Advice on getting more out of my high end rig

Discussion in 'Newbie Lounge' started by Sploody, Sep 14, 2018.

  1. Sploody

    Sploody New Member

    Joined:
    Sep 11, 2018
    Messages:
    13
    Location:
    Brisbane
    Hi all, first post - woot! I have been reading through a couple threads and wow, there is some serious knowledge on here, it makes my head hurt! Also, prepare thyself for an essay.... Here goes....

    Anyhow this pc i have built will be my third self built PC and so far it has been my proudest, but I am having quite a few issues with it at the moment...

    My aim for this PC was to build a watercooled, 4K gaming beast with everything RGB because RGB makes games run better (sarcasm) - along with the power to do some recording and video editing for when i get back into recording YouTube videos. I am also wanting to delve deeper into the world of overclocking hence all the watercooling.

    I have dealt with a major issue so far of constant crashes during games (PUBG a lot), to which I upgraded from a Corsair AX760 PSU to the 850W below, and then everything seemed to be dandy for a while, with except PUBG still crashing, and slowly crashing more frequently. (It's my main game played generally on 4K and Ultra - constantly getting 50-60FPS). No BSOD, just straight freeze for a second or else a straight reset.

    I fiddled around (re-seated GPU, changed RAM slots and ran two seperate PCIe power cables to the 1080ti rather than the single PSU slot - dual PCIe connector style) and it now seems to be pretty good and I rarely get crashing.

    Only thing is that I wanted a bit of smoother gameplay in PUBG so reduced my res to 2K and left everything on ultra, only to be chatting to my game mate who has MUCH older internals (3rd gen intel build but with a 1080ti) and hes getting better FPS than myself in my games.

    I dont want to sound like a - "oh i spent all the money therefore it will run fast" - kind of guy, but shit, this thing has cost me thousands, and I could have basically put a 1080ti in my previous build and would have gotten the same results, only to pocket the other few grand.

    Specs of my PC are as follows:
    Acer 28" 4K G-Sync monitor
    Intel i7 8700k
    Gigabyte Z370 Aorus Gaming 7 Mobo (F7 Firmware i believe?)
    G.Skill 2x8GB 4266MHz GTZR RGB DDR4
    Gigabyte GTX 1080 Ti Aorus Xtreme Edition
    Corsair HX850i 850W 80Plus Platinum PSU
    Full EKWB Kit - CPU + GPU Coolers with P-Series 360mm Front / S-Series 240mm Top Radiators
    ALL THE CORSAIR HD120 RGB FANS
    Corsair Crystal 570X ATX Case

    One other surprise i am getting is that every time I enable XMP in BIOS, my computer pops an error on startup that there was a BIOS startup error and I can only get the PC to run if I run off defaults.

    My CPU is OC between 4.5GHz-4.8GHz through Intel Extreme Tuning Utility and my GPU through MSI Afterburner (+400MHz clock speed, 150% power limit). I haven't touched RAM timings or changed my CPU/GPU voltages because I simply don't understand what I need to increase to get this better performance, and I dont want to fry my shit because I cant afford to replace any mistakes at this point in time.

    I would love for any insight in how I can go about getting the most out of my internals, getting XMP working again, and a basic guide as to which voltages I could touch and which ones I should leave out, and whether to simply use software to guide the OC or if i should use the BIOS to achieve this.

    If you read down to here, I congratulate you, and I seriously thank anybody who can give me any information as to how to get this machine to run that bit better.

    Cheers everyone and miles of smiles


    TL;DR - guy with too much money buys top end components, has stability issues, sorts some, but needs to know what is the best way to OC the components so he can play PUBG like Shroud.
     
  2. Ratzz

    Ratzz Member

    Joined:
    Mar 13, 2013
    Messages:
    6,997
    Location:
    Cheltenham East 3192
    4266 is tough. Dial up XMP, leave everything at XMP settings, but reduce the frequency to 4000. Very few motherboards can cope with 4266 XMP. Its the ram that is causing the crashes IMHO, or more specifically the motherboards inability to cope with it at 4266.
     
    Sploody and self_slaughter like this.
  3. self_slaughter

    self_slaughter Member

    Joined:
    Dec 12, 2010
    Messages:
    1,210
    Location:
    4165
    Ouch, yeah 4266 on a board like that is going to be virtually impossible to get stable, Ratzz is bang on about dropping to 4000Mhz with XMP rated timings.
    If that's stable then try lean on the timings a bit, but really it's going to make piss all difference anyway.

    Once that's sorted and stable, try push the CPU some more. Unfortunately Intel kept most of the good 8700ks to jack the price and sell as 8086ks, but if there's any more gaming performance to be had it's there.
    I only managed to get 4.7 on all 6 cores with my 8700k with 1.32V, but the 8086k happily runs 5.2 all day every day. Albeit at ~1.4V lol.

    Bloke on here put up a reasonably good video on clocking them up and basic rundown of some of the settings. That'd probably be a good starting point for you. But 8700ks are fairly simple beasts and vcore is generally all you really need to change. They have a bit of a wall where it takes a bootload of voltage to get past, so it's just a matter of finding that wall and generally best just staying behind it. :p

    https://forums.overclockers.com.au/threads/8700k-8086k-overclocking-guide.1248160/

    Give it hell and hope for the best. :)

    Also, have you delidded the CPU? They can run pretty hot normally. If you wanna push it hard de-lidding and slapping some liquid metal under there can usually knock 15-20'C off your temps, stop the temps spiking so quick and give you a lot more room to play. Headin's probably your man for doing that. On the GC so close enough.
     
    Last edited: Sep 14, 2018
    Sploody likes this.
  4. OP
    OP
    Sploody

    Sploody New Member

    Joined:
    Sep 11, 2018
    Messages:
    13
    Location:
    Brisbane
    Thanks heaps my dudes, Ratzz and self_slaughter! :thumbup::thumbup:

    I just looked up Gigabyte's specs on the board and you're both correct - support only up to 4133MHz so thats fine, i'll lower the DRAM clock manually. I also did the smrt thing and ended up watching some Linus and J2Cents vids about RAM speed and if memory serves correct, both say that above 3600MHz, you see fk all difference - really I just saw the fastest stuff and I wanted to 'future proof' my build, hence the 4266 purchase because previously I did my last build on a 7600K and 980Ti... only to hear that Intel were releasing the 8th gen soon after along with like a couple weeks passing and the announcement of the 10 series Nvidia cards.... But i digress...

    Second, I have looked into the delidding of CPUs and although they show HEAPS of potential (totally correct in 15-20'C LESS after the process) along with people having the knowledge in doing this ( nice and close in GC - again thanks self_slaughter! ) and the little jigs available on these OC forums and stores, its just something I'm not comfortable in doing.

    Ive had all my stuff (except GPU) reset to default, CPU running at 4.3GHz with all 6 cores, its a 26'C day (about 28'c in the room) and my CPU temp using the MSI Kombustor CPU Burn utility and Intel ETU stress test, max temps were 62'C (Kombustor).
    I am more than happy to take that up to the 70-80'C range when under full stress because TBH, only when I will be converting videos in the future will I ever come close to using 100% CPU load.

    I'm having a watch of that video you linked me, first thing i've taken, is to do Fan Profiling and Pump speeds from BIOS along with OC settings from BIOS rather than using software (This i found very important). This makes complete sense as I was experiencing some difficulties with stability having i guess the BIOS, Intel Tuning Utility along with MSI Afterburner, probably all fighting with eachother.

    Again guys, thanks heaps for the advice, Im also giving HUGE props for explaining all the OC meanings even though its on an ASUS mobo to the_sentinel and thanks self_slaughter for the numbers. I'll be using those as a bit of a guide to see where I can get to!
    You get a shout too ratzz for the fast first reply explaining why my XMP profile settings were failing

    Now to watch a few more videos about what software to use to check OC stability! Taking suggestions ;)

    Cheers guys!
     
    self_slaughter likes this.
  5. self_slaughter

    self_slaughter Member

    Joined:
    Dec 12, 2010
    Messages:
    1,210
    Location:
    4165
    Good luck with it all. Worst case scenario when the i9 9900k 8 core CPUs arrive, it should drop straight into your current motherboard and hopefully offer pretty good clocks compared to an 8700k.

    They are also meant to have the heat spreader soldered to the CPU again, meaning temperatures will behave more like a delidded CPU rather than the hot mess they are currently when you try push~ 1.3V (I wouldn't push too far past that voltage without delidding)

    For reference my 8086k delidded and under water with fans set to ~900rpm for near silence and running the absolute max voltage I would recommend pushing these to under water (~1.4V under avx load according to hwinfo64) runs about 66'C with prime95 stress testing.

    As for stress testing, I've found Intel burn test etc is great for quickly honing in your multiplier / vcore, and prime95 is great for really stressing it hard once you think you have it nailed. I've found most clocks that appear to be stable, but arent quite will fail in prime95 around the one hour mark.

    So that's usually my rough benchmark for stability. Past that if you get any weird issues crop up later on just up the vcore ever so slightly and resume normal use and see how it goes.
     
  6. OP
    OP
    Sploody

    Sploody New Member

    Joined:
    Sep 11, 2018
    Messages:
    13
    Location:
    Brisbane
    Thanks for the notes on VCores as well - I am confident in my current CPU and cooling to start around the 1.3v mark, and I will go from there.
    Ive ended up downloading Cinebench, Intel burn test, Prime95 along with RealBench and Unigine Heaven. basically grabbing everything noted in that vid lol.

    Any suggestions on a good temp/volt/clockspeed monitoring program?

    now that it's being said over and over again...maybe i should delid the cpu ;)
     
  7. self_slaughter

    self_slaughter Member

    Joined:
    Dec 12, 2010
    Messages:
    1,210
    Location:
    4165
    hwinfo64 is great for monitoring your current vcore / temps etc. Load line calibration will skew your real vcore from what is set in BIOS and depending on your board is a bit of a balancing act to get things to sit where you want them.
    ie. Just because you set say 1.3V in bios, doesn't mean under load it wont get a potentially lethal dose of voltage (1.5V+) if you have LLC set too high.

    Shows more sensors in your system then you would of thought even existed. lol
    ~1.3V vcore is normally where I'd max out, but with this CPU I decided screw it and just fed it to it. Only time will tell how long the clocks will hold up before the CPU starts to degrade. :p
     
    cmi83 likes this.
  8. OP
    OP
    Sploody

    Sploody New Member

    Joined:
    Sep 11, 2018
    Messages:
    13
    Location:
    Brisbane
    So i nearly bricked my PC... like... no kidding (grab some popcorn - long post)

    I watched that vid on the OC for the ASUS mobo, understood pretty well most of what he was saying, only to go into my Gigabyte BIOS to see that there are about a thousand different settings/names for similar settings so that fucked me up pretty good. Could understand the VCore settings, multipliers, etc. but then there were all these extra settings.

    So I got onto the net looking for a guide for the Gigabyte Z370 board, and found one claiming easy 5GHz OC on the 8700K with very little voltage (automatically seemed a bit sus to me). Anywho I went diving and got into the BIOS with a bunch of notes taken from the video and this guide about the gigabyte board : https://overclocking.guide/gigabyte-z370-overclocking-coffee-lake/
    Followed the guide much closer than I did to Sentinel's video as there was heaps that did not seem to carry over.
    Since using software OC (intel XTU) i was sitting at 4.7ghz with very little to complain about in terms of how able the PC was at doing what I wanted to do, I only went to the 4.8GHz, increasing the VCore by 0.01V to compensate, along with allowing XMP (limiting my ram to 3800MHz rather than the full 4133MHz specified for the board), and bang, pre-BIOS restart loop.

    I stopped the loop, and back into bios, bumped VCore up from i think 1.28V -> 1.30V and leaving the x48 factor, and tried again. This time it loaded through BIOS and instant BSoD... I dont remember the error unfortunately (some Kernel error), but I went back to BIOS, restarted at 3600MHz in the XMP profile, along with going back to x47 factor.
    This time it showed up under the AORUS logo - "Repairing your PC"....................."Diagnosing your PC".................. then repaired and bang, loaded to Win10 - thought "Fuck yeah! XMP finally works, core is happy with my voltage changes at 4.7GHz! All on BIOS too, no janky software to argue with itself!" only to realise that a few of my apps weren't installed (as mentioned before - Cinebench, Prime95, IntelBurnTest etc), and the old Intel XTU was back o_O

    Then the PC just hung.... and suuuuper slow - MSI Afterburner (which opens as soon as i start Windows) showing some heavy CPU work was happening in the background - so I waited a bit, before opening a folder of all these documents I need as my partner and I have LITERALLY just signed for our first house as of last week - instinct to see if it was all still there.
    This then created another BSoD - popping out WHEA UNCORRECTABLE ERROR(fuck); cause: NTFS (double fuck)

    So, restart - BIOS reset to defaults - tried to load - "Diagnosing your PC" forrevveerrrr - reset PC again and reset the CMOS, opened BIOS, accepted CMOS reset, and started with basic defaults - did the loop with that message again.... At this point I am freaking out.

    Reset - BIOS - Defaults - Diagnosing your PC - at least another 4-5 times, with a second CMOS clear and full power downs between, drained mobo of all power etc...

    Last hope - THE WINDOWS 10 USB I MADE! - Revert to a previous state - to just 2 days ago (fuck yeah not far at all! and no real important forms filled out on this PC since then!) - couldnt find some Microsoft Edge file so it canned the whole repair. (fuck you Edge... shitty browser)

    Tried the Repair Start Up - said it couldn't - restarted and booted from the USB again - tried to repair start up... Diagnosing your PC...........Repairing your PC (yessssssss), Restarting...... Boom! Windows 10 and here I am!


    TL;DR - back up all your shit before attempting overclocking - always expect the worst (re-installation of windows) and hope you get lucky. And overclocking is terrifying. Solid 2-out-of-7 experience.

    ... Will try again (just for XMP to work)
     
  9. Ratzz

    Ratzz Member

    Joined:
    Mar 13, 2013
    Messages:
    6,997
    Location:
    Cheltenham East 3192
    If you really want to play with overclocking, its a lot of fun. It is pretty easy to fuck windows up though. CLR CMOS can get back original BIOS settings for you, but windows is a fragile animal.

    Stick a second boot drive in there. Make your main OS drive your daily, and select your 'overclocking' drive' when you just wanna get down and dirty... or at least until you have sorted a stable overclock you are happy with and plan on keeping as a 24/7.

    You should never have anything but a minimum on your OS drive anyway.
    I have a 'data' drive, effectively a pair of sata HDD's mirrored with raid 1. This one contains pix, documents, anything important, and raid 1 means that if either of those two drives shit themselves, I'll always have a backup.
    I also have a 'movies' sata HDD, a 'TV' sata HDD, etc..

    On my last rig I had an 'overclocking' drive too, which I used to boot into when I wanted to play. Both daily and 'overclocking' OS drives were M.2 NVME. That was an Intel (Z270) ITX which had a pair of M.2 slots. I've since moved to AMD though.

    Mind you, I'm not a gamer, I just overclock for fun, and its just not the same experience with Ryzen, so I no longer overclock much since I shifted to the red team. My Ryzen is perfect for CAD and video conversions but not so much fun for overclocking as my Intel rigs were. I did still have to find the ideal settings first though, for which I initially had an 'overclocking' drive, but once I'd found the best compromise, I repurposed that drive as a cache drive. My current ITX rig only has one M.2, so the overclock/cache drive is just a sata SSD.

    If however you were a gamer, and not into overclocking, or you had already decided on your 24/7 overclock and had no further plans to experiment, you'd use/repurpose that second M.2 (or SSD) to load your favourite games from, or as a cache drive, or... the list is endless.

    Finally.. read the thread in my sig re: liquid metal :thumbup:.
     
    Last edited: Sep 15, 2018
  10. self_slaughter

    self_slaughter Member

    Joined:
    Dec 12, 2010
    Messages:
    1,210
    Location:
    4165
    hah, good times indeed. Tis very easy to kill windows after dozens of failed boots. Can sometimes take several re-installs of windows before you manage to get your settings right.
    If you only have one drive, plan on a fresh install once you have things dialed in anyway as your almost guaranteed to kill windows, especially with ram overclocking at that's got the most potential to corrupt data.

    Fortunately killing hardware is rather hard now days, although occasionally it will seem as if everything has gone to hell and give your heart a good work out.
    Kinda half the fun I guess. :)

    Also remember with those super high ram speeds, not only does your motherboard play a big role in it, the memory controller is built into the cpu, not part of the motherboard and may or may not be able to handle super high speeds.
    There's a reason there's a * that says overclocking next to those speeds on the motherboard list. So just because your motherboard says it can possibly handle 4133Mhz or something, doesn't by any means that your cpu can actually handle it.

    Another setting to play with to help that is VTT. Which is the voltage for the memory controller in the cpu itself.
    It's usually best to work on overclocking one thing at a time, ie.

    * CPU multi/vcore/llc
    * CPU uncore
    * RAM

    Then using that knowledge try to combine the lot as best you can and if you then run into instability try clock one down till they balance out. Saves wasting time on say feeding the cpu more voltage and countless restarts if its actually something like the memory controller crapping out at higher core/ram speeds/temps etc. Takes some patience, especially on new platforms where you don't really know the base settings that should work with most setups.

    De-lidding is so easy now days, some of the tools are just easy as pie. I've got an aquacomputer delidding tool here and its as easy as locking your cpu into the device, turning a screw a few times which slowly and precisely twists the heatspreader just enough to crack the glue and she's done. Clean off the crappy paste, apply some liquid metal and bobs your aunty. Far less dangerous than the old knife or whacking it in a vice methods of old.
     
  11. OP
    OP
    Sploody

    Sploody New Member

    Joined:
    Sep 11, 2018
    Messages:
    13
    Location:
    Brisbane
    Yeah i have the USB as a backup Win10, can also boot from it.

    My m.2 drive is the OS plus any games i want to run from it for smeuuth load times (i think its pubg and GTA5) along with drivers.
    The HDD is a 6TB so I have all my Vids, Movies, a whole backup of my main programs etc on it. Along with a 2TB external with again the same movies and drivers etc.
    Just stupid on my behalf to have tried to start the OC'ing without backing up those important documents (which ive since done) - also lucky that 99% of them live in all my email attachments to all the parties involved so in worst-case scenario, there would have been maybe 1 or 2 documents lost which I really care about.

    Other than that, I dont care about reinstalling all my games etc if I lose them, theyre all through Steam etc. so all save files I think are linked to the Steam Cloud save shiznit.

    Would be a thing to consider in getting a second HDD to run a mirror though. A 2nd M.2 drive has always been on the cards since day dot since I had to cheap out and get the Samsung 950 Pro 256gb rather than the 512GB - but u know, I have 3 M.2 drives on this mobo so heaps of options for some SSD speed and stability.


    Anyhow, im going to go slow this time, as on your advice self_slaughter - starting with increasing my multi, then Vcore the LLC - correct me if im wrong but LLC is the compensation the mobo uses when the CPU is put under load by passing higher V into the core to account for the sudden drop in V throughout the capacitors to the CPU???
     
  12. OP
    OP
    Sploody

    Sploody New Member

    Joined:
    Sep 11, 2018
    Messages:
    13
    Location:
    Brisbane
    I jumped right into it (again) really only using auto settings in the BIOS and activating XMP and the "Automatic CPU Upgrade" setting that the Gigabyte BIOS has - "4.8GHz 8700k" setting, and RAM to 2666MHz.
    Ran Kombustor CPU Burner for around 15mins (not long, I know) - 100% CPU load achieved max CoreV of 1.325V and temps of max 86'C which I am happy with...
    Then I loaded Prime95 and on the Small FFT test, CoreV jumped to a max range 1.348V - 1.360V on HWinfo , but CPU temps skyrocketed to 98'C - stopped that quick smart.

    Although using HWinfo - Am I meant to be looking at the V through the individual cores? or going through Vcore on the motherboard? - - - this is confusing me as the measurements given above are on the individual cores, but when looking at the mobo values, Vcore got to a max of 1.416V which I am not that happy about.

    Cheers
     
  13. self_slaughter

    self_slaughter Member

    Joined:
    Dec 12, 2010
    Messages:
    1,210
    Location:
    4165
    Ahh, this is where things get a little tricky.... Prime95 by default will use AVX instructions, which will use more voltage than normal. You can set prime95 to not use AVX instructions by adding a line in local.txt (inside the prime95 folder) that says: CpuSupportsAVX=0

    However, most motherboards have an option "AVX Offset" which you can use to drop the multiplier by a certain amount when AVX instructions are running to compensate for the extra load/voltage required. Usually just things like video encoding etc. that will use AVX. So if you say set AVX Offset to -2 and your cpu runs at 4.8GHz normally, it will apply that offset and drop by 2 multipliers to 4.6GHz when AVX instructions are in use.

    So it's probably a good idea to test with AVX support enabled and disabled in prime95, just to see how things will handle in either situation.

    As for looking at your actual voltage, its normally a bit of a scroll down in sensor mode to your motherboard, then there will be a "Vcore" line. That's the ACTUAL voltage your cpu is currently using.
    Near the top of the sensor page will be VID. Which is the amount of voltage the cores are requesting, not what they are actually getting once you take LLC and manual voltage settings into account etc...

    So if your real voltage is shooting way past your set voltage in bios under load, it might be worth dropping LLC down a level. As you load up your CPU your voltage will normally droop to a lower level, which is where it gets tricky when overclocking. Normally set LLC to the 2nd or 3rd highest level on your motherboard to try keep your actual voltage closer to the voltage you set in BIOS when they are under load. If you set LLC too high the voltage will actually rise above your set vcore under load which can be very,very bad! :p

    They are designed to droop underload to protect themselves... But it gets in the way when trying to push chips to their limits.. It's a balancing act to find the right vcore/llc settings to get voltages to stay where they need to be under load, without dropping so low that it may crash under high loads, and not so high that voltage/temps get out of hand underload.

    Edit: In saying that, prime95 is absolutely brutal on most newer cpus. Especially if you aren't de-lidded. That's why I normally just leave that till last as an extreme worst case scenario "every things crashed and gone to hell" type situation. If you are really struggling to get it under control, just skip it and work with less stressful software and deal with any issues if they arise later.
     
    Last edited: Sep 15, 2018
  14. OP
    OP
    Sploody

    Sploody New Member

    Joined:
    Sep 11, 2018
    Messages:
    13
    Location:
    Brisbane
    Ahhh AVX is the culprit. No worries then, I'll drop the AVX ratio I reckon as you said, - 2 because it srs scared the shit out of me hahah

    I just ran intelburntest on high and screen blacked for about 5 seconds or so during, and then came back just dandy. So I think this is nearing the top of the OC using mostly auto settings.

    So yeah I'm seeing VID up to 1.350v max, and Vcore at 1.416v max (I have two Vcore readings coming up... Dunno why) - the second Vcore reading was 1.386v.

    I feel 1.416v is a little high for my liking, I'll reduce my LLC because I think it's at Turbo setting or something.
    I'll keep tweaking and I'll post results on what I get a little later tonight.
     
  15. self_slaughter

    self_slaughter Member

    Joined:
    Dec 12, 2010
    Messages:
    1,210
    Location:
    4165
    yah without de-lidding, I'd be aiming to keep your actual vcore somewhere around the 1.32 mark outside of AVX tests. Auto tune overclocks always feed way too much voltage, but does give you an idea roughly what multiplier things really go to hell at.

    These things really do hit a hard wall at a certain multiplier. Might scale well with minor voltage bumps up to say 48x, but then 49x might take 1.45V vs 1.3V to achieve. Once you work out where your cpus wall is at you can get voltage surprisingly low. 1.4V+ is probably on the verge of slowly degrading your cpu. Especially at high temperatures.

    LLC is a weird thing, one voltage and one llc setting might droop from say 1.32V to 1.28V under load. up vcore to 1.325 then it might droop to 1.31 or something instead of 1.285 like you would expect. The main thing is just to get your loaded vcore close to your set one without varying by huge margins.

    Edit: I edit too much. :p

    Edit 2: To clarify on the vdroop thing, Setting say 1.32V in bios may not actually give you 1.32V, It might give you say 1.30V. That's not vdroop, thats just the nature of the software/hardware. Vdroop is where that 1.30V you are actually getting (assuming you are using a set voltage, rather than adaptive that moves up and down according to your VID) then drops down under heavy load to say 1.25V.
     
    Last edited: Sep 15, 2018
  16. OP
    OP
    Sploody

    Sploody New Member

    Joined:
    Sep 11, 2018
    Messages:
    13
    Location:
    Brisbane
    Right, so im now recording all my results and going to start an excel spreadsheet to make things a bit easier - also writing down all my changes to BIOS and test conditions etc (instead of relying on memory)

    Current BIOS settings using Intel Burn Test - x48 mult, VCore = 1.340V, AVX Offset = -2, LLC = Medium.

    Ran the test on standard when the room was pretty cool (been afk for a couple hrs) and its now at least 30'C in here - Vcore dropped to lowest 1.248V/1.254V (remember I have two readings for some odd reason).

    I'll run 3 more tests, another Standard (because PC has been on for a little bit and its now pretty warm in here) while recording VCore and temps etc. and ill write all this down in the spreadsheet - recording only the minimum and maximum Time(s), GFlops, and make sure no weird results.

    Does this sound like an accurate way to determine what my LLC settings need to be? Also, I notice when idling, my Vcore on the motherboard can fluctuate from 1.320v up to 1.356v - is there some reason why this is happening? Do i need to check a BIOS setting so its more stable?
    Second question along with the last - do your Vcore settings remain pretty stable when idling and when under load? - obviously not accounting for drop/increase in V when under load due to compensation from LLC.

    Thanks

    Also - the edits are SUPER helpful, please... keep editing! hahaha
     
  17. OP
    OP
    Sploody

    Sploody New Member

    Joined:
    Sep 11, 2018
    Messages:
    13
    Location:
    Brisbane
    I just noticed something concerning when running burn test on High - my front x3 radiator fans stopped spinning o_O
    They are hooked into my HPWR Pump header whos fan curve is copied to my top rad fans and exhaust fan... Any reason for this? kinda concerning if under large load, my main rad stops doing its job...

    Ive noticed this before about x1 of my top rad fans - during Kombustor CPU Burn, one of them may stop - I believe he is plugged into my standard SYS-3 Fan port, but all others kept working then, and now with front 3 not working, the top two were working. Any thoughts? Bad port to have a fan plugged into?
     
  18. zx_spectrum

    zx_spectrum Member

    Joined:
    Oct 13, 2010
    Messages:
    40
    On many previous generation Gigabyte boards, the best LLC setting was always 'High'. The High setting should give you minimal Vdroop with very little overshoot. Go above that to Turbo or Extreme and you will have significant overshoot that can shorten the life of your processor. I have an old X79 Gigabyte board overclocked to the max. that the CPU can support and have run LLC on High for about 6 years now. Still going strong.
     
  19. self_slaughter

    self_slaughter Member

    Joined:
    Dec 12, 2010
    Messages:
    1,210
    Location:
    4165
    Ahh I've been so lazy with this cpu, only spent about an hour on it then let it be. lol
    I was using a solid vcore at ~1.408V with extremely minor fluctuations.

    All this talk has got me playing with offset voltages now and trying to set up a better long term clock. Still at 5.2GHz with 0 avx offset. Idling at 0.592V, fully loaded up in prime95, large ffts... without AVX instructions at 1.360V, and with AVX instructions at 1.408V. Will take more testing to ensure it's still truly stable, but turns out this is quite the nice CPU. I got lucky for a change, will have to try push it harder. lol

    As for the fans, extreme benches kinda hog ALL the cpu which can cause some things to act funny, Allthough I have all my fans running through a corsair commander pro manually set to a static speed to reduce noise. So not too much idea on whats going on there for you. :/

    Edit: I've been a bit out of the loop oc wise for a while, but AVX loads seem to use whatever voltage they feel the need for, so seems best to test without avx, using a manual voltage so you can work out the bare minimum it needs to stay alive at your given clocks. Then adjust your avx offset till that runs at a reasonable voltage. Then experiment with offset clocks to get idle voltage down low.

    Current testing on prime95, large ffts, avx disabled.
    [​IMG]

    Edit 2: and trust the vcore to spike a little higher when I was capturing and uploading the image. lol. With just prime95 running it was bouncing between 1.344 and 1.360.
    Still a nice drop and got my temps down even further. Loving it.
     
    Last edited: Sep 16, 2018
  20. OP
    OP
    Sploody

    Sploody New Member

    Joined:
    Sep 11, 2018
    Messages:
    13
    Location:
    Brisbane
    Wow dude those are awesome numbers! Wish I had known about the 8086K :(

    Ive been working on my voltages all day today - have figured out my ideal settings:
    LLC = Turbo
    Vcore = 1.290 (so far, yet to drop more as i started at 1.350v)
    still 4.8GHz
    Always testing between in IBT (standard and High) and CineR15 - ive managed to drop my temps from 69*C pkg, 80*C CoreMax --> 60*c / 72*c on the High setting. Ill give Prime95 a go in a second - im interested to see what that does to the CPU and temps at this voltage.

    If it can handle it, I might consider moving on up to 4.9GHz to see how that takes 1.290v because at this rate, I feel i can either keep dropping my V or raise the mult.

    Also - what would you have enabled to allow those idle voltages to go down to 0.6v? I will have to have another look - im pretty sure I disabled it (some power saving feature) so my stuff is constantly at 1.29v and 4.8ghz now

    Edit: I also found this page - super helpful as it goes through my combo of the Aorus Gaming 7 / 8700K https://www.overclock.net/forum/6-i...-gaming-7-8700k-overclocking-settings-65.html
     
    Last edited: Sep 16, 2018

Share This Page