Ofpmark 2.0 and -nomap command

[cornhelium 0]

Hi,

I'm using the -nomap command and wondered if it affects performance much (Suma stated that OFP used less memory when run with -nomap, but with a slight loss in performance).

I just ran OFPMark 2.0 using recommended .cfg and settings, four tests each (straight from a clean boot), with and without -nomap (no other differences in settings/background programs).

Results:

Totals without -nomap: 820.429, 826.425, 820.823, 848.37 Average 829.01175

Totals with -nomap: 854.684, 860.617, 818.642, 848.948 Average 845.72275

I'm pretty sure I ran the tests the right way around, so it seems that the -nomap command actually returns a slightly increased OFPMark score.

I did notice a very slightly lower score on tests 2 and 4 with -nomap (both heavy on camera panning), but test 1 seems to return a consistently lower score without -nomap.

Can anyone corroborate pls?

Either way, it seems that -nomap is definitely the  way to go if you have any crashes/freezes with 1.94b  

Cheers - Cornhelium

Asus A7M266, BIOS 1008b, AGP4x, AGP aperture set to 128mb

512mb Crucial pc2100 (266mhz) RAM

XP2100+ (Palomino core)

WinME

GF4 ti4600

Audigy2

[suma 8]

The performance problem with -nomap is that when opening large files (from a pbo or from a standalone file), the whole file is loaded into a memory - which can be slow on some systems.

On the other hand with memory mapping (which is used by default when -nomap is not used) the file is only mapped and then loaded as needed. The problem is that mapping is done at .pbo level and therefore it costs a lot of virtual memory address range when using lot of large pbos (but it uses very little physical memory).

The performance issues seen with -nomap (if any) you should expect is not worse average performance (which is what benchmarks usually measure), but a short drops in performance (when big files are opened).

[cornhelium 0]

Thanks for clarifying Suma.

OFP Forever!