4770k good (ish) temps at 1.39v

Hi everyone,

I'm a bit of a n00b, so please bear with me.

I'm trying to overclock my new pc. I have a 4770k, corsair vengeance pro 16gb 2133, asus maximus vi extreme, corsair 760i psu, asus gtx 780ti ocii, corsair h100i cooler with 4 noctua nf12 fans in push-pull, all in a HAF X case, running win 7 home premium 64bit.

I keep reading that 1.25v for the CPU should be the max voltage to use, with maybe 1.30 on some custom liquid cooling rig, which obviously I haven't got. My progress so far:

RAM is set to xmp (I know high RAM speeds can limit oc stability but ideally I would like to run it at it's rated speed)
The cpu at 4.5gig on all 4 cores will boot to windows but any stress tests (have used realbench 2 and aida64) will go to a blue screen almost instantly. This is at 1.39 volts.

Cutting my losses I set all 4 cores to 4.4 and started playing with the voltage, starting at 1.25. I am up to 1.39 again. This now boots fine and completes the realbench 2 benchmark, but the system freezes after about 20 minutes on aida64. I think I can't be far off getting this to work, so will start reducing some of the individual core speeds to 4.3. The temps using coretemp and the asus oc panel that came with the mobo max out at 98 while running aida 64 and hover around the mid 80s most of the time while running this. That sounds too high, so I will be reducing the voltage a bit before moving on.

What really puzzles me, though, is how at 1.37 volts I get max temps of 92 while running aida64. 100% cpu load at other times seems to hover around the mid 80s. My cooling is good, but not that good, surely. Have I just ended up with a weird CPU that has rubbish overclocking, but will cope with silly amounts of voltage without heating up? Also, if I end up with a stable overclock that has acceptable temps at 1.37v, will the high voltage sonehow reduce the lifespan of the cpu independently of the temps? As in the cpu gets damaged by the high voltage despite not running unacceptably hot?

Any advice at all on this would be greatly appreciated. Apologies for the long post, but I'm probably being over inclusive, because I don't really know what I'm talking about .

Thank you all in advance.

1.3v is fine for 24/7 assuming the cooler can keep up (it may). I wouldn't go much over that.

that 'low' ram clock will not affect CPU clocks.

1.39v is WAY too high for that clockspeed unless you have a TERRIBLE trip.

You need to keep that CPU under 90C while stress testing. You are WAY too hot.

Thanks for the reply.

I'm now at 1.3v again after realising that I hadn't increased cache voltage too, and going strong on aida64 at 3 hours now. Got the cores on 4.5,4.5,4.4 and 4.3. Had to dial down the RAM to 2000, though. Crash at 2 hours on aida64 when I had it on 2133.

Highest temp on stress testing has been 85, with everything hovering around the mid 70s or so most of the time.

You should have no reason to be messing with the cache voltage or speed. Leave it at auto.

Why do you not have all cores at the same speed?

How do you know it was the ram that caused the problem? The way to isolate that is to lower the overclock to stock and then test, otherwise, you have an overclocked CPU as a variable.

Well, ram was set to 1600, core voltage to 1.3 and I started playing with multipliers. 4.5 on all cores booted but didn't do a few seconds of testing before crashing. Increased Cache Voltage to 1.3 and it crashed after about an hour of testing. Started reducing the two cores bit by bit until the thing was stable after 4 hours of testing. ( I read most of the time you tend to use one core anyway, so without doing this it would all be at 4.4 and I'd miss out on two cores at 4.5). I then put the ram back to 2133 and it crashed after an hour. Increasing vccsa offset to 0.9 something (can't remember now) made it work for 2 hours plus, but then crashed at some point in the night. Hence why I blame the RAM which apparently reduces oc potential on Haswell, which is why I'm now at 2000.

Like I said, where your ram speeds are at, will not affect the CPU clocks also being so slow. The phenomenon you are mentioning is found with MUCH higher clock and ram speeds. I run 2666 all day long at 4.9Ghz to 6Ghz+(under LN2).

Hmmm. 4.5 hours and it's crashed again.

Didn't realise the ram issue only happens at higher clock speeds. Ram is stable at 2133 with stock cpu speed as the first thing I did was to check stability at the kit's rated speed and stock cpu before doing any overclocking.

Looks like I'll be dialling all the cores back to 4.4 and trying again.

I hope 4.4GHz works..

Sadly no. I have given up.

Couldn't be bothered to start dialling down some of the cores to get at least 2 of them on 4.4. The testing was just taking too long before each crash. All cores on 4.3 now and runs like a peach.

I'm happy (ish) just sad the cpu fairy brought me a crap chip. Ah well. Better luck in a few years when it's time to upgrade again.

Don't feel bad I have been on this since Haswell was released and allot of people share the same clocking speed.

Could be worse. I'm struggling to get past 4.0 on air.

Personally, if I was you, I would start from scratch. You adjusted things that do not need to be adjusted (ring speed/voltage for one example). Reset the bios, disable the power saving features. Set vcore to 1.3 and try again at 44/45.