Overclocking is for those who hate stability - period.

[dansicotte]

Constantly trying to get another 16, 32, 48, 64, 120, or 300 MHz out of a chip seems to be fruitless to me. Expecially if you understand how manufacturers like Intel, AMD, and lots of other semiconductor manufacturers do "speed binning", what it means, and why it's done.

They usually produce one part (CPU) and then they test the part in several ways, hot, cold, and high and low voltage thresholds, and then based on all the test results - they label the part for speed.

When they label a CPU 1.4GHz, what they are saying is, "this particular part (CPU) passes 85% of our tests at 1.6GHZ, 95% at 1.5GHz and 100% at 1.4GHz". So you can overclock the chip to 1.6GHz, but it will only be reliable and execute instructions correctly, 85% of the time. So 15% of the time - you are generating bad data, corrupting good data, and in general suffering from a "sick" computer.

When you overclock, what you don't realize is that the part you are overclocking (CPU usually) was already overclocked in the "speed binning" QA process and found to be unreliable at anything faster than it's labeled speed.

Think about it this way. What would the manufacturers have to gain by labeling a chip slower than it should be labeled. Nothing. They label the chips with the "proper" speed based on the results from the QA and Speed Binning process.

Take Care

 

[ninja fetus]

Now thats good info!!!! Wow that makes me think twice about OC'ing my 2600+.
Very good thread. Thumbs up!!!

 

[harrack52]

What are some of the consequences of having a sick computer ?

 

[ninja fetus]

- you are generating bad data, corrupting good data,

i think that

 

[Big B]

Depends on what the system is for. For a server, not ever, and you're a fool if you're going to OC anything on that at all.

Data corruption: With the nForce 2 and K8T800 Pro chipsets for AMD and the i875/865 and 9x5 chipsets having PCI (and AGP locks in the 8x5's) locks, this allows overclocking the CPU without OC'ing the PCI and AGP busses. This keeps the PCI devices from having a cow. In the past, yes, this has been an issue, but the newer chipsets have gotten around this.

If you're overclocking, you know that there are limitations, which is why you take the OC up in increments and test. You're right in that OCing to the point of instability is stupid. However, most people do educate themselves on overclocking before trying it.
If you're doing it right, you'll do enough testing yourself to find what OC is working fine. Running Prime 95 to tax the CPU, running Memtest to check your progress is helpful. Of course it going to cause errors if you try some insane OC right off the bat. That's why you test if you decide to OC.

I've got a 1.4GHz chip that does 1.7GHz with stock voltage and air cooling with a less than spectacular cooler. And it's stable.

It's a hobby for people. If you don't want to do it, fine. I'm not going to be insulted. I agree, it is a risky venture, and if you have a PC that requires stability, you shouldn't OC it. However, if people want to try it, that's fine, as long as it's your system.

Intel and AMD could just go sell only the fastest speed CPU's, but not everyone wants to pay $1k for a CPU. Yes, part of it is stability, but they'd loose a lot of money not selling slower and less expensive chips. It's more profitable to do the speed binning process.

What it comes down to is choice. This isn't a religious or moral decision, and it's not a sin to OC. If you bought the CPU, it's your perogative if you want to OC it.

Anymore OC'ing isn't necessarily done to save money (although it never hurts), but also simply for the hell of it. I really don't get spending $5k on a rig thats LN2 cooled just to run 3dmark a few times to gain the highest score and then have the system go to hell, but there are people who want to spend their money like that. Fine, it's there money, and as long as it's not illegal---I really don't care.

 

[dansicotte]

I won't waste my time attempting to destabilize any of my systems by overclocking.

Depends on what the system is for. For a server, not ever, and you're a fool if you're going to OC anything on that at all..

- I totally agree. I view all my PC's as "servers", no mater what their use - as I've passed my days of time-luxury where I could constantly spend hours trying to run diagnostics to find the happy medium for overclocking.

I've got a 1.4GHz chip that does 1.7GHz with stock voltage and air cooling with a less than spectacular cooler. And it's stable.

- Remember what I said about stability? Maybe your just one of the lucky ones.

It's a hobby for people. If you don't want to do it, fine. I'm not going to be insulted. I agree, it is a risky venture, and if you have a PC that requires stability, you shouldn't OC it. However, if people want to try it, that's fine, as long as it's your system..

- Doesn't running an electronic device that operates on either a 1 or a 0 in iteslf enough to warrant the need for stability?

What it comes down to is choice. This isn't a religious or moral decision, and it's not a sin to OC. If you bought the CPU, it's your perogative if you want to OC it.

- I'd rather not get in the moral discussion of overclocking, I'm only making the point that you will in fact suffer from some form of instability, some time when you run faster than the manufacturer's labeled "proper" speed. With all the hours of running diagnostics, and all the "overclocking education" you can get. Plain and simple. You will, at some point, drop some ones-and-zeros, or those ones will turn into zeros or vice versa. Bitflip becomes a big problem when you push things.

- Bottom line, if you want to mess around and wast time attempting to squeeze blood from a turnip - then go ahead. I have more to do with my time, then to be constantly messing with my equipment to make it stable.

- Now don't get me wrong - this is not meant to be a personal attack on any who has overclocked, or is currently overclocking. And you may also be having "acceptable" results. All I'm saying is that my definition of "acceptable" is probably much different than anyone who overclocks.

Take Care

 

[Big B]

I hear ya. Personally, as much as I do support overclocking, it's not the biggest priority in my purchasing decisions. I'm more intrested in what I can get for my money. If it turns out to have a big OC, great, but that's not my first thought. If you want faster, the only way to guarantee that is to buy it. I've only had a couple of good OC'ing CPU's, and both were P4's (which tend to scale like mad). Getting massive OC's is 5% good hardware, and 95% luck.

I hope that clarifies my position on the subject.

 

[dansicotte]

Overclocking will become harder as chips get more and more complex.

- For the most part, Moore's law still holds true; Moore's law boils down to the doubling of the number of transistors packed into a chip, using new technology and developments gained over the course of each new year.

- As this trend continues, and as chips become more and more complex and internally modular - the more room for failure and errors from overclocking, and the harder it will be to get a chip to overclock with positive results.

- Remember, manufacturing a "system on a chip", such as a CPU, using "Deep Submicron" physical design rules is very difficult and is in fact a science. When working on things this small, physics plays a profound role in what you can and can't do.


From Intel's web site:
-------------------------
"Gordon Moore made his famous observation in 1965, just four years after the first planar integrated circuit was discovered. The press called it "Moore's Law" and the name has stuck. In his original paper, Moore observed an exponential growth in the number of transistors per integrated circuit and predicted that this trend would continue. Through Intel's relentless technology advances, Moore's Law, the doubling of transistors every couple of years, has been maintained, and still holds true today. Intel expects that it will continue at least through the end of this decade. The mission of Intel's technology development team is to continue to break down barriers to Moore's Law."

Gordon's Original Paper:
--------------------------------
ftp://download.intel.com/research/silicon/moorespaper.pdf


Take Care.

 

[Big B]

I have no evidence to back this up, but it just struck me now. With the i915/i925 chipsets, Intel has instituted a 10% OC limit. Now based on the immediate above post, that could be another reason that they are doing this. Fool proof, not at all, since Asus and others have already gotten around it. However, it is possible Intel is limiting the OC with their chipsets to 10% for this reason. Again, no proof, but just a thought.