Intel Haswell i7-4770K IPC and Overclocking Review @ [H]

Prime95 is crazy on these processors. Even with manually controlled voltage (i.e. not adaptive etc) my CPU will hit 100*C within minutes (if not seconds). I have resorted to using the IETU and AIDA64.

At 4.4GHz AIDA64 fell over on me. I am now down to 4.3GHz. Going to run AIDA overnight. If it's still standing in the morning then I think I will accept my fate and leave it at that.

Can't win them all I guess...
 
Syphon Filter said:
Can't win them all I guess...
Click to expand...
I went through 8 4770Ks before I found this one that runs rock solid (prime95 28.5stable) at 4.6Ghz. Most of them were only stable at 4.2Ghz or 4.3. You could tell right away this one was different because the first time it posted I went into the BIOS and it was at a much lower default voltage than I was used to seeing. The closer to 1.0v when you first your chip in the better, and this one runs at below that. like .976 or something iirc.
 
I don't think that's a reliable indicator. Pretty sure mine was 1.05v or so on the first boot.

@rasta - the OCN community average is 4.5 at 1.3v (I'm rounding of course). If you beat that, then yeah you're doing well. 4.6 at 1.2v is ridiculous though, not even mine does that.

My scaling is 1.188, 1.235, 1.284 (for 4.5, 4.6, and 4.7Ghz respectively). I suppose if people were reporting their BIOS settings or VID, maybe it'd be a bit more deceiving (1.175, 1.215, 1.265 respectively, in my case). Anyone who does that isn't being honest though, since you typically see a 0.02v bump at load anyway.
 
Mav451 said:
I don't think that's a reliable indicator. Pretty sure mine was 1.05v or so on the first boot.

@rasta - the OCN community average is 4.5 at 1.3v (I'm rounding of course). If you beat that, then yeah you're doing well. 4.6 at 1.2v is ridiculous though, not even mine does that.

My scaling is 1.188, 1.235, 1.284 (for 4.5, 4.6, and 4.7Ghz respectively). I suppose if people were reporting their BIOS settings or VID, maybe it'd be a bit more deceiving (1.175, 1.215, 1.265 respectively, in my case). Anyone who does that isn't being honest though, since you typically see a 0.02v bump at load anyway.
Click to expand...

I thought it was an established "ball-park" indicator. But comparing initial UEFI voltage from an i5 and an i7 seems flawed; more cache, higher base clock... I actually use HT, and the thing is prime95 28.5 stable. Most of the other chips I've tried have been junk compared to this one.
 
Last edited:
I'll have to try out 1.3V some time then, but I'm afraid my cooler won't be able to handle it. How are you guys hitting 100C? My system always crashes way before that point.

Also, how important is VRing in the OC?
My current setup is:

BCLK 100
CPU multi 43 dynamic
Ring ratio 41 (chosen arbitrarily)
VCore 1.155 adaptive
VRing 1.115 adaptive
HT on
SpeedStep on
RAM at spec (specs in sig)

I've hit 85C on P95 v27.9. I'm quite afraid of going higher.
 
rastaban said:
I'll have to try out 1.3V some time then, but I'm afraid my cooler won't be able to handle it. How are you guys hitting 100C? My system always crashes way before that point.

Also, how important is VRing in the OC?
My current setup is:

BCLK 100
CPU multi 43 dynamic
Ring ratio 41 (chosen arbitrarily)
VCore 1.155 adaptive
VRing 1.115 adaptive
HT on
SpeedStep on
RAM at spec (specs in sig)

I've hit 85C on P95 v27.9. I'm quite afraid of going higher.
Click to expand...

You're using adaptive voltage with prime 95?
Only use manual with stress tests. I use manual voltage 24/7, but I figured out how to make it so it still downclocks and undervolts as if it were in adaptive state. Z87 is dumb. I'm curious if z97 will have this crap fixed. I doubt it. I think it's a problem with the chip architecture itself.
 
Ah good point. I thought adaptive would run cooler with normal use so kept it at that while doing the test. I'll probably get around to retesting at the end of next week.
 
Well I set changed the voltage controls to Override instead of Adaptive, but the results are the same as far as I can tell.
 
Syphon Filter said:
It's not MORE heat though it's just concentrated/distributed differently.

Kor is 100% right. You CANNOT produce more heat from less Watts. How that heat is dissipated is the issue here.
Click to expand...

That is it.

The die size is smaller and less contact area with heat sink/water block to dissipate heat from the die to the heat sink/block. That is why the smaller die that uses less volts and amps under load heats up more... under load the heat gets trapped like a sponge and quickly over heats because the heat transfer area is to small for heat sinks or water blocks to quickly transfer the heat.

To tackle this problem Intel will use an advanced Next Generation Polymer Thermal Insulate Material and I read this
NGPTIM is backstopped by a slight silicon layout revision which works to spread hot-running transistors across a slightly broader area. These two additions should guarantee lower temperatures relative to first generation Haswell processors.
Click to expand...
From here: http://www.hardwarecanucks.com/foru...-devils-canyon-i7-4790k-i5-4690k-preview.html

I think Devils Canyon will not have the transistors sticking out on the top surface of the substrate like the 4770ks had... And the Die will be more flush with this back layer of silicon to help increase the area size of thermal transfer.

I plan on delidding DC as soon as I get one and it will be fun to see how the design has changed. It think these claimed design improvements for cooling will also help for bare die cooling which is what I plan on doing. I hope hardwarecanucks is correct.
 
Last edited:
Someone should type up a thread with first post describing rate of heat vs temp vs die size vs etc, so we can avoiding this discussion for the 1000th time ...
 
I just want everyone to know here that my 4770k wasn't stable over 4.2ghz on my Asus Z87 but on my new Z97 mobo I can now do 4.5ghz stable.


strange..............:confused:
 
TroyX said:
I just want everyone to know here that my 4770k wasn't stable over 4.2ghz on my Asus Z87 but on my new Z97 mobo I can now do 4.5ghz stable.


strange..............:confused:
Click to expand...

What brand for the Z97, also Asus?
 
Better late than never...

My i4770k (had it for ~ 9 months), on an Asus Z-87 Pro, NZXT Kraken (280mm) AIO, just using Asus AI Suite III to do the overclocking (stop laughing!!), 2 cores at 4.4, 2 at 4.5. Vcore at 1.26.

Rock solid. Stressing the CPU's, the max temp I've ever seen is 61C. (I delidded, and used liquid metal pro.)

I could possibly run it higher, but this is good enough for my purposes.

Ken
 
I run my chip at 4.3Ghz constantly (according to CPU-Z) - just from selecting the "Performance" basic setting in the BIOS. No other tweaking needed.
 
c3k said:
(I delidded, and used liquid metal pro.)
Click to expand...

Hey hey hey, I got this stuff, and need to know, does it last forever? Or will it need to be taken apart and redone in the future?

Wolfdale75 said:
I run my chip at 4.3Ghz constantly (according to CPU-Z) - just from selecting the "Performance" basic setting in the BIOS. No other tweaking needed.
Click to expand...

The important setting is the CPU minimum state. I use 100% for gaming (4.5), and 75% for desktop (3.1) save 10 watts or so, more importantly lowing voltage from 1.21 to 0.95 for idling. Alter the balanced 5% default for convenience.

I wish there was a button on my keyboard to quickly toggle that. (And I web searched, there is!)
http://www.howtogeek.com/howto/windows-vista/create-a-shortcut-or-hotkey-to-switch-power-plans/
 
Last edited:
Syphon Filter said:
I'm about ready to throw the towel in with my 4770k.

I can't get stable at 4.4GHz with the following settings

VCore - 1.30V
VCache - 1.25V
SA - 1.2V
DIO - 1.2V
AIO - 1.2V
SVID - Disabled

Speed-step disabled
LLC of about 50%
Phase control set to extreme
Current capability of 120%

I think 4.3GHz is going to be the limit of my chip and I believe it will sustain that at about 1.2875V (give or take). Failing that, 4.2GHz it is :(

Lost the lottery this time I guess.
Click to expand...


CPU Uncore 40 / 1.2v [wont need more than 1.10 final]

SA .95
AIO 1.015
DIO 1.075

Svid - Auto


_________________

Put your VCCIN to 1.9v

Make sure your ram is at xmp or lower , and make sure all spread spectrum & speedstep & power C1 C 7 etc on your cpu are off.

Up your line level calibration to one notch under max - which is 100%

Make sure your cpu thermal limit is at 120/130%


on my asus board its level 9 level 8 is 100%

Your VTT must be HALF of what your ram volts are which on auto it should auto do. When running my ram at 2133 /1.5v it is .7550 VTT

Make sure your gpu mult is at 22-24 [dunno if it shows if not using builtin igp, set it anyways if it does]..





And yeah skip prime, only use Intel burn test v2 on standard for starters then higher as stability is established. If you can run burntest on standard drop vcore down .15v each time, then try a higher heat setting
-Raise it back .15 if it fails of course.
Final only change may be your vccin from 1.95 to 1.90 to 1.875 to 1.85 > > >> Depends what is stable once all else is set write the numbers down once figured out what settings are it for your board/cpu.

Only your vCore and cache will be lowered if these settings stick for 4.4 / 4.5.
Only VTT and VCCIN will maybe need changed after such, possibly up your uncore if you want later.
Is your cooler and grease seated and done right. . Disable hyperthreading
If you cant run 4.4 at 1.25 then you cant run your cache at 44x 1.25 do 40 or 42 you want it at least .5v less than your vcore max .25v less

These settings have my 4670 at 4.5ghz w/1.2v on air
(1.215 actual)

Yeah I delidded , but did it before that too just hotter

Report Back how it went.
 
Last edited:
You must log in or register to reply here.
Back
Top