Lets say I get my super liquid nitro cool-it-now cpu refridgeration unit so that heat is no longer a factor.
Then what are the limits to over clocking.
On a related note, what are the different ways to OC?
Lets say I get my super liquid nitro cool-it-now cpu refridgeration unit so that heat is no longer a factor.
Then what are the limits to over clocking.
On a related note, what are the different ways to OC?
I really don’t know what I’m talking about, so I’m just taking a stab here.
Your CPU, on the theoretical end, is hampered by distance.
A given electrical current has to physically travel the span of a processor to get processed in different segments of the physical CPU. Now, electrical currents move pretty damn fast. However, we’re also talking about millions (or billions) of cycles per second. And so the processor can only cycle as fast as the electricity can physically travel to different components of the CPU.
That’s the only immeadiate thing I can think of - and it’s also why you hear about CPUs advancing to smaller and smaller dies.
And… as for overclocking…
Your computer has an internal pacemaker of sorts that cycles X times per second. On modern computers, this can vary - but for simplicities sake, let’s say it ‘ticks’ 100 million times per second. 100 Megahertz.
Now, this regulates the system bus - which handles things like memory and such. PCI devices and such usually operate at a fraction of this speed.
Anyway… Your CPU has to operate a lot faster than this (modern systems getting up to 2 ghz or so…), and so there is a multiplier that regulates the CPU speed. It takes the bus “pacemaker” ticking away and multiples the rate by X to generate your CPU speed.
So if you have a 100 mhz bus, and your CPU is running at 800 mhz, you have a multiplier setting of 8.
Now, there are 2 ways to overclock this. FYI, they don’t make “800 mhz cpus” and “850 mhz cpus”, they simply make a whole bunch of CPUs of the same exact type. Due to manufacturing differences, some are better than others - so they test a processor, and if it can handle 900 mhz, and they have a demand for 900 mhz cpus, they’ll sell it as that. (If there’s a bigger demand, for whatever reason, for 800 mhz cpus, they’ll just take the same chips and mark them down to 800…)
So this means, for us overclockers, that your CPU has the potential to be as high as the physical limitations of that CPU - it’s just that they set it to work at X speed.
So there’s 2 ways we can go about getting more performance out of it:
We take the system bus (the pacemaker), and make it go faster. Since CPU is a multiple of the bus speed, this in turn will increase that speed. If we have a 100 mhz bus at an 8x multiplier, and we crank the bus up to 110 mhz - then we have a processor running at 880 mhz.
We alter the multiplier number. (This isn’t possible on some chips, and/or may take some work with a soldering gun). But we could take that same CPU, and increase the multiplier to 10. Then we’ve got a 1 ghz (100 x 10) processor.
Hope that helps.
Not sure if this is really related, but several years ago I was running a Pentium 90 at 100 MHz. Everything was working fine until I tried to install Windows 95. I kept getting installation errors about halfway through the process, and could not figure out why. I eventually returned the processor to 90 MHz and the installation worked fine! I don’t know why. But I do know that the 11% increase in processor speed wasn’t really noticeable.
SenorBeef pretty much has it.
Note, however, that of the two overclocking methods not all are necessarily available.
As mentioned all CPU’s of a given type are identical. Early on in the manufacturing cycle all the kinks haven’t been worked out so some chips off the line will only be able to handle (say) a 400MHz clock while some others might be able to scale to 500MHz. Eventually the manufacturing line gets everything pinned down and all processors might be able to clock to (say) 800MHz…the absolute limit for that particular model. They still have to sell slower, and therefore cheaper, chips for people on a budget. For those people the manufacturer simply locks in a lower core multiplier so instead of having a multiplier of 8 on a 100MHz bus (for an 800Mhz clock speed) they set the chip’s multiplier to 5 so you get a 500MHz clock speed.
You might be asking yourself why the hell you would pay the big bucks for an 800Mhz processor when you could buy a cheaper 500MHz and just crank up the speed.
First of all, there is no guarantee you will get the 800Mhz speed. Secondly you just voided your warranty on the chip. Nevertheless many people took exactly this route and needless to say the manufacturers were none too pleased. In Intel’s case, around the time of the P-II, they locked the core multiplier into the chip and the consumer could not undo those changes. As a result the only choice for overclocking an Intel CPU was to crank up the Front Side Bus speed (e.g. a 100MHz bus to a 110MHz bus so a multiplier of 5 goes from being a 500MHz chip to a 550MHz chip). Unfortunately there are limits to how fast the FSB can be cranked up. Not only does your CPU run faster but so does your memory and PCI bus that connects all of your peripheral cards (it is probably the memory or an overstressed card that flaked out ski’s PC).
CPU’s from AMD can have their core multiplier adjusted. However, AMD doesn’t make it easy on you and has also locked the multiplier onto the chip. Depending on the model of Athlon chip you have there are various methods for circumventing that lock and adjusting the speed. I’m sure this is intentionally left this way by AMD and may partly explain the esteem computer geeks hold for AMD over Intel.
In the case of AMD you now get the best of both worlds. You can pump the FSB as well as adjust the core multiplier allowing for a wide range of overclocking possibilities. I have done this to my PC and I love it and have no problems.
All of that said let me re-iterate one more time that overclocking will void your warranty. It is definitely possible to permanently damage your equipment when doing this. That said it isn’t rocket science either and if you read up a bit on it you can somewhat safely travel that road. Just don’t come crying if something gets hosed.
Finally, you need a motherboard that allows overclocking. Most motherboards you get in store-bought machines are cheap and have no features that allow you to fiddle with your system (which is partly intentional beyond cost so people don’t crew-up their PC’s then come crying back to the store). There are, however, plenty of motherboards that do allow tweaking to greater or lesser degrees. Some allow only FSB tweaking, others allow both types and some allow you to even set the voltage going to the CPU (possibly important for stable overclocking). ABIT, ASUS and MSI are a few mobo manufacturers I can think of off of the top of my head that have products that allow you to overclock.
Yep, the practical limit to overclocking is how fast you can get the FSB to run. While a super-fast memory-starved CPU is not entirely useless, it nevertheless isn’t as effective as a slower CPU with a better data flow. AMD tweakers will gladly forego the top attainable processor speed in order to kick the bus speed up as high as it will go. Intel tweakers don’t have a choice, unless they’ve happened upon an unlocked engineering sample.
Here is a dated but still perfect example:
Why 702 MHz Is Faster Than 728 MHz
There are other considerations, such as the amount of on-die cache on the CPU. Those of us who scored the Celeron 300A were delighted to discover that when overclocked to 450 MHz, its 256K on-die cache often exceeded the perfomance of the Pentium II 450 with 528K of half-speed off-die cache. (And my old housemate is still smarting over that one–heh, heh.) Similar results were discovered when the AMD Athlon moved to the Thunderbird core.
I have argued that the theoretical top speed of a processor increases when the manufacturing process undergoes a die-shrink, as mainstream processors did a few years back when they went from 0.25 microns to 0.18, simply because the pathways are shorter. Others more in the know inform me that this is the least of the benefits that a die-shrink confers upon a chip, and that it is not really measurable in the real world. Heat reduction and more, better quality samples are what really makes a shrunken CPU “faster,” and cheaper, too.
The answer is, “it depends”. It actually depends on many things, and can even vary between different units in the same lot. Tom’s Hardware Guide had an article from February where he was able to overclock an Athlon 1200 to 1.6 Ghz using Asetek’s Vapochill, which is a refrigerant-based cooling unit. So don’t expect miracle speeds - don’t think you can double-clock any CPU even if you could cool it down to below freezing.
Yeah, from experience I can tell you this: If you overclock a CPU, and it’s not happy with it’s new speed, it can give you really bizarre errors that you wouldn’t connect with an OC’d CPU. Whenever you start overclocking, always keep an eye out for things that don’t seem to work quite right. Usually, they can be fixed by returning it back to normal clock.
At the risk of reiterating what’s already be said for my own personal amusement, the limits to overclocking are determinded by the physical properties of a chip. A chip functions by performing predefined functions know as instructions. Each of these instructions takes a certian amount of time to compute. More specifically the logic paths required to execute an instruction require a certain minimum time before the output can be guaranteed to be the correct value. This is analogus to someone sliding beads around on an abacus for a few moments before the correct answer is reached and the abacus ‘settles’ into a stable state.
The clock pulse of the computer (after multiplying) is responsible for telling each instruction when to start execution. As you can’t begin the next instruction before the last a finished, the clock must be at least slower than the slowest datapath.
The chip manufacturing process is much as what Whack-a-Mole said. They make a bunch of chips, test them out, and see how fast they can get them to run. The slower ones they sell cheaper as slower chips. The faster ones they pump up and sell as top-of-the-line. The whole overclocking thing is possible because chip manufacturers are cautious. They say “this chip can run at 800MHz, but to be careful we’re going to sell it at 700Mhz”, meaning that you get an extra-stable processor, however it’s not running quite up to it’s capacity. If you choose you can bump it up to 800Mhz, however it’s not guaranteed to function the way that the manufaturer says it will.