1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Intel Haswell Overclocking Clubhouse.

Discussion in 'techPowerUp! Club Forum' started by cadaveca, Jun 9, 2013.

  1. TheHunter

    TheHunter

    Joined:
    Apr 10, 2012
    Messages:
    980 (0.99/day)
    Thanks Received:
    373
    Location:
    Europa
    Nice temps in Cinebench11.5, I get that @ 4.7ghz 1.28v if its cool enough, ie up to ~ 25C ambient. Score I think 10.39:)


    What temps did you get in this x264fhd?

    I score 31.3fps @ 4.7ghz, 42x, 2400 here.. max temp per core 75-73-71-67
     
    Last edited: Jul 26, 2014
  2. fullinfusion

    fullinfusion 1.21 Gigawatts

    Joined:
    Jan 11, 2008
    Messages:
    8,473 (3.33/day)
    Thanks Received:
    2,212
    71c was the hottest core, the rest were 65-67c
     
    TheHunter says thanks.
  3. erixx

    erixx

    Joined:
    Mar 24, 2010
    Messages:
    3,491 (2.01/day)
    Thanks Received:
    500
    I am just beginging with this chip. At stock it is amazing how fluid and how cool it stays during BF4 game at max settings.
    Have done some test runs at higher clocks and also the auto OC thing, but not much.

    Thanks for all the info full infusion1
     
    fullinfusion says thanks.
  4. MxPhenom 216

    MxPhenom 216 Corsair Fanboy

    Joined:
    Aug 31, 2010
    Messages:
    10,216 (6.47/day)
    Thanks Received:
    2,386
    Location:
    Seattle, WA
    Appears my chip wasn't as stable as I thought at 4.5ghz. Dropped it down to 4.4ghz for the time being. Was starting to get a lot more lock ups in game. Multiple times a day in Smite mostly, few in Battlefield 4.
     
  5. TheHunter

    TheHunter

    Joined:
    Apr 10, 2012
    Messages:
    980 (0.99/day)
    Thanks Received:
    373
    Location:
    Europa
    If you had that 4.5ghz @ 1.350v did you try 1.354v? and if you had cpu system agent @ stock or idk 0.020v did you try 0.025v+ offset.
    Also input voltage probably needs to be ~ 1.83v with LLC - auto and in DIGI+ 120 or 130% cpu current.

    I noticed input voltage is ok with ~ 0.45v gap, anything less and its unpredictable.
    Im my case 4.6ghz @ 1.234v >> input voltage 1.72v and LLC - auto, anything less and it can bsod or hard reset.

    Imo you're almost stable just minor tweaking to stabilize to max :)
     
    cadaveca and MxPhenom 216 say thanks.
  6. MxPhenom 216

    MxPhenom 216 Corsair Fanboy

    Joined:
    Aug 31, 2010
    Messages:
    10,216 (6.47/day)
    Thanks Received:
    2,386
    Location:
    Seattle, WA
    I was at 1.34v actually. 1.9v input. +0.045 vccsa and current at 130% i think.
     
    TheHunter says thanks.
  7. TheHunter

    TheHunter

    Joined:
    Apr 10, 2012
    Messages:
    980 (0.99/day)
    Thanks Received:
    373
    Location:
    Europa
    I see, well when you're in a tweak mood again maybe try 1.344v, lower input to 1.85v and vccsa maybe 0.050 or 0.055v


    btw I edited my post a little :]
     
    MxPhenom 216 says thanks.
  8. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    14,149 (4.45/day)
    Thanks Received:
    7,354
    Location:
    Edmonton, Alberta
    I'd agree with TheHunter here and say that vInput might be a bit high, and that you should be able to pull everything in line with a bit of time spent tweaking things. I'd also suggest boosting DIO a little bit, and increasing vCache a big chunk., especially if you've updated to one of the BIOSes that came out since the Devil's Canyon launch.

    Also, I've been talking wit ha few other users of ASUS Z87/Z97 boards and there seems to be some issues still rpesent with BIOS corruption when crashing on OC, so flashing a fresh copy of your current BIOS may prove to fix thing all on it's own... USB BIOS Flashback feature being built in now seems like it was there for a reason. :p
     
    TheHunter and MxPhenom 216 say thanks.
  9. MxPhenom 216

    MxPhenom 216 Corsair Fanboy

    Joined:
    Aug 31, 2010
    Messages:
    10,216 (6.47/day)
    Thanks Received:
    2,386
    Location:
    Seattle, WA
    Ill probably mess with it a bit later on. Thanks for the info. my vcache is at 1.15v for 39x multi.

    Wierd thing is, im not quite sure the bugcheck code for the hard lock/BSOD i got last night is due to cpu/memory instability.

    0x0000001e

    I have been getting some Nvidia driver crashes through too, when I play the game Smite. I remember getting driver crashes too, when my CPU wasnt 100% stable a while back when I first started overclocking. the crashes would go away when CPU was at stock.

    EDIT: Sounds like that code means more vcore.
     
    Last edited: Jul 27, 2014
  10. TheHunter

    TheHunter

    Joined:
    Apr 10, 2012
    Messages:
    980 (0.99/day)
    Thanks Received:
    373
    Location:
    Europa
    Yes its either this or this

    0x1E = increase vcore, or lower RAM frequency/ increase RAM voltage


    0x1E = add more vcore


    Although if you OC your ram then do use 120% ram power current in digi+.
     
  11. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    14,149 (4.45/day)
    Thanks Received:
    7,354
    Location:
    Edmonton, Alberta
    he's got 1.5V 2133 Dominator platinum sticks, shouldn't need any digi+ boost with those ones in particular, although boosting them up to 1.575 might make sure that it's not a weird ram thing that is the problem.
     
  12. MxPhenom 216

    MxPhenom 216 Corsair Fanboy

    Joined:
    Aug 31, 2010
    Messages:
    10,216 (6.47/day)
    Thanks Received:
    2,386
    Location:
    Seattle, WA
    I have them clocked at 2400. Voltage is at 1.65v
     
  13. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    14,149 (4.45/day)
    Thanks Received:
    7,354
    Location:
    Edmonton, Alberta

    shouldn't matter much, even then, but you can push up the current limit no problem. I just don't think you'd need to. you could probably push those to 2666...vDIO is what needs the boost to be able to boot that speed, and what might be something to look at rather than current limits. ;) When we went through the OC on your sticks, seemed to me that your CPU was the limit preventing you from getting higher.
     
  14. mstenholm

    Joined:
    Nov 7, 2009
    Messages:
    2,141 (1.14/day)
    Thanks Received:
    1,276
    I got a L336D105 but didn't test it a lot. It does 4.59 @ 1.273 V.
     
    Crunching for Team TPU More than 25k PPD
  15. TheHunter

    TheHunter

    Joined:
    Apr 10, 2012
    Messages:
    980 (0.99/day)
    Thanks Received:
    373
    Location:
    Europa

    When I ran mine at stock 2133mhz I left this ram current alone, @ 2400mhz I had 120% for a while and it was rock solid, then I changed to 110% and after some time (~2 weeks) it started to act weird, right before logon it would bsod or rest pc.. After I put back to 120% it stopped and haven't seen it since.
    But now its using more wattage, always min 20w, at stock 100% I think it was a lot lower, 5-6w or so.


    btw whats vDIO, digital IO cpu voltage? :)


    EDIT:
    So I could run mine at ie 110% too, but I would need to adjust this vDIO?

    Although I already played with this DIO once, but idk how to set properly, I set it 50mv more then analog but it didnt help much.
    Im also using vccsa 0.045v atm @ 4.6ghz, had 0.040v but now at 0.045v i noticed i get better fps by cpu bound game benchmarks.
     
    Last edited: Jul 28, 2014
  16. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    14,149 (4.45/day)
    Thanks Received:
    7,354
    Location:
    Edmonton, Alberta
    yep, digital I/O voltage. stock is usually around 1.02-1.05 for most chips I've seen, but I've seen BIOS default to higher. I know that for me to be able to push those 3000MHz + clocks, 1.2V DIO is essential.

    So, I tell you my story about this:


    I am testing 4790K test further, and find that up to 2400 Mhz boot, no problem. 2600/2666, no go, and I got ram rated for 3100 MHz...WTF!?! Try other memories, nothing over 2400 MHz is working, and when I do get 2666 working, bandwidth is garbage...just worse than 1600 basically, but with OK latency...OMG why is this so hard when it was so easy a year ago? Swap in a few boards, try other things, pull down MSI Z97 Mpower, start ticking off the voltages, one by one...

    So, I check back in my notebook as to what I was setting way back when I was pushing 3400-3600 MHz on my review 4770K, since I decided to use 4770K and see what's what, since my ES 4790K has some damage, and remembered after some time of no luck (like a week off playing with settings) that IO's on 4770K needed 1.25V, or no boot...but here I am @ 2600 with no boot? So I push up to 1.25V...it works fine.

    Slowly lowered it until no boot re-appeared, and this began @ 1.125V. But, I could warm boot, just not cold boot. So push a bit higher, 1.15V, cold boot, warm boot, both no problem, from 2600-3200 MHz. But full stability... that took just a wee bit more, up to 1.185V....


    I am not sure what this DIO, what it's for. Like, I got some ideas, but documentation and such is so sketchy... So that's where I'm at right now. almost every board, no matter the brans, over 2400 MHz..there's ZERO bios tuning. MSI has some nice profiles to use, they work great. ASUS, tertiary timings are garbage, and they say "oh, you try set 8 for everything, and then lower and test". Yeah, right. If that worked, BIOS would already been tuned like it was over a year ago, new BIOS programmer isn't doing it fully, or is a noob. Check Gigabyte.. same story. ASRock...yeah, they need some work too.

    Anything up to 2400 MHz, no problems.

    That said, with al lteh testing I've done, I'm getting pretty close to being ready to write my next guide. :roll:
     
  17. erixx

    erixx

    Joined:
    Mar 24, 2010
    Messages:
    3,491 (2.01/day)
    Thanks Received:
    500
    There are not enough "Thanks" buttons to thank those contributions like the above. :toast:
    I am not sure if this thread includes the 4790K (yes, is a Haswell, but ...)

    It looks like this cpu or the z97 platform is not turn multis up and ready...
     
  18. TheHunter

    TheHunter

    Joined:
    Apr 10, 2012
    Messages:
    980 (0.99/day)
    Thanks Received:
    373
    Location:
    Europa
    Ok so I tested like so and it got a bit hot..

    4.6ghz @ 1.234v
    LLC auto, cpu current 120%, ram current 110%
    SVID 1.72v
    vccsa 0.020
    vDIO 0.125v

    HOT.jpg

    before I had max 60C on core0,

    @ 4.7ghz it was max max 65C, usually 62C if its not to hot inside.



    Although when I ran cinebench15 it was 2C cooler on core0 (max 67C) then at vDIO 0.150v (max 69C), and I got better score with 0.125v (935cb) then at 0.150v (933cb)




    EDIT: ok must have been cpu vrm Powerin/powerout glitch. I ran 3dmark2011 physics test and it was cool like it suppose to be and with correct score.

    3dmakr.png

    By RE5 I see it happened in 1st level (1min later later), its usually where it gets the hottest now with new gpu.. Before with my old 580gtx it was at 3rd part of bench this as on pic above.



    Will try to lower some more :D
     
    Last edited: Jul 29, 2014
  19. TheHunter

    TheHunter

    Joined:
    Apr 10, 2012
    Messages:
    980 (0.99/day)
    Thanks Received:
    373
    Location:
    Europa
    Sorry for spam xD

    This atm looks more promising :)

    4.6ghz - 1.230v
    vDIO 0.100v+ offset
    cpu Power Decay - enabled (just in case si it doesnt overshot when going form idle to max load)

    rest the same
    SVID 1.72v, cpu 120%, ram 110%
    vccsa 0.020v+ offset (0.832v)


    and I got this in RE5
    vDIO 0.100 4.6ghz.jpg


    Cinebench15 got 937cb core0 68C, which is also almost the best score @ 4.6ghz (pb was 939cb)


    Hopefully its ok by cold boot too, tnx Dave :)
     
    Last edited: Jul 29, 2014
  20. fullinfusion

    fullinfusion 1.21 Gigawatts

    Joined:
    Jan 11, 2008
    Messages:
    8,473 (3.33/day)
    Thanks Received:
    2,212
    You guys and your offset lol.. I don't have offset so I can't advise. Headhunter you need to get a DC so you can stop comparing but my chip gets this at that lol..
    I ran a pile of tests last week and in fact these new hasswell chips are different then the older chips. I have given Dave all my info and I guess we're all going to have to wait till he gets back with his new updated oc hand book ;)
     
  21. TheHunter

    TheHunter

    Joined:
    Apr 10, 2012
    Messages:
    980 (0.99/day)
    Thanks Received:
    373
    Location:
    Europa
    ^^
    Its not that different to DC from what I saw. Guess mine also has a smaller gap between glue and cpu making it little cooler then the most average Joe..
    I only showed those various benches because I showed them here before with other temps. C, now im just repiling how these new settings work :)


    btw Im still ok at those settings from yesterday @ cold boot.


    Will try to lower vDIO some more :D

    edit: atm 0.090v is still ok
     
    Last edited: Aug 1, 2014
    fullinfusion says thanks.
  22. TheHunter

    TheHunter

    Joined:
    Apr 10, 2012
    Messages:
    980 (0.99/day)
    Thanks Received:
    373
    Location:
    Europa
    Ok for 4.7ghz I had to leave vDIO @ 0.090 and bump up vccsa from 0.20 to 0.030 (stabler fps), rest the same cpu power 120, ram power 110%.

    Much much better then old 4.7ghz+ 2400mhz ram OC settings,. Its 5-7C cooler too in something like x264 encoder @ adaptive voltage with avx spikes, thanks again Dave :)
     
    cadaveca says thanks.
  23. fullinfusion

    fullinfusion 1.21 Gigawatts

    Joined:
    Jan 11, 2008
    Messages:
    8,473 (3.33/day)
    Thanks Received:
    2,212
    This is working pretty well :rolleyes:

    This DC has so many options for a OC!

    [​IMG]
     
  24. LagunaX

    LagunaX

    Joined:
    Mar 8, 2009
    Messages:
    1,072 (0.51/day)
    Thanks Received:
    303
    I settled for 4.7ghz core 4.4ghz uncore on my delidded Devil's Canyon as far as vcore and temps for 24/7.

    [​IMG]

    One thing I must say is that Prime95 v285 is worthless unless you want to fry an egg for Haswell.

    Funny thing is that I thought I was stable after passing IBT on max and AIDA64 for 1.5 hrs but crashed on the second pass of the x264 HD v.5.0.1 benchmark.

    Upped my adaptive offset from 0.8 to 1.0 and input voltage from 1.83 to 1.85v and was stable after that.

    x264 seems the definitive stability benchmark for Haswell and maybe Ivy too:
    http://www.techarp.com/showarticle.aspx?artno=520
    link to a modified looped version for haswell testing below:
    http://www.overclock.net/t/1487922/going-deeper-on-the-x264-v2-stress-test

    Delidding and build pics here:
    http://www.overclock.net/t/1313179/official-delidded-club-guide/25750#post_22687530

    @cadaveca : yes, running the Samsung greens at 11-11-11-28 1T DDR3 1100 1.45v above ;)
     
    Last edited: Aug 13, 2014
    TheHunter, cadaveca and ...PACMAN... say thanks.
  25. LagunaX

    LagunaX

    Joined:
    Mar 8, 2009
    Messages:
    1,072 (0.51/day)
    Thanks Received:
    303
    Oh yeah, just for everyone's FYI I delidded using the hammerless vice method, which I believe to be the most brainless and safest, where once you have the padded green chip PCB against one edge and the top shelf of the metal IHS against the other edge, gradually little twists of the vice will give you your delid without the potential of cutting into the PCB or knocking a resistor off.
     
    Last edited: Aug 14, 2014

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)

Share This Page