>overclocking my new 6700k >stable at 4.6ghz, vcore set

>>55803897
>6700k @stock isn't fast enough for my autistic needs
Fedora.tip

>>55804125
free performance famalam

>>55804290
How to avoid warranty for free*

>>55804558
When was the last time you had a CPU die on you?

T. Different guy

>>55803897
This is normal. It might kill your processor pretty quick, but it's set to increase vcore if AVX instructions are used.

>>55804290
>"free"

Used to run my old Deneb @ 4.1ghz 1.63v and it lasted a solid 5 years til i replaced it.

I mean, surely your fancy intel silicon can do better than 2008 AMD, right? Don't worry about it.

>>55804632
>if AVX instructions are used

This is what I don't understand: with manual vcore at 1.32v, OCCT is fine (ran it for 2 hours) - but the board pumps 1.5v in to it on offset mode.

Doesn't seem that clever. I will stick to manual vcore like I used to, offset/adaptive mode doesn't seem like it provides a massive improvement.

>>55803897
babby's first überclock

>>55804704
holy shit nice.

>you will never see the release of a cpu as capable and worthy of its price as the phenom II line
why even buy CPUs anymore?

>>55804786
It's still pretty capable, loses to modern FX processors mostly due to memory bandwidth related issues and obviously gets smoked by intel but you'd be hard pressed to find something that it can't passably handle.

>>55804762
You're right, with my 3570k I just set a manual vcore and multiplier

Never tried all of these fancy settings until now

>>55803897
>vcore set to manual at 1.32v
Anything 1.3 and under is "safe". However anything over 1.3 begins to degrade the silicon.

1.32 is definitely causing your chip to die a slow death.

>>55804755
You don't need to understand it, that's just how it works because fuck you. Maybe it's related to motherboard somehow, I know Haswell is supposed to have that problem too but I get normal voltage load with prime95 despite AVX.

>>55804755
Is boost still on? Could be causing issues. Stick to the offset and turn off boost.

>>55804704
Old processors naturally used more voltage. You can't compare the OC voltage of a 2009 CPU to one released this year.

>>55804857
I think it's a matter of tradeoffs. I could probably back down to 1.3v and 4.5 GHz but by the time the CPU degrades enough to be a problem I would have most likely moved on to something else.

>>55804704
My 80286 ran at 5V, surely your new fancy pants AMD processor can do that much.

>>55805020
I mean maybe, but god knows the VRM's wouldn't get close to that

>>55803897
Haswell does that too, there's no real way to disable it as far as I know, although funnily enough when I updated my Z87 board to the latest BIOS for 4790K support, it stopped doing that. The voltage control on my mobo is really fucking odd though and doesn't work at all like it used to in the older BIOS, or in the way every single guide I've read describes it to.

Anyway, you have 2 options:
1. Use constant voltage, but get higher power consumption and higher temps in idle/low load
2. Use adaptive, get lower power consumption, better temps in idle/low load and refrain from running tests that are very likely completely unrealistic for your workload and will literally never be encountered during the CPU's entire useful lifetime

After getting over my autism about running programs that I literally never run when the PC is actually doing something useful to me, I went with #2.

>>55805104
It won't, btw. It would instagib the processor.

>>55805115
>refrain from running tests that are very likely completely unrealistic for your workload and will literally never be encountered during the CPU's entire useful lifetime

For that reason I'm tempted to switch over to the x264 encoding test, but I've read that Handbrake does use AVX instructions these days.

How large is the difference in power consumption between fixed/adaptive?

>>55804558
>warranty is important
>cpu

>>55805398
I don't know, but my 4790K steps down to ~0.755V in idle, compared to 1.305V in load, so the voltage difference is large. It's idle anyway so the power consumption can't be that big of a deal, not something you'd notice in bills I imagine, but in the end I see no reason to run my CPU at higher voltage than required most of the time.

Most things like web browsing, programming, playing music and even videos won't keep a modern i7 in its maximum performance state except for very brief "spikes" and as such it won't be spending much time at all at full voltage. I even play software-decoded 1080p 10b files and it barely if ever jumps up to full voltage, so I really see no point in keeping it at full juice when it's clearly not required most of the time.

Also, in my own experience with Haswell, AVX torture tests aren't actually indicative of stability in all situations. I've had Haswell CPUs that easily passed extended runs IBT or prime95 small FFT only to crash in <20m in a game like BF4, which is much, much less harsh.

tl;dr: I don't see much of a point in AVX torture tests on Haswell, as such I see no point in configuring my system for them.

>>55803897
Overclocking CPUs really confuses me these days. I'm just used to dancing with base clocks and vcore. I gave up a while ago.
Bought a 6700K and let it auto-overclock itself with some fancy option in my BIOS, it hit 4.56GHz on its own and I'm happy enough with that.

>>55805768
same here, somehow overclocking got pretty retarded when it became a standard "feature"

>>55805944
I still manually mess with GPU overclocking, which is much simpler. GTX 1070 running at 2GHz is a fun numbers game I'm impressed the card can handle.

>>55805768
It's a lot less straight forward since Haswell and more difficult to ascertain absolute, full stability, partly due to the automatic voltage fuckery that can't be quite disabled and partly due to super hard stress tests not providing conclusive results anymore. That's what it looks like to me anyway.

The way I do it nowadays is to push it with fairly "normal" loads and not extreme stress tests. Intel's Extreme Tuning Utility or w/e it's called is quite useful in that regard I find, as are things like video encoding, Cinebench and even good ol' prime95 but on Large FFT, which despite being AVX doesn't hit nearly as hard as small FFT or IBT. After the settings seem stable, I simply run my real-world applications and adjust if needed. If something crashes quickly and often I use that to tweak things further.

Hitting my 4.7GHz 4790K with small FFT at 1.3V is fucking insane, temps go up 15-16C compared to large and after I find it to be "stable" it crashes within minutes of opening a game, fuck that shit.

>>55805768

I'd stay away from auto-oc for two reasons primarily. The first is that it often sets the voltage higher than necessary, and the second is that it rarely achieves high overclocks. It may not be impossible for you to reach a higher clock speed with the same, or possible even lower voltage if you do it manually.

>>55804558

what decade are you living in? overclocking has been covered under warranty for most consumer products since at least 2010

>>55804755

generally 1.25-1.3v is pretty damn high voltage and not recommended for a permanent overclock unless you don't mind selling it off in 1-2 years to a poor chump when the next line comes along

I mean it could be fine but if it dies you'll be shit out of luck as finding old boards/cpus for old intel lines is impossible for sane prices

>>55806013
>and after I find it to be "stable" it crashes within minutes of opening a game, fuck that shit.

I've yet to bother finding a good overclock for this exact reason, spend a few hours tweaking settings to get sane voltages for a clock and having it be stable in stress tests and within minutes of opening a game it'll crash

>>55806078
>generally 1.25-1.3v is pretty damn high voltage and not recommended for a permanent overclock unless you don't mind selling it off in 1-2 years to a poor chump when the next line comes along

is it? i'm running at 4.7, 1.35v. stock was 1.25v for me

pretty much everywhere else said to keep it below 1.4

>>55806053
Sure, but I've been running this overclock for almost a year now and temperatures never rise above 50C either, so as far as I'm concerned, this is solid. I'm sure it could be achieved with less voltage but perhaps questionable stability. I've yet to experience any hiccups with this.

>>>/v/

>>55806260
because there's no possible reason anyone would want more free performance besides for playing video games.

>>55806078
>generally 1.25-1.3v is pretty damn high voltage

i didnt even do OC and the voltage automatically goes up to 1.27 suddenly so....

>>55806078
>I've yet to bother finding a good overclock for this exact reason, spend a few hours tweaking settings to get sane voltages for a clock and having it be stable in stress tests and within minutes of opening a game it'll crash
It's not so bad in the end, you just have to accept that you either need more voltage or lower the clock. On Haswell the CPU vreg input voltage is also quite important, play with that too. These days I only use "stress tests" to weed out extremely unstable settings and sometimes to do more of a worst-case cooling test rather than a stability test.

>>55806142
I don't think it's that bad either, especially if you're on adaptive voltage and have EIST and all that enabled, it won't actually sit at that voltage most of the time. Temperature also matters though.

I find it a bit ridiculous how everyone worries about their OC'd CPU hitting 80C or something, meanwhile Intel CPUs in some high-end (business even) laptops pretty much hit 100C and thermal throttle by design and nobody gives a fuck.

>>55806272
Only a gamer would have a 6700K. There are better CPUs out there for professional applications.

>>55806272
Well, not that this discussion isn't obviously related to technology, but OCing really is especially suited to gaming. It's an application that requires as much performance as you can get but hardcore 100% stability isn't really required. There's a reason why it's generally the consumer stuff that gets OC'd and comes with OC features.

I phrased it a bit dramatically, should have read something like >1.25-1.3v

>>55806142
>pretty much everywhere else said to keep it below 1.4

people like to push boundaries, it has been 1.3v tops for safe usage for a very long time but given how well haswell, broadwell and skylake chips tended to do on 1.4v you'll find more and more less informed people recommending 1.4v as it's very likely to get a stable-ish 4.6-4.7ghz

fwiw on a 5820k I get a more stable overclock with a slightly higher than average results when using >1.3-1.4v than if I'm just using 1.2-1.3v and a slightly below average clock(but still a pretty respectable 4.2~ghz), that might have something to do with it but I'm not sure how indicative a single chip is

of course 1.4v~ is not an instant guarantee of failure and it's a relatively mild overclock compared to some custom loop setups so it'll very likely handle quite a while of it but 1.25-1.3v on the high end is generally considered the safe limit

>>55806341
>i didnt even do OC and the voltage automatically goes up to 1.27 suddenly so....

you can thank mobo manufacturers ignoring intel spec so they get higher fps benchmarks on reviews for that, you may also find you have a higher turbo frequency on some of the first cores for a similar reason

>>55806475
hyperthreading is garbage for gaming, it's better for workstation tasks.

>>55806776
Actua extral cores are even better.

Is there a reference voltage for a 4790K, I find that 1.231v for stock 4.4GHz is high. I'm thinking the Gigabyte motherboard overvolts it a bit to ensure that they operate stable.

I'm gonna try 1.15v with 4.4GHz.

There's no way around it, just use adaptive voltage, add an extra 0.003 to it, enable all power savings, C states etcetera and hope it don't crashes under load.

You should take in consideration even if you passed stress test at 4.6ghz 1.32v it could still crash when going down to 4.5Ghz 1.XXv.

AVX will always use a higher voltage, thats how it works on stock settings too.

Intel CPUs use a voltage table for each core speed, you can't manually edit it you can only add + or - to it so since AVX already uses a higher voltage on stock now that you have a higher custom voltage that voltage table will only increase thus using even more voltage.


Until Intel adds a way to customize individual voltages per core clock or their AVX instruction sets become more efficient this problem will keep happening.

>>55803897

whats ur core multiplier set to?

>>55805239
inb4 someone sets their house on fire doing this

>>55806848
Well just because you're performing a task besides gaming doesn't mean money is automatically not an object. i'd think a 6700k would be a fine choice for something like video rendering