![[H]ardOCP](/images/blank.gif){kind=small}
- Date:
- Thursday , March 08, 2001
- Author:
- Kyle Bennett
- Editor:
- Sean Quinn
- Google +1
GeForce3 HRAA High Resolution Anti Aliasing
What can the GeForce3 do for you TODAY?
Below are some pictures of what no AA, 2XAA, QCAA, and 4XAA can deliver to you inside of the Quake3 engine. There is no doubt in my mind that QCAA looks better than 2X sampling AA. Also, in most situations that I saw, QCAA looks as good, if not better, than 4X sampling AA. Vid Card reviews are about to have new levels of subjectivity in them for sure. It is going to be hard to tell the difference from QCAA and 4XAA while you are playing a game of Quake3, unless you are "standing around" simply admiring the maps. Keep in mind that the pictures here are JPeGs that have been compressed, which compromises quality to some extent. However, they really did not effect the AA edges you will be looking at. We have magnified these sets of pictures so you could see the edges better.
NoAA - 2XAA - QCAA - 4XAA
The four opened up side by side, but the pixel size is reduced so that you can view it easier.
It is arguable that in the above shots that the QCAA is the best looking. When viewing QCAA, the edge is a bit "softer" as was stated to us by an NVIDIA engineer. This actually gives what I have seen as a better look than 4XAA. Certainly some of this is going to come down to your personal likes and dislikes and will possibly vary from game to game. All of the above shots where taken at 800x600 with 32bit color on the VisionTek GeForce3. All in all, there is no doubt that AA gives you a much better image than without. I wanted to show you some Giants screen shots, but due to some driver conflicts I could not get Giants to run. That is a real bummer too, as I think that game would benefit tremendously from AA. The entire game pretty much takes place outside and the environments are huge; there are tons of jaggies to contend with. In fact, as compared to Quake 3, where the jaggies don't really detract from the gameplay, the jaggies certainly did detract from the game for me while playing Giants. As more and more games come along that utilize "outdoor" environments of huge proportions, AA is going to become much more wanted.
But where would we be without performance with all this beauty? Kinda like the guy in high school that drove the very fast '69 Camaro but the car looked like total ass. On the flip side of the coin, there is not a lot of respect out there for a car that looks great but delivers nothing in terms of horsepower or handling. So what did NVIDIA give us? Do we have power and beauty? Keep in mind that these results are likely to change, as they were taken using very unofficial NEDs, and NVIDIA is still working on drivers as we speak. I would venture to say that what we are showing you here will get even better.
The first benchmark we are going to cover is a synthetic 3DMark/MadOnioin-type of benchmark. This benchmark is done using a proprietary OpenGL engine. That is right, NO DX7 or DX8 being used here. Thought we would lead off with something different? OpenGL was NEVER mentioned in anything I saw that came from NVIDIA during their launch of the GeForce3. Does the GeForce3 support OpenGL? Damn skippy it does, and I have been told by some folks that the OpenGL features that can be programmed using the GeForce3 chipset will blow DX8 out of the water. Now, whether or not there is any truth to that, I have no idea personally, but I am sure those opinions will be forthcoming from the gaming/hardware community.
GLmark HRAA Benchmarks
Vulpine, a company in Germany, is responsible for this benchmark and it can be downloaded from their site here. There is also a pretty good bit about their engine on the site, so I suggest you explore it if you have questions about the engine itself. I have seen the benchmark run numerous times over the past few days and I have to say that I don't think it even comes close to exploiting the abilities that I have been told lie within the GeForce3 chipset. That is not to say that the benchmark does not tax the power of the chipset, it just does not show off all the features that it could. Looks good? Yes. Utilizes all the goodies that the GeForce3 has? No.
Here are a few screen shots of the benchmark running. Purty stuffs. I would like to say thanks to Vulpine for making this freely available to us. The dl is only 57MB if you have the time or a fat pipe.
Benchmark System Specs: VisionTek GeForce2 Ultra or VisionTek GeForce3, 750MHz TBird OCed to 1088MHz (courtesy of PCNut), 128Megs of Mushkin Rev.3 SDRam, ThermoEngine (courtesy of AZZO), 10.5 NEDs.
Benchmark Setup: Both Cards are run with No AA, then we compare the GeForce2 using 2XAA to the GeForce3 using the Quincunx method (QCAA). Advanced or GeForce3 Features ON, High Detail ON, VSync OFF, OpenGL Lighting ON, Vertex Array Range ON, Texture Compression ON, 32bit Textures ON, Fullscreen 32bit. Below is a screen shot of the configuration screen for GLmark.
At 640x480 resolution, we see where the GF2 was slightly outpaced by the GF3 using NO Anti-Aliasing at all. With the GF2 running 2XAA and the GF3 in Quincunx mode we see the GF2 take about a 50% reduction in frame rate while the GeForce3 held its ground losing only a bit over one frame.
Similar pattern at 1024x768, but the GF2 takes an even bigger hit at 2XAA. The GeForce3 loses 9.9 frames to Quincunx Anti-Aliasing. It reasons that if we were actually playing a game on this engine that it would be unplayable on a GeForce2 Ultra if you used 2XAA.
Wow, I cant believe it got a ZERO! No, not really. The GeForce2 will not do 2XAA at such a high resolution due to the way the sampling works. Still worthy of showing off is the 50% faster frame rate that the GeForce3 exhibits under the Vulpine engine when compared to the GeForce2. While the gap percentage is a bit wider this time, the QCAA shows a 12 FPS hit, which equates to roughly a 33% slower score.