[Why Buy a Vérité] [Why FPS Doesn't Matter]
Recommendations
Turn back now if you have no wish to listen to me ramble. On this page you will find out what I really think about Rendition chips and Quake. Of course, these are only my opinions, so why not just consider them for their entertainment value?
Why Buy a Vérité |
There are a lot of people engaged in the war of "my video card is better than your video card." Pretty childish, actually. In fact, I believe that there is no such thing as a "best" video card. As soon as you declare something the "best", I have to ask, "Best for what? And for whom?" Every one has different needs and different resources. Those factors have to taken into consideration to determine what is best for you.
I bought a Sierra Screamin' 3D back when Rendition was the only game in town for accelerated Quake. I bought a Monster 3D when GLQuakeWorld came around and Voodoo Graphics was the only choice. For me, those were easy decisions. Since then, I bought a Diamond Stealth II when I needed a new 2D card to power the larger monitor I purchased. I quickly exchanged it for a Hercules Thriller 3D with the faster RAMDAC and higher refresh rates. At that point, I found the Monster 3D to be redundant, so I sold it. Then came Voodoo2. I had the money in my computer account and 800x600 at super speeds just seemed too hard to pass up. I've loved every one of those video cards, but each of them had serious limitations. And I certainly would not recommend them to everyone.
I would recommend the Rendition v2x00 cards to people with the following needs:
1) Top visual quality
2) Good-excellent 3D performance
3) Decent 2D performance
4) Money considerations. Remember, unless you already have a good 2D card, a 3D-only card can become very expensive when you add in the cost of a 2D card.
5) Good GL gaming capability
6) The ability to run full OpenGL apps (in NT, soon in 95)
7) Low- to mid-range systems (Socket 7)
I would not recommend the Rendition v2x00 cards to people with the following concerns:
1) Already possess an excellent 2D card and need only 3D-acceleration
2) Desire the best possible 3D performance
3) Want best degree of game compatibility (GLIDE is everywhere)
4) And perhaps owners of fast Pentium-II systems (the v2x00 doesn't "scale" as well as some other chipsets and the AGP functionality is limited).
I would also no longer recommend a Voodoo1 card to most people. It requires an additional PCI slot, still costs at least $135 mail-order and usually $179 or more retail, and provides performance barely distinguishable from a Hercules Thriller 3D. It also cannot do 3D in a window, which will become a more important concern for people, even people who care only for games. It is also limited to 640x480 for almost all games.
Although I have a Voodoo2, I would also not recommend that card to everyone. For one thing, it cost me $299! That's not pocket change. If one were building a new system, add another $100-$250 for a 2D video card and the cost becomes even more considerable. Like Voodoo1, it cannot do 3D in a window and eats up an addition PCI slot (which are scarce in my system). But Voodoo2 is great for people with a wad of expendable cash and the desire for the fastest performance available to the consumer.
All things considered, there are many good reasons to own a Vérité card. It is a good card and certainly adequate for the current generation of GL games. For the price, it's possibly untouchable (a Stealth II can be had for $99, which becomes $79 after a $20 rebate). When I build my Pentium II (or AMD K6-3D?) system, my Thriller will make the migration. However, I eagerly await the announcement of Rendition's v3300...
Why FPS Doesn't Matter |
People are benchmark-crazy. They can get so focussed on benchmarks that they lose sight of what matters - playability. Don't get me wrong, benchmarks like Quake timedemos are very useful, but they are not the last word in the value of a particular video accelerator.
One thing that benchmarks obviously do not indicate is visual quality. Some areas of visual quality are obvious, even objective. The lack of full colored lighting (src*dest blending) in the PowerVR is a good example. RIVA 128 is limited to square textures and its dithering algorithm is sub-par, even with the newest drivers. Voodoo1's filtering is over-the-top and leaves textures too blurry. Rendition v1000 and v2x00 cards lack per-pixel mip-mapping, which can result in texture "popping". All things considered, the Rendition cards have excellent visual quality, superior even to Voodoo1 in my opinion. Voodoo2 is another subject.
But some areas of visual quality are far more subjective. In the end, you really have to see it with your own eyes in order to judge. Screen shots can be helpful, but still shots do not truly indicate what the game will look like in full motion. That's where FPS (frames-per-second) comes into play.
"Hey, wait a minute," you say. "Didn't you just say FPS doesn't matter?" Good question. The current benchmarks test average FPS. But average FPS is not necessarily a good indicator of playability. Consider this scenario. Video card X can generate a timedemo score of 50fps in Quake. Card Y can muster only 30fps. Image quality is equivalent. Card X must be better, right? Not necessarily. What if card X achieves 50fps average by accelerating to 100fps in areas with little activity, but chokes down to 5fps when shots are fired? What if card Y, which only generates a 30fps benchmark, never slows down below 20fps? Suddenly card Y seems like a much better option. In other words, no one cares how fast your FPS is while staring at a wall. What matters is what frame rate your card can sustain under stress. Currently, no benchmark will tell you that. You will have to rely on your eyes.
Of course it is likely that a card generating a higher benchmark will sustain higher minimum frame rates, but that is not necessarily true. So rather than simply trusting timedemos, I recommend running a demo like Bigass1 for Quake1 or Massive1 for Quake2. But instead of using timedemo, use playdemo to watch the demo at normal speed. If under those conditions, your display remains smooth, then your video card is certainly sufficient, regardless of what your timedemo score is.
Many people have already experienced this phenomenon just in upgrading to Quake2. In Quake1 a person might have gotten 25fps in demo2 for instance, but in Quake2, perhaps the timedemo score drops to 17. But for some reason, the person notices that it looks smoother in Quake2. Bigass1 in Quake1 generates a higher timedemo score than Massive1 in Quake2 on my system (about 4fps difference). Yet Massive1 in Quake2 seems much smoother to me, almost devoid of the choppiness that is constant throughout Bigass1.
I have no way to verify this, but I believe that Quake2 has been optimized in ways that prevent many of the visual stalls that occurred in Quake1. So despite the fact that Quake2 requires a lot more computational and rendering power, it ends up seeming smoother to the eye. In other words, FPS doesn't matter. Well, sort of. What would be really helpful is a timedemo that also reports back the maximum delay between screen updates.