i5 6600K / i7 6700K OC results

[Woomack]

Many users never test at stock as they assume it's stable but later bump clocks and try stability at overclocked settings. We look at the hardware in a bit different way

 

[EarthDog]

Meh, Realbench is still plenty viable.

 

[wingman99]

I used Realbench stress test and lowered my core voltage - 0.020v from the original +0.075v and it passed, however prime95 failed in round off checking. So much for Realbench for me.

 

[Kenrou]

Strange, Prime95 passes for me ?

 

[EarthDog]

I used Realbench stress test and lowered my core voltage - 0.020v from the original +0.075v and it passed, however prime95 failed in round off checking. So much for Realbench for me.

so, you actually prime then so you need prime stability? Does your system freeze for what you do/how you use your pc?

Personally, thats how i choose my program. I dont judge its effectiveness based off another application.

 

[wingman99]

so, you actually prime then so you need prime stability? Does your system freeze for what you do/how you use your pc?

Personally, thats how i choose my program. I dont judge its effectiveness based off another application.

I use my PC for everything gaming, Encoding video streaming, web surfing, E mail, shopping, it is on six hours a day. I'm having a data corruption problem, I'm trying different memory now.

It is overclocked past Intel specification, so the best I can do for absolute stability is use prime95 with the option of round off checking enabled. Realbench does not test well enough for me.

I don't want my CPU to write to a incorrect memory block or change the data, intern then corrupting my SSD data, the PC will still run however the Data can round off incorrectly and not lockup.

A CPU core is dumb all it does is a binary numbering system Ones and Zeros with boolean logic, to simplify it only works with numbers. So you need a program that can really load all the transistors at the same time to save time testing and prime95 does real well.

 

[SalemSaberhagen]

Did some heavy testing this morning.

Every thing went fine except prime 95

After a few times having one cpu bow out I incrementally increased the voltage. I got stable again at 1.4 but at ten minutes in to testing one of my cores hit 92c for a second.

Seeing that I decided more testing needed to be done at a lower OC.

I gave up on the FSB OC and lowered my voltage, all is testing perfectly now with just the 4.6ghz.

Thought I had a real sweet air OC but I guess I can settle with pretty good.

- - - Updated - - -

I may gain some nerve and see of I can get the 47 multi once I get over seeing 92c

 

[Kenrou]

Why does everyone try FSB overclock if they have unlocked CPU's

 

[SalemSaberhagen]

Old habit I guess, the bus speeds up everything.

- - - Updated - - -

It really isn't translatable t to today's multiplier options.

 

[Deleted member 117404]

running x 45 @ 1.37 V, didnt won the lottery this time @ tempas max 75 degrees. Is this good or should i try x46 ?

 

[Kenrou]

running x 45 @ 1.37 V, didnt won the lottery this time @ tempas max 75 degrees. Is this good or should i try x46 ?

As long as its below 1.42v

 

[Deleted member 117404]

1.42 for daily use? Then I can go x47 proberly but I don't want to damage my processor faster then needit

 

[Kenrou]

If you're running 4.5ghz @1.37v i think the best you could hope for is 4.6ghz @1.41v/1.42v if you can get it stable

 

[Deleted member 117404]

If you're running 4.5ghz @1.37v i think the best you could hope for is 4.6ghz @1.41v/1.42v if you can get it stable

i think youre right, but even with x46 @ 1.4V aidia giving me general hw failure at system stability ttest i check off cpu, fpu, cache and stress system memory. aida fails, but i did just played bf4 3 hours without any crash..

I ran a aida test for 30 minutes and it was ok, now i put the voltage down to 1.41 from a 1.42, and aida cpuid still shows me it uses 1.440 V. but HWinfo shows me a 1.39 up to 1.41
which one should i trust?

btw you have 4.7 on 1.39, thats amazing.

thank you

 

[Woomack]

There is one thing which I've noticed. Depends from microcode ( BIOS version ) and used motherboard, CPU may require higher or lower voltage so if your CPU needs something like 1.3V at 4.5GHz then on other motherboard it can be 1.25V or 1.35V. On MSI Z170I Gaming Pro AC non-K BIOS I have to set about ~1.45V at 4.7GHz on the same CPU as on ASRock OCF requires ~1.35V but on the same MSI with latest official BIOS ~1.38V is enough for stability at 4.7GHz.
Why I'm saying that is because I see all comparing voltages on various motherboards. Not always your CPU is automatically weak if it needs higher voltage, maybe it's something with motherboard or BIOS version.

 

[Kenrou]

Also depends on LLC settings, I got mine as tight as it goes, so I got at most 0.03v variation

I got lucky with this CPU, with my current cooler I do 4.7/4.7 1.39v AVX or 4.8/4.7 1.39v non-AVX (CPU/uncore).

 

[Woomack]

I run my pretty average 6700K @4.2/4.2GHz ~1.18V and don't need more for games It's also frequency for my reviews and I try to keep it this way so can compare all my memory results ( and readers can easily set this frequency ).

 

[Kenrou]

Never tried going below 1.24v (4.4/4.4) i used the ASUS guide to start at 4.6ghz 1.35v and kept going from there thankfully my CPU seems to be fairly predictable, climbs in 0.05v increments till 1.4v.

 

[Woomack]

If you care about high clock then it doesn't matter much how high voltage CPU requires at low clock but how it's scalling at higher clocks. Like if your CPU requires +0.05V each 100MHz then it's not the best chip. If it requires 0.03V but starts at higher voltage then can be better in general.
Something like 4GHz = 1.10V vs 1.20V can be:
4.2GHz = 1.15V -- 4.2GHz 1.20V
4.3GHz = 1.20V -- 4.3GHz 1.23V
4.4GHz = 1.25V -- 4.4GHz 1.26V
4.5GHz = 1.30V -- 4.5GHz 1.29V
4.5GHz = 1.35V -- 4.5GHz 1.32V
4.6GHz = 1.40V -- 4.6GHz 1.35V <=== here you already see big difference

Next thing is that always at some clock and temperature CPU have that higher voltage step and since then you have to bump voltage much more like in Skylake it usually appears at 4.7-4.8GHz. Till this clock most CPUs require about +0.04V each 100MHz but since this point it may require +0.06V or more. Low temp is helping but not always.

So far my best Skylake is i3 6320 and it was good for benching at 4.9GHz ~1.42-1.45V. 6600K and 6700K couldn't run on air/water at 4.9GHz.

 

[Kenrou]

I would need to test below 1.2v to make sure, this is my table so far with BIOS 1902 :

Prime95 28.9 / Realbench

4.3ghz 1.2v
4.4ghz 1.24v
4.5ghz 1.29v
4.6ghz 1.34v
4.7ghz 1.39v

Prime95 26.6 / Aida64

4.4ghz 1.2v
4.5ghz 1.24v
4.6ghz 1.29v
4.7ghz 1.34v
4.8ghz 1.39v

HT off

4.5ghz 1.2v
4.6ghz 1.24v
4.7ghz 1.29v
4.8ghz 1.34v


This has to be the most "linear" CPU i have ever had, never saw voltage go up like clockwork. 4.9ghz+ is unstable/freezes regardless of voltage even with HT off, maybe with custom water/LN2/DIce i could get better results ?