Discussion in 'TechPowerUp Club Forum' started by cadaveca, Jun 9, 2013.
Yes imo anything over 2.0v sounds dangerous
I can confirm it is.
I had a much better experience with per-core mode either because of the BIOS updates, or because I'm using adaptive mode. Although, it has one interesting behavior- it seems to stay .025v under the target turbo voltage. For example, I set:
Core total adaptive mode voltage: 1.3v
1 core: 48x
2 cores: 48x
3 cores: 47x
4 cores: 47x
If I use the IET stress test, it will choose 47x and 1.275v. Occasionally, especially when I stop the test, there will be a spike to 1.3v as the cores drop to idle ratios. Similarly, if I set Prime95 to blend with 2 workers, the CPU will choose 48x with the same 1.275v (and surprisingly not BSOD, at least in the short-term).
Then, if I set 1 and 2 cores to 47x also, the 1.3v turbo voltage is used the whole time. I'm not sure why it drops load voltage just because you change the turbo for a certain number of cores active.
Also, it's worth mentioning that Windows is still in charge of which CPU cores to use. The power plan in use, such as Balanced or Performance, will affect how CPUs and hyper-thread slots (or whatever) are used. With the above 48x/48x/47x/47x setup, I ran Prime95 with 3 workers and a Balanced power plan. Win7 sees 8 CPUs, and it will bounce those worker threads from CPU to CPU. The effect is that sometimes only 2 cores will be used, allowing for 48x operation, and other times 3 cores will be used at 47x. So...
1. You will constantly fluctuate between 47x and 48x as the OS shifts the workers around.
2. You cannot control which particular cores run at the faster ratio. The Asus BIOS, for one, might seem to present the Per-Core mode as a way to set a particular ratio on a particular core. In reality, the settings represent "ratio for 1 core active," "ratio for 2 cores active," and so on. So there's no way to bump up the ratio on just that cooler-running core #4.
Sounds like a good idea... I'm also testing an adaptive per core overclock using some "stable" 46x settings... Might as well try and get the single threaded boost if I happen to not be "crunching".
Cores 1&2 @47x
Cores 3&4 @46x
When I first OC'd this chip I set the voltage, didn't use adaptive. Left all power savings on. When idle the CPU volts would drop off. No need to use adaptive. Later I had a play with adaptive. And now, I have to use adaptive if I want the volts to drop off. C3 and EIST alone won't lower the volts when idle anymore.
haha i knew it
And I noticed something with 0x124 bsod.,
Cache voltage is very sensitive to it, I tested 4.5Ghz at 1.18v (cache min 35x; max 40x) and i knew this was stable and yet it gave me random bsod while encoding videos using x264 codec - this uses all my cpu and sometimes spikes to 1.25v because of AVX.
I raised adaptive cache voltage from 1.115v to 1.125v and behold all bsod stopped, i monitored cache voltage and noticed it jumped from 1.079 up to 1.14v.
I will test 4.6ghz and lower cpu voltage to see if i can get away with it. 1.216v would be nice vs 1.235v.
Manual a.k.a. static voltage doesn't change with lower power states right? Is it possible you were using an older version of CPU-Z that was reading the voltage incorrectly?
If you used OFFSET mode the voltages will still move around, but will be increased or decreased as a whole by the variable you set it to.
Adaptive is much better IMHO. It can go to minimal STOCK idle voltage and applies the OC voltage when it needs to be there. Leave the Offset at AUTO, then set the Adaptive voltage below that to the 1.xxx you need.
Manual voltage will NOT drop off...at least in my experiences with this and other Z87 boards and previous generations. If it does that may be due to vDroop but not a reduction idle situation. I've used all 3 modes while playing with my crabby 4770k.
Speaking of...my chip decided to have a good night last night and then today it's a pain in the ass! I am almost frustrated back to the point of going stock...which defeats the purpose of buying this damn K chip in the first place! Does Intel warranty a crappy clocker under it's OC warranty? LOL!
Last night I dunno what happened...all of a sudden things were lining up, and stability was easy to find, I kept dropping votlages. I had 1.24v on the CPU, +.025v on the ring (I tried adaptive on the Ring, it locks voltage..using offset allows idle voltage), and 1.75v on VRM, was looking good and passing all tests. But now today...not stable.
Though ambient temps are higher today, I am peaking 75C under load tests this afternoon which is much cooler than when I was pushing for 4.4Ghz. Still just trying for 4.3Ghz...I am just convinced this should not be THIS hard to attain with stability, decent voltages and acceptable temps...I am apparently WRONG. Anyone wanna trade?
EDIT: Reading this guy's OP where he had similar issues I did, though his chip ran hotter and got it RMA'd from Intel. Maybe I have a faulty chip? Gonna download Linx and IBT and see what my GFlops are measuring out to be and research into that method. His replacement chip is a sweet OC-er so maybe...just maybe I have a chance. This current chip is beyond frustratingly inconsistent.
Any thoughts Dave? I'm curious what your take is...am I wasting my time? Do I just have a crap clocker? Maybe I'm doing something wrong. I dunno...it just annoys me that something that I can make stable one day...is crap the next...hell I actually had a BSOD booting into windows at 1.24v this morning...that insta-pisssed me off after dedicating so many hours of yesterday to trying to find stability with this thing!
Yeah, that's what I think. In order to get CPU speed, you should drop cache to 35, maybe, and maybe lower memory a little bit, run 1600 or 1333 if you aren't already.
Your chip is 1.070 V @ stock..to me, that's 4.3 GHz chip. 1.060 V is 4.4 ghz, 1.050 is 4.5 GHz, on down to 0.950 V.
But then, I see that PolRoger, who should have a 4.9/5.0 GHz chip, isn't getting that clock yet.
Of course, I came to this "bin by BIOS V" from having 15 chips only, maybe I haven't had enough chips yet. Seems we are missing something still, or pushing the Cache is the issue, since PolRoger is pushing cache higher than what I think it should run at.
If you run cache/CPU at 1:1, or like cache 3x multi down from CPU, my "rules" don't apply, since that changes the loading characteristics of the chip.
Overclocking while running reboots, rather than shut-down, will lead to false stables, IMHO. I ALWAYS fully shut down a stable clock, leave the rig for a couple of hours, then come back and boot it up to see if it works. Cold boots and reboots get slightly different results, has been that way for a few generations now. I have a video on my youtube channel back form 775 days, on the Foxconn BlackOps board, showing how timings can change in reboots when system warms up, and that was many many years ago. I still look for this behavior, and I see discussions with the extreme guys dealing with setting RTLs properly. These chips are easy to a point, then OC becomes MASSIVELY more complex.
Does memory affect core clocking that much?
Also I'm gonna catch the occasion to thank you guys, this thread is really a SOURCE
not really at low clocks, but as has been said, core speed is king, so why not try? It really seems that to max out each area requires dropping the others lower at the high end, so might as well try at the low-end with a poopy chip...
It sure is! Thanks to all that have contributed thus far and also to Dave for all of his insight
Is that stock voltage thing more of an exponential scale, where the difference between 1.010v and 0.990v is only like 25MHz, whereas 1.050v to 1.030v might get you 200MHz? And I'm pretty sure everyone hits a wall at 4.8 or 4.9GHz- the voltage requirements jump to 1.4v range and then 1.5/1.6v range. I haven't heard anyone claim 4.9 or 5.0GHz 24/7 stable, and not at anything less than 1.4v even for unstable.
The memory controller, cache, and core all scale exponentially also, one would assume. Additionally, as you approach aggressive settings on any one of these factors, it will increase the slope of the other two factors.
This probably varies chip to chip, but again exponential. There probably isn't much difference in core clocking stability between 1333 and 1600 for all chips. And from what I've seen, 1333 to 1600 is a big difference in performance across the board, but speeds higher than 1600 see much smaller gains.
Honestly, I am sure you are right, but the range where that starts to apply...is unknown. Maybe PolRoger has that chip...maybe it's @ 1.000 V. and maybe 1.075V it increases...my chips have only been within those ranges, as are most here. That's honestly why I started this thread, so we could find these things out.
I just realized...I'm totally geeking out on this stuff.
I guess we can safely say that unless Intel does some radical changes to the CPU pipeline there won't be an air/water clocker that goes much higher than what we've been seeing from Sandy to Haswell.
This chip does scale insanely good with COLD though, I've seen some 4770k fly really high on phase.
That's one thing I haven't done in a while is drop the cache to 35.. been running 8 Min and 39 Max. Memory I have backed off from 2133 to 1600.
Well maybe 4.3 is all I'll get...maybe I need water to attain 4.3? I can't imagine 70-75C loads on air with a higher ambient temp on this summer day can really be killing it seeing as I've read some guys are running right at the thermal throttle with OC's...though whether or not they're stable I don't recall reading...
Damn this CHIP! lol
I tried doing a 1:1 ratio a couple times with HORRID results...if you thought I was pissed from instability at 43 CPU/39 Cache...man-o-man!
I can't say my chip has been easy on any OC point...at least beyond 4.2GHz which was a simple set it to 1.20v and go...it gamed okay, but I never did back all the way to that and test for stability. I am at stock now and going to run those tests as that guy detailed in the link I listed in my last post to see if I have a GFlop issue maybe leading to a faulty chip. Honestly I'm kinda hoping for that to have a chance at a different chip at this point. It would make a lot more sense for the random stability issues I've had at all levels of operation...except stock from what I recall tho. With a K chip I was assuming a little more OC compliance with 4.3 being a realistic goal and 4.4-4.5 being the fight. Wishing there was a microcenter nearby and I had $350 burning a hole in my pocket!
Yeah, overclocking on Intel is back with a vengeance. Phase cooling is high only my list of priority purchases.
Dude describes throttling issues, gets RMA replacement. Go figure. I wouldn't read too much into that whole thread.
If you were closer I could have helped you hook with an awesome Phase builder...
Not that you miss good Phase builders over there, naturally.
I'm just tempted to pick a 4770k chip and see how it fares on my single stage, the only issue is that it'd probably die in a matter of weeks even on cold.
Oh guys look at what appeared on Newegg :ASUS MAXIMUS VI FORMULA LGA 1150 Intel Z87 HDMI SA...
Yeah, I wonder what Dumo thinks about that, he's been running water/phase a lot with memory clocking.
And yeah, there are some good phase builders, but they all seem to keep pretty busy. I'm going to school for HVAC, so eventually I'll just build my own.
I posted a pic of mine a while ago... It's been calling my name from my shelf ever since.
Ya I'd say it's back! LOL. Man I remember when the first Core 2's showed up and I was soooo happy! I could attain my e6300 1.8Ghz to 3.5Ghz and it was STABLE and on an AC F7P none-the-less! It made my hot running P4 630 setup seem like a total joke! And all the other P4's and P3's and Athlons before that too lol.
Ya I'm not gonna read too much into that thread but one of the last posts I read this excerpt which might be interesting...I think I may contact Intel anyways to see what they have to say. But I'm gonna look into that OC warranty some more...
But if that's the case, which that was the first thing I did after buying it was getting that performance tuning plan...DO I want to use it on this chip now or "if" it fails within the next 3 years?
This chip is good stock...the second I OC it, it's shit with an intermittent case of "okay" that fools me into thinking that maybe, just maybe this time it'll stick and be stable! Ha!
Well gonna go play some games that I can't otherwise play while overclocking anymore. I've spent too many hours for finding stability on what is otherwise a fairly weak OC in the first place to not enjoy this build. Owning this cpu has made it hard for me to want to sell them to my customers.
That's what I don't like/like about Haswell.
CPUs aren't good (tons of duds between chips) this gen but motherboards are basically nutella, they are selling the CPUs this gen, and not the other way around.
Also Dave how many motherboards does the Formula have ahead of it in getting a review?
Yeah, the Tuning Plan, for OC, does get you another chip for $25, no questions asked. DOM said he got a crappy chip on his RMA though...so..
I don't see any other reviewer, period, hyping the Tuning Plan like I do, in every board review. It's hard for the Plan to work without everyone supporting it fully. Not too much else I can say about that...I've definitely done my part. If we support it properly, this could be a great program for those of us wanting to OC, but not many see it in the same light, unfortunately.
I prefer that there is big difficulty in OC and that good OC costs in cooling. It sucks that Intel chips cost so much, but I sold the chips I did buy extra already, no problem. for normal users needing a new PC, buying anything else doesn't really make much sense in my books.
Meh. I have Sabertooth Z87 and Z87-A waiting as well as the Z87I-PRO, IMPACT, and FORMULA. Extreme is on my motherboard testbench right this moment, review to be done this weekend along with the Z87X-UD3H, but the Gigabyte is first. I got a memory review I am working on right now, (testing max OC right this moment) and another to start on Monday.
So, I can fit any product in after the Z87X-UD3H. I kind of want to cover the ECS GANK MACHINE though. So maybe I'll do the ECS and then an ASUS next week(what I do and publishing are a bit delayed, proof reader and W1zz still need to do their parts for each and every review, not just mine.).
You think I should cover the Formula? or maybe one of the others?
I also have several ASRock boards, the G1.Sniper5, hmm, maybe a couple others... MSI MPOWER MAX review is waiting for the front page, too....that's the next you will see.
Meh to be honest I think you could even skip it or do it at a later stage, shouldn't be much different than the Hero or Extreme.
RoG line seems to be pretty much on the same level, their ITX is interesting though.
Anyway I'm very happy that ASUS has decided to give you the rog line to review, it was a shame of them not to give you a RIVE.
I agree the Tuning Plan needs to be put out there more. Maybe a Techpowerup article? Maybe a Intel Tuning Plan TPU Club? I bought it without question after reading your OP...actually I read this thread to where it was at around mid-late June prior to ordering my hardware.
I do agree that an OC should be a challenge...but how many days, weeks, months for a 800MHz OC on a CPU until it can be called to attention of something beyond just an OC challenge. Wanna buy my chip and give it a shot?
I have read the Tuning Plan is transferable.
EDIT: Well decided tonight would be a good time to try, try, try again. I may have found stability with CPU 4.3GHz @ 1.270v, 8-39X Cache @ 1.20v, CPU VRM 1.80v, Memory at 2133 @ 1.60v. At least it passed AIDA64 for a couple hours...temps were pretty warm..but if I drop on any of those voltages I will get a BSOD. Still need to give it a longer term test to see if it's truly stable. Not amazing voltage for that OC, but not the worst I suppose.
May see if I can't get the Cache to 40-42X, maybe..we'll see. I'm gonna make sure this is stable then I may shoot for 4.4GHz again just to see what it may take. If this chip continues to play nice it can stick around for while.
EDIT Part Deux: Well in that time (what maybe 30 minutes?) Wargames: EE caused a 0x124 BSOD....so while AIDA was more stable...gaming sure isn't...
I would try cache like so,
adaptive cache offset 1.12v (this could overvolt to 1.148v max)
And keep memory at 1.65v if its a 1.65v just to be sure its not a combo of both..
As for cpu use what you think its ok.
Separate names with a comma.