why do hardware sites benchmarks with q3a and UT

[hans007]

I was thinking there are about twice as many people playing counterstrike online as q3a and UT. Granted UT and q3a are more "advanced" engines, but more people play CS and you never get benchmarks of it except rarely (i think aceshardware does them under halflife engine). Anyways i'd much rather have CS benchmarks because during the rare times that i do play games its CS, not q3a or UT. I mean if i was buying a car I wouldn't want to know how well it drives on a dirt road in africa, i want to see how it handles in los angeles, on the freeway (yeah yeah, it goes 4 mph on the 405). Anyways ANAND if you see this start running CS benchmarks. And why do they run it with MDK2 for that matter? who plays that game?

 

[ahfung]

There are more people working with Word and Excel than any gamers combined. Do we need to benchmark them for that matter? 😉

Who plays MDK2? Good question. There are a lot of them.

 

[xtreme2k]

Benchmarking CS is more like benchmarking GLQuake1. There is really no point since most card will do it at an acceptable speed.

 

[BenSkywalker]

Counterstrike- A) is very CPU dependant, and hence doesn't scale well with video power and B) is horrible for benching to begin with. They break the benching code with nearly every patch, and is simply a major pain to deal with(This is a HL engine in general problem, not CS specific).

Why use Quake3 and UT?

Well Quake3 is pretty much the *perfect* OpenGL bench for gaming right now. It uses many advanced rendering techniques(large textures, real 32bit support, scaleable geometry) and pretty much everything has an impact on FPS in one bench or another(be it 640x480 fastest or 1600x1200UHQ or somewhere in between). Not only that, but the benches are very repeatable and are easily created(Timedemo1 may be a bit lacking, but benches like Q3Crush and Quaver are much more vid card intensive). It does not, however, tilt the scales too much in favor of any one board. For isntance, it doesn't fully take advanatage of hardware T&L, it doesn't have EMBM or Dot3 support, and it doesn't use any volumetric textures or cube mapping. A very good representation of current games.

UT svcks as a bench, the only reason I can think of why reviewers use it is because it is a good looking D3D game. Unfortunately, UT is best running glide on 3dfx boards, OpenGL on Radeon and particularly GF based boards, and best running MetaL on S3 boards. Why it has become such a popular bench is beyond me. It is too CPU dependant to scale well, though Rev's bench helps solve at least some of that. One of the plusses is that it does report the minimum FPS, a nice advantage over Quake3.

"And why do they run it with MDK2 for that matter? who plays that game?"

MDK2 is a very serious contender, if not the flat out winner, for game of the year. If you need to ask that, I have to assume that you haven't played, though I suppose you may have and didn't enjoy the game, anything is possible🙂

MDK2 is a good example of an OpenGL title that is moving in the direction of games that we are going to see with increasing frequency. Besides scaling well with fillrate, it also supports hardware lighting for noticably superior visuals. This trend is starting to become a more common factor. The fact that it is easy to bench, and produces easily reproducable results makes for a good bench.

Evolva I think should replace UT as a D3D bench. It is fillrate intensive at the higher resolutions, CPU intensive at the lower resolutions, and has the option to enable Dot3 or not for the up to date boards that support the feature(nearly all of them released in the last 15 months).

You are looking at this from how many people are playing games on line, which I don't think is a very good indicator. I don't play CS online, over LAN or single player HL is pretty much it, but I have to say that as of late I've been playing a lot more Q3 and UT then HL or CS. None of that will show up in on line statistics, and I know there are many many more people who can say the same. Broadband isn't available to everyone, what people are playing on line isn't always a great indicator of what people are playing in general.

 

[hans007]

ahfung, they do benchmark for word and excel its called winstone and content creation.

 

[ahfung]

Both Winstone and content creation are about as CPU intensive as CS.

 

[Sunner]

I agree with Ben about CS, its WAY too CPU intensive to be used as a vid card benchmark, and someone from Valve even confirmed that the timedemo code is broken, and will produce unreliable results, though this was a very long time ago.

As for Evolva, I dont agree.
Sure, technically its a better benchmark, but really, its the same case as 3DMark, you dont buy a GFX card to do demos, you buy it to play games, and noone plays Evolva(and frankly, for a good reason).

I dont think UT is a very good benchmark either for that matter, but at least its something thats playable(though personally Im not too fond of UT either🙂).

 

[Dulanic]

Also Q3 is such a reliable benchmark. I can run Q3 Demo001 12 times in a row and each one will get the exact same FPS +/- .1FPS.... Ive never had my system vary more than .2 FPS unless I was running other programs.