Idea: Stand-alone Q3Map2 BENCHMARK!

(Matt) #1

Ydnar… Check this out…

How much time, or how hard do you think it would be for you to create a benchmark using a complex sample map that you or someone else makes (with gazillions of brushes and lights), and make a small gui for q3map2, that’s as simple as, “CLICK TO BENCHMARK” and when the compile is complete, it gives a time, which can be used in benchmarks and graphs in comparisons of systems…

Most people in the hardware industry benchmark the quake 3 engine, and I’m sure they’d use and like this sort of benchmark, as it would be fresh and new, and would really show the number crunching of a given processor.

Do you think it would be possible to do something like this without quake 3 installed on the persons computer so the benchmark would be a stand-alone, without the person having to get the q3demo or buy it?

(system) #2

I had this idea some time ago for RTCW. in RTCw there is no Timedemo-Map like in Q3 to get a proper Benchmark.
I wanted to make a terrain with fur-shading grass and a small village in it, placed in ww2.
I planned the village with an eye on architecture styles.
But i have other things to do then making a non-playable rtcw-map. :wink:
But the idea is quite nice, and it would look great.

Other advantage would be that we have a benchmark and a complete samplemap for shaders and other q3map2 supported stuff.

But ydanr is a fast guy and the map would be really soon outdated. :slight_smile:

(Matt) #3

No, that is the exact opposite of what my idea is.

I’m talking about a q3map2 benchmark, that measures the time it takes to COMPILE an ultra-complex map.

New q3map2BENCH versions can be distributed with NEWER more complex maps to compile, every 6 months or so, so the benchmarks get more complex as the CPUs get faster.

It wouldn’t have to be a real or playable map either. It should be something that fills the entire world, and contains thousands of columns and patches shaped like corkscrews, and as many lights as possible.

(ydnar) #4

It already exists. It’s called Q3M17.

Imagine I’ve compiled that map somewhere in the neighborhood of 5000 times. Down to about 1 minute for a full compile on my primary machine… :slight_smile:

y

(Emon) #5

That’s a very bad idea. If you keep upping the amount of data to be compiled, it’s going to seem like the CPUs in July are the same speed as CPUs in January. Perhaps every 2-3 years would be more appropriate.

The Q3 engine is already getting old and outdated, it’s not really good for benchmarking new hardware, not as good as something that runs on the Unreal Warfare engine (what UT2K3 and UII and others use).

(Emon) #6

That’s a very bad idea. If you keep upping the amount of data to be compiled, it’s going to seem like the CPUs in July are the same speed as CPUs in January. Perhaps every 2-3 years would be more appropriate.

The Q3 engine is already getting old and outdated, it’s not really good for benchmarking new hardware, not as good as something that runs on the Unreal Warfare engine (what UT2K3 and UII and others use).

(Emon) #7

That’s a very bad idea. If you keep upping the amount of data to be compiled, it’s going to seem like the CPUs in July are the same speed as CPUs in January. Perhaps every 2-3 years would be more appropriate.

The Q3 engine is already getting old and outdated, it’s not really good for benchmarking new hardware, not as good as something that runs on the Unreal Warfare engine (what UT2K3 and UII and others use).

(Matt) #8

Thank you for saying that 3 times, genius. Another genius point goes to you for not knowing that there is hardly any BSP process for Unreal2 maps. I map for UT2003, and it takes only a few seconds on the most complex compiles.

Quake 3 is “Old and outdated,” but complex Q3Map2 compiles still take anywhere up to a few hours on my machine, and it uses 99% of the CPU. My system is an AMD AthlonXP 2400+ with 1.5GB DDR ram… I’d like to see how badly it beats an Intel system in such a benchmark. (Since AMD normally beats Intel CPUs even when the AMD CPUs are clocked 1Ghz slower.)

(Emon) #9

Sorry about the triple post, I’m not sure why it happened.

I know that Q3 compiling is a good test for CPU power, I was just pointing out that, in general, Q3 isn’t a good benchmark as some newer games.

(Matt) #10

Yeah it’s hard to say what is better when all the top systems get 400 frames per second in a game that doesn’t do the newest stuff.

I guess the next benchmark games will be Doom 3, Unreal 2, and maybe some of the DirectX 9 games we’ll see after the GeForceFX arrives. For some reason I don’t really consider Unreal Tournament 2003 as a valid benchmark… Something about how my GeForce DDR performs just as well as my friend’s GeForce4 4400, and my other friend’s Radeon 9000, on high settings… That doesn’t seem like correct math to me, but whatever.

Maybe SquareSoft will release something along with nVidia, or perhaps ILM and nVidia? SquareSoft has already released a Final Fantasy 11 benchmark (but it sucks, because it looks like crappy playStation2 graphics.) and nVidia has the actual models used in Jurassic Park, and they said at the press release of the GeForceFX that it can display 100 JP T-Rex’s at 100FPS on the screen at once in real time. Hmmmm.

I still think that big number crunching is where benchmarks are at. Q3Map2 would be marvelous for this if a standalone could be created.