Clock for Clock: Skylake, Broadwell, Haswell, Ivy Bridge, Sandy Bridge

The last bit of information I wanted to share before getting to our standard benchmarks was to see how the most recent Intel architectures are scaling in terms of IPC – instructions per clock.  To test this, I set our Skylake, Broadwell, Haswell, Ivy Bridge and Sandy Bridge CPUs to a static clock rate of 3.5 GHz and ran through a handful of tests.  This allows us to compare the last five generations of Intel CPUs equally.

With all five CPUs running at 3.5 GHz, Skylake is 21% faster in single threaded results and 25% faster in multi-threaded results when compared to Sandy Bridge. For current Haswell CPU users, that IPC improvement is 10% in single threaded and 11% in multi-threaded.

First interesting point here is that Broadwell seems to have a lower level of AES acceleration than Haswell or Skylake. Instead, it is more in line with Ivy Bridge and Sandy Bridge results. In our average results over several encryption tests, we see the only instance of a backwards step, again with Broadwell. For Sandy Bridge users looking to upgrade, you'll see a 33% increase in encryption capability.

In x264 encoding tests, we see very solid scaling from Sandy Bridge up through Skylake. There is 28% additional performance to be gained moving from the oldest to newest design.

« PreviousNext »