Overclocking thread?

>>56002113
>a stable clock of 4.4ghz at 1.152v
I doubt that's very stable besides desktop use.

Try H264 encoding or some CPU rendering.

4.5-4.7GHz is usually possible with 1.25-1.3v

>>56002113
Also any idea why it isn't possible to download the vcore/clocks on a manual setting? It did so with my old Gigabyte Z87 motherboard, it doesn't with an Asus Z170.

>>56002137
I ran the Aida64 stability benchmark and the CPU stress benchmark on CPU-Z.

>>56002167
neither of those are actual stress tests.

Intel burn test, Prime95, etc. Those are stress tests.

>>56002154
*downclock

>>56002180
thanks anon, will test

[email protected] / Stable 1.290V

>>56002154
Using Z170-A, manual vcore changes work fine here

File: Untitled.png (357KB, 1185x889px) Image search: [Google]

357KB, 1185x889px

>>56002180
I ran 3 passes but stopped after. I've set it to 1.15v in the bios

>>56002324
I'm using a Z170 VII Hero and it stays constant at max clock. Unsure what to do senpai

>>56002359
oh wait it downclocks if I put it on power saver

File: cpuz.png (32KB, 399x401px) Image search: [Google]

32KB, 399x401px

>>56002359
if you mean that through CPU-Z the vcore isn't changing, I noticed something similar. Mine's set in BIOS to 1.290V but sits at 1.280V when I check.

>>56002444
My clock speed wasn't changing either, I just changed my windows power setting to 'power saver' and it downclocks itself when idle

Hey, does anyone here use Linux? I need a program to test my CPU frequency.

>>56002492
cat /proc/cpuinfo

>>56002113
>Just recieved the 6700k I ordered to replace my 4790k
How does it feel knowing you spent twice as much as me on a CPU with half the performance?

>>56002338
Lol most people declare something IBT stable if it lasts 24 hours. Some people even do 48hours.

>>56002568
Please elaborate.

>>56002547
Oh, thanks

>>56002630
2x Xeon E5 2670 for $70 each

>>56002655
No one cares about a poorfag, kys

>>56002666
>poorfag
say that to my 32 cores when I'm laughing my ass off at your overpriced “consumer” CPU struggling to match it

>>56002655
>eon E5 2670

Man what's up with you dissing his i7k ? Fucking poorfag much? Can your shit hyperthread? Fuck off.

>>56002676

>32 cores
>0 girlfriend

>>56002682
>can your shit hyperthread
yes.. i7s are literally gimped xeons. Enjoy your consumer-tier binning, no ECC support, etc.

i7s are marketed for people too poor to afford xeons. They're in a higher price class.

>>56002700

It's still pretty good considering 95% of /g/ only uses their rig to play 18+ moded Skyrim

>>56002700
>shit single threaded performance
no thanks.

>>56002676
>posting non-overclocking CPU
>CPU with laughable single core clocks

Literally a waste outside of the 0.1% of programs that can make use of more than 8 cores, even including many professional ones.
There is a limit to how much clock speed you should sacrifice for more cores.

>>56002758
>>56002766
>single-core performance
admit it, you just need it for shitty video games you cancerous >>>/v/tard

>>56002338
You need to test that shit for at least an hour, but for true stability do as
>>56002616 says.

>>56002700
i7's are higher clocked though

>>56002700
>Enjoy your consumer-tier binning
Who cares about binning on a voltage-locked CPU? Do you even know what you're talking about?
>ECC
because everyone on /g/ runs their systems 24/7 with 128GB of RAM and cannot possibly allow their systems to suffer a crash or data corruption

>>56002180
should be careful with prime95 on intel cpus, avx instructions make the chips go up to insane thermals like 90c+

>>56002811
>because everyone on /g/ runs their systems 24/7 with 128GB of RAM and cannot possibly allow their systems to suffer a crash or data corruption
If you don't, you're a fucking casual and don't belong on /g/

>>56002700
ES Xeons are literally cheaper on ebay. You can put together a 14 core (with a shitty clock speed, 2.2 w/ turbo to 3.2 IIRC) x99 build for the price you'd pay for a Z170 + 6700K.

>>56002692
you totally rekt him m9

File: 1322799779183.jpg (83KB, 715x759px) Image search: [Google]

83KB, 715x759px

>>56002783
>using a personal computer for anything other than video games and personal enjoyment in the <CURRENT YEAR>

Literally end yourself. I have a Haswell Xeon running on my server AT WORK as well as an Ivy Bridge hexacore that I have to fire up for a few hours once every other day for CAD-CAM applications AT WORK. If you own a Xeon for multithreaded usage at home you are permavirgin status.

>>56002700
They are literally diffrent products for diffrent market areas. There is nothing gimped about them.

>>56002783
You seem to mis-understand. It's fine to have up to 8 cores because it's been tested to be the apex of the frequency/core curve for most high-end programs like Adobe Premiere and such, which is still a lot for a typical PC.
Unless you're running a server, doing 3D rendering, or After Effects, there is no point in having more than that in this current day and age. And even then, for rendering, GPUs are starting to take over as CUDA processing is vastly more efficient, on the order of 6x compared to a dual-CPU system.

>>56002881
They are literally the same silicon with features disabled (ECC, remote management shit, etc)

Hey man I have a question for you, if you have kik message me at iranherover

How the fuck did you fry your motherboard

>>56002866
>Xeon
>For multithreaded usage
More like for everything usage. cant think of many single threaded applications that would actually be bottlenecked by a xeon. Especially something like a 2670.


>>56002893
Xeons have a locked multiplier / base clock though. Not everything is solvable by running a gillion cores.

>>56002885
>high-end programs like Adobe Premiere and such
lol, I don't care about adobe's incompetent piece of shit software

show some x264 bench scores

>>56002825
Haswell CPUs in particular don't like AVX. It's an irrelevant option for stress testing anyway, there is nothing that will ever push your system that hard.

>>56002893
Not always, Xeons have specialty chips that have no real equivalent in the consumer segment, like quad-core 2011-3 socket CPUs. The 16xx series is often the one closest to the actual consumer chips, for instance the 1660 v3 is the exact same thing as the 5960x but with ECC, VT-d, etc.

>>56002953
>Xeons have a locked multiplier / base clock though. Not everything is solvable by running a gillion cores.
It's artificial lock though. They could be offering the whole package but they intentionally separate the markets.

File: 4.8.jpg (676KB, 1920x1080px) Image search: [Google]

676KB, 1920x1080px

Easy Test : 20 run @ very high with IBT
Hard Test : 12h Prime95 @ blend

Anything below that wont be taken seriously

>>56002974
>It's artificial lock though. They could be offering the whole package but they intentionally separate the markets.
Yes. The same way the ECC and neat features are locked away on Consumer parts.

Unfortunately Intel is the devil and unless someone provides a bunch of microcode updates to unlock fucking everything it doesnt matter whether the locks are artifical or not if they cant be bypassed.

>>56002893
They are exactilly not the same silicon. You really think you have any idea what you are talking about?

>>56002113
Just dropping by to say LGA1151 CPUs look fucking sexy

>>56003001
How's that work since they have the same core count, architecture and features?

The only thing differentiating them is how they are binned and what microcode is on the chip.

>>56002113
I'm running a 4790K @ 4.7, 1.303V and I'm quite happy with it, 2400MHz RAM too. I don't really do anything stressful (CPU or GPU) on my desktop other than playing games (at 4K) and I'm quite happy with the CPU.

I don't plan on upgrading until I can get something with 6+ cores that can beat it in single-threaded performance too.

File: 2016-06-28 22_45_41-CINEBENCH R11.5 (64 Bit).png (258KB, 1020x552px) Image search: [Google]

258KB, 1020x552px

>>56003106
I can run my 5820k at 4.7GHz with 1.34v which should beat or at least match your 4790k in single core.

File: test.png (152KB, 1213x890px) Image search: [Google]

152KB, 1213x890px

>>56002977
new anon here, testing rn, have only done standard so I decided to push a little harder. waiting for a new gpu so I just have stability testing to do

>>56002977
IBT is borderline useless on Haswell (and maybe anything newer too, I don't know). I had a stable OC in IBT that never crashed or failed no matter after how many runs I tried, but then fucked up in under 10 minutes every single time I started up BF4. I assume this is due to the voltage fuckery Haswell does with AVX. The CPU runs at a higher voltage during the super stress test and doesn't crash, but then under normal load and lower (actually normal) voltage it crashes.

>>56003129
Eh, sure, but X99 is pretty old now and going out the door, yet it's still expensive and there's no guarantee I'll get a 5820K that would hit 4.7. It also doesn't come with enough lanes for x16/x16, which apparently can matter in some cases (like DOOM). I'd rather wait to see what else comes out, maybe Zen is going to live up to the hype, or at least force Intel to lower prices on their enthusiast line.

>>56003129
Quit being a whiny faggot and run cpumark and post results then

>>56003199
>Eh, sure, but X99 is pretty old now and going out the door
doesn't even really matter because the performance increase between each intel cpu generation since sandy vagina is so fucking minimal unless you actually need the extra processing power for something that isn't games

File: test2.png (167KB, 1219x902px) Image search: [Google]

167KB, 1219x902px

>>56003146
>>56002977
looks stable to me

>>56003264
>the performance increase between each intel cpu generation since sandy vagina is so fucking minimal
Exactly, so why even bother upgrading Haswell->Haswell when I can very well wait a bit more for some more shit to come out and get a few generations of improvements at once, along with a chipset supporting all the modern shit?

Probably off-topic... Is RAMDISK a good idea? Games and programs don't seem to be faster even though the read/write speed is supposedly 25x faster.

>buying a 6700k when Kaby Lake is comign in just a few months

>>56002825
Depends on CPU. My Haswell CPU runs just normally with Prime95.

>not buying http://www.ebay.com/itm/IBM-x3850-M2-4x-E7450-2-4ghz-Six-Core-Processors-128gb-RAM-No-HDD-/172299240453?hash=item281dd5c405:g:96oAAOSwYGFUsDOG

cucked

>>56004243
Good for productivity not necessarily games (they tend to saturate memory bandwidth)

I recommend ramdisk plus from superspeed, the dataram and softperfect offerings are buggy and have horrendous interfaces

Is there a decent overclocking micro ATX board?

By "decent overlocking" I mean 8 or 10 power phases and a discrete voltage controller.

Fuckin EVGA and MSI sell z170 boards with 4 phazes. Have they no shame?

File: cpuz.jpg (3MB, 10000x10000px) Image search: [Google]

3MB, 10000x10000px

OCCT: CPU for 2 hours is better than running any stress test for a week.

If you are truly confident about your overclock (caring about actual stability and not just online dick waving) then run occt for at least an hour. if your system is unstable it will show within the first 30 minutes.

It sounds like bullshit but it really works.

>>56002113
4790k @ 4.977MHz / Stable 1.335V

File: 29234_PE116289_S5.jpg (468KB, 2000x2000px) Image search: [Google]

468KB, 2000x2000px

4770k @ stock clocks because after being in the pro overclocking scene for a bit, I quit

>>56007825
do you know der8auer ?

>>56002113
I overclocked the q6600 of my browsing PC (not main machine) from 2.4ghz to 3.15ghz @ 1.3 volts. I ran 10 passes of intel burn in test on maximum on it and it didn't crash so it's stable.

Could someone please explain how to properly determine the fsb termination and mch voltage? I increased both of those gradually as I increased the CPU's and the RAM's clock speed.

>>56007834
Not him, but fuck that faggot. Sells shitty 3d printed delidding and lapping tools for €60.

>>56007994
He is just a smart jew. OC fags who are too pussy to delid stuff the classic way get jewed hard. I see no wrong.

>>56007770
>OCCT: CPU
Is that another AVX test or something else?

>>56008030
You can't really delid them the classic way anymore, because the jews at Intel made the PCB as thin as possible, so they crack under pressure and the newer batches come sealed as tight as possible, so you have a good chance of damaging the pcb with a blade.

There are free 3d printer templates for the tools, similar to the ones he sells. I just don't have a 3d printer. I would buy a tool for €15, but €60 is just another level of kikery.

>>56005411
>implying kaby is worth waiting for

File: 1262163212155.jpg (11KB, 220x168px) Image search: [Google]

11KB, 220x168px

>about to upgrade to a 2600k for only 110 euro
>gonna oc the shit out of it
>2500k will probably still sell for 70 euro because Sandy bridge is still competitive at higher clocks

Feelsgoodbrah

I want to overclock my i5-6600k to 4.0GHz and I have a hyper 212 evo with a noctua nf-f12. Can I do it? and what would be a good voltage?

File: 1421917648115.jpg (51KB, 670x472px) Image search: [Google]

51KB, 670x472px

>>56007770

>>56002338
Far from stable. You should be able to run IBT(linpack) for hours and preferably for 24h for absolute stability. Try running Prime95 but watch out for temps.

>>56008196
Make sure to get a decent z77 motherboard while you're at it

>>56008356
am I the only person that just does 10mins blend, 10mins small fft, then uses my PC as normal until it crashes?

excluding reboots for updates, I've got like 2 months uptime now

>>56008536
I do that too, especially since I got Haswell because stability in stress tests doesn't actually guarantee stability in real applications, which is to say games in my case. As such I only use stress test tools to weed out highly unstable settings.

>>56008565
idk about haswell but for sandy there are some "sandy-killer" FFT sizes that will crash a lot of 'stable' setups

How do i overclock safely?
I've been running my 4690k stock because i'm afraid that i will fry it with too much voltage. Can i just up the multiplier without touching the voltage?

>>56008732
Increasing the multiplier will increase the voltage if it's set to auto.

What you could do is check the current voltage it's at under load, then use voltage offset to keep it under that while you're overclocking.

1. increase multiplier by 100-200MHz
2. check how much voltage goes up by in cpu-z (while it's under load)
3. set a negative voltage offset to counteract that
4. check it worked
5. repeat steps 1 through 4 until your system is nolonger stable, then back off to the last stable stage.

Also if you're using negative offset, make sure your system is actually stable at idle, as it might turn out you're undervolting too much.

If your mobo doesn't have offset voltage control, you got fucked because it's by far the best way to overclock for normal people.

Anyone here got a Skylake non k CPU and overclocked via baseclock?

>>56006368
>Depends on CPU. My Haswell CPU runs just normally with Prime95.
Really now? In every other benchmark like Linpack my CPU stays below 80c, mostly at 70c after OC, but in Prime95 temps can spike up to 100c which is just damn scary.

>>56008919
No but I have an S processor which for some reason has an unlocked multiplier.

>>56002928
Did you mean to quote someone?

I managed to fry my motherboard while overclocking. It was a Z87 """""""""""""""""""""ultra durable""""""""""""""""""""" Gigabyte mobo. Sure, it wasn't a board designed for heavy overclocking which what I did on it. I think it fried while running linpack with 1.46v core and 5.1 GHz on a i7 4970K. That drew north of 200w from the CPU socket. Once it crashed and I rebooted with normal clocks, USB started to malfunction, devices reconnecting themselves, dGPU would sometimes disappear for a second or two until the drivers recovered, DPC latency was in milliseconds(even spiked up to 10ms+) and any kind of real time multimedia usage was impossible due to this.

>>56007701
Asus Maximus VIII Gene? It has 10-phase power design with discrete digital voltage control.

>>56002113
Running i7 4970k with 1.30Vcore, 1.900V input and 4,90 GHz. I bought it as a binned for 4.8 GHz but seems to be fully stable at these settings on 100MHz higher clock. Delid with Liquid Pro on top and under the IHS with my AIO H110 water cooler keeps the temperatures in check.

>>56008717
Sandy doesn't do integrated VRM voltage fuckery for AVX like Haswell, so I wouldn't expect AVX-based tests to be as useless. IBT was great for my Q6600 and i5 750.

I have an ancient i3 330UM @1.20Ghz, I read somewhere it could be OCd to 1.60Ghz, but I haven't really found how. Anyone got some idea?

might be a dumb question but ill ask anyways. i have my i5 overclocked to 4.4ghz running at max clock with all power saving settings turned off, does the actual cpu usage % make a difference as to how much power it's actually using?

>>56012124
>does the actual cpu usage % make a difference as to how much power it's actually using?
Yes.