For all the people who like to overclock processors
without a clue about chip design, this is a good read :
For all the people who like to overclock processors
I take some of his points, but he left out the fact that when processors are manufactured to a certain spec, and fall below that in testing, they’re classified as a lesser processor.
When you take that into account, and the fact that human error can come into play during that test, attempting an overclock doesn’t seem so crazy.
In most overclocking, you do need an idea of where the extra speed is coming from.
generally, this is from one of a few areas…
- Thermal limitations, by applying extreme cooling - if the main liniting factor is heat generation at higher clock speeds
- Bump the voltage, raise the speed - switching will normally get faster, but heat generation increases (see 1.) - some older CPU series actually used this, the higher speed running at higher voltage
- Batch selection - the CPU’s in your batch are actually better than they needed to be - once the required amount of “top end” CPU’s have been taken out, the next speed down will include those that FAILED the higher speed, and those that would have passed if they were tested at it - this is likely in late production.
- Pushing your luck - infringing the reliabilty margins - that’s why you NEVER overclock a critical system.
Totally agree!! Reformatted too many times for O/C ing to be worth the extra 500 to 600 Mhz.
Typical “engineer article”. All the lines of thinking are sound, conservative and verifiable. But it completely misses the point, as engineers often do.
O/C’ing is FUN , pure and simple. A reasonably well-informed enthusiast with the proper tools and testing proceedures can reasonably expect to tweek an O/C into improved performance. The results can be expected to be stable and secure as long as the right methods are followed and hardware is willing.
Most serious O/C’ers are not concerned about frying something, if they do they just replace it. (often, it’s an excuse to try a new one) They also know that no 2 CPU’s will perform the same and tweaking is required.
I agree that it’s not for the faint-hearted or ignorant. Spath has often applied the same line of thinking to the practice of modding burner firmware, and while it’s logical, it also misses the point. Getting that extra 5% of performance is not the point, exactly. It’s satisfying and fun to do it. Man over machine.
I also think that any resource making such things available to the public is responsible for educating them in the use and consequences of use.
If you can’t afford to lose it, don’t abuse it.
All things considered, computer building and tweaking is a reasonably cheap hobby, much more so than hot-rodding. And as hot-rodders are only too well aware, sooner or later everything breaks.
I pretty much agree with rdgrimes here.
…Only thing different for me is, I don’t do overclocking for fun, but for the performance increase.
My P4 2.4C GHz@3.42GHz provides a performance increase of 10% if not more.
My Radeon 9800 non-pro@9800XT provides a performance increase of almost 100%, if not more.
My systems are running prime stable, and temps are excellent.
If you aren’t confident or your system can’t handle it, then don’t do it.
Same with firmware crossflashing on burners…I don’t do it for fun, but for the benefits/features.
Why anyone would go around crossflashing a burner for fun is beyond me
Yes, so ? Chips are sold for the speed they have been succesfully tested
at, what’s you point ?
Chips are not tested by hand, get a clue.
Overclockers can have fun and play dicksize games, I couldn’t care less.
But they should stop pretending to really understand what you’re doing or
to give lessons to professionals. It’s just guessing and trial and errors, and
luckily that’s not how chips are made.
Lol, oops! Oh well, never lost any components / data to OCing yet touch-wood.
Anyway, the safety margins on today’s CPUs etc… ensure that someone that knows what they are doing can safetly overclock a system just shy of the limits of the individual components without adverse affects.
Take the lower end A64 Winchesters for example (3000+, 3200+)…the majority of these CPUs will do 2.4Ghz (3500+ speed) on stock voltages / cooling…of course there are a few exceptions, but why not overclock to these speeds if the capability is there?
Hardly! If you know which components to get, OCing is as safe as running at stock speeds. You just have to be willing to do a lot of rigorous stress testing to ensure stability. Manufacturers stress test to a far higher standard than most PCs will ever be used at…so if a chip narrowly fails their testing for a particular speed (and gets binned to run at a lower speed), that doesn’t mean to say that the chip will not run at that speed safetly. Manufacturers could test to a lower standard and still have no problems with reguards to chip failures…but they have to be 110% sure!
Wow, was I just flamed by a mod? I gotta call my friends…
I never said I was an OC’er, I was just pointing out a fact.
Sometimes you shouldn’t read between the lines, because there’s nothing there…
Overclocking is a guess of trial and errors?
Wow, this thread is getting funnier everytime I visit it.
The actual rigorous speed tests are the ones run at the factory, which
you don’t have. What you can do yourself is to increase the speed and
run a few programs, maybe increase it again if it seems ok, or decrease
it if not. That’s the trial and error method I was talking about.
How would you explain Mobile Athlons then?
They come with multiplier unlocked. That’s right.
One would think that if they can’t reach certain speeds or pass rigorous speed tests that AMD would lock the multiplier.
A chip may be labeled from highend to low if demand of it is high.
Take the 2.4GHz for example…The most sought after P4’s. Intel couldn’t keep up with the demand, so they just took some 2.8/3.0/3.2GHz chips that passed tests at those speeds and labeled them as 2.4GHz.
Graphics industry is the same.
They make the highest end of graphic cards and sell them at $500. When they find out that most people only want a $300, rather than developing a new process, they just downclock the $500 card, disable some pipelines and keep everything else the same(Lazyness).
My card had half its piplines disabled, and I doubled its bandwith by unlocking its pipelines…Going from unplayable 9800SE like performance to higher than 9800XT performance playable is a big difference in DOOM III. I just saved myself $300+ from NOT having to buy a new card.
No way was what I did a comparison of dicksized games.
My Radeon 9800 non-pro is still alive and well since May 2003, and I’m not changing cards until I see Longhorn and DX10
Bah…the “dangers” of overclocking are far too over-rated; i’m more than happy for other people to buy a shiny new Venice A64 and keep it at stock, while i run merrily along at 2.8Ghz + with one.
Maybe one without a clue. Or one who cannot read english, since the
reason of multiplier locking was given in the article. Thanks for illustrating
the kind of people I described in my second post.
The reason for multiplier locking was given in the article(b/c of grey market scams), but the reason for AMD leaving the multiplier UNLOCKED on some of their chips was NOT given in the article. Yes, there is a difference between the 2.
You seem to have forgotten the "UN"locked part in my previous post.
I wanted a reason for "UN"locked and you, nor the author of the article have provided a reason for that.
If I wanted to be like the people in your 2nd post, I’ll get myself liquid nitrogen or vaporchill and run my CPU at 5.x+ GHz.
There are those who don’t care about Tmpegenc taking 10hrs to encode, and there are those who care about cutting the time to less than 8hrs, and use the remaining 2hrs to do something else valuable with the extra time.
I just love how the author seems to think that overclocking is mainly a gaming issue and seems to only mention gaming all the time.
Speaking of “Nitro breathing, flame sporting…”, my CPU runs cooler than ALL Prescott CPUs on the market today.
The author also seems to be forgetting my analogy…
If Dell requests for 10,000 2.4GHz CPU’s in 2 weeks, and Intel manufacturers can only manufacture 4,000 per week and no extras are available, Intel has no choice but to take some of their available 2.8 and 3.0GHz parts and relabel them 2.4GHz. There’s no reason why those extra 2.4GHz parts won’t be capable of running at 2.8 and 3.0GHz stable.
Hard to find a good reason to have unlocked multipliers, other than to allow a single CPU version to be used at two different FSB speeds - remember the 200/266 “B” and “C” Athlon Thunderbirds, and the later 266/333 FSB and 333/400 FSB for the last of the Bartons? - with an unlocked multiplier, no need for multiple CPU versions
A free choice of multiplier would also allow full speed running on some older systems