New Halo Patch, bench numbers up

[BenSkywalker]

Link.Old numbers first, then new numbers.

Date / Time: 10/9/2003 12:10:35 AM (476237453ms)
1900MHz, 512MB, 128M nVidia GeForce4 Ti4200 (DeviceID=0x0253) Driver=6.14.1.4345 Shader=1.3
C:\Program Files\Microsoft Games\Halo\halo.exe -timedemo (Version=1.0.1.580)
Frames=4700
Total Time=218.88s
Average frame rate=21.47fps
Below 5fps= 10% (time) 0% (frames) (23.041s spent in 16 frames)
Below 10fps= 12% (time) 1% (frames)
Below 15fps= 28% (time) 10% (frames)
Below 20fps= 41% (time) 20% (frames)
Below 25fps= 58% (time) 38% (frames)
Below 30fps= 78% (time) 64% (frames)
Below 40fps= 96% (time) 91% (frames)
Below 50fps= 99% (time) 96% (frames)
Below 60fps= 99% (time) 97% (frames)
Memory used Max=162MB, Min=130MB, Ave=150MB

For further information, please visit the timedemo FAQ at: http://halo.bungie.net/site/halo/features/hpcperformancefaq.html
Date / Time: 10/9/2003 1:16:35 AM (726609ms)
1900MHz, 512MB, 128M nVidia GeForce4 Ti4200 (DeviceID=0x0253) Driver=6.14.1.4345 Shader=1.3
C:\Program Files\Microsoft Games\Halo\halo.exe -timedemo (Version=1.0.2.581)
Frames=4700
Total Time=160.51s
Average frame rate=29.28fps
Below 5fps= 6% (time) 0% (frames) (10.488s spent in 12 frames)
Below 10fps= 7% (time) 0% (frames)
Below 15fps= 27% (time) 9% (frames)
Below 20fps= 29% (time) 10% (frames)
Below 25fps= 37% (time) 16% (frames)
Below 30fps= 49% (time) 27% (frames)
Below 40fps= 74% (time) 57% (frames)
Below 50fps= 93% (time) 86% (frames)
Below 60fps= 98% (time) 94% (frames)
###Sound Options###
Hardware Acceleration= Yes
Sound Quality= High
Environmental Sound= EAX
Sound Variety= High
###Video Options###
Resolution= 1024 x 768
Refresh rate= 85 Hz
Framerate throttle= No Vsync
Specular= Yes
Shadows= Yes
Decals= Yes
Particles= Off
Texture Quality= High

For further information, please visit the timedemo FAQ at: http://halo.bungie.net/site/halo/features/hpcperformancefaq.html

Don't know why it says Particles are off, they are on High and are clearly working in game and in bench.

Edit- They are still couting loading times, so the benches still don't jibe with in game, but they disabled debug mode which still impacts performance and gives a better idea at least-

Yes, we are doing two things:
- For every frame we render, we will not look at the memory footprint anymore.
- We are reformatting the output to display which graphical / audio settings are enabled / disabled. (so timedemo.txt will tell you if shadows were on or not, for example).

These are the only two changes. Any changes in the average frames-per-second numbers you use to see with halo.exe version 1.00 or 1.01 is only the result of not looking at the memory consumption anymore.

We had a debate about whether we should factor in loading times or not in the timedemo results and we have decided to keep loading times in the results. Actually, we don?t really have a choice. For any given frame, we have no trivial way of separating loading time from actual rendering time. This means that a given Windows 98 system running with DMA enabled on the IDE hard drive will render the whole sequence faster than the exact same machine running with DMA disabled. Similarly, two high-end Windows XP systems running with different performing hard disks (5400 rpm VS 7200 rpm, for example) will produce different results.

We think that it is fair for a benchmark to factor these in. After all, benchmarking is not exclusively about raw rendering power ? the performance of the other components in your system should also be taken into account.

Link.