Official Intel Haswell (K) Overclocking thread

VRIN LLC = VRIN Loadline Calibration.

VRIN is set to 1.800V manually.

The mV for CPU and RAM was applied 7 weeks ago (upped by 50mV), hasn't made a difference.

I'm using G.Skill ARES C9 2133 MHz (blue kit, 2x4GB)

Forgot to add, i did switch to Enhanced Memory Stability aswell ^^

I will still swap Memory if it turns out to be the issue, i had 3x4GB Mushkin Redline in my old X58 setup (Gigabyte aswell) and it was rock solid.

It's kinda hard to test this issue, as it only happens every 6 weeks or so.

 

Something weird as **** happened.

I upgraded my bios, it resetted all my settings, so far so good, I thought I would just set them back.

Now for some reasons, the voltage during full load is 1.194v by default (Auto). I used to run an offset of +0.025v before, which gave me 1.16v total.

Now, I would have to offset it again with -0.034v to get to 1.16v but my idle voltage is too low and would simply make my computer BSOD. I am now running an offset of +0.010v because it still BSOD at idle, which means my full load voltage is 1.204v.

I reverted back to my previous BIOS and same thing happens. I am not sure what to do and why this happens. Its the summer and my temps are higher than usual, so the extra voltage is not really welcomed.

Since I run my computer 24/7, I dont want to set it manually. The idle voltage will always stay high if I do.

You know, previously when I booted into the BIOS, I had the CPU voltage that was detected by the BIOS (on auto), so I could just add the offset.

Now the BIOS tells me 1.154v on Auto. If I add 0.006v for the 1.16v I need, it doesn't matter. It still says 1.194v (+ the offset I set it at) during full load.

Another thing thats weird. Depending on the voltage settings, manual, offset or whatever and LLC set to Level 1 (0%) in the BIOS, HWMonitor reports either 0v, 1.229v or 2.017v.

No way to fix it?

 

I replaced my htpc mobo and cpu.

From i5 4440 3.1ghz, h87i-plus to i5 4690k, h97i-plus.
Asus stated on their site that my h87 can use 4690k through a bios update. I updated it, installed the new cpu and wouldnt boot... was annoyed. So I got the h97 board and everythings fine.

Im really fkn impressed with the temps of DC. i5 4440 @ 3.1ghz using a h100i stress testing would hit max 49c... now with the 4690k which is 400mhz faster would max out at 36c... thats 13c cooler! Thats insane... both were just over 1volt... pretty sweet... just thought I let everyone know incase they wanted to know.

Now to OC it

 

Fully Manual Mode : Enabled
Voltage on Auto - 1.154v

Fully Manual Mode : Disabled
Voltage on Auto - 1.194v

What the ****? I am so pissed. I can't even overclock anymore, I get BSOD even at 4.2ghz. Even downgrading BIOS doesn't do a thing. I don't understand.

 

is this normal for 4690k?

just testing temps at stock, it seems under full load the cpu voltage is always 1.007v... is this normal?

 

Why do you tweak with offset?

Use adaptive voltage and by total turbo voltage manually enter 1.165v. Keep LLC - auto.

 

I gave up playing OC'ing after 2 days trying to stabilize 4.6 GHZ and set everything back to default to keep the PC cool for now.

I noticed CPUZ, HWINFO64 and HWMONITOR reads my vid and vcore wrong when compared to Asrock OC Formula software since when the system is on load, there are voltage fluctuations yet its not shown in the mentioned apps.

My CPU specs:

Stock vid 1.088
4.4 GHZ Stock Vid 1.188
4.5 GHZ all on auto vid 1.230 - this is rock stable on x264 benchmark
4.6 GHZ all on auto vid 1.268
4.6 GHZ x264 testing BSOD x101 - needs more vcore or vccin

So I slowly up vcore and vccin on adaptive (so I can see voltage fluctuations in the Asrock app).

On load vcore goes from 1.268 to 1.288, vccin 1.82x - BSOD x101 as I get near the 50% mark on x264 benchmark
As I up vccin from 1.8 to 2.0 to 2.1, I can see vcore pulling more voltage 1.268 to 1.328 core 1/3 and 1.344 core 0/2 to 1.344 on all 4 cores.

Temps go up to mid 80s but it then crashes on the second round of x264 bench near the 50% mark. Corsair H50 is running out of capacity here!

I think its not the cpu but my cooling. I just saw the Swiftech H220x review on XS and it was able to keep a 4.9 GHZ stable vs a Corsair H105 BSOD x101 @ 4.8 GHZ at 78/79 + C temps.

Gonna revisit OC this cpu after I pull together some money for a better case to fit better cooling.

Good news is that HASWELL rocks all over LYNNFIELD for SVP 60 FPS movie watching. Previously my system would max out all 4 cores with CFX OpenCL with 1080p video streams, but this system @ 4.4 GHZ Turbo is super smooth @ ~ 40 % cpu usage watching the same video streams @ 60 FPS.

Well worth the force replacement just for 60 FPS movie/video watching.

Plugged my CFX back in - colors and clarity are way better than the onchip Intel GPU.

 

Thats not really how it works on Asus boards.

When fully manual mode is disabled, the voltage is 1.194v

Offset allows me to reduce it to 1.16v but also reduce the idle voltage by that much, causing me to BSOD.

Adaptive voltage is the same, I get to input the offset. Additional turbo mode voltage does nothing if its lower than the 1.194v (or I cant get it work at all, no idea if its an issue with asus)

Besides, its already at 1.194v, so additional voltage doesnt seem good to me (since it doesnt reduce it, like I said, it seems to only work past 1.194v).

My LLC is at Auto. I also tried Level 1 to disable it.

EDIT: Tested again. Additional turbo mode voltage doesnt do anything when the value is not above the detected full load default voltage (1.194v).

I wonder if its a sign my cpu could be going bad? The voltage was previously detected as something like 1.13v so i could use like +0.03v offset and now after a bios update, its detected as something like 1.194v and have to use a -0.034v offset. Reverting to an older bios version doesn't fix it. So why would my motherboard detect higher full load voltage by default?

Yet, if I manually set it to 1.16v its stable. So I don't think its my CPU going bad?

EDIT2: From Praz on the Asus forums : When performing a UEFI update there are times that one or more blocks may be updated that will not be rolled back when reflashing to a previous UEFI version. In these cases to get back to where one was before the update the flash needs to be performed with an EPROM programmer.

So I am out of luck. I am guessing this bios update messed up the voltage detection.

 

I just got a noctua dh-15 to replace my megahalems. Funny thing is that when I first started it up the temps were, in intel burn test, around 74 max. I then reseated the non-middle fan to sit closest to the outtake and got ~ 89 max. Stuff like that always happens to me. I even reseated the cooler twice but I cant get it below 89 as a max. Still a decent improvement cause my old one would go to 94-95.

Well this is going to be nice for a while but I will probably be getting the swiftech h220x when it arrives at stores here, its currently 55$ shipping here and Im not going to pay for that. Also they are having some issues with the refill screw and the batch that will be shipped here will probably have that fixed.

I see you have been following the thread on oc´s too tj. I have been wanting to get one ever since the h220.

Ok so now to my question. My temp never goes above 75 in normal usage, ie games and whatnot. However in ibt it goes up to 89, but its stable. Would you say thats acceptable results?

My vcore is 1.283(+-0.001) @ 4.6ghz, im pumping in 1.475 into the ram (1.5 -+0.075 acceptable according to kingstons specs). This is on a 4790.

 

^
IBT is HOT HOT, imo those temps are fine.. What kind of temps do you get in something like Cinebench15 or x264fhd benchmark?


@MasterBash

I mean like so, set offset to + and leave it at auto, now enter only 1.165v like on pic bellow



And are you sure its not cpuV and not cpuVID that goes to 1.19v?

 

Maybe a better case? That dh-15 should be even better than my h100 to be honest

 

71 c on the hottest core with cinebench15. Was a bit surprised cause it was over so fast. Thats with a cpu fan curve that gives me 90c in ibt instead of 89. Needs a little tweaking I was hoping it would lower the temp. Going to try a bit of bf4, which is a decent test here since I´ve lowered all gfx settings (including fairly thorough user.cfg) and that means it will tax the cpu a bit extra. Will try divinity original sin also since it pretty much only taxes the cpu, and not my gpu so much .

 

Case is open atm. Its rather hot here ~25c.

Update: temp is up on 91 now. Since theres not much of a difference between that and 94 I think I will be returning this and wait till the h220x is available. My prolimatech will do the job just fine until then.

 

Thanks for replying. It is the CPU Core Voltage (as you see in your screenshot) and VID under HW Monitor and CPU-Z next to Core Voltage. I know what you mean with that screenshot too. Unfortunately, Additional Turbo Mode Voltage does not work for me if I set a value of less than 1.194v (In fact, anything less than what is being detected by the motherboard as the required voltage). I looked it up on the Asus forums and apparently, it appears to be that way with a few people.

EDIT: Apologies. It works, thanks a lot. The reason why it seems confusing (and probably to some people on the Asus forums) is that extra voltage seems to kick in during AVX stress test. Why is that? I know for a fact that it didn't happen before, even during prime95.

Usually it kicks in when you use Auto voltage, not a set voltage. Weird... Or maybe I am wrong, I haven't messed with my OC for a while until now.

 

Cool, btw its how it is.. Its a offset & Adaptive voltage thing.

http://rog.asus.com/forum/showthrea...r-Overclocking&p=269776&viewfull=1#post269776


If using adaptive try to avoid those stress tests/apps.

 

Been running some tests and have found that a small core clock hit is worth it for raising uncore clocks. For example 4.6/4.6 core/uncore is quite a bit better in overall speeds and system feel than 4.7/4.2 core/uncore.
My ssd test seems to really like the higher uncore speeds;

 

^
Are you sure its not C-states conflicting with IO?


Higher Cache freq. is here mostly for memory communication and L3 speeds.
But its minimal ie 4.2 vs 4.6ghz, 5-7% at best in raw cpu performance.

I tested this in Cinebench15.. I gained total 4points with 400mhz higher Cache muti..But I didnt saw its worth it because it needs more volts and extra heat.

 

^ same as I found as well. For my daily oc it's not worth the extra voltage and heat to raise it from it's default - 39.

 

Im keeping mine at min 35x (idle), max 42x as it still needs relatively low voltage ~ 1.137v.

Also anything higher can be very picky at higher cpu OC (4.7ghz+) and higher ram OC (ie 2400mhz OC), needs way more volts then usual or its not stable at all..

The max I tried was cache multi x45 @ 1.20v (cpu multi @ 45x needs only 1.17v) and yeah it still wasnt enough, could 0x124 randomly. Didnt try higher though, returned to cache 4.2ghz.

 

do i get 7% going from 39x max cash freq to 42x ? it didnt over volt that
much on auto , only 1.230 v still in yellow zone and doing that much heat .
i mean 7 % is worth it . its like going 4.8 .