 |
|
|
|
Kingpin benchmarks, Technical
[Aoril 18, 1999]
Here are more benchmarks for Kingpin. It is basically a continuation of the benchmarks posted in my previous article. The machine stats are the same as before (e.g. 128 megs of RAM, PII 400Mhz, etc.). As mentioned in my previous article, the settings are defined as follows:
As you can see, the Voodoo III: 2000 provides a very slight improvement over the Voodoo II: SLI in performance. The Voodoo III's biggest advantage, however, is its ability to provide higher resolutions--if only my monitor can keep up with such resolutions. Looks like my next upgrade will not be a graphics card, but a new monitor. Boy, that will be very, very, very expensive. And, of course, the "sync every frame" and "fog" has absolutely no effect on the 3dfx cards using the miniGL. Someone suggested that the "sync" issue maybe related to the high CPU requirements of Kingpin. I pre-ordered the TNT2. That is the only card coming out in the near future that has enough horsepower to provide performance numbers that is fairly close or higher than the Voodoo II: SLI and Voodoo III:2000. If CPU is the cause for the "sync" issue, then we should see a very similar effect with the TNT2. The TNT results seem to suggest that this maybe the case. Disabling "sync every frame" only has a small impact on the TNT (~5%). It will probably have even a much smaller impact on the TNT2. As for the fog "issue", it would appear that I have to keep repeating this until it actually registers into the zealot's brain. The miniGL does not have any OpenGL fog functions. That is why there is no performance effect with fog being disabled or enabled. OpenGL compliants cards such as the ATi Rage Fury and the Creative Labs Graphics Blaster TNT both support OpenGL fog. Therefore, it stands to reason that enabling fog will hurt the performance numbers in comparison to the 3dfx cards using the miniGL. And you can see that the performance issue with fog is pretty standard--there is a very significant performance penality in using fog. Since the miniGL does not support OpenGL fog, it will never ever see a performance penalty from OpenGL fog. Yet, we still will see zealots proclaiming that such and such a card is the fastest card based on unfair comparisons. Here's something easy to understand: if you want to see which cards are the fastest card, then you have ALL the cards be treated the same. So if you have two OpenGL ICD cards, then all the other cards MUST ALSO be OpenGL ICD compliant cards. Moreover, you make sure that all the settings are the same for all of the cards. This means that if fog is enabled for one card, it must be enabled for all. The 3dfx cards do not support OpenGL fog. Duh! Therefore, it is NEVER enabled for these 3dfx cards. Duh! I see people saying that fog was enabled for their Voodoo II/III benchmarks as if it actually means something. It does not. If you want to see what kind of frame rates you'll get in a particular game or a class of games (e.g. Unreal-engine games, Quake II-engine games), then you post up a table of frame rates like the one above. It is not a table to see which card is the fastest card in the world, it is a table to see what kind of frame rates you'll get with certain settings and with all the advantages you can give that particular card (e.g. using a "miniGL" instead of a full OpenGL ICD, using OpenGL instead of Direct3D, using Glide instead of OpenGL, etc., etc.).
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
