Basically, there are three PAR2 apps that are optimized for multi-core processors for Windows.
I tried these particular versions:
MultiPar v1.1.7.6 - (I believe it uses speedups of both phpar2 and TBB. Includes full GUI too! Maker is currently working on PAR3 specification.)
phpar2 v1.3 - (Based on original par2cmdline 0.4 with hand coded optimizations, and bug fixes)
Par2+TBB v20100203 64-Bit Edition - (Based on original par2cmdline 0.4 and makes use of Intel Threading Building Blocks 2.2)
There have been numerous debates on which is the fastest for par creation. I spent few hours researching and ran my own benchmark using a stop watch.
System Rig:
i7 920 @ Stock
6GB DDR3 1600
Gigabyte EX58-UD4P Motherboard
150GB VelociRaptor (System Drive)
WD 750GB AAKS (Drive I used for Par creation)
Windows 7 Pro x64 w/ SP1
Benchmark Procedure:
- I defragged the test drive three times before starting the benchmarks. All unnecessary applications were turned off, and I did not use the computer while running the tests.
- I used a stop watch from the moment of starting the batch script to point of completion.
- I then deleted the newly created parset and ran the batch again.
- I used the batch script three times per application and made an average of the three times.
- I used the following settings: -s1920000 -r10 -l (10% redundancy and 1, 920, 000 block size)
- The source of the parsets were one 1080P scene release that is roughly 8.5GB in size and created parset for only the RARs.
Benchmark Results:
MultiPar: ~11:49.50
Par2 + TBB: ~11:23.79
phpar2: ~10:29.16
There you have it. phpar2 is the fastest! Almost a 1.2 minute difference from MultiPar, which makes a noticeable difference when one is parring up over 200GB. Please make note that phpar2 is especially optimized for Intel Core 2 Duo architecture.
I hope this information will be helpful for others. Also, I still use MultiPar for checking/repairing when needed since it's GUI is awesome and supports file association with par2 files.
