- Last Updated on Thursday, 04 September 2014 07:20
Note that in this Benchmark Results menu, there is a menu item "Old Benchmark Results" which shows the results for CS6, using the older PPBM5 Benchmark test. There is another menu item called "PPBM6 Results", which shows the results for CS6, CC and CC2014, using the new PPBM6 Benchmark test, but you have to login first to see that option.
The PPBM5 Benchmark Test is available for all versions of Premiere Pro CS5+, including CS6, so go ahead and use that for the time being. The Top 25 Charts and MPE Gain Charts from the PPBM5 Benchmark as well as new Top 26-50 charts are available on the next pages, but since this is all about CS6 and CC, we only include CS6 and CC results and the Relative Performance Index, RPI, has been recalculated for CS6 and CC results only. To refresh your memory, the RPI is a balanced weighting of the different tests to compensate for different durations of individual tests and is based on the average of the Top-10 observed values.
We have added information about the distribution of observed values as shown below.
Statistically these indicators mean the following:
- D9: upper decile, 10% of the observations are better, 90% are worse.
- Q3: upper quartile, 25% of the observations are better, 75% are worse.
- Med: median, 50% of the observations are better and 50% are worse.
- Q1: lower quartile, 75% of the observations are better, 25% are worse.
- D1: lower decile, 90% of the observations are better, 10% are worse.
When interpreting results from these charts, keep in mind that a 'balanced' system is one where all the scores are around the same percentile level. If all scores are around Med (median) this means a balanced rather average system, if all scores are around Q3 (upper quartile) this means a well performing balanced system, clearly above average, but if one test is clearly different from the other tests, say all tests are around the Q3 level, but one test is way lower than Med, that is a clear indication something is causing these disappointing results and needs further investigation.
If the Total Time score and the RPI (Relative Performance Index) score differ widely in numerical value, that is a clear indication of an unbalanced system. For instance 200 versus 400. A balanced system would have values that are very close.
These are the various percentile scores we found on the different tests:
The following chart is especially instructive, as it shows all the main tests in one overview, the Relative Performance Index or RPI as columns, the Total Time as the blue area and the individual tests as lines with a different scale on the right hand side.
What is noteworthy is that H.264-BR performance does not degrade as much as Disk I/O performance when the overall performance goes down, because CPU performance, even with fewer cores and lower clock speed, has less impact than fewer disks. Most of the lower performing systems are limited to a single project disk and at the other end of the scale are the raid arrays with multiple disks. Also note the grey columns (they appear a darker blue when overlayed with the Total Time area) that show what the RPI would be if the system is nicely balanced. The Top 25% of the systems are nicely balanced, but the bottom 25% are not very well balanced.
To give you a rough feel of where your system should ideally be in Total Time, based on indicators like `Monster`, `Warrior`, `Economical` etc. and assuming it is properly balanced in all components, each task being performed about equally well, this chart emerges:
In Relative Performance Index terms, the ideal balanced system should look like this:
If your score does not measure up to these expectations, have a careful look at your setup and tuning. If your score does not even reach the D1 level, and is amongst the slowest 10% of systems, it is a hopeless case. Only a new and faster system will improve your editing experience.
We have also added an overview graph with the popularity distribution of PC Brands. It is obvious that Asus and Gigabyte account for around 75% of the systems, a clear indicator that most systems are DIY ones and the tier 1 brand names like Dell and HP are not very popular.