As of this writing, there is no known method for
recording a timedemo in F.E.A.R. So, I followed a set path through the demo that
begins just before you enter the first room containing two guards. My path ended
by shooting the guard on the top railing that is soon followed by the heavier
armored guards that burst through the door. You can see a "mini-newsreel" I put
together below showing my path.
F.E.A.R. Benchmark Path
Each individual benchmark result below is based on me repeating this path six
times with Fraps
and recording the minimum (MIN) and average (AVG) framerates. The first
iteration was not used in the calculation though as F.E.A.R. runs much better
once the level has initially been cached (similar to Doom 3 and other games). I
decided to do so many iterations due to how varied the AI in F.E.A.R. can be at
times. Depending on your timing, the enemy AI may take completely different
approaches than the time before.
I was very CPU bound (and too with just 1GB of memory instead of 2GB) with
F.E.A.R. so my overclocking gains were negligible. The 7800 GTX will scale like
crazy which gives you a great reason to upgrade your CPU. From the low end of
1024x768 with no antialiasing to 1600x1200 with 4xAA transparency antialiasing
my delta was within 1~2 frames per second. That is why I just give you the overclocked (495/1370) results below to show that running at these speeds proved
stable in F.E.A.R.
F.E.A.R. Benchmark Results
Overclocked @ 495/1370
Athlon 64 3500+, MSI K8N Neo4/SLI, 1GB RAM
NVCPL Antialiasing and Application Anisotropic Filtering
Results are in Minium / Average Frame Rate
TMS = Transparency Multisampling
TSS = Transparency Supersampling
I was impressed with these results considering the bottleneck I am
currently under with my CPU. There is little doubt that I could at least reach the
desired 30FPS MIN and a respectable AVG at settings of 1600x1200 @ 2X AA TSS.
Enough conjecture, though. I did achieve my goal by reaching a very
playable experience of 1280x960 @ 2X AA TSS. On top of that, I was in the same
playability neighborhood under 1600x1200 with no antialiasing. For someone on a
19" CRT the aliasing effects are obviously diminished anyway at 1600x1200 so I
was pretty happy.
Granted, these frame rates are not in the almost silly stratosphere that can
be found in Doom 3 and Half-Life 2 with a 7800 GTX. Again, though, my goal here
was to really stress the card with a game engine itself and not by just piling
on 8XS antialiasing and ever higher resolutions.
Some may wonder why I didn't benchmark with Gamma Correct antialiasing. The
reason is because the performance hit was basically non-existent. You
practically get Gamma Correct benefits for free. This is evidenced in forum
ChrisRay's excellent 7800 GTX antialiasing investigation.