Unreal 2 Demo Benchmarks

[HendrixFan]

While looking at Anand's review of the new Geforce 4, the benchmarks seemed a little off on the Unreal 2 Demo from the article just a little while back. I understand that the tests use different builds, but some of the increases seem extreme.

In the older review with build 848, we get 1024X768 32bits with:

Radeon 8500 - 51.7 fps
GF3 Ti500 - 48.6 fps
GF3 - 41.0 fps

Found here

With the review today, using build 856, at 1024X768 32bits we have:

GF4 Ti4600 - 85.6 fps
GF3 Ti500 - 65.3 fps
GF3 - 62.8 fps
Radeon 8500 - 58.7 fps

Found here

So the GF3 goes up 22 fps, over a 50% performance boost. The Ti500 goes up 16 fps, about a 33% performance boost. Anand explains that the newer Radeon drivers decrease performance in the benchmark, but the Radeon went up 7 fps.

I cant imagine the Unreal engine going through such a huge optimization for the GF3 in just a few builds, especially so close to the release date of Unreal 2, and I dont think card specific optimizations could result in a 50% boost in performance, unless there were absolutely no optimizations in the older build. And why the Radeon takes a performance "hit" with the new drivers but still gets higher fps, I dont know.

Ive read other talk about the GF3 Ti500 showing up with nearly 100% performance gains over the GF3, as seen in the Unreal 2 Quincuxx, and the Unreal 2 Anisotropic Filtering test.

Is it the benchmark test (Unreal 2 engine) thats screwy, or just some other issue?

 

[BD231]

Those are some pretty big jumps but im sure your over looking something, I have no clue what though, maybe the new 27.30 drivers?. Anand said that the new Radeon drivers took a hit in VillageMark performance but he dident say anything a performance hit in any other app.

 

[BrunoPuntzJones]

interesting, 50% increase in a week, sounds good to me

i assume it's the exact same demo, just different builds. now that you point it out, and that it's only been a week, that seems kinda high. unless of course the 848 is 2 months old and the 856 was just released.

 

[Def]

I think the demo is actually different. As in the difference between the Quake 3 1.00 demo and 1.03 demo. Same game engine, different FPS for each card because you are rendering a totally different scene.

The important thing to look at is the relative standings. Therefore the 8500's performance hit puts it below a regular GF3(which is pretty bad IMO considering the clock speed advantage it has).

 

[Mavrick007]

The new Radeon 8500 drivers are supposed to make it more stable, but takes a hit in speed. I've seen other benches in the comparisons with the Gf4 on other games and it does seem to occur that now the 8500 is slower. If the build reports such a big improvement in the other Gf cards, then it probably would make an improvement with the 8500 as well, just not as much.

This could very well be a good example of how different builds can implement features supporting the cards out now. Afterall, we haven't seen any finished product or even a beta demo. The real test is going to be somewhere in the not so distant future I hope when I see a test demo. Sometimes a few tweaks is all your program needs to make it more effecient, as I've seen in my own programs. Who's to say that there's not like 8 builds in between.