Root Overclocking Sweet Spot

[mitchellvii]

In the world of overclocking faster isn't always better. Your CPU may be capable of a high overclock but may actually run faster at a lower overclock.

With the new Snap 8.1 kernel, I am able to run my EVO at 1.22 stable, however, after much benchmarking, I would say it seems to run a bit faster at 1.15. So at this point I would say 1.15 is the sweet spot for this CPU. Opinions?

 

[exBBuser]

^^^^^^ I agree, it all depends on how well ur phone takes the new kernel. I been able to OC @1.26 while still maintaining performance,but with 8.1 once past 1.15 everything goes down the toilet. But still a kick ass Kernel,well put together by Cyanogen team snap have become and addiction to me.

 

[uminchu]

In the world of overclocking faster isn't always better. Your CPU may be capable of a high overclock but may actually run faster at a lower overclock.

With the new Snap 8.1 kernel, I am able to run my EVO at 1.22 stable, however, after much benchmarking, I would say it seems to run a bit faster at 1.15. So at this point I would say 1.15 is the sweet spot for this CPU. Opinions?

Weren't you through with rooting and OCing, back to stock forever?

*whisper* pssst - welcome back to the dark side *whisper*

 

[EarlyMon]

First, I think it's important to understand a few things, so here's some background info.

Overclocking - Wikipedia, the free encyclopedia

Next, know the risks -

Understanding CPU Overclocking

Next - let's look at WHY you can overclock in the first place.

The wikipedia link (given first) explains that it has to do with manufacturing margins in the clock circuit.

Instead of focusing on the clock - the favorite subject of overclockers - I'd like to contribute on the manufacturing margin part.

Here's what most everyone thinks a close-up of a semiconductor looks like:

Stock Photos: Chip

Here's a popular block diagram (ok to just ignore the words) of a mosfet - a kind of transistor that you can consider one of THE fundamental building blocks of a semiconductor (aka computer chip) -

The interface between silicon and a high-k oxide

Now - I've set the link to scroll down on this page, tap the thumbnails to enlarge - here's what semiconductors really look like -

Home Page

Not clean, not razor sharp - lumpy and fuzzy.

And those are the GOOD examples.

Here are little videos - with more close-ups at each spot - of the generic semiconductor manufacturing process -

How Semiconductors Are Made | Renesas Electronics

In that, they constant mention lithography and etching.

For practical purposes of this discussion - semiconductors are made with a high-tech printing process (with some cutting / etching).

If you've seen a lot of Sunday funnies with lots of variation on the colors lining up - you've got the right idea of what's really going on. Especially if you think in terms of printing with metal on a microscopic scale.

System speeds are the designed-for, rated speeds. Overclocking is only possible because often that printing process is even better than needed.

But that amount of _better_ varies insanely from lot to lot, and even from one side of the wafer to the other. (You've looked at the little vids, so you know what a wafer is.)

This is why the kernel devs supporting overclocking won't commit to any one number - it's unpossible to find, specify or characterize.

When your phone ran faster with a slower clock speed, it simply meant that you stopped stressing some devices or class of devices somewhere on the motherboard. And that was without question the kind of stress that lowers your devices' reliability and lifetime.

It's entirely possible that your overclocking experiments - similar I am sure to what so many others are doing - may have no long-term adverse effects.

But being a reliability freak, you couldn't pay me to overclock.

And to repeat the answer to original call for opinion - without a doubt, there's no one right number, but I've seen others advocate 1.15 as safe. How they arrived at that, I don't know, but I have seen it.

 

[mitchellvii]

Weren't you through with rooting and OCing, back to stock forever?

*whisper* pssst - welcome back to the dark side *whisper*

And your point is?

Seriously though I was. However, the vanilla froyo dialer allows me to turn Bluetooth on and off during a call on the fly - sense does not. I NEED this feature. Hence I had to root again despite the possibility I have a bad mini USB on this thing.

 

[mitchellvii]

Regarding the poster who mentioned the dangers of overclocking to your CPU. I have been overclocking CPU's on my PC's for years and NEVER once cooked one. Modern CPU's will pretty much shut themselves off if you attempt to push them too far.

As a matter of fact, I would not be surprised to learn that on a statistical basis, the failure rate of overclocked processors was no greater than the population of all CPU's in general. Just because you overclocked a CPU and it failed does not necessarily mean that the overclocking caused the failure. It might have failed anyway.

 

[uminchu]

And your point is?

Seriously though I was. However, the vanilla froyo dialer allows me to turn Bluetooth on and off during a call on the fly - sense does not. I NEED this feature. Hence I had to root again despite the possibility I have a bad mini USB on this thing.

I was just kidding around with you. I guess I missed all the 'hinky' posts at the CM6 forums, heh. I also grok that this stuff is more an addiction than a hobby, so it wasn't about if you'd be back, it was all about when (I gave it a week).

Anyways, [/HIJACK], and welcome back.

 

[EarlyMon]

Regarding the poster who mentioned the dangers of overclocking to your CPU. I have been overclocking CPU's on my PC's for years and NEVER once cooked one. Modern CPU's will pretty much shut themselves off if you attempt to push them too far.

As a matter of fact, I would not be surprised to learn that on a statistical basis, the failure rate of overclocked processors was no greater than the population of all CPU's in general. Just because you overclocked a CPU and it failed does not necessarily mean that the overclocking caused the failure. It might have failed anyway.

And I'm ok with the idea that you've not cooked one.

As far as statistics of failure - nearly impossible to gather. The hard damage will be likely caused by the overclocking experiments - one article I supplied pointed that out.

And, so we're clear - I was not limiting to CPU failure, the usual suspect of overclockers. I was referring to potential lifetime shortening on any devices in the system resulting from the system clocking faster.

FWIW - I work in semiconductor reliability. I don't pretend to know everything, but it is quite true that overstress leads to reduced lifetime.

Whether or not overclocking leads to overstress is the 64 dollar question.

When you reduce clock and got higher overall response - then that higher clock stressed the system. How much and how severe is a product of how much and how severe you used the system in that state.

Maybe you knocked a week off of a 15-year part. Maybe you knocked a year off a 3-year part. No way to know without x-ray vision.