• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

3570k absolutely not stable at 4.5 GHz???

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

SPL Tech

Member
Joined
Nov 28, 2006
I've had by 3570 for about two years. At first I had it clocked at 4.7, then eventually over the year I've had to back down to 4.6, then 4.5 and reticently 4.4. I have an Evo 212 and the CPU is delided w/ liquid metal TIM so it stays cool enough regardless of how high I crank the clock freq or vcore. Anyway, the computer will randomly lock up at 4.5 GHz. even with a whopping +013.5 vcore offset and CPU voltage line level at 1 (maximum)! Most people seem to get close to 4.8 GHz with that much voltage, but not me. However, when I lower it to 4.4 Ghz. it only needs +0.050 vcore. So a 100 MHz increase seems to need more than +0.065 more vcore and it's still not stable... That doesent make sense as you dont typically need that much vcore for such a small increase. What's stranger is that the frequency of how often it locks up doesent change that much with voltage. I can run 4.5 GHz. at +0.10 or +0.135, regardless it still only locks up about once every ten hours, and it's typically not during game playing but rather during Internet browsing.
 

Tech Tweaker

Member
Joined
Dec 14, 2010
This sounds sort of like my 4690K (minus the random lock ups). I can run 4.2GHz at 1.18v, but for 4.3GHz I need 1.25v (a +0.07v increase).

It is sort of common in my experience to need a massive voltage increase if you are near the maximum of what your CPU can handle speed-wise, for even a small 100MHz increase.

It sounds like your CPU's overclocking ability is degrading if you are continually having to decrease your overclock over time, including the recent decrease to 4.4GHz.
 
Last edited:

Tír na nÓg

Member
Joined
Sep 19, 2015
^this.

My 3770k was running [email protected], but needed 1.43v for 4.9.

The 5820k in sig needs 1.275v for 4.7GHz, 1.32 for 4.8GHz and more than 1.4v for 4.9GHz.

That is just the vcore wall.

Sometimes higher, sometimes lower. In you case, quite low it seems.

What was the vCore you were using when running 4.7GHz?
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
Too earl;y for math in the morning.. just give people the voltage instead of offsets and forcing simple addition. :p

That said, those CPUs are good to 1.4V+ so... if I can do math, 1.23+ .13V = 1.36V. Try running straight up manual voltage instead of reindeer games with the offset. If you want power savings, leave it on in Windows and it will throttle things back. But being delidded can be another problem showing up as well...
 
Last edited:

Silver_Pharaoh

Likes the big ones n00b Member
Joined
Sep 7, 2013
Lucky. Mine won't even do 4.2Ghz.

The fact that you had to back off the OC a few times leads me to think it was either never stable above 4.4Ghz or it has somehow degraded. Perhaps you pushed the voltage to high in the past?
 
OP
S

SPL Tech

Member
Joined
Nov 28, 2006
Lucky. Mine won't even do 4.2Ghz.

The fact that you had to back off the OC a few times leads me to think it was either never stable above 4.4Ghz or it has somehow degraded. Perhaps you pushed the voltage to high in the past?
I had manual voltage at 4.6, and I was at about 1.33 if I recall right. Now I can run 1.35 and I am not stable at 4.5 let alone 4.6... I've ran into this issue a million times with GPUs and other CPUs over the years. The computer passes a stress test for several hours, but then it doesent work in one particular game or app at some point in the future. I've ran into this with my GTX 970. I am running +100 and +450 RAM, and the card is stable in every game I have except Planetside 2, even though PS2 puts the lowest load on the GPU of any of my games. In PS2 my driver crashes at those OC clocks, even though no other game crashes and the card and I can run heaven benchmark on loops for two hours no problem. This leads me to believe that stress tests dont absolutely guarantee stability in every app. Over time, it's possible that one particular app might still cause the computer to crash at a particular OC clock, regardless if your computer can pass a stress test. I say this as I've seen this with four different video cards (8800, 295, 660, 970) and two CPUs (e6600, 3570k) over the years. The chips always pass a stress test when I first install them, but at some point, possibly weeks or months later, problems start showing up. Driver crashes, random lockups, BSODs, ect, and the only way to fix it is to drop the OC. This has always been my experience and I've been overclocking for almost 8 years. I've even seen cases where I can run Prime95 for two hours no problem, but less than 15 minutes into gaming and the computer freezes until I drop the CPU OC.
 
Last edited:

Silver_Pharaoh

Likes the big ones n00b Member
Joined
Sep 7, 2013
I had manual voltage at 4.6, and I was at about 1.33 if I recall right. Now I can run 1.35 and I am not stable at 4.5 let alone 4.6... I've ran into this issue a million times with GPUs and other CPUs over the years. The computer passes a stress test for several hours, but then it doesent work in one particular game or app at some point in the future. I've ran into this with my GTX 970. I am running +100 and +450 RAM, and the card is stable in every game I have except Planetside 2, even though PS2 puts the lowest load on the GPU of any of my games. In PS2 my driver crashes at those OC clocks, even though no other game crashes and the card and I can run heaven benchmark on loops for two hours no problem. This leads me to believe that stress tests dont absolutely guarantee stability in every app. Over time, it's possible that one particular app might still cause the computer to crash at a particular OC clock, regardless if your computer can pass a stress test. I say this as I've seen this with four different video cards (8800, 295, 660, 970) and two CPUs (e6600, 3570k) over the years. The chips always pass a stress test when I first install them, but at some point, possibly weeks or months later, problems start showing up. Driver crashes, random lockups, BSODs, ect, and the only way to fix it is to drop the OC. This has always been my experience and I've been overclocking for almost 8 years. I've even seen cases where I can run Prime95 for two hours no problem, but less than 15 minutes into gaming and the computer freezes until I drop the CPU OC.

Okay so you didn't jack up the voltage that's good.

I understand what you experience, in fact, it just happened to my Phenom II yesterday. I woke up to a BSOD. I knew core #5 was very weak and can be unstable and prime95 proved that so I disabled it and ran Prime95 again for 9 hours. Core #4 failed and I was getting WHEA errors. I dropped the FSB to 229 from 243 and now I'm Prime95 stable 9 hours.

Basically I always run P95 overnight once I think I found a good OC. So I guess in our cases we both thought our overclocks are stable, and for 99% of the time they are it's just that one program that cripples us, but in reality they weren't.



Now, maybe try upping some other voltages too. IIRC PLL overvoltage helps with stability as well as the system agent voltage. VCCM I think? It's been a while since I overclocked Sandy bridge... Perhaps the RAM is a bit too tight and it's a bit unstable too I always set the RAM voltage to 1.65 or 1.72v regardless of timings. I back the voltage down after I know it's stable.

Other option is to start OCing from scratch again but run Prime95 longer than 2 hours.
 
OP
S

SPL Tech

Member
Joined
Nov 28, 2006
With the exception of the vcore, I have all voltages set to auto. Accordingly, shouldent the mobo increase the voltages as needed? One thing I always thought was interesting is auto works on vcore up to 4.5 GHz. After that, auto wont jack the voltage up enough to make it stable, which is a bit weird. It really would be nice if all overclocking entailed was setting the clock speed and all the voltage stuff was handled by the mobo. That way you could get a guaranteed stable clock without any testing.
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
Stress tests NEVER guaranteed stability... it's the best we have though.

PLL voltage likely won't help. It's PLL override voltage (enable) that can help but not typically that low of clocks. System Agent voltage is for the memory.
 

Silver_Pharaoh

Likes the big ones n00b Member
Joined
Sep 7, 2013
Stress tests NEVER guaranteed stability... it's the best we have though.

PLL voltage likely won't help. It's PLL override voltage (enable) that can help but not typically that low of clocks. System Agent voltage is for the memory.

Which is why i increae the system agent voltage. Just incase the RAM becimes unstable while i tweak the OC. Of course, i messed with the BCLK on every Ivy bridge and Sandy Bridge chip i owned so thats why i set it a bit higher to start...
 

Soulcatcher668

Member
Joined
Jan 2, 2012
I have had to drop my 2600K as well. I figured that was just a natural progression over time.
I have one of the rare chips that refused 4.5 gigs at 1.35 volts. I originally needed 1.38 for a stable 4.5. :(
Now I run 4.4 at 1.38. I still feel it has been a good run for this rig. Almost five years.