Just thought I'd put a shout out for some comparisons of temps/voltage and what people are deeming stable on their Haswell overclocks.

 

Hoping that westman, HLJames and Techguymax might chime in with your responses as I know you've all been playing around with 4770K's.

 

I think I have a pretty decent chip. Currently I've deemed a 4.5Ghz overclock stable with uncore at 42x, vcore 1.22, vring 1.15 and CPU input voltage 1.8. This is with HT on and RAM at XMP setting for 2400Mhz.

 

For stability I've run prime for 15 hours, IBT on max, AIDA for 12 hours (with combinations of just FPU or all tests checked), Cinebench, Handbrake. Plus just normal usage. Highest temp I saw was about 93C in the AIDA FPU only test but Prime and IBT were all around 82C max. Normal usage is obviously a fair bit lower but haven't actually run FSX yet.

 

This is with specs as per my profile and a H110 for cooling in a Carbide 500R case.

 

Now pushing for 4.6Ghz at 1.26V.

 

So what has your experience been?

 

What tests do run to call yourself stable?

 

Another interesting thing I noted was that with stock clocks the vcore the system applied was higher with RAM at XMP 2400Mhz compared to the default 1333 or 1600Mhz the motherboard applies by default. As a result temps were also a little higher.

 

Based on this I'm thinking I could probably trade ram speed for a little extra clock speed, but I think the benefit of the extra ram speed might be better. Not sure.

Behind with computer tech lately - haswell is the next offering after ivy bridge?

Spot on. Been out since the beginning of June.

4.9GHz @ 1.44 - 1.456V

core temp 68* C max FSX

Stability: max cfg + building storms + LAX to JFK

 

HLJAMES

does it have heat issues like IB did?

Stability: max cfg + building storms + LAX to JFK

 

HLJames do you use anything else for testing stability? eg. Prime/IBT/AIDA etc?

 

Having trouble stabilising above 4.5Ghz. I've been reading up that tweaking +0.1V on System Agent, CPU IO Analog and Digital may help above 4.5.

 

What was your experience?

 

 

does it have heat issues like IB did?

It does run quite hot, but if you get a decent chip and have good cooling it's manageable. Intel used the same goop/adhesive as on Ivy.

I think I have a pretty decent chip. Currently I've deemed a 4.5Ghz overclock stable with uncore at 42x, vcore 1.22, vring 1.15 and CPU input voltage 1.8. This is with HT on and RAM at XMP setting for 2400Mhz.

 

For stability I've run prime for 15 hours, IBT on max, AIDA for 12 hours (with combinations of just FPU or all tests checked), Cinebench, Handbrake. Plus just normal usage. Highest temp I saw was about 93C in the AIDA FPU only test but Prime and IBT were all around 82C max. Normal usage is obviously a fair bit lower but haven't actually run FSX yet.

 

Your experience with voltages vs temperatures mirror mine, although my CPU is 1 tier below yours (needs 1.22V for 4.3 GHz, 1.24V for 4.4 GHz and 1.32V for 4.5 GHz). With Haswell, the voltage, not the clock speed will determine your temps. Without de-lid, I think you'll find 1.26V produces too high temps with Aida64 FPU-only test. Whether that's a problem is up to you to decide. You will *never* come anywhere close to the Aida64 FPU temps in any "real" application.

 

I know many overclockers simply ignore outliners like Aida64 FPU. Technically it's "cheating", but if the CPU does what *you* need it to do, no one can say it's wrong... It's however important to mention this when comparing overclocks, because you're comparing apples and oranges.

 

I'm currently running my delidded CPU at 4.5 GHz and 1.32V. This illustrates the huge difference between FSX and Aida64 FPU. I first ran FSX for a few minutes while logging temps, then quit the sim and started the Aida64 FPU test.

 

In absolute numbers the difference in peak temperature between FSX and Aida64 is about 36C.

 

What tests do run to call yourself stable?

 

My criteria tends to be: No throttling in the Aida64 FPU test and stability for at least 12 hours of Prime95 and Aida64.

For the first weeks afterwards, check for any stability issues during normal use and tweak the settings if necessary.

IIRC the Haswells do not o/c as easily or well as the Ivys...is this not so?

 

Reason is I too am looking to upgrade from my i7920 and am looking at an i5-3570K.

JimmiG, seems like we have pretty similar methods to the way we like to work, although I'm not planning to delid.

 

Had to bump vcore up further to 1.28 in my attempts at 4.6Ghz and temps are getting up there on the stress tests but I know they'll be a lot lower in regular use.

 

My main trouble is with Prime. I can pass Aida all day long but Prime is another beast for me. Longest run at 4.6Ghz so far is about 4 hours.

 

I've been looking for tips/reading a lot over at overclock.net trying to really nail down haswell overclocking, but it seems so many people aren't really stressing their systems and are just calling it stable. That may be fine for them but I'd rather know for sure. It also leads to a lot of inconsistent information and so called "load" temps etc.

 

Helo_head - it's not so much that they don't overclock as easily or well. There's a lot more variables on Haswell to play around with. I think i would say its more time consuming but not necessarily harder. Then again it depends on your goals.

 

They also aren't reaching the clockspeeds you can get on Ivy, but the thing to remember is it's faster clock for clock (so just for example a 4.3Ghz Haswell might equal a 4.8Ghz Ivy or something).

 

In my opinion if you're upgrading now, go for Haswell.

 

 

HLJames do you use anything else for testing stability? eg. Prime/IBT/AIDA etc?

Before the Digital VRM we used those floating point programs to find the proper core voltage. Sandy Bridge, Ivy Bridge, Haswell up to 4.6 GHz the digital VRM will give you a precise and accurate core voltage for the application it sees..

When your cooling can no longer dissipate the heat (radiator, heatsink) then it brings the heat back to the CPU...instability.

Hi

My 4770 is a really bad chip. Not delidded yet

Its stable 4.6 1.25 it booth to win at 4.7 not stable

H110 4.6 1.3v stable Aida and prime little to high temps

LC 2xD5 560 rad 4.8 Aida prime high temps fsx 4.9 rockstable 1.42v

SS -40C 5.2 ghz 1.48v Aida prime you name it 100% stable

 

Run my new mems 2666 8 12 12 31 1T 1.7v they do 2800 8 12 12 31 1T 1.9v

H110 4.6 1.3v stable Aida and prime little to high temps

 

Westman, what do you call a little too high temps? Were your RAM speeds constant at those clocks?

 

Unless I delid (which I'm not) I think H110 might not be enough to push much past 4.6 if I'm concerned with temps in Prime/AIDA etc. Temps would be fine for normal use though I'm sure.

 

What other changes are you making in BIOS apart from vcore? What is your uncore/cache set at?

 

I know I was not following what most recommendations say since I was pushing for my RAM at 2400Mhz and not trying at lower speed first, so I'm testing at the moment if I can do higher clocks at lower RAM speed. Currently trying 4.6Ghz with RAM at 1600Mhz, since I'm just curious. I can immediately see temps are lower.

 

But even if I can do say 4.7-4.8Ghz with RAM at 1600/1866 or 2133Mhz, I'm thinking it may be better to stick to 4.5Ghz on the CPU and 2400Mhz RAM. What do you think?

 

Also need to be cautious of temps in the coming months as we come into Summer here in Australia.

My main trouble is with Prime. I can pass Aida all day long but Prime is another beast for me. Longest run at 4.6Ghz so far is about 4 hours.

 

Same here, Prime95, especially with Small FFT's, always finds errors that Aida64 and others fail to detect, even though temps are lower. I wonder how many overclockers actually run it for more than an hour or so.

 

While testing overclocks I actually had it stable for 18 hours and as I got home from work and was just going to shut it down then *poof*, BSOD and 0x124 error just as I was going to click exit  :mad: Increased the Vcore by another 0.005V and re-ran it for 19 hours just to be sure. The good thing is that once you get it stable for 8-10 hours, it takes very small adjustments of VCore to achieve stability for longer. It's just that I want to find the minimum Vcore that is still stable, in order to reduce temps and increase CPU life span.

 

Remember to try increasing your VRM Input voltage (the one that's at 1.8V by default).This may or may not help. I had to bump it to 1.92V for VCore voltages above 1.22-ish. I know many are even using 2.0V for VRM input.

Little to high 90C

Ram and ramspeed, general Samsung based chips is best for benches and games (FSX)

I have at least 10 different 8gb ram sets to play with the best do 2800 cl8 @ -60C 2.05v

For normal allday with stock ram cooling a run 2666 cl9 (9-12-12-31-1T) @ 1.7v thats the lowest setting i use.

VRM 2.0 - 2,10

CPU @4.8 uncoreratio 44 , CPU @ 5.2 uncoreratio 47

 

I little nice benchmark for IB and Haswell based CPU :s is Intel XTU benchmark you can

Download it on Intel site

 

 

Remember to try increasing your VRM Input voltage (the one that's at 1.8V by default).

 

Thanks, I had increased that as well and been playing around with it, Thanks for the tip.

 

Changing RAM speed really had no effect on stability so I think it will just come down to bumping up voltages. Trouble is sometimes it's hard to know whether it's vcore, uncore voltage or VRM input voltage.

 

Just for fun I had a play around at 1.3 vcore. could run cinebench up to 4.7Ghz. At 4.8 it would hard lock looking for more voltage.

 

Running 1.3V and 4.6Ghz, during Prime I wasn't happy with the temps. One core hit 100C after 30 mins and started to throttle so I stopped it there. I think for the time being I may just call it a day at 4.5Ghz HT on, although I may have a play around with pushing higher, maybe even with HT off.

 

 

 

VRM 2.0 - 2,10
CPU @4.8 uncoreratio 44 , CPU @ 5.2 uncoreratio 47

I little nice benchmark for IB and Haswell based CPU :s is Intel XTU benchmark you can
Download it on Intel site

 

Westman are you using that VRM at 4.6 aswell, or just at higher clocks?

 

what about uncore voltage, VCCSA/IOA/IOD?

 

Seems our CPU's may be very similar as our voltages/temps are pretty much identical

 

I've actually been using that Intel XTU since I like the monitoring tools/graphs.

 

Have you used it much for stress testing/benchmarks?