Theoretical Bottleneck between Memory and CPU

[randomface]

I've been looking at building a new rig (my current rig is 6 years old now) and I was doing research to determine which processor I should choose and I noticed that all of the 1150 Socket core i processors have a memory speed or 1600, and all of the 2011 socket core i processors have a memory speed of 1833. The question that this raises for me is: If this number is the CPU's max memory speed then what good does it do to buy memory that is faster or for that matter what is the point in overclocking memory beyond the CPU's max speed? Shouldn't the memory bottleneck at the processor if its running faster than the CPU can handle? If thats true then I can imagine that the data coming through the RAM would do so faster, however it shouldn't make any difference after it leaves the RAM, correct?

 

[Pinhedd]

That's the microprocessor's highest tested data rate. It is certainly possible to exceed that value up to some configurable maximum but operation outside the tested parameters is not guaranteed by Intel and won't be covered under warranty if it causes damage.

For example, my laptop has a i7-2820QM which is rated at DDR3-1333 using standard timings and voltages. I'm running DDR3-1600 at standard timings and standard voltages.

My desktop has an i7-3960x which is rated at DDR3-1600 using standard timings and voltages. I'm running DDR3-2133 at non-standard timings and non-standard voltages.

Both work fine, but my desktop requires a little bit of extra coercing to obtain full stability.

 

[randomface]

So what your saying is that the value provided by Intel isn't the maximum speed the processor can handle it is simply the maximum value they tested it with? Or at least the maximum guaranteed stable value? In other words its not a limit (physically or logically) it is simply a safe figure?

 

[Pinhedd]

So what your saying is that the value provided by Intel isn't the maximum speed the processor can handle it is simply the maximum value they tested it with? Or at least the maximum guaranteed stable value? In other words its not a limit (physically or logically) it is simply a safe figure?

That's correct. There is of course an absolute maximum that can be configured (usually based on the motherboard's clock generating capabilities) and a statistical distribution of stability at various speeds above standard