Doom3 couldn´t miss from an update, since it wasn´t released when
I started the review months back. Despite the discrepancies it has created
in numerous of debates lately, it´s still amongst the key applications
for testing/evaluation purposes.
While I like the game, I wouldn´t say that it lives up to the myth it
has created through the years it was under development. I expected a by far
scarier atmosphere than I experienced, or I´ve simply watched too many
good horror movies through the years.
On Page 4 of this write-up, where I tested various synthetic applications,
3dmark03´s GameTest2 / Battle of Proxycon and PowerVR´s FableMark
are two of the applications that have at least some similarities with Doom3´s
rendering method. Despite the differences between those and Doom3, I´d
say that both proved to be (in relative terms) adequate indications.
All tests conducted in high quality mode, which includes 8x sample anisotropic
filtering. I used the game´s own demo1.
1024*768*32 = 57,8 fps
1280*1024*32 = 54,9 fps
1600*1200*32 = 45,8 fps
It doesn´t come as a surprise that the system is still CPU bound, even
with a XP3000+.
1024*768*32 = _57,8 fps
1280*1024*32 = 48,6 fps
1600*1200*32 = 37,8 fps
In low resolutions 2xAA comes essentially for free, yet here probably the CPU
limitation is playing a role too. Performance penalties start to kick in
at 1280 though, to reach 17% in 1600. The game here is playable up to 1280*1024.
1024*768*32 = 50,2 fps
1280*1024*32 = 36,3 fps
1600*1200*32 = slideshow
Of course only 1024 is playable with 4xAA. In 1280 the performance drop equals
~34%. 1600 with 4xAA is one of the cases that seem to puzzle me for quite
some time now. Granted I shouldn´t expect much from a 128MB frame buffer
and 4xAA in 1600, yet the sub-5fps average score, seems extremely low. I
don´t know what could be wrong, especially since I can see close to
20+fps scores in 1600 with 4xAA on other websites/reviews. Either way the
game isn´t and won´t be playable with 4xAA in that resolution.
Let´s have a look how enabling the 4th quad / 6th VS unit affects the
game´s performance (percentages are in comparison to default hardware
configuration and frequencies):
1024*768*32 = 60,2 fps (+4,15%)
1280*1024*32 = 51,3 fps (+5,55%)
1600*1200*32 = 41,0 fps (+8,46%)
1024*768*32 = _61,7 fps (+6,74%)
1280*1024*32 = 55,6 fps (+14,40%)
1600*1200*32 = 46,3 fps (+22,40%)
2xAA / 8x AF (high quality) became playable in 1600*1200. Being able to increase
one resolution in such a demanding game is a very welcome benefit for me.
I got curious to see how 4xAA/8xAF behaves in 1280*1024 now: 43,3 fps (+19,28%).
It´s more or less as playable as 2xAA in 1600; I personally though
prefer the latter case since resolutions can never be high enough on that
specific monitor J.