Retail 3770k overclocking

[TidusZ]

Will 4.5GHz be your final overclock?

So I just woke up. I'm already at 4.7 with better temps than my q9550 in linx. What people here probably don't realize is that my 83c in linx is 68-70 c in prime 95, and all the prior results of temps on the Hong Kong Ivy's were in prime 95, so my temps are actually a lot better than those. I will probably end up at 4.8 as a daily clock.

The design of some heatsinks goes counter to the idea of lapping.

http://www.xbitlabs.com/articles/coolers/display/thermalright-silver-arrow-sb-e_2.html#sect0

Look half way down the page.

That's assuming you don't lap the processor. A lot of people bash on lapping but it gives a pretty good temperature benefit combined with using less TIM and heating the TIM so it spreads like butter. I'm using chill factor III.

What board are you using?

Gigabyte Z77 UD5H with F5 bios

Very nice, good luck with 4.8. Any pictures of the "lap job?" 🙂

I'm afraid not, but it was as close to perfect as I could get with 2000 grit as the last polish. It was pretty much mirror on mirror, and I heated the chill factor 3 up in near boiling water for around 5 minutes before applying a small amount. I did the small dab in middle and let the heatsink smooth it out as I tightened it down, and I tightened about as much as I could.

I also ran crystal disk mark yesterday, I have 2 mushkin deluxe 240 gb in raid0:

The computer feels really fast but I'm not done with the overclocking yet, and still need to o/c the memory.

 

[Tempered81]

Nice results, TidusZ. Keep pushing it. Still got 22C to TjMax - might be able to prime 4900

 

[amenx]

I have a feeling a new stepping for the 3570/3770 is in the cards. I give it 2-3 months max.

 

[evangelionstar]

I have a feeling a new stepping for the 3570/3770 is in the cards. I give it 2-3 months max.

Considering that IVB-E is expected around end of year and completely new Haswell platform early next year? Not a chance.

It is what it is and it aint changing.

 

[meloz]

4.5 ghz

This is what I've got so far, 4.5 ghz 1.200 V at load (2nd highest LLC on gigabyte UD5H), ~20C idle, 72/75/76/71 highest after 1h30 of linx with HT.

This is rather decent, actually. :thumbsup:

People are just too greedy and were expecting an unreasonable amount of overclock headroom. IVB is not the calamity some are proclaiming it is, you just have to ramp up the voltage gently and keep and eye on temperatures you are comfortable with.

Many so called 'experts' at other reviewing sites have not yet grasped this simple concept, they are jumping straight to 1.3 volt and then acting shocked at the temperatures.

 

[TidusZ]

I flashed my Gigabyte UD5H bios to F5 and found my overclock unstable, even after adding a bit more voltage. I rolled back to F4 bios and am running linx again and it appears stable at the old voltage once again. I would have to say sticking with F4 over F5 on UD5H is the way to go right now. I'll go to 4.8 in an hour or so and see what voltage I need and how bad temps get in Linx/Prime. Highest prime temperature I've seen at 4.7 is 75C and highest linx is 84c.

The q9550 in my sig reached high 80s in Linx but never gave me a problem in 3 years, so I'm not afraid of pushing temps in linx pretty high since its unrealistic for temps to ever get that high anyway.

 

[LOL_Wut_Axel]

So I just woke up. I'm already at 4.7 with better temps than my q9550 in linx. What people here probably don't realize is that my 83c in linx is 68-70 c in prime 95, and all the prior results of temps on the Hong Kong Ivy's were in prime 95, so my temps are actually a lot better than those. I will probably end up at 4.8 as a daily clock.

What voltage are you running now?

 

[TidusZ]

What voltage are you running now?

This is the lowest voltage I've been able to get it stable for more than 10 minutes of Linx. The temperatures are as high as they are going to go before throttling and so 4.8 seems to be the cap for linx stability testing. I will probably just stabilize this 4.8 ghz, o/c my ram, and leave this as my clock for the next few years.

Once people figure out a good way to remove the IHS and deal with that, I'll be all over that, as my heatsink is only a little warm which is almost certainly due to the crappy paste/IHS interface. I have a feeling that once that IHS is either removed or filled with some non conductive solder-like material that IVY will o/c much higher/cooler.

Imo, there is no way that a 95 C processor wouldn't make the heatsink very warm or hot to the touch, even if it is a bit smaller physically than SB.

 

[Rvenger]

I just found this:

http://www.techpowerup.com/164858/I...IM-Inside-Integrated-Heatspreader-Report.html

I wonder why Intel would do that. It could partially explain the higher temps.

If would instantly void the Intel protection plan and any other warranties. It would also limit overclocking under the Intel protection plan. Nice work Intel.

 

[LOL_Wut_Axel]

This is the lowest voltage I've been able to get it stable for more than 10 minutes of Linx. The temperatures are as high as they are going to go before throttling and so 4.8 seems to be the cap for linx stability testing. I will probably just stabilize this 4.8 ghz, o/c my ram, and leave this as my clock for the next few years.

Once people figure out a good way to remove the IHS and deal with that, I'll be all over that, as my heatsink is only a little warm which is almost certainly due to the crappy paste/IHS interface. I have a feeling that once that IHS is either removed or filled with some non conductive solder-like material that IVY will o/c much higher/cooler.

Imo, there is no way that a 95 C processor wouldn't make the heatsink very warm or hot to the touch, even if it is a bit smaller physically than SB.

Ivy Bridge should suffer from more degradation than Sandy Bridge. If I were you I wouldn't run 1.3V 24/7; only for benchmarking runs. Temperature is only one part of the equation. But if you don't care about having to lower your overclock or raise voltage in two years or so then you'll be fine.

 

[TidusZ]

I don't think 1.3 V degradation will be much of an issue. If the o/c becomes unstable a couple years down the line it wouldn't bother me that much and I'd likely just throw a 3880k in or something. I don't think 1.3 V is going to prove to degrade things that fast though. If I were really worried I'd just use turbo with voltage offset but I am not worried. 1.4 V on my q9550 is fairly high voltage and when I stability tested over 3 years ago it became unstable at about 4.3, and its still stable as a rock at 4.25. I did have bad degradation on a E4300 in the past, but that was doing 3.2 ghz from 1.8 stock and I don't even remember the voltage (probably pretty damn high) and it was good for 3.2 for about 2 years. I just put it at stock after since I didn't care to stability test it, and right now its not being used anymore anyhow.

Basically I've seen people say IB is worse than SB for degradation and I've seen people say its better. Nobody really knows outside of Intel. It is probably worse, but at 1.3 which is not an insane voltage it will last long enough that I'm not going to be disappointed (i.e. if 4.8 ghz lasted a couple months).

 

[blackened23]

Nowadays, I like using load line calibration so that my CPU's only overvolt at 100% load situations - that definitely helps prevent degradation. All the new asus motherboards have load line calibration settings where you can tweak your voltages, your voltage will be low at stock non turbo and will go up at 100% load, definitely gives me peace of mind.

I had a lynnfield 870 die from degradation, basically I pushed the heck out of it though. REALLY overvolted it and left it on 24/7....and it died 1.5 years later, constantly BSOD'ed at stock speeds - and symptoms went away by replacing CPU.....by the time they died I didn't really care anyway, I like getting shiny new toys.

 

[Rvenger]

Ivy Bridge should suffer from more degradation than Sandy Bridge. If I were you I wouldn't run 1.3V 24/7; only for benchmarking runs. Temperature is only one part of the equation. But if you don't care about having to lower your overclock or raise voltage in two years or so then you'll be fine.

I don't think voltage directly correlates with the die shrink process. Others were pumping 2v into this chip to get it over 6ghz and it was still surviving.

Does anyone know the idle and load stock voltages of Sandybridge vs Ivy?

Bulldozer's max safe is 1.5v, the same as its older brother Phenom 2.

 

[LOL_Wut_Axel]

I don't think 1.3 V degradation will be much of an issue. If the o/c becomes unstable a couple years down the line it wouldn't bother me that much and I'd likely just throw a 3880k in or something. I don't think 1.3 V is going to prove to degrade things that fast though. If I were really worried I'd just use turbo with voltage offset but I am not worried. 1.4 V on my q9550 is fairly high voltage and when I stability tested over 3 years ago it became unstable at about 4.3, and its still stable as a rock at 4.25. I did have bad degradation on a E4300 in the past, but that was doing 3.2 ghz from 1.8 stock and I don't even remember the voltage (probably pretty damn high) and it was good for 3.2 for about 2 years. I just put it at stock after since I didn't care to stability test it, and right now its not being used anymore anyhow.

Basically I've seen people say IB is worse than SB for degradation and I've seen people say its better. Nobody really knows outside of Intel. It is probably worse, but at 1.3 which is not an insane voltage it will last long enough that I'm not going to be disappointed (i.e. if 4.8 ghz lasted a couple months).

Except the newer architectures and CPUs that have them aren't as resilient as old ones. As you go down process nodes, degradation from overvolting increases.

IB IS worse than SB for degradation, no doubt. That's how its been in all past "ticks" in comparison to their predecessors, including Penryn and Westmere. Having higher temperatures than Sandy Bridge doesn't help, either.

But again, if you're gonna replace the platform in two years, you'll be fine. Just remember that Ivy Bridge is end of the line for LGA 1155.

 

[ehume]

I think the answer is on pp 8-10 here. Intel essentially ran into problems with lead-free solder, found their indium solder deteriorating with the stress of heat cycling.

"Typical failure modes encountered in STIM joints relate
to thermal fatigue cracking of Indium close to the
IHS/Indium interface . . ."

"Figure 20: White CSAM image showing delamination
at the Indium/IHS interface and corresponding crosssection SEM image showing cracking post reliability
testing."

So it appears even mighty Intel runs up against the limitations imposed by Ma Nature. Using traditional TIM rather than indium solder TIM allows their chips to be more reliable rather than passing more heat through. At stock Voltages, the traditional TIM is Good Enough.

What we really need is an Enthusiast Line of i7 3770K that explicitly states the chips won't live as long, but they have the old-style solder TIM and will run cooler with a good heatsink or WC.

 

[blackened23]

I don't think voltage directly correlates with the die shrink process. Others were pumping 2v into this chip to get it over 6ghz and it was still surviving.

Does anyone know the idle and load stock voltages of Sandybridge vs Ivy?

Bulldozer's max safe is 1.5v, the same as its older brother Phenom 2.

Well degradation doesn't happen overnight, it takes a period of months and years to become symptomatic. Best bet is to play it safe by using the LLC feature of your motherboard BIOS!

 

[LOL_Wut_Axel]

I don't think voltage directly correlates with the die shrink process. Others were pumping 2v into this chip to get it over 6ghz and it was still surviving.

Does anyone know the idle and load stock voltages of Sandybridge vs Ivy?

Bulldozer's max safe is 1.5v, the same as its older brother Phenom 2.

No, it is not.

And 1.5V was too high for Phenom II unlike you're like the people with SB running at 1.4V that were gonna replace it later anyway. I've seen plenty of samples suffering from degradation two years after running 955 and 965s at 4GHz or 4.2GHz with it, needing another voltage boost to become stable again.

Degradation is something that happens gradually; how much of it happens depends on voltage and temperatures. As you raise voltage higher, degradation becomes higher. What they call "suicide runs", or running a very unsafe amount of voltage to get a very high overclock, won't necessarily kill your CPU right away because, like I said, degradation is a gradual process. It's if you run your chip like that for 3 months or more that's when you get into trouble.

 

[Don Karnage]

No, it is not.

And 1.5V was too high for Phenom II unlike you're like the people with SB running at 1.4V that were gonna replace it later anyway. I've seen plenty of samples suffering from degradation two years after running 955 and 965s at 4GHz or 4.2GHz with it, needing another voltage boost to become stable again.

Degradation is something that happens gradually; how much of it happens depends on voltage and temperatures. As you raise voltage higher, degradation becomes higher. What they call "suicide runs", or running a very unsafe amount of voltage to get a very high overclock, won't necessarily kill your CPU right away because, like I said, degradation is a gradual process. It's if you run your chip like that for 3 months or more that's when you get into trouble.

Phenom 2s was 1.5v? I thought it was 1.55

 

[LOL_Wut_Axel]

Phenom 2s was 1.5v? I thought it was 1.55

1.55... if you're running water.

1.5V for PII is like 1.4V for SB. It'll run fine for two years, but then you'll need to up the voltage again or lower the clock speed.

I personally think of CPUs as an investment for the time I have them, and I will keep my 2600K for a total of 3-4 years or perhaps even more. If you're gonna replace the chip in 2 years high voltage and clock speed is fine, but I'd rather go for efficiency and draw a line as to how much I'm willing to push. 1.3V is like 1.4V on Sandy Bridge, but it's up to you if you want to run that much voltage.

 

[TidusZ]

I'm curious what sort of voltage other people are requiring with their 3770k's.

 

[LagunaX]

You'll know later today and tomorrow.
Microcenter is selling today for $289 plus tax.

 

[LagunaX]

Crap...looks like 4.6ghz may be where I'm at.
Prime95 v.27.6 x64 with AVX needs more volts than regular Prime95 and hits 89c (stopped after 30min) on air with a Prolimatech Armageddon and 2 140mm fans.

Might go back to Prime95 v.26.6 x64 without AVX...

EDIT: 30 min as above no AVX v.26.6 9c difference @ 4.6ghz:

 

[Don Karnage]

Crap...looks like 4.6ghz may be where I'm at.
Prime95 v.27.6 x64 with AVX needs more volts than regular Prime95 and hits 89c (stopped after 30min) on air with a Prolimatech Armageddon and 2 140mm fans.

Might go back to Prime95 v.26.6 x64 without AVX...

89C isn't that bad. 😉 Start worrying at 99C 😀

 

[JAG87]

I'm at 4.3ghz at stock voltage (which is around 1.080v under load), 8+ hours Prime95 and Linx stable. Anything higher requires more than stock voltage. I just don't see the value in the last 300-400 mhz when I have to jolt an extra 0.2V into it.

 

[LBXAC20]

I think the answer is on pp 8-10 here. Intel essentially ran into problems with lead-free solder, found their indium solder deteriorating with the stress of heat cycling.

"Typical failure modes encountered in STIM joints relate
to thermal fatigue cracking of Indium close to the
IHS/Indium interface . . ."

"Figure 20: White CSAM image showing delamination
at the Indium/IHS interface and corresponding crosssection SEM image showing cracking post reliability
testing."

So it appears even mighty Intel runs up against the limitations imposed by Ma Nature. Using traditional TIM rather than indium solder TIM allows their chips to be more reliable rather than passing more heat through. At stock Voltages, the traditional TIM is Good Enough.

What we really need is an Enthusiast Line of i7 3770K that explicitly states the chips won't live as long, but they have the old-style solder TIM and will run cooler with a good heatsink or WC.

Man dude...very interesting reading to say the least :thumbsup:

If Intel's decision was indeed motivated by the above, chances are TIM is here to stay... perhaps with Haswell too