noob to overclocking but surely somethings wrong i5 3570k

Christurbo

Christurbo

Christurbo

Associate
Joined
3 Apr 2013
Posts
190
Location
Northwest uk
hi guys this is what i have atm.

VTX3D*HD 7870 Black Boost Edition 2048MB GDDR5 PCI-Express Graphics Card
Intel*Core i5-3570K 3.40GHz (Ivybridge)
Kingston*120GB SSDNow V+200 Drive SATA 6Gb/s Solid State Hard Drive
Cooler Master*CM-690 II Advanced USB3.0 Windowed Dominator Case
Gigabyte*Z77-D3H Intel Z77 (Socket 1155) DDR3 Motherboard
Western Digital*Caviar Red 1TB SATA 6Gb/s
Corsair*Builder Series CX 500w Modular '80 Plus Bronze' Power Supply
Adata*XPG Gaming v1.0 8GB (2x4GB) DDR3 PC3-12800C9 1600MHz
Alpenföhn*Matterhorn*cpu cooler

Right guys useing
coretemp
Prime95
And cpu id"I think its called that"

Whilst doing a stress test I noticed
it was at roughly 3595 clock on coretemp
at around 1.225v
and was getting into the 70s c

And also on cpuid I noticed the clock was set to 3.6ghz

Know I always thought the 3570k was a turbo model thing
3.4 normal and on turbo was 3.8

Do you recon my settings are right and is it easy enough to xhange just the turbo settings to lets say 4.2ghz whilst putting the voltage on auto?

As I dont game all the time and watch a lot off films and browse ect do dont want it running at 4.2ghz all the time
it usually sits around 1.6-1.8ghz on idle usually

sorry for all the questions

Kind regards chris
 
Last edited:
Don't let the voltage on Auto, it is usually unnecessary high. With this cooler 1.25-1.3V and frequency from 4.4 to 4.6GHz should be no problem whatsoever.
Start with manual fixed voltage and then play with the LLC settings, get a feel for them and set the voltage using the offset option when you find your stable settings. Double check for stability afterwards because sometimes it wants a bit more voltage using offset vs fixed as the voltage fluctuates and sometimes does not climb high enough to keep the system stable.

I would start with a frequency as low as 4.2GHz and find the lowest voltage it is stable with (say 12-24 hours of Prime95 on blend, or OCCT Linpack+AVX but the second one will be too hot and still might not catch errors). Then try and find the next stable voltage for 4.3GHz. Usually the increase needed for each additional 100MHz is somewhat linear, say +0.06V. It can vary from chip to chip, say 0.05, or 0.07 but it is roughly linear up to one point. It is still a nice target to check so you do not go through every voltage step you have at your disposal. From that point onwards it starts to need exponentially higher step for the next 100MHz and this is where you have to stop looking for more as there is none left.

Temperatures you have to be aiming at (on my opinion) are high 90s for stress testing - under 100 in general. This will give you 60-80 degrees maximum under normal applications - gaming, rendering, anything else you use daily.
Safe voltage levels are very high - everything under 1.45V is save if you can keep the temps down.
P.S. Regarding your initial question about the weird frequencies, try loading optimised defaults, it might be bugged auto settings of the UEFI.
 
Last edited:
Have you scouted your BIOS at all? :) SB/IB/Haswell overclock is dead simple. You have only a few things you can fiddle with - multiplier, voltage (which has two options - fixed manual mode in which you set up say 1.2V and it will stay at 1.2V all the time, or offset mode which is set with +/- values from the default VID of the processor, and will boost up to your desired voltage only on load), LLC (load line calibration which is an option for the offset mode of voltage control and although it is implemented differently by all the manufacturers its purpose is to try and stabilise the voltage upon load and bring it as close to the setting you want to achieve) and that's pretty much it. You have a lot of other stuff that do not come handy until extreme overclocking so no need to discuss them.

First step would be to set up manually the voltage at a fixed starting point, say 1.15V or 1.2V and crank the multiplier to 42. Then you need Prime95 or OCCT to test for stability. There is a very nice setting you can use for fast error detection with prime95 but it works only with version 27.7 beta2 and it is the following. Set up the test on custom with min size 480 and max size 480, check the run ffts in place field and that's pretty much it. If the test does not fail in any way (one of the threads stopping, the system freezing/becoming inresponsive, restarting, or in any way behaving differently from normal) after 30 minutes, it is most probably stable at those settings. Once you find settings that are rock solid for 30 minutes you can go and test for a prolonged period of time - 12+ hours with Prime95 blend or OCCT Linpack+AVX with 90% memory usage.
In order to have the multiplier still dropping while in idle you need C1E and Intel Speedstep enabled.

After you have found some voltage and frequency settings you like with fixed voltage you can start playing with the offset so you get the same voltage under load and play with the LLC levels. It is a lot easier to do than to describe actually. The difference between 4.2 and 4.5 if your particular chip is good is marginal in terms of voltage increase, themperature increase and power consumption increase but I agree 4.2GHz is a good starting point. If your chip can do 4.2GHz at 1.1V you are obviously going to want to get more frequency. :)

If you are feeling even lazier than this you can give some BIOS screenshots and we can give you settings to try out.
 
Last edited:
Thanks for being very detailed
I watch a youtube vid last night with my mb and chip so think I have a rough idea , think ill play around and use prime 95 think I have the one u mentioned
currently trying to sort a few thinks out aswell as on my mb z77 d3h
For some reason I cant load the bios whilst the gpu is plugged in, will only load bios from pc lead from the motherboard
Ive read that the bios might need updating but in other hand some are saying I need to set the pci as default load up
So going to sort that first as will make overclocking a lot easyer instead off keep having to take power off the gpu and rebooting it.
 
If the graphics is set to Auto, upon booting it should select the device wherever you have plugged your display. I would leave it on Auto as if something goes wrong with your video card you can always plug it into the iGPU output of the motherboard and manage to boot into the bios easily without clearing the CMOS. Similarly if you unplug the graphics card from the PCIe slot and plug your display into the mobo it will boot no problem from the iGPU when left on Auto. If it ain't broken, don't fix it. :) Things do not stand the same way for voltage settings though, there you need as much control as you can get.
P.S. There is a newer version of Prime95 that has AVX so I would stick to it for the prolonged stability tests as the OCCT is really really hot, I would say unnecessary hot, not to comment the Linx, IBT and the rest stuff that sets the cpu on fire without being that specific about catching errors.
 
Last edited:
Well ive found what I think here sorry to link to another site mods
But he has the same cpu , gfx card motherboard ect http://forums.****.co.uk/showthread.php/125513-PC-wont-boot-when-GPU-connected
A
R
I
A
had to update bios

Ill try sorting this tonight and then ill try sorting the overclocking
Going to be a lonngggg night haha
 
Last edited:
Obviously you don't have to do it at once. Try fiddling with the settings so you know what they do. After you have partial stability with the 15-30min 480-480 method you can leave a long stress test and go to bed. :) And iteratively continue tomorrow. Try not to copy other people's settings as every chip is different and they will most probably not work on your system. Get to know the chip and the OC is very very easy.
 
right well ,
went round to my mates after it not loading back up after turning off internal graphics
and wouldnt boot up on gfx card
went his booted up straight away on his monitors
so we have come to a decition that its my so called smart tv thats messing everything up
any how overclocking wise
turbo clocks on 4.2 each core,
turbo power limit to 250, core current limit to 250.
cpu pll @ 1.75v
Dynamic vcore(DVID): +0.050
everything else i left alone on out i think

only ran prime 95 for about an hour with 0 errors and 0 warnings

However
the voltage on coretemp was around 1.33-1.35vs
and the Max temps in this hour were roughly off top off my head
85
87 -- This being the concerning one
79
83

any advice or anything
first time doing it so was happy i suppose that it didnt crash ect

also the ram was set to 1333
which i changed to match mine @ 1600mhz

kind regards chris
 
Have you tried lowering the voltage on the same frequency? 87 degrees is not something dramatically high but still I am sure you can do 4.2GHz below 1.2V. My chip is quite bad and at 1.3V I am on 4.5-4.6GHz already. Also 1 hour of prime is not enough to be sure that you're not gonna get blue screens while gaming, have it in mind.
 
Oh yer I know an hours not enough
What should I lower then ?
the cpu pll from 1.75v to 1.65v?
It used to run on idle about 1.15vs @1600
Now its about 1.09 @1600
 
Leave the PLL at default, and I mean Auto, do not touch it, you can kill the CPU if you increase it too much. You need to lower the offset voltage, you have described it as Dynamic Vcore (DVID). Also I recommend you to try and set the voltage to manual fixed value, it's easier when you are not sure what you're doing. Load the XMP profiles for your memory and leave it on default for now.
 
Well the cpu pll on auto was set at 1.8vs so really ive lowered that.
Ill just have a play around and what not
Dont really want to set a voltage as then it will be at that the whole time instead off lowering and rising to suit gaming or idle / movies ect
 
Christurbo said:
Well the cpu pll on auto was set at 1.8vs so really ive lowered that.
Ill just have a play around and what not
Dont really want to set a voltage as then it will be at that the whole time instead off lowering and rising to suit gaming or idle / movies ect
Click to expand...

If you set it manually while finding the optimum one, there is a chance it will take you less time/nerves. Overall it makes no difference except the energy saving and potentially putting less strain on your chip in the long term. Leave the pll on 1.8, it might induce instability if lowered. If I recall correctly the maximum recommended for the pll was 1.9V. You can read more details about Ivy Bridge overclocking here. The values in the table are from Intel documentation and can be trusted fully about what is safe and what is not.
 
Last edited:
just try lowering the dvid more so maybe +0.040v

power limits should be fine on auto aswell,ive never had to mess with those

so just stress test,if it passes lower dvid a few clicks more and play some games as it can pass stress tests and crash to desktop in gaming if cpu v is too low

keep looking in windows event viewer for any warnings/errors that are whea parity 19 (cpu)

raise dvid/cpu v till theres none
 
Ill give it a go 2night , its mainly getting thos temps down
I cant really see 4.2ghz being a problrm really not like im trying for 4.6-5 ect lol
 
Dunno if it got addressed in above as theres quite a bit to read but regarding the 3.6GHz thing on default settings the clock speed will vary between the base and boost clock depending on load and how its spread over the cores it won't just be at either 3.4 or 3.8 but will also use values between them.
 
Rroff said:
Dunno if it got addressed in above as theres quite a bit to read but regarding the 3.6GHz thing on default settings the clock speed will vary between the base and boost clock depending on load and how its spread over the cores it won't just be at either 3.4 or 3.8 but will also use values between them.
Click to expand...

True but most of the manufacturers (MSI and Asus for sure) cheat and put the turbo boost clocks at all the cores and are quite aggressive on the load-clock speed relationship. I though Gigabyte is in the game as well but obviously they aren't.
 
welllll guys just noticed something :s
on
- coretemp 1.0rc6 my VID is @ 1.341v's which is what i stated earlyer i think however i didnt notice but on
-CPU-Z it shows it @ 1.272v's

now im really confused :s
i also lowered from +0.050 to +0.040
didnt seem to make any difference but i did change that before i notices the above on cpuz

any light?
 
You must log in or register to reply here.
Back
Top Bottom