Just an observation. I have an impression that AMD and Intel are not eager to break the 4 Ghz mark. Instead of producing processors that can run internally at 4 ghz, it looks like the trend nowadays is to increase the number of cores and the amount of L2 cache and even adding another cache (L3). Any explanation guys?

With current technology and materials the 4GHz mark is possible but impractical. A dual-core CPU can often easily outperform a sing-core CPU even if the speed of the single-core chip is much higher than the cores of the dual-core chip.

You are correct in assuming that manufacturers today are shying from pure speed and are instead ramping up the number of cores. Speed can only take you so far.

Hi,

One of the most common misconceptions is that clock and performance are the same thing. They aren't.

Clock is a signal used to synchronize things inside the system.

If you have a CPU that is capable of doing things using less clock cycles that means that this CPU will be faster.

Just one simple example. If you take a CPU that processes one instruction at one clock cycle, it will have double the performance of another CPU that processes the same instruction using two clock cycles instead, assuming that both CPUs have the same clock rate. So here you can see how even running at the same clock two CPUs can have totally different performance levels.

The problem is that nowadays CPUs are very complex, with several different circuits that change the whole dynamics of the CPU and affect performance, like the memory cache.

But to make my answer short, manufacturers instead of wasting a lot of money researching materials to solve the several problems that a higher 4 GHz clock would bring (e.g. EMI), they decided to optimize the architecture and gain performance on the architecture instead of on the clock increase.

It is always interesting to remember that Pentium 4 and Pentium D architecture, Netburst, had a major flaw for performance: this architecture only achieves a higher performance by increasing a lot the clock rate, which brings heat problems and also a lot of manufacturing challenges (like EMI, as mentioned).

In fact when the Pentium 4 was release it was slower than a Pentium III with the same clock rate, as reported by Tom's Hardware Guide at the time.

This is the reason that Core 2 Duo uses Pentium III architecture and not Pentium 4's. Core 2 Duo is faster than a Pentium D even running with a far lower clock rate.

On our CPU section we have all those "Inside xxx Architecture" tutorials that explains all of this in depth.

Cheers,
Gabriel.