4.7 ghz vs 4.8 ghz

[tigersty1e]

So I can hit 4.7 with 1.327 vcore and hit gaming temps of 62c, or 4.8 with 1.389 vcore and gaming temps of 66c.

I like the big e-peen feeling of 4.8. Will I noticeable extend the life of my cpu or noticeable avoid degradation if I run 4.7 vs 4.8?

 

[alcoholbob]

4.7 to 4.8 ghz is probably not even 1 fps in 99.99% of games.

 

[pandemonium]

The biggest factor for degrading CPU life expectancy is the V-core and temperature. They're conservatively set by Intel, V-core found here (Section 7.10.1, page 84) and thermal limits found here (Section 6-2, page 43 & 44).

In short, Intel has defined (for the 3570k) the maximum V-core to be 1.52 and the maximum temperature to be 72.6ºC @95W (over TDP). See IDC's reponse.

The biggest consideration is going to be your individual results from the extra .1GHz. Run benchmarks before and after in your games of choice and make that decision for yourself. Each CPU and system is slightly varied, so the only way to know is to do it yourself. Certain games can be more or less CPU bound than others.

>>This article on IEEE is pretty cool to read if you're at all interested in the detailed dynamics of transistor degradation.

 

[aaksheytalwar]

Take it to 4.5 and call it a day.

 

[2is]

Thats a lot of volts for a 22nm CPU, regardless of temps IMO.

I agree with aakshytalwar

 

[mikeymikec]

I like the big e-peen feeling of 4.8.

I had to google that to find out what it meant, which answers my next question (Why).

 

[Idontcare]

The biggest factor for degrading CPU life expectancy is the V-core and temperature. They're conservatively set by Intel, V-core found here (Section 7.10.1, page 84) and thermal limits found here (Section 6-2, page 43 & 44).

In short, Intel has defined (for the 3570k) the maximum V-core to be 1.52 and the maximum temperature to be 72.6ºC @95W (over TDP).

I think you are confusing the intentions and meanings of those specific specification parameters in the Intel documentation.

Max "safe" Vcore is not 1.52V, the max value in the VID table is 1.52V. Those two parameters are not to be confused as being one in the same.

At 32nm the max safe Vcore was 1.35V for gulftown CPUs. That was the last time Intel published a max allowed (or max safe) operating Vcc. At 22nm we all just take guesses based on the 32nm spec value and assume the 22nm max safe voltage must be less than the 32nm max safe voltage value.

Likewise the max Tcase temperature is not to be confused with the maximum allowed operating junction temperature (the temperature of the silicon itself).

TJmax is 105C for IB. There is a reason the CPU is allowed to get that hot before it throttles. If 72.6C was the limit then Intel would have set TJmax to be 72.6C.

Tcase is specified so that OEMs and system builders don't slap an inadequate HSF onto the CPU, or shove it all into an inadequate air-flow case, and cause the CPU junction temperature to hit 105C such that the CPU is needlessly throttling under normal operating conditions.

That said, neither the safe max operating temperature (TJmax) nor the safe max operating voltage (Vcc_max) mean much once you start overclocking the clockspeed itself as they are both defined with respect to stock clockspeeds.

Raising your voltage while raising your clockspeed and raising your operating temperatures all conspire to rapidly reduce the expected operating lifetime of the CPU.

But only Intel's engineers who characterized the silicon and determined the spec limits based on data would know just how much and how quickly the lifetime is being reduced when we over-volt to say 1.4V and operate at 80C with 4.8GHz OC's. Could be we can run that way for years, or maybe only for a year.

 

[pandemonium]

Thanks for clarifying that, IDC. It was the closest thing I could find from Intel - while not defined as the absolutes - with respects to "safe" guidelines.

 

[BrightCandle]

The safe guidelines are really hard to determine. Unlike Intel we haven't been subjecting this process to incredibly stringent and high quality engineering QA. We don't have any clue where the breaking voltage lies or the long term impact of the voltages we run.

 

[john3850]

For what ever reason the 45k seems to be the best setting for the lowest volts and heat.
I ran 47k for a few months till I found 1000 Whea errors.
To help get rid the Whea errors you raise the vcore which gives you more heat and more possible errors.
Two days back I updated the bios and went back to 47k this morning I found 6 Whea events not bad for 48 hrs uptime.
I will need to max out the llc to stop the errors which will raise the volts to over 1.3v.

 

[tweakboy]

No difference in gaming ....

 

[guskline]

4.4Ghz works for me at auto vcore setting without WHEA errors. Stable no matter what benchmark test I throw at it.

In my case it's so fast I can't outplay the hardware. Haven't had Det. DeWitt in BioShock Infinite scream at me that my 3770k has to run at 4.8Ghz not my "lowly" 4.4Ghz or Comstock will prevail!

 

[Rakehellion]

So I can hit 4.7 with 1.327 vcore and hit gaming temps of 62c, or 4.8 with 1.389 vcore and gaming temps of 66c.

I like the big e-peen feeling of 4.8. Will I noticeable extend the life of my cpu or noticeable avoid degradation if I run 4.7 vs 4.8?

I think someone who is knowledgeable enough about electronics to understand functional limits to keep theirs in working order would have a bigger e-peen than someone who thinks blue LEDs keep his case cool.

 

[Idontcare]

Thanks for clarifying that, IDC. It was the closest thing I could find from Intel - while not defined as the absolutes - with respects to "safe" guidelines.

No problem, the only reason I know anything about it is because of my own frustrations in attempting to divine/find/locate a max safe operating voltage for Intel's 22nm. That journey has led me down many a rabbit-hole, but no concrete data/specs from Intel

AMD is even more difficult, not only can you not get a max safe voltage spec from them, you also can't get a max safe/allowed operating temperature either (for their latest piledriver chips). Double

 

[BTRY B 529th FA BN]

Where does one check for WHEA errors?

 

[Face2Face]

Where does one check for WHEA errors?

Use this. It works well.

http://forums.anandtech.com/showthread.php?t=2312638

So I can hit 4.7 with 1.327 vcore and hit gaming temps of 62c, or 4.8 with 1.389 vcore and gaming temps of 66c.

I like the big e-peen feeling of 4.8. Will I noticeable extend the life of my cpu or noticeable avoid degradation if I run 4.7 vs 4.8?

JJ @ Asus said that anything at or under 1.35v is safe in one of his Ivy overclocking videos (Newegg, I think?).
Where does he get his info from? Hell if I know....

 

[MrDudeMan]

But only Intel's engineers who characterized the silicon and determined the spec limits based on data would know just how much and how quickly the lifetime is being reduced when we over-volt to say 1.4V and operate at 80C with 4.8GHz OC's. Could be we can run that way for years, or maybe only for a year.

Much of my job (I'm an Intel I/O designer) is related to this topic. I'm obviously not going to give you details so I don't get fired, but I'll try to help as much as I can. The degradation imposed by increasing voltage, temperature, and frequency is severe (as in extreme lifespan reduction) depending on the different combinations. There are times when temperature is a much bigger knob than anything else, but there are also times when voltage could be far more damaging. It also depends on how often your CPU is busy. Power gating, adaptive frequency, etc. all provide relief, so there really isn't a magic formula.

 

[guskline]

Take it to 4.5 and call it a day.

I called it a day at 4.4Ghz even though 4.5 was solid. I benchmarked with Intel's own Overclocking utility and 4.5 Ghz passed but the temps were just a bit too high for me. 4.4Ghz worked better. Still a very decent OC.

 

[BrightCandle]

What we could really do with is a max voltage and frequency and temperature that is very likely to give us something like 4 years operation.

 

[Face2Face]

Much of my job (I'm an Intel I/O designer) is related to this topic. I'm obviously not going to give you details so I don't get fired, but I'll try to help as much as I can. The degradation imposed by increasing voltage, temperature, and frequency is severe (as in extreme lifespan reduction) depending on the different combinations. There are times when temperature is a much bigger knob than anything else, but there are also times when voltage could be far more damaging. It also depends on how often your CPU is busy. Power gating, adaptive frequency, etc. all provide relief, so there really isn't a magic formula.

Can you tell us if there is a certain voltage us Overclockers should be staying at or under? Assuming our temps are a bit below TJmax?

 

[Idontcare]

What we could really do with is a max voltage and frequency and temperature that is very likely to give us something like 4 years operation.

Yep, and they already have this data as it is generated in the process of qualifying a fab for production of any given IC fabbed in it. There will be Weibull plots from bake-test data and accelerated voltage testing galore.

Very simple for them to pick the point on the Weibull plot for which temperature/voltage/clockspeed can be expected (its all statistics after all, so the 50% percentile would do us just fine) to survive 4yrs at X% duty cycle (say 100% full load, 50% of the day).

The data exists, no fab gets qualified for production of an IC without the qual data being generated and vetted. But there is huge implied liability if that data became public domain as it can get intentionally contorted into all sorts of class-action stuff if unscrupulous lawyers decided to make a go at it.

You'll never see the weibull plots and qual data from TI for the same reason, likewise for TSMC and so on. In a cost/benefits analysis, the risk vs. reward for making that info public simply doesn't justify making it public, and so it never will be.

 

[MrDudeMan]

Yep, and they already have this data as it is generated in the process of qualifying a fab for production of any given IC fabbed in it. There will be Weibull plots from bake-test data and accelerated voltage testing galore.

Very simple for them to pick the point on the Weibull plot for which temperature/voltage/clockspeed can be expected (its all statistics after all, so the 50% percentile would do us just fine) to survive 4yrs at X% duty cycle (say 100% full load, 50% of the day).

The data exists, no fab gets qualified for production of an IC without the qual data being generated and vetted. But there is huge implied liability if that data became public domain as it can get intentionally contorted into all sorts of class-action stuff if unscrupulous lawyers decided to make a go at it.

You'll never see the weibull plots and qual data from TI for the same reason, likewise for TSMC and so on. In a cost/benefits analysis, the risk vs. reward for making that info public simply doesn't justify making it public, and so it never will be.

There's way more that goes into it other than a plot from a fab and the activity factor. You're forgetting about the design aspect. As a designer, I can choose to use a specific width of wire on a specific routing level to minimize routing congestion while maintaining acceptable degradation at the specified operating parameters. You have no idea which circuits will cease to function first or how long that will be simply by looking at process information and you also can't say you'll be operating in the 50th percentile. It could be far, far worse than that. CPUs can fail to operate correctly after hours of stress testing. Without that kind of data about your particular CPU skew parameters, you're shooting in the dark once you exceed the recommended limits. Unfortunately, I can't give you any information about yield and process beyond what I've already said.

The direction of the current through a node also makes a huge difference. Vias, for example, wear differently when comparing DC vs AC current. Unidirectional nodes have to be designed differently than bidirectional nodes. If you overclock and happen to run the same group of instructions far more than others, it's possible that the vias will open much faster than expected. This is detailed in a few IEEE articles IIRC.

 

[MrDudeMan]

Can you tell us if there is a certain voltage us Overclockers should be staying at or under? Assuming our temps are a bit below TJmax?

Sorry, no. I probably couldn't even if I had the authority to do so. It's a complex relationship that isn't easily defined.

 

[Face2Face]

Sorry, no. I probably couldn't even if I had the authority to do so. It's a complex relationship that isn't easily defined.

No problem, thanks for your reply. Great to have you on here

BTW: I am running my 3570K @ 4.8Ghz with only 1.3v - VDroops down to 1.272. No LLC on this mobo

 

[BrightCandle]

I am kind of hoping there is something useful and simple you can tell us. Ever since Intel stopped giving us safeish voltages we have been totally in the dark and unable to really say anything about safe voltages. Obviously its complicated and confidential but its not like we need the process testing details or the design schematics. We just want to set basic limits like 1.3V so that its unlikely that there will be a problem.

That is how we frame it, how do you as an expert in the process and these concerns distill down to the parameters that we have? We are looking for advice not the word of Intel. We don't need official, we just want some guidance.