i5-3570k overclock temps

[GamingDaemon]

Good stuff. I am planning on getting one of these. I am not a huge overclocker. I have a CM 212+, and would like to be able to overclock to about 4.0GHz with the 3570. Is that do-able? Will I need to up the voltage? What temps, roughly, should I expect?

 

[Borealis7]

Here's to hoping Haswell overclocks without expelling more heat per square mm than the sun!

 

[darkfalz]

Put on an event popup for WHEA errors. This will find overclocking instability long before relying on system crashes.

 

[darkfalz]

Good stuff. I am planning on getting one of these. I am not a huge overclocker. I have a CM 212+, and would like to be able to overclock to about 4.0GHz with the 3570. Is that do-able? Will I need to up the voltage? What temps, roughly, should I expect?

I do 4 on stock cooler an undervolted (-0.2v). It should be pretty easy. Prime hits around 70 after extended period.

 

[Tigashark]

"How high is too high?" really comes down to "How long do you want your CPU to last?"
Rule of thumb is every 10C you can lower your CPU temp will increase the lifespan of the CPU by 1 year.

TJMax on IVB is 105c, TJMax is the point where the CPU will be permanently/instantly damaged or degraded and CPU throttling should begin at approx 10c lower than that (95c) to ensure the CPU never ever reaches 105c (assuming the thermal protection on your CPU is in good shape)

However all silicon degrades slowly over time, and this occurs faster at higher temps, even when those temps are below "TJMax"..basically, cooler=longer life

For my CPUs..I typically buy higher end parts and expect 3+ years useful life out of them.

Personally id be concerned at anything over a maximum of 75c for 24/7 operation IF you intend to be running the part for 3-5 years at the same overclock... higher is OK for short term, but for 24/7 operation , constantly hitting within 10c of TJmax will shorten the lifespan of your CPU by a significant margin.

The main thing with Ivy is a small bump in VOLTAGE can have a fairly large impact on temps and subsequent lifespan, ... for my CPU 4.5 @ 1.235 volts is doable with max temps ~65c, but 4.9 was barely stable even at 1.45v with max temps ~88c, so I dialed it back to keep temps sane...



My IVB system temps :-

Ambient temps ~23c
Idle ~23-25c

I7 3770k (OC 4.5 ghz, 1.240 volts) ~ 67c running prime 95 blend with 8 threads over 10+ hours, I see around~62c in AIDA64 stability test

Gaming I rarely see over 55c even with the extra GPU heat being dumped into the water loop, the 3770k just isn't being pushed that hard in gaming and most of the heat in the situation is coming from the GPU.

Cooling is a custom water loop based on the Swiftech H20-220/Apogee GTz kit but with a few additions and upgrades over the years, CPU/GPU/Chipset blocks are in series.

1/2" ID tubing
Current CPU block = Swiftech Apogee HD
Pump is a Laing D5//MCP655
Radiator is Swiftech MCR-220 2x120mm, externally mounted

"watercooling" should always be qualified with part number/brand/type as some of the low end closed loops are worse than mid range air cooling.

 

[GamingDaemon]

I do 4 on stock cooler an undervolted (-0.2v). It should be pretty easy. Prime hits around 70 after extended period.

Wow, awesome. That would be perfect for me. Thanks.

 

[Idontcare]

Wecome to the forums Tigashark

Rule of thumb is every 10C you can lower your CPU temp will increase the lifespan of the CPU by 1 year.

The rule of thumb is actually that every 10C you lower your CPU temp the expected lifespan doubles.

It comes from the Arrhenius equation and the fact that the dominant degradation mechanisms in IC's are kinetically limited - thermodynamically favorable at all operating temperatures but are thermally limited by an activation barrier.

Voltage increases the rate of degradation because it lowers the requisite amount of thermal energy needed to surmount the activation barrier (the electric field that arises from the applied voltage contributes energy to the activated complex, enabling degradation with lower thermal energy).

But, all else held constant, at the same voltage and so on, if you reduce your operating temperature 10°C then the kinetic rates of reaction are reduced by 50% and as such the rate of degradation slows down by a factor of ~0.5x. Your chip takes twice as long to accrue the same amount of degradation at the lower temperatures versus the higher temperature.

Conversely, heat up your chip by 10°C and the rate of degradation doubles if you keep voltage the same. Increase voltage, to counteract the thermal noise that comes with more blackbody radiation, and your are increasing the rate of degradation even more so.

 

[Tigashark]

Wecome to the forums Tigashark



The rule of thumb is actually that every 10C you lower your CPU temp the expected lifespan doubles.

It comes from the Arrhenius equation and the fact that the dominant degradation mechanisms in IC's are kinetically limited - thermodynamically favorable at all operating temperatures but are thermally limited by an activation barrier.

Voltage increases the rate of degradation because it lowers the requisite amount of thermal energy needed to surmount the activation barrier (the electric field that arises from the applied voltage contributes energy to the activated complex, enabling degradation with lower thermal energy).

But, all else held constant, at the same voltage and so on, if you reduce your operating temperature 10°C then the kinetic rates of reaction are reduced by 50% and as such the rate of degradation slows down by a factor of ~0.5x. Your chip takes twice as long to accrue the same amount of degradation at the lower temperatures versus the higher temperature.

Conversely, heat up your chip by 10°C and the rate of degradation doubles if you keep voltage the same. Increase voltage, to counteract the thermal noise that comes with more blackbody radiation, and your are increasing the rate of degradation even more so.

TLDR: Don't Assume TJMax (or even close to it) is a "Safe" temperature for 24/7 operation, its not.

The lesson here is simple...more volts and higher temps = shorter lifespan.

Long version:-

Good info... and this is also the reason why there have been quite a few reports of people degrading their IVB within 3-6 months when they are hitting 95c and pushing north of 1.35v for attempts at 24/7 OC..

If your going to do that, buy the Intel Performance Tuning Protection Plan

Some examples Ive experienced 1st hand
My C2Q9550 is still able to hit ~4.2 after ~3 1/2 years of 24/7/365 3.84ghz @1.275v overclock.... at worst its max overclock is only 100mhz lower after 3 years (originally it would hit 4.3ghz)...why? because I had it watercooled and dialed the voltage back to "Sane" voltage levels and max temps of ~55-65c.. in fact its still a viable CPU because it retains that overclock ability, its OC performance is still comparable to an I7-920.

My C2D E8400 on the other hand started out being able to hit 3.2ghz @1.35v, but it gradually degraded to the point where it will only do 2.8ghz stable at the same voltage. That CPU needed to be replaced after ~1 1/2 years because I figured 85c-90c was "OK" and still a ways away from TJMax, my CPU never throttled, it never shut down...but it DID degrade faster.


The lesson here is simple... higher temps, more volts = shorter lifespan.

 

[domjam]

hay just saw this thread and ur post. I was wondering if you could explain to me a bit more.when u say ''4.3ghz is stable with no voltage offset..what is the actule offset value set to..and what settings in bios do u have turnd on and of.because for me im using an asus z77 sabertooth and when i changed the turbo clock from auto to 44 and in HWM I was gettin a max value vcore of 1.42v without running any stress testing programs...my temps were well bellow 85c though but regardles of that i manule changed the turbo from to 38 and my cpu offset value from auto to 0.010 and now HWM is givin the max value of 1.2v but shouldent my volt be lower then this? im a bit lost and could do with some guidence on this.thanks

 

[lu.nemec]

Hey guys, I've recently bought this CPU, and now I changed stock cooler for
Cooler Master Vortex 211Q

the stock was too loud, and this works great when idle and gaming, but when I run intel burn test (stock freqs) , it goes often as high as 107 C, I can't have bigger cooler, because my case is super small (I tried opening it up so there is more room for air to go to cpu cooler but no change) what would you recommend to lower these temps, or is that normal with intel burn since it is a 'burn' test anyway?


Thx

 

[cmdrdredd]

I guarantee bit isn't 107c. Isn't the shutoff temp 105?

What are you using to monitor temps? Try real temp. It will read the core temps and report it in real time. With IBT (Intel burn test) I get into the 80s and as high as 85 with a 4.5ghz 3570k and a noctua d14

 

[lu.nemec]

Yep well, I'm using utility from MSI clickbiosII that reads tems from bios directly, but I'll try real temp and post results here

105 should be throttling temp, I think shutoff is about 110 or more..

I'll post it in 4 hours, I'm still at work

 

[blackened23]

What software are you guys using for the temps? I noticed that RealTemp and the software by the mobo (Asus in my case) varies way too big for the load temps... I'm not quite sure how Asus computes for its CPU temp (since it is only showing one entry and not the per core basis).

I was only getting 65C to 70C for the Asus software (AI Probe) while I am already hitting 95C (highest core temp) with RealTemp though on the average it hovers 85C to 90C.

That by the way is just on stock speed, almost "auto" everything, and on the stock Intel HSF. Stress tested it via IBT (standard test only).

I just hooked up my NH-U12P SE2 yesterday and set RAM to XMP profile (I only have one option for my G Skill Ares 4x2 1600). Unfortunately, I still have to find time to benchmark my rig...

Real temp is accurate. Asus software probe from AI suite is NOT accurate. It under-reports temps by quite a bit at times.

 

[lu.nemec]

I guarantee bit isn't 107c. Isn't the shutoff temp 105?

What are you using to monitor temps? Try real temp. It will read the core temps and report it in real time. With IBT (Intel burn test) I get into the 80s and as high as 85 with a 4.5ghz 3570k and a noctua d14

Well, I just tested it on standart intel burn test and 2 cores got 105C, that is TJ max, si throttling is in effect, but it is extremely strange, on idle I have about 45-50C, but when burning it jumps by 50 C!! what the hell is that? and on normal clock that is 3.8GHz

EDIT:
but quite nice is that on whatever load on intel burn it holds on cca 100C and nothing happens.. no reboot, nothing and that is not reason to return the cpu ... so I guess there is nothing I can do about it, maybe change the thermal past for that liquid metal I used a while ago, that could help quite a bit, also there is a possibility I have too much thermal-conducting shit between cpu and cooler .. smart people tell me whyyyyyy

 

[coffeejunkee]

Hey guys, I've recently bought this CPU, and now I changed stock cooler for
Cooler Master Vortex 211Q

the stock was too loud, and this works great when idle and gaming, but when I run intel burn test (stock freqs) , it goes often as high as 107 C, I can't have bigger cooler, because my case is super small (I tried opening it up so there is more room for air to go to cpu cooler but no change) what would you recommend to lower these temps, or is that normal with intel burn since it is a 'burn' test anyway?


Thx

You probably didn't install the cooler correctly. Check to see if all push-pins are inserted properly.

The stock cooler does something like 80 to 90 on Linx, yours should do quite a bit better. 50 idle is also way too high, should be more between 20-30.

OT: I like to stay below 95 during stresstesting. During regular use you won't come near this anyway.

 

[lu.nemec]

You probably didn't install the cooler correctly. Check to see if all push-pins are inserted properly.

The stock cooler does something like 80 to 90 on Linx, yours should do quite a bit better. 50 idle is also way too high, should be more between 20-30.

OT: I like to stay below 95 during stresstesting. During regular use you won't come near this anyway.

Well this isn't my first cpu cooler, I did study 4 years of IT, so HW and SW is not new to me, however I think I might have put on too much of the past, I ordered collaboratory liquid pro, and will try it tomorrow, so lets hope it will go better, if not, well I might try to return this piece, because this is first time any cpu does this strange stuff.. and I had about hundred of cpu coolers changed since my career, that is why I'm writing it here, because it never happened to me and I did this quite a lot, and only on my pc it will go wrong ...

 

[coffeejunkee]

A bit too much paste won't give you 100+ temps. A loose pushpin will.

 

[Idontcare]

Well, I just tested it on standart intel burn test and 2 cores got 105C, that is TJ max, si throttling is in effect, but it is extremely strange, on idle I have about 45-50C, but when burning it jumps by 50 C!! what the hell is that? and on normal clock that is 3.8GHz

EDIT:
but quite nice is that on whatever load on intel burn it holds on cca 100C and nothing happens.. no reboot, nothing and that is not reason to return the cpu ... so I guess there is nothing I can do about it, maybe change the thermal past for that liquid metal I used a while ago, that could help quite a bit, also there is a possibility I have too much thermal-conducting shit between cpu and cooler .. smart people tell me whyyyyyy

You aren't hitting TJmax because of too much TIM. Some jackass around here put this much on their 3770k and it only raised temperatures by 2-3C at mild OC's.

D:

Even with the stock HSF you won't hit TJmax when running the Intel burntest at stock clocks.

Something is wrong with your HSF mount. Pushpins are notorious for causing this kind of issue, textbook notorious.

 

[lu.nemec]

Yeah, I didn't put that much of the paste, just small blob 4mm in diameter and 3mm high in center of the cooler and then pushed it onto the cpu, but I checked all the pins so they'd be really clicked in, as I said, it is not my first, but I'll get that collaboratory liquid pro today, so I'll check the how the past was spread on the cpu and I hope will learn why is it going so high .. the funniest thing is that the temp jumps 40C in about 1.5s and that is really strange...

I'll let you know once I change the liquid

P.S. that picture made my day

 

[casophylip]

Great thread. Thanks for all the info cmdrdredd!

 

[john3850]

Right now my temps are 85-90c in ibt at 4600 with 1.256v with a room temp of 65f.
I am runnig W/C on the cpu only.
Now when the summer comes that room temp will be 75f which will be way to hot.
I need raise or disable the LLC next to get ready for the added summer heat.

 

[lu.nemec]

Well, I just changed the paste (it is the same because the collaboratory liquid pro corodes aluminium) and now I'm absolutely certain the past and the cooler are correctly, I didn't pimp with cooler this much ever.

And temps didn't change, on idle it's like 36C, but when I burn it, it rises to 80C in about 10s and after while it goes to 105C again .. so no change

the pins are correctly all of them clicked and it doesn't move at all..

I had a suspicion that it could be because PS is right above it and the heat would just stack up, but now PS is lying next to the PC and nothing changed, and worst is that the heatsink is not that much warm, I noticed this before on the stock cooler, it is just slightly warm, and really it can't have 100C or even 80C, because I can with no problems hold my hand on it and it just doesn't burn, as it should ...

that is really weird..

EDIT: maybe the sensors are wrong, I'll check BIOS update if it'll help

EDIT: well isn't it just awesome... the BIOS update worked! but as a side effect windows will not boot up and end with BSOD!!! temps didn't change, still the same, I think wrong piece of cpu, but there is nothing I can do since it is stable without OC, and warranty does not apply to that. So I'll leave it as it is, it is working and performance is good, it does not restart and it is quiet..

Thx guys for your help

 

[Vectronic]

For whatever use this may be to anyone:

Turbo V: auto
PLL Overvolt: disabled
LLC: 1 (for offset), and auto for fixed (which shows as 1)... on this board 1= 100% (supposedly, lol)
RAM: 1.450v (from 1.5, otherwise stock clock/timings 1600/9/9/9/24)
PLL voltage: 1.603 (from 1.832, does have an effect but not much so I undervolt)
C1E enabled, C3/6/States disabled
IGPU stuff all disabled
Otherwise defaults (for this board)

22C ambient... ignore the "watts" (was going to, didn't bother including) although load voltage does almost equate to Package watts.

CPU & Exhaust (140) fans are set to 25% (62 of 255 on this board, lowest they'll go and still spin up from "off") with target temp 45C, takes about 3 minutes for this board to spin them up to full/100%... intakes (2x140 front, 1x120 bottom) are set to a constant 5V... so temps could be much lower (-6 at idle, -2 at load or something, not really enough to make up for the noise)

My 24/7 config is 4.2GHz with 0.01v less than seen here, VTT & PCH are slightly undervolted, and RAM is overclocked to 1800 otherwise same settings as here.

Regarding TIM, I've done it 3 times now with this CPU/Heatsink, right now I'm using about this much:

(like, a cylinder of that width/length at the scale of the image not actual scale - "rice grain")

First time I used about 3 times that much, second about 2.5 times... can't really say between those 3 amounts it's made much of an effect.

 

[cmdrdredd]

I still think anything close to 100c is crazy temps. That said, when I ran 1.32v for 4.6ghz I was up in the 90s after prolonged IBT running. So...I dunno.


Could be your case airflow somewhat as well

 

[amenx]

I have a feeling with some CPUs there may be inconsistencies under the heat spreader where the distance between the HS and chip as well as spread of TIM is not uniform. Hence you see wide variations in temps between chips, even using same coolers (ie, stock cooler). Seen one fellow with a 3570k OC'd to 4.2 on stock cooler and with no temp troubles at all (80-90c P95 I believe) while a few cant do that with much better coolers. In such cases, a delidding may be in order. I'm about to pull the trigger on a 3570k and hope I dont have to do that.