Sign in with
Sign up | Sign in
Your question

Do cpu's use a crystal oscillator?

Last response: in Overclocking
Share
June 19, 2012 6:56:45 AM

do cpu's use a crystal oscillator?
is this why overclocking a cpu can damage it? because it causes the crystal to fracture/crack or change frequency?
is this why some cpu's overclock better then others even tho the're the same model?
or does overclocking damage the metal or silicone in the chip?
i know graphics cards sometimes break from electromigration is this the case with cpu's somehow?

im trying to learn more about computers and their parts if you have any useful easily understandable links to this i'd appreciate it.

thanks :hello: 

More about : cpu crystal oscillator

a b à CPUs
June 19, 2012 7:03:26 AM

I think they just break because they get too hot and the silicon gets damaged, no idea about oscillators, I thought they were a radio thing
m
0
l
a c 283 à CPUs
a c 110 K Overclocking
June 19, 2012 7:07:25 AM

The crystals never change frequency, they're just there for a reference frequency. What kills CPU's from OC'ing is too much voltage and too much heat.

And yes, electromigration is a major problem for CPU's as well.
m
0
l
Related resources
a c 185 à CPUs
a c 150 K Overclocking
June 19, 2012 7:08:43 AM

Yes, they have crystal oscillators.

No.

No.
m
0
l
June 19, 2012 7:15:16 AM

The clock for the CPU is supplied by the mainboard. There is a (or several - no idea) oscillator there, but it's practically indestructible. (They are designed to supply clocks way in excess of what the CPU is capable of taking)

What "breaks" inside a CPU are the silicon transistors it is made up off. Lot's of complicated physics here :D 

The reason some CPUs overclock better then others are simply variances in manufacturing quality.
m
0
l
June 19, 2012 7:19:59 AM

The crystal oscilator is on the MB. Overclocking usually breaks the CPU because to achieve it, you need to go to higher voltages, which will lead to higher currents in the transistors. That current generates heat, and locally (transistor level) it could overheat because of internal thermal resistance (cannot disperse the heat fast enough), thus breaking the transistor (1 damaged transistor in billions that current CPUs are made of could render it unusable). This can happen even with the CPU temp monitors at good levels.

In other words, even with ok temps (so powerful enough cooling) you could still damage the CPU. The higher you push it, the lower temps you need to achieve though better cooling.

As for how the oscillator works, it's all about PLLs (phase locked-loop). A PLL generates internally a frequency (which can vary a lot with temperature) and feeds it to a timer to compare it to an external reference clock (at much lower frequencies). Once stable values are found for the counter (no to very little drifting from the external clock), the rest of the circuits are activated, based on that frequency.

Now... the MB oscilator frequency is fed to the MB PLL, thus generating the external CPU frequency. The CPU itself has another PLL which uses the CPU ext. frequency as reference.
So I explained all this to tell you that with one crystal oscillator you can generate almost any higher frequency.
Also, according to wikipedia, few oscillators are over 30MHz.
m
0
l
a c 169 à CPUs
a b K Overclocking
June 19, 2012 7:21:31 AM

CPUs do not use crystal oscillators internally to generate their clock (although there is no technical reason that they cannot). Typically the clock source comes from a pulse generator of some fixed low frequency (sometimes a crystal but it can be a purely electronic circuit created using analog components) and this low frequency pulse train is pumped into what's known as a PLL or Phase-Locked Loop. The PLL is responsible for multiplying the input signal into a number of configurable output signals. The output signals are a function of the input signal's period/frequency, a scalar multiple and a phase angle. All parameters need to be known or found by comparison in order to avoid nasty causality problems.

If a PLL is configured to create two output signals with a frequency 10 times that of a 50Mhz input signal, with one signal Pi radians (180 degrees) out of phase with respect to the source signal then the PLL will create two 500Mhz output signals, with one trailing the other by half a period.

The reason some CPUs overclock better than others has purely to do with silicon quality. Not all CPUs are manufactured equally, however CPUs within a family are manufactured identically. The best manufactured ones are taken and packaged as the top end CPUs and downward from there. CPUs that have physical defects which would cause them to malfunction when operated at full capacity may have part of their functionality cut and then be sold as a lesser model. The famous Sandybridge-E 3960x and 3930k are actually both fabricated as 8 core processors with 20MB of L3 cache. Both processors have 2 cores disabled for yield reasons (the 8 core Xeon models cost thousands of dollars) and between 5 and 8 MB of L3 cache disabled.

What sometimes happens is that silicon yields and market demand diverge. Either good silicon yields are low and top end products cannot be manufactured in sufficient quantities (this is what's holding back the GTX 680 right now, yet is resulting in tons of GTX 670s) or silicon yields are high and far more top end products can be manufactured than necessary (this tends to happen as a manufacturing process becomes more mature). In the latter case the manufacturers take perfectly good high quality chips and package them as weaker processors to meet market demand. This is why some AMD X2,X3, and X4 processors can be unlocked to X4 and X6 processors yet others cannot. Some were binned down due to the presence of defects, others were binned down for marketing reasons.

So that explains why some chips overclock better than others, but what about the effects of overclocking?

All electrical circuits undergo a process called electromigration. Electromigration occurs when current flow cause by an electric field physically damages the conductive path that the current flows through. This doesn't only happen in CPUs, it happens in electrical wires everywhere. At low temperatures the crystal structure of the conductive paths in CPUs is strong enough that high current wont dislodge the atoms from the structure. At higher temperatures, or at higher field strengths, the flow of current places a much greater force on the conductive path and degrades it quicker. Eventually this can degrade the conductive paths to the point where the CMOS logic network cannot switch states fast enough and the only way to rectify this is to feed it more voltage which in turn furthers electromigration. Eventually, the conductive path will break completely and render the circuit useless (burned out). Electromigration happens whether or not you are overclocking; overclocking will simply worsen it exponentially. So yes, overclocking increases the rate at which the silicon sustains damage.

I hope this helped :) 
m
0
l
June 19, 2012 7:30:30 AM

ok, these answers are AMAZINGLY helpful ive googled this a lot and almost no website says theres even a crytsal in them or any of the things im asking :pfff: 
anyway electromigration happened to anvidia 8500gt which i was actualy able to fix by the popular
"bake it in an oven" technique, would this work on a cpu? i mean it would reflow the metal connections so it would work again and the graphics card had a gpu on it so it cant really damage it too mutch. also if its broken doesnt matter if it breaks anymore
m
0
l
a c 169 à CPUs
a b K Overclocking
June 19, 2012 7:33:14 AM

The "bake it in an oven" technique is used to fix solder points that have melted and broken due to excessive heat buildup and warping. The baking heats the board up evenly which allows the solder to flow back in place.

Electronmigration happens at the molecular level, you can't really see it
m
0
l
June 19, 2012 7:35:20 AM

oh so it wouldnt change a thing?
m
0
l
a c 283 à CPUs
a c 110 K Overclocking
June 19, 2012 7:37:47 AM

fraza said:
oh so it wouldnt change a thing?


Not with a CPU because there's no solder to "reflow" (even if that WERE the problem). Pinhedd explained it very well, actually.
m
0
l
a c 169 à CPUs
a b K Overclocking
June 19, 2012 7:39:16 AM

fraza said:
oh so it wouldnt change a thing?


CPUs should never warp due to the socket design, so no not really
m
0
l
October 18, 2012 1:34:22 AM

Yes the CPU uses a crystal oscillator but the frequency of this oscillator is fixed meaning that it will never be change. The changing of frequency is due to the breakdown of the components inside the CPU and it is also because of too much heat the silicon gets damage.

pcb assembly
m
0
l
!