X86-64 and SSE2 performance on John the ripper

Written by Gionatan Danti on . Posted in Hardware analysis

User Rating:  / 4
PoorBest 

Processors generations: a comparison

So, its time to go pragmatic: what is the real performance differences between the examined processors? Well, the next graph show you this thing:

Processors speed compared

The green line represent the averaged performance improvement as measured by the three DES benchmarks.

You can see that the line has a steady trend. This means that, in 2006/2007, by using a medium-high class processors as Core2 T7200 clocked at 2.0 GHz, you had 3X the performance of an 2004-released equivalent medium-high class P4 clocked at 3.0 GHz. Some years later, the 2009-released Lynnfield Core i7 860 processor clocked at 2.8 GHz give us another 1.7X performance speedup, for a grand total of over 5.2X our original Pentium 4.

So, we have very high performance gains, and all that using only a single computational core. If you think that the Core2 has two cores and Core i7 has four core, well, the potential performance advantage over P4 became embarrassingly high.

What about processors clock-for-clock efficiency? The next graph tell us about efficiency...

Processors clock-for-clock efficiency

 

Clock-for-clock, the Core2 is a whopping 4.5X faster than Pentium 4! While this was expected (the P4 was designed with clock speed in mind, rather that efficiency) it is a impressive result. The Core i7 processor improve another 1.4X over the Core2, for a total of 6.3X over our old P4.

The interesting thing is that, speaking about clock-for-clock efficiency, the curve is not more linear: it clearly break down at the Core2 / Core i7 switch. This normal: while the Core2 was a revolutionary change over the disappointing Netburst microarchitecture, the Core i7 was an evolutionary change over the Core2: for example, the execution back-end (the numbers of interger and floating point numbers) remain the same as with Core2. The added efficiency is a results of others important improvements, as:

  • integrated memory controller

  • trilevel cache design

  • improved branch predictions

  • wider out of order execution window (Intel claims a 33% more microinstruction “in-flight”)

However, while these changes are very important, the results Core i7 processor give us a lower total speedup compared to Core2 vs P4.

You have no rights to post comments