. Sandy is indeed safe up to 1.4v if you keep your temps under control and thats not hard with this chip. Ivy and Haswell are much hotter.

I'll say try how high can you go and find that wall. Also you try reducing PLL from 1.8v to 1.7/1.6. It should help with temps and maybe allow you reduce vcore little more.

UPDATE:I raised my Offset to -0.020 because blend was stable but small fft was giving me errors.
I have a question,should I really be basing my evaluations on Prime? I mainly game on this PC and I will never use 100% of my CPU so I don't see the point in basing everything on a load that I will never achieve,as I said before Battlefield ran fine...

 

[Undying89]

Its up to you. If you think you can afford not having 100% stable system and crash at some point without loosing your daily work then no need.

 

[ale91]

Its up to you. If you think you can afford not having 100% stable system and crash at some point without loosing your daily work then no need.

Thanks undying,you are very helpful,do you have any idea why enabling C states helped so much reducing Vdroop?

 

[Undying89]

Its up to you. If you think you can afford not having 100% stable system and crash at some point without loosing your daily work then no need.

Thanks undying,you are very helpful,do you have any idea why enabling C states helped so much reducing Vdroop?

Well offset overclocking kinda requires c states to be disabled so i have no idea...

It may couse bsod while idle when enabled with offset.

 

[chris987]

chris987, Ok, lets get something clear. Sandy is 32nm chip and it can go even beyond 1.4v. For instance, my 2500k stock voltage is 1.243v.

I'm at 4.8ghz at 1.4v on air for last 2 years. This isnt an Ivy that runs hoter and limit at 1.3v. Stop scaring him, lol.

hhmm, correct, that may have sounded scary afterall and yes sandy bridge is a way better overclocker! i should mention something higher as default voltage.! however was trying to focus on the process and not the actual numbers not kwowing anything about the cpu cooler! having a good starting point will eventually lead into more fine tweaking!