No_Style

Since the release of the first official Doom 3 benchmarks shortly before the games release, the gaming and video card enthusiast community has been awash with what has basically amounted to a post mortem... The question on everybody's lips - What's wrong with ATIs Doom 3 performance? The answers for this have been far flung, from detailed technical evaluations through to tinfoil hat-worthy conspiracy theories.

One thing is for sure though - The board which suffered the worst in these benchmarks was the Radeon X800 Pro. Despite the X800 XT PEs defeats at the hands of the GeForce 6800GT (and in some circumstances the 'vanilla' 6800), performance was at least at a reasonable level across the board. The same couldn't really be said of the X800 Pro - As we saw in our recently published Radeon X800 Pro review, although performance at 1024x768 was good enough even with anti-aliasing enabled, anything much above that became a struggle for the card.

Since then, two things have come to light. Firstly, ATI have taken the bold step of releasing a set of 'Doom 3 drivers', in the form of a beta release of their Catalyst 4.9 set. Secondly, we saw an unofficial tweak to some of Doom 3s code see the light of day, courtesy of the ATI employee 'Humus'. After a vast amount of discussion regarding the said tweak, it seemed to come to light that the performance improvements seen were in some way tied to ATIs anisotropic filtering, and more precisely how it behaved when forced through the driver control panel compared to when it was set by the application.

So, with this in mind, this article sets out to investigate several things. Firstly, the relative performance of Doom 3 using the latest current WHQL ATI drivers (Catalyst 4.7) against the beta Catalyst 4.9 drivers, to see what improvements have been made. Secondly, we'll be examining and discussing the performance delta between using application and driver control panel anisotropic filtering in the game, as well as taking a quick peek at image quality with both settings. Finally, we'll be testing Humus' tweak to the Doom 3 code to see what it has to offer from a performance perspective.
_______________
Read More/Source: Elite Bastards

__________________

BWX

Ok, great idea: try to figure out the deal with the game/ driver/ graphics card/ IQ settings, etc...

"All testing seen here has been run on the following system:

- AMD AthlonXP 2500+
- 512Mb PC3500 DDR-RAM
- ABIT NF7-S v2.0 nForce 2 motherboard
- IBM 40Gb hard drive
- Pioneer 16x DVD-ROM
- 350W power supply"

Again I have to say just like a couple reviews I saw a few weeks ago, what the heck is wrong with these people? You can't get accurate results
benchmarking new games like DoomIII, Farcry, Etc with 512 mb of ram!

As soon as I saw that, I know the rest of the article is flawed, so probably not worth reading... What is wrong with people? Don't they know new games require 1024mb of ram to obtain consistent, accurate results?

I know how D3 runs with 512mb of ram, it runs like crap....

I guess the article kind if serves it's purpose, but would have been more accurate with more ram...

__________________

No_Style

While benchmarking, all that really matters is that it's consistent all the way through. They were comparing drivers and as long as they used the same amount RAM throughout the tests then the results will be correct. The only thing more RAM would do is provide higher scores for both driver results...

As for consistency, I don't think that's an issue if you repeat the benchmark at least 3 times.

It really depends on what you're benchmarking for.. if you're trying to maximize the game's performance, then yeah I would agree with you. But if you're just trying to compare driver differences.. it's not an issue.

My bro is playing with a 9500 PRO with 512 MB RAM @ 800 x 600 with medium settings. It's very much playable, but more RAM would help.

__________________

Vampyromaniac

not according to Anandtech, they found 512, 1024, 2048 all the same if I remember right

__________________

BWX

It ran like crap on my rig with 512mb, so does Farcry. I still think they should use 1 gig of ram while doing any benchmarking that involves Doom III or Farcry. Especially when trying to compare drivers and graphical settings...

I mean even with 1 gig ram you get a faster bench score after running through it once (timedemo 1) the second time... If you use the command to show mem activity- - set com_showMemoryUsage "1" - you can see that it uses much more than 512mb of ram, on the timedemo, so they must get a lot of disk activity. I'd just like to see the same test with more ram..

__________________

Vampyromaniac

here's the comparison I saw: http://www.anandtech.com/guides/showdoc.aspx?i=2156&p=2

__________________

BWX

Yeah, that's the time demo- which doesn't calculate much of what is calculated in-game, but still, Take 512mb out of your machine (assuming you have 1gig) and run the game, it's not pretty.

(like they did at E. bastard's) to compare drivers only gives you an idea of what the real performance would be anyway, because of the fact it doesn't calculate everything the real game does.. It's too bad someone can't figure out a way to record a demo that when played back does all the calculations the game does while it is running...

__________________

Vampyromaniac

I think what we need is not a demo, which just says "render this", but rather a full sequence that just provides a standard set of inputs and lets things "just happen", as it were. dunno if that even makes sense . . . oh dear, this thinking has turned me crosseyed again

__________________

The_Neon_Cowboy

I have an I am

both games run fine

__________________

pr0digal jenius

do'nt people also whine about nforce chipsets and ATi cards too???

BWX

No it makes perfect sense, that is what you need t0o really see what's going on.

Yeah after you turn all the IQ off and at low resolution.. w/o a gig of ram both games hitched and stuttered on my rig so I stopped playing them until I got another 512mb of ram. Now they run silky smooth.

__________________