Max safe 3770K voltage & temp for running 24/7 at 100% load?

[LMF5000]

What is the maximum "safe" voltage and temperature for an i7-3770K that runs 24/7 at 100% load?

I know "safe" means different things to different people, but in this case I'll be a bit conservative and ask for values that give a good lifetime (say 5 years).

I've heard all sorts of values from different sources. Intel specs give an absolute maximum of about 1.5V with air cooling, and a lot of reviewers are afraid to go over 1.35. Similarly the CPU throttles at 105C but many forum members tell you to keep it below 90C.

Can someone clear this up please?

My 3770K currently runs at 4.22GHz @ 1.05V at 70 deg C with a CoolerMaster 212 Evo, but I'm considering whether I can overclock further.

P.S. In case you're wondering about the load, this CPU is used for crunching with BOINC full time.

And if you can recommend a maximum safe GPU temperature, that'd be great. My stock gigabyte GTX670 runs at 66C at full load.

 

[ShintaiDK]

Under 105C for temps.

Where do you see Intel says 1.5V for aircooling on Intel specs?

GPU wise its under 96C if I recall right.

 

[coffeejunkee]

Where do you see Intel says 1.5V for aircooling on Intel specs?

According to the datasheet the maximum value for the cpu's VID is 1.52V. I don't really understand why this value is so high, I'm pretty sure it's impossible to find an Ivy Bridge cpu that actually uses this.

Personally I think a voltage like that will cause degradation quickly, but on the other hand if Intel specifies it, who am I to argue?

But I don't think anyone besides Intel can actually answer TS question. All we have is a TJmax of 105c but noone knows what the actual life expectancy is when running at this temperature.

 

[dma0991]

My personal limit would be no more than 1.3V on a 24/7 basis, especially when there are temp variations from night to day. It wouldn't be a problem running F@H @ 4.5GHz all night as it is cool but during the day, an additional 5C is all it takes for a stable system to be an unstable one. Crashed F@H client ain't fun when you have to redo what could be hours of work. Keep it more stable, not more performance.

 

[pantsaregood]

I wouldn't bother worrying about degradation on Ivy Bridge. You'll be thermally limited long before you're pushing enough voltage to degrade significantly faster.

1.3v, as stated before, is probably around the point degradation begins. Most Ivy Bridge CPUs can probably hit in the 4.7-5.0 GHz range at that voltage, but a 212+ probably won't be enough.

 

[LMF5000]

Under 105C for temps.

Where do you see Intel says 1.5V for aircooling on Intel specs?

GPU wise its under 96C if I recall right.

No, no, I don't mean the temperature limits beyond which the CPU will refuse to operate. I'm looking more for the temperature and voltage that it'll be comfortable running at without degradation over time (or at least without degrading much faster than it does at stock settings).

 

[Meghan54]

My personal limit would be no more than 1.3V on a 24/7 basis, especially when there are temp variations from night to day. It wouldn't be a problem running F@H @ 4.5GHz all night as it is cool but during the day, an additional 5C is all it takes for a stable system to be an unstable one.

If just a 5C temp. rise is all it takes to make a "stable" system unstable, it really never was a stable system in the first place.

 

[coffeejunkee]

No, no, I don't mean the temperature limits beyond which the CPU will refuse to operate. I'm looking more for the temperature and voltage that it'll be comfortable running at without degradation over time (or at least without degrading much faster than it does at stock settings).

What if I tell you running at 104c 24/7 doesn't cause much faster degradation? Will it make you feel comfortable?

Anyway, what pantsaregood said.

 

[dma0991]

If just a 5C temp. rise is all it takes to make a "stable" system unstable, it really never was a stable system in the first place.

It is stable for everything else that I do and I'm aiming for the lowest voltage that I could do for the given speed so it isn't exactly rock solid stability that you would expect at stock settings. Sure, I could bump the voltage >1.3V to give it more stability but it is a tradeoff that I'm willing to make if it means that my chip is able to survive longer than my expected usage of 5 years. 5C is quite a conservative figure that I'm giving. It could be anywhere from 5-10C difference actually.

 

[LMF5000]

What if I tell you running at 104c 24/7 doesn't cause much faster degradation? Will it make you feel comfortable?

Anyway, what pantsaregood said.

Well, yes actually :biggrin:. But is it really true?

I used to run BOINC on a laptop. The poor CPU spent two years constantly at 92-95 deg. C. Eventually the plastic chassis cracked at the place where the holes for the heatsink exhaust are, so I stopped running BOINC. Then a year later I bought my current desktop. Laptop still runs fine though.

 

[LMF5000]

If just a 5C temp. rise is all it takes to make a "stable" system unstable, it really never was a stable system in the first place.

That's a good point. I'm always baffled by people on the BOINC forums who say that they don't run BOINC during the day (or in some cases during the whole summer!) because it's too hot. Surely their output could be much improved by using a less aggressive overclock, or a better cooling system, and then being able to run BOINC at all times regardless of the weather...

Fortunately (or unfortunately), in Malta the room temperature in my room is 30-31*C all day and night during the summer. Or at least that's what my el-cheapo eBay alarm clock with built-in thermometer has been saying for the last three months.

 

[coffeejunkee]

Well, yes actually :biggrin:. But is it really true?

Wish I knew

I understand your question, but it's easy to find someone who will say 'no more than 60!' and equally easy to find someone who will say '80 degrees is no problem'. Also, it's often said lifespan of ic's halves with every 10c increase in temperatures, but what is the lifespan of a cpu? I'm pretty sure some 20 year old chips still work fine...

So I can add my own totally arbitrary limits which would be about 80c max for 24/7. For voltage I use a simple rule too: 0.05 less for every dieshrink. Starting from 65nm: 1.45, 45nm: 1.4, 32nm: 1.35 and 22nm: 1.30.

 

[n0x1ous]

Your not going to find Exact answers from Intel or anything, but I think the concensus would point to 1.3V or Less and 80 degrees C as your caps for a good long lifetime.

The EVO probably isnt your best bet in this situation. NH-D14 or H100 would be better.

 

[BonzaiDuck]

An interesting situation -- 24/7 at full loading. As with others here, that's more or less the consensus from "skinny on the street" -- 1.30V. The ongoing thread about "de-lidding" addresses the thermal problem, and may eventually yield some answers about choices of TIMs, their permanence and how to apply the TIM to the die. If it's something you won't want to do, then you will live with the temperatures and accept the OC limitations of the processor.

 

[LMF5000]

I moved the CM212 fan from "push" to "pull" and it seemed to reduce temperature by 1-2C, so I tried to go for 4.5GHz. So far it survives intel burn test. Will check Prime95 soon. Here's the data:

Room temperature 30C
Base clock: 100.5MHz (thanks to Asus "auto" profile. It cuts two minutes off 40-minute GPU projects in BOINC compared to 100.0MHz so I left it at 100.5)
Multiplier: 45 (gives 4522.5 MHz)
Core voltage: Set to 1.155V in motherboard (manual mode, not offset mode). Gives 1.144V in CPU-Z under load.
Maximum temperature with intel burn test: 89C hottest core
Avg. temperature: 80C hottest core
Constant 24/7 temperature with BOINC: 70C

The temperature is acting weird. The instant I apply a heavy load (like starting burn test or BONIC or Prime95), the temperature jumps to a certain high value, like 80C. Then, two seconds later, it decreases to a lower value, like 70C, and settles there. Is anyone noticing this during the first few seconds of high load? Can anyone explain it?

 

[EpicSurvivor]

I hope I'm not to out of place posting this here but I didn't want to start a new thread.... I got a simple question and its somewhat in topic. I'm just learning about computers and I don't know much so please be nice to me. I have the i7 3770k (Specs Below) none OC at the moment. My idle temps is 32c but while gaming for example on Battlefield 3, I recall my temps going up to 45c. So I was wondering if this is good base temperature. What is the average temp for the 3770K idle and under load?

PS.
What software could I use to view my CPU (NOT GPU) Temps while in game?

 

[LMF5000]

I hope I'm not to out of place posting this here but I didn't want to start a new thread.... I got a simple question and its somewhat in topic. I'm just learning about computers and I don't know much so please be nice to me. I have the i7 3770k (Specs Below) none OC at the moment. My idle temps is 32c but while gaming for example on Battlefield 3, I recall my temps going up to 45c. So I was wondering if this is good base temperature. What is the average temp for the 3770K idle and under load?

PS.
What software could I use to view my CPU (NOT GPU) Temps while in game?

To measure CPU temperature I think the best free program is RealTemp - it lets you put an icon in the system tray to read the temperature live at all times. It also supports logging data to a text file, and tells you the highest and lowest temperatures reached as long as it was running.

45c is a bit low, but we don't know what cooler you have. You need to determine maximum temperature using a program that loads all cores to 100% - IntelBurnTest is a good (free) program that does this. Games don't load the cpu to 100% in most cases.

 

[BonzaiDuck]

Wish I knew

I understand your question, but it's easy to find someone who will say 'no more than 60!' and equally easy to find someone who will say '80 degrees is no problem'. Also, it's often said lifespan of ic's halves with every 10c increase in temperatures, but what is the lifespan of a cpu? I'm pretty sure some 20 year old chips still work fine...

So I can add my own totally arbitrary limits which would be about 80c max for 24/7. For voltage I use a simple rule too: 0.05 less for every dieshrink. Starting from 65nm: 1.45, 45nm: 1.4, 32nm: 1.35 and 22nm: 1.30.

I must have missed your summary here before I posted my own "skinny on the street." That's a good summary, which I either followed myself through the processor generations or would recommend.

Somewhere around the 65nm to 45nm transition, I think we lost any assurances that Intel was previously providing, so our recommendations get close to the demarcation between fact and folklore. But there is just too much evidence (or lack of people reporting failure) to suggest that the folklore is out of bounds.

Also, it has been pointed out that the early-release sample-sizes of Intel's own testing were small, and the sample size grew over a CPU's marketing life-cycle -- giving more optimistic results for higher voltages. At least, so one of our senior and well-regarded members here told me about the Kentsfield processor.

I still have questions as to how "momentary" or brief increases in voltage affect the degradation issue under a situation where EIST is enabled and the over-clocking is the "Turbo," "Offset" or "Turbo-voltage" variety. You would expect the effects to be lessened in comparison to "fixed VCORE" approaches which leave your processor at an idle maximum.

 

[EpicSurvivor]

45c is a bit low, but we don't know what cooler you have.

Stock cooler, but my case has 5 fans well 6 if you count the CPU Stock fan.

 

[Sunny101]

1st u need to know the basics of over clocking
Only ramping up the core speed won't do much because it acts as a bottleneck
U need to increase the corevolt aswell
Ok soo u haven't really mentioned what motherboard ur using,that is important because cooling and efficiency of ur chipset is a lot dependent on the motherboard
2nd u cooling tower isn't gonna help in the long run, u need to upgrade that immediately cuz 212 evo isn't that good for the extreme or high over clocking
That is the reason ur temp are high.....at ideal it should be around the 50s and load about 70s if u want it to last long
Also ur core voltage and speed ratio is totally wrong
It should be 4.2ghz with about 1.235v that should do it...if u wanna increase the sipped more then u have to increase the volt aswell and test it until it chashes...I would recomend nothing above 1.350v
But there is a catch...u see ur processor is a ivy bridge that get hotter than the older ones when over locked....soo liquid cooling is a must....probably corsair h100i or similar long tower radiater
And also u asked about gpu temp...max for nvidia is 80 and for amd it's about 110
That is because of the different architecture of the products
I myself am a hardware tester and also to beta testing for amd and asus soo yeah I know...
I guess u should be good wid this info....ask me for further dilemma

 

[Bubbleawsome]

Yeah, you're totally wrong and replied to a 2 year old thread. Smooth.