4.8Ghz at 1.4v what settings?

[borisvodofsky]

Yeah I did about a rice grain but stretched it out vertically along the whole cpu.

It's freezing outside but I have all the windows closed so I've opened them and it will be much cooler.

If you have a box fan. use it like a case fan, but for your room

 

[0dervish0]

Arhhhh!! So I have all the case fans maxed out, cou fan full speed, case fully open, all my windows in my room open, door open, two external fans and the temp next to the case was 22 C. Then I run IBT and 1 minute in I get BSOD??? I'm so confused now I'm at 1.44v and still getting problems while 1.38v passed 6 hours of prime95. I'm thinking maybe my RAM need more voltage? As IBT tested the RAM pretty heavily too not just the CPU, while small fft in prime95 works on the CPU mainly.

 

[Kenmitch]

Arhhhh!! So I have all the case fans maxed out, cou fan full speed, case fully open, all my windows in my room open, door open, two external fans and the temp next to the case was 22 C. Then I run IBT and 1 minute in I get BSOD??? I'm so confused now I'm at 1.44v and still getting problems while 1.38v passed 6 hours of prime95. I'm thinking maybe my RAM need more voltage? As IBT tested the RAM pretty heavily too not just the CPU, while small fft in prime95 works on the CPU mainly.

I pre-tested my 2550k with the stock cooler at 4.8ghz using Intel Burn Test. It throttled as one would expect but lives to tell about it. When you don't play around with all the c-states the chip will just throttle down speed and voltage to save itself from damage or early demise! I've had my chip all the up to 5.5ghz and haven't touched some of the settings you seem to be thinking you need to use.

I'd suggest starting over from scratch if you want to get 4.8ghz stable.

Just save your current profile so you don't loose nothing but some time.

If your goal is to fry your chip then carry on

If you want to start over and try again. Then I'll make some suggestions for you.

Uninstal the AXTU software as it's useless in my testing. Makes my RealTemp readings whacky!

Try latest version of RealTemp to monitor temps. (Doesn't like AXTU on my rig) If you click on the time box it will show watts which will give you an idea of how much power your waisting stress testing

In the following screen shot the setting can be at defaults for 4.8ghz. Internal PLL overvoltage may or may not be needed for your chip. If it doesn't need it to boot at 48x then it won't do much for stability if you have it enabled. At least from my testing. The short/long duration and current limit don't need to be set that high for only 4.8ghz but doubt it would hurt anything as your chip won't be pumping that much juice thru it at 4.8ghz anyways.

Take the additional turbo voltage off auto and give it a value of 0.050v's or so to start. This is how you'll want to take control of your loaded vcore. Trial and error.

This page needs some work it looks like. The hard locked overclock of the past isn't needed for 4.8ghz unless your chip is a dud or possibly damaged from your adventure so far.

I'd suggest the following modifications to your settings as shown in the photo as a starting point for 4.8ghz to see what happens. Just change what I say otherwise leave at settings in photos.

Switch vcore to offset mode. Set value at 0.005v's
Bump your memory voltage up to about 1.54-1.55v's as your slots are full
Try VTT around 1.077v's
PLL change to closest value around 1.8v's for now. You can play with this later if needed.
Load-Line try around level 3 for now. Seems to be all thats needed until 5ghz or higher for me anyways.

Raping your chip won't help you achieve your overclocking goal

Change only the following settings.

C1E to enabled
Thermal throttling to enabled

Keeping the C3 and C6 states disabled seems to help out with the single, dual core load bsods that alot of users get using offset vcore voltage. Strange the way it works but can see why it does. By default the chip will want a vcore based not only on load but how many cores are loaded. If you have a single core load the vcore will be way lower than with a 4 core load. This is what causes the random bsods doing things like surfing the web, email, etc....But on the other had rig will be stable during stress testing like prime, IBT, etc.

Doing the above will only take you a couple of minutes for a try and see if it'll work for you or not. Saving your original overclock profile gets you back to where you started with a reboot.

I'd think it'll boot into windows with the suggested settings I've given you. Guess the important thing is the vcore as I'm not sure how yours will react to these settings. If by chance you bsod just bump up the additonal turbo voltage about 0.010v's or so and try again.

Once in windows you'll need to do a quick check using cpu-z and Intel Burn Test to see what vcore is being fed to your chip under load. Just use IBT's default settings to start and make sure you monitor your temps!

 

[Don Karnage]

I thought sandy bridge overclocked so much better then Ivy Bridge?

 

[borisvodofsky]

I thought sandy bridge overclocked so much better then Ivy Bridge?

It does !

y u selling ur ivy then hun....?

bet ur just itching to get some sandy action

 

[borisvodofsky]

Increasing the Vccio to 1.55 is a good idea. You may want to try 1.7125 and 1.8125 for cpu pll voltage... Some people reported lowered vcore required for their overclocks when they used a "lower" cpu pll voltage.

It makes no sense and but yea. that's what I read yesturday in another forum.

 

[Kenmitch]

Increasing the Vccio to 1.55 is a good idea. You may want to try 1.7125 and 1.8125 for cpu pll voltage... Some people reported lowered vcore required for their overclocks when they used a "lower" cpu pll voltage.

It makes no sense and but yea. that's what I read yesturday in another forum.

Vccio,VTT,PLL are variables just like vcore. Some chips need adjustments after a certain clockspeed some don't. Best starting point is auto to get an idea. If bsods don't respond to vcore bumps looking at these voltages should be next. Sometimes a bump in VTT is all it takes to stabilize an over clock. The chip will have a sweet spot for all of them which can very by overclocked speed. Trial and error is the key.

 

[guskline]

I use Intel Burn test instead of Prime 95 first. I run IBT extreme 10 times and if it passes I go on to Prime95.

It's one thing to tweak the H - E double hockey sticks out of these CPUs to see how far they will go. It's another thing to get a solid OC that you can run 24/7/365. I opted for a decent OC that I could run IBT, Prime95, OCCT, or you name it, forever and the CPU would run solid with good temps. My 2500ks are set at 4532 (103 x44) and give me great performance. I could tweak some more but would rather spend the time gaming.

BTW 0dervish0, your computer specs in your sig are impressive.

 

[0dervish0]

Kenmitch:

Ok I will try punching those settings in and see how it works. So should I start with a low vcore again such as 1.37v in the bios? and if that doesn't work try adjusting the Vccio,VTT,PLL voltages? What temps should I be trying to stay under in IBT?

I might sound like a noob and thats because I am haha, I only built this system about 6 months ago and before that I never built my own desktop so trying to learn as much as I can. Built it mainly for photoshop and now trying to push a tad more out of it as I've ordered a Nikon D800E and those 36.3mp raw files need a lot of juice to push around. I can always just put it back to 4.5Ghz were I had it stable before. If you want a laugh watch this timelapse I made of me building my system:

http://youtu.be/A05Wg_ADIzg?hd=1

guskline:

Yeah i'm probably wasting my time trying to get 4.8Ghz and I should just put it back at 4.5Ghz were it was running fine. I need to be spending this time studying for my exams :/

Is that sarcasm? If not thanks.

Thanks everyone for all your great advice, I really appreciate it.

 

[Don Karnage]

It does !

y u selling ur ivy then hun....?

I can get a 3930K for 197.00 in 2 days thats why

 

[borisvodofsky]

I can get a 3930K for 197.00 in 2 days thats why

the point is,, you're jumping ship because the one you're on was leaking from the start

And I just wanted to point that out to make you feel bad.

 

[Don Karnage]

the point is,, you're jumping ship because the one you're on was leaking from the start

And I just wanted to point that out to make you feel bad.

Not really. I just bought into the hype thats all. Thought it was going to be this amazing 5.2+ghz monster when its more of a kitty.

 

[Kenmitch]

Kenmitch:

Ok I will try punching those settings in and see how it works. So should I start with a low vcore again such as 1.37v in the bios? and if that doesn't work try adjusting the Vccio,VTT,PLL voltages? What temps should I be trying to stay under in IBT?

I might sound like a noob and thats because I am haha, I only built this system about 6 months ago and before that I never built my own desktop so trying to learn as much as I can. Built it mainly for photoshop and now trying to push a tad more out of it as I've ordered a Nikon D800E and those 36.3mp raw files need a lot of juice to push around. I can always just put it back to 4.5Ghz were I had it stable before. If you want a laugh watch this timelapse I made of me building my system:

http://youtu.be/A05Wg_ADIzg?hd=1

guskline:

Yeah i'm probably wasting my time trying to get 4.8Ghz and I should just put it back at 4.5Ghz were it was running fine. I need to be spending this time studying for my exams :/

Is that sarcasm? If not thanks.

Thanks everyone for all your great advice, I really appreciate it.

Try the offset voltage way. Not sure how much additional Turbo voltage you'll need to get 1.xx you'll have to try what I stated and hit it with IBT to see what you get. Then just up or down the offset accordingly. Try stay under 80*c Max for IBT. Normal use won't even be close to IBT temps.

 

[0dervish0]

Kenmitch:

So I did exactly what you said and these are the settings I changed:

Additional Turbo Voltage: + 0.051v
OFFSET VOLTAGE: + 0.005V
DRAM: 1.545V
VTT: 1.077V
PLL: 1.791V
CPU Load-Line Calibration: Level 3
Enhanced Halt State (CLE): Enabled

and uninstalled axtu and installed the latest version of Realtemp.

So windows booted fine, temps were around 38 degrees max, then I ran IBT with the standard settings (10 passes) and I did 6 passes till the temps reached 87 degrees and I stopped it. The thing is the vcore in OCCT was showing 1.38v-1.4v which is a lot higher then before.

Here is a screen shot of OCCT after I stopped IBT but you can see the max temps for all the cores.

 

[0dervish0]

So pretty much should I just lower the Turbo Voltage?

 

[borisvodofsky]

So pretty much should I just lower the Turbo Voltage?

yes, if you want lower voltages with this llc, you would lower the turbo.

the issue is, you may infact need up to 1.40 to hit 4.8. which is normal.

my voltage goes up to 1.36 - maybe for small durations to 1.37 for 4.8 during ibt

 

[0dervish0]

So I lowered the Additional Turbo Voltage to 0.043v from 0.051v and it passed 10 runs. The temps eventually reached 85 degrees.

So should I now continue to lower the turbo voltage and see what is the lowest I can get? Should I play with the voltage offset (-0.005v?).

 

[0dervish0]

+ 0.027v (turbo) passed IBT with OCCT vcore readings of 1.37v-1.38v at a max temp of 81 C. Should I just keep on reducing the turbo voltage until it BSOD?

 

[0dervish0]

0.012v (turbo) passed IBT in standard settings but the temps aren't really changing still maxing out around 80 degrees and cvore is still around 1.37v.

 

[Kenmitch]

So I lowered the Additional Turbo Voltage to 0.043v from 0.051v and it passed 10 runs. The temps eventually reached 85 degrees.

So should I now continue to lower the turbo voltage and see what is the lowest I can get? Should I play with the voltage offset (-0.005v?).

Did you enable the CPU Thermal Throttling setting? You need this enabled as a just in case your cooler can't hang.

Your on the right track now it looks like. The additional turbo voltage value I gave you was just to see what your rig would get with loaded vcore. Dropping this value will reduce your loaded vcore. Dropping the value 0.010v's doesn't always get you 0.010v's less vcore tho. Trial and error will get you were you need to be.

The voltage offset value is for idle voltage. Your rig does throttle down to 1.6ghz with vcore around 0.980v's or so now doesn't it? Changing this value will only drop the idle voltage.

The AXTU software was most likely messing with the OCCT readings. The values OCCT is giving you seem to be correct now.

Intel Burn Test is killer on the temps as you've seen. In the low 80's would be fine during IBT. You should test with more memory tho at least go up to the 4096mb setting as you can get bsods when using more memory. Not sure about max memory as you have 16gb's so maybe try it with 8096mb or so....Unless you can get the temps inline as it'll take a long time to complete even 10 runs.

Try some gaming out and everday use and see what your max temps are. Most likely they will be about 15-20 degrees lower or possibly more anyways.

You'll wanna do some more stress testing before you start dropping the vcore or alter other voltages. Try for a overnight prime run and see.

You might be able to drop the PLL voltage and vcore a little bit more and maybe shave off a couple more degrees during IBT testing. Once you make sure your stable then you could drop PLL down to the next setting and test again.

When playing around with voltages you can somewhat get an idea of it hurts or helps performance by using IBT and watching the GFlop output. Using the high or very high settings seem to be more consistant than the standard setting. For example you may get 117Gflops at 1.36v's but get 119Gflops at 1.38v's. If you do the standard, high, very high settings you should see something like 115Gflops, 119Gflops, 124Gflop for example. Sometimes just a little bit more juice to the memory will make it scale better. I only run with 2 4gb sticks as I don't need more and it seems like all slots full gives more issues with max overclocking anyways.

If you decide your cooler can't hang at 4.8ghz and go for a lower overclock it would be best to base it off the settings your using now. Just drop the multiplier and see what your vcore is. Drop the additional turbo value to lower it. If too high still then you could try using a -0.0xx value on the vcore also. Only other option is dropping the LLC down a notch. If your ASrock board is like mine I get no -0.xxx values for additional turbo voltage which comes in handy at lower overclocks.

 

[borisvodofsky]

No, do not lower your turbo voltage drastically, because the chip is passing IBT does not mean it can pass Long Prime sessions.

For 4.8ghz, you're not going to go much below 1.35 on PRIME 95 loads.

So if you're heat bound at "that" voltage for IBT, that's ok, because IBT doesn't represent real life.

So, now you know you need at least xxx voltage to do the IBT, that is the preliminary baseline we have.


From here you have to use Prime 95 custom test.

under custom, for minimum use 1792k, and leave max alone, put it 35min in the test length.

Chances are you have to start increasing voltages from here, instead of decreasing them.

THIS is if you're still pushing 48,, because if we're going back to 45, then you may need significantly less voltage.


I am considering this move myself, because I have some render work coming in, and the 300 mhz was really just for show n pride. :biggrin:

 

[0dervish0]

Thanks so much for your help everyone! I will continue stress testing till I find the sweet spot

 

[BD231]

Kenmitch:

So I did exactly what you said and these are the settings I changed:

Additional Turbo Voltage: + 0.051v
OFFSET VOLTAGE: + 0.005V
DRAM: 1.545V
VTT: 1.077V
PLL: 1.791V
CPU Load-Line Calibration: Level 3
Enhanced Halt State (CLE): Enabled

and uninstalled axtu and installed the latest version of Realtemp.

So windows booted fine, temps were around 38 degrees max, then I ran IBT with the standard settings (10 passes) and I did 6 passes till the temps reached 87 degrees and I stopped it. The thing is the vcore in OCCT was showing 1.38v-1.4v which is a lot higher then before.

Here is a screen shot of OCCT after I stopped IBT but you can see the max temps for all the cores.

Ratio=47x
PLL Overvoltage=Enabled
CPU Voltage= 1.35v to 1.375v *ACTUAL*
PLL Voltage=1.7v
CPU Spread Spectrum=Disabled
V-Core PWM Mode=Extreme
VRM fixed=350
Phase Control=Extreme
LLC Calibration = 75%
All c-states=On

That should at least get you 4.6-4.7 stable and bring your temps down a ton.

Dumas

 

[borisvodofsky]

Dumas

ehh... if all the power state settings were enabled one may well need a really high llc to get the voltages up there.

That said,, I despise those c1,c3,c6 states. diaf :biggrin:

 

[0dervish0]

I haven't had time to check for proper stability with prime95 but I recently got a random BSOD when browsing the net with the code 0x50. So I think this means "RAM timings/Frequency or uncore multi unstable, increase RAM voltage or adjust QPI/VTT, or lower uncore if you're higher than 2x". Any advice what I should do?

Thanks