FFmpeg x264 encoding; More cores or faster cores?

Something different this time, testing with a faster CPU that has less cores. Is having more cores or faster cores an obvious difference for FFmpeg video encoding?

Last time the test went through all the FFmpeg x264 encode presets to compare the outcomes.

The first test CPU was 2x E5-2650 v2 @ 2.60GHz (16 cores, 32 threads), This test uses an  E3-1270 v2 @ 3.50Ghz (boosting to 3.90Ghz) but only 4 cores 8 threads. 32 GB of ram and a HDD not a SSD like the first time.

The same media is also being used, an unnamed large 1920 x 1080p video ~ 24GB in size.

Preset comparisons

Similar results in the outcome size but the speed in particular the slow based presets are definetly slower.

More cores or faster cores?

TEST A: 16 cores @ 2.60Ghz (E5-2650 v2) TEST B: 4 cores @ 3.90Ghz (E3-1270 v2)

Geek bench single and multi core test has the E3-1270 V2 beating the slower clocked E5-2650 v2 considerably. However as the chart above shows if you pair two E5-2650 v2 CPU’s together it will turn the tables and encode video much quicker than a fast CPU.

Test A veryslow was 4 hours and 47 minutes quicker thanTest B, For slower it was 2 hours and 11 minutes quicker. Again at the slow preset Test A finished 1 hour 24 minutes before the Test B time.

Check out the slowbased presets. The dual E5-2650 v2 system hammered the E3-1270 v2. Its easy to say that more cores are better for FFmpeg encoding, however the system with more cores did have an SSD which would favour it.

Another aspect being the E5-2650 v2 is a work horse CPU, at 8 cores and 16 threads its built for demanding tasks like this. It is also obviously designed to perform along side many CPU’s in the same motherboard. Does it scale and spread load better?

Had i been able to run 8 cores @ 3.90Ghz the margin would have closed and even likely surpassed the 16 cores @ 2.60GHz, its rather hard to say.

I thought the faster CPU would have performed better alas it shows that more cores can spread the load and encode video quicker.