GoldenEye 007 Nintendo 64 Community, GoldenEye X, Nintendo 64 Games Discussion

[GoldeneyeGuru]

I'm trying (and failing so far) to get the Setup Editor working but I suspect that someone here has already tried the hack that I was thinking of anyway...do we know what the practical limit is of the amount of enemies that the N64 (i.e. not on emulation) can handle at once?

I'm sure you've seen these insane DOOM levels with loads of enemies:

https://www.youtube.com/watch?v=b2Hx8RGxiPI

Has anyone tried putting loads of enemies on say the Runway level and see how many it takes for the console to crash?

[HackBond]

We have rough triangle counts for frame rates. I haven't tried actually flooding the game with guards. For reference, boiler suit trevelyan (the best comparison to a guard with a head I can think of) has 734 triangles.

Quote from Trevor

Quote:

F3D Optimum Throughput (1 Cycle Z-Less) 416 @ 60FPS 4p 555 @ 60FPS 3p 833 @ 60FPS 2p (4p @ 30FPS) 1,666 @ 60FPS (4P @ 15FPS) 3,333 @ 30FPS 6,666 @ 15FPS 14,285 @ 7FPS 25,000 @ 4FPS 50,000 @ 2FPS 100,000 @ 1FPS GE throughput test results (with z-buffer) 1 Cycle 2048 = 20FPS - 3 (2.2) Frames (930 tris/frame) 5056 = 12 fps - 5 (4.2) Frames (1,203 Tris/frame) 11,200 = 6FPS - 9 (8.5) Frames (1,317 Tris/frame) 1 Cycle EnvMap 2048 = 20FPS - 3 (2.2) Frames (930 tris/frame) 5056 = 12 fps - 5 (4.3) Frames (1,175 Tris/Frame) 11,200 = 6FPS - 10 (9.1) Frames (1,230 Tris/frame) 2 Cycle 2048 = 20FPS - 3 (2.5) Frames (819 Tris/Frame) 5056 = 12 fps - 5 (4.5) Frames (1,123 Tris/Frame) 11,200 = 6FPS - 10 (9.2) Frames (1,217 Tris/Frame) 1,200*60 = 72,000 Triangles per second

_________________
Also known as Spyster or Nyxem
[Youtube]

[Decoy] Antenna | Control | Silo | Escape

[Wreck]

Besides the polygons currently being rendered, there are other processes going on with the AI. Enemies can easily start to bog down the performance when a cluster of them are actively on screen. In my personal opinion, I'd rather buff (within reason) the stats on a handful of guards, rather than flood the map with tons of them. I see a lot of enemy spam in mods, in a wide range of games.

[GoldeneyeGuru]

HackBond wrote:

We have rough triangle counts for frame rates. I haven't tried actually flooding the game with guards. For reference, boiler suit trevelyan (the best comparison to a guard with a head I can think of) has 734 triangles. Quote from Trevor Quote: F3D Optimum Throughput (1 Cycle Z-Less) 416 @ 60FPS 4p 555 @ 60FPS 3p 833 @ 60FPS 2p (4p @ 30FPS) 1,666 @ 60FPS (4P @ 15FPS) 3,333 @ 30FPS 6,666 @ 15FPS 14,285 @ 7FPS 25,000 @ 4FPS 50,000 @ 2FPS 100,000 @ 1FPS GE throughput test results (with z-buffer) 1 Cycle 2048 = 20FPS - 3 (2.2) Frames (930 tris/frame) 5056 = 12 fps - 5 (4.2) Frames (1,203 Tris/frame) 11,200 = 6FPS - 9 (8.5) Frames (1,317 Tris/frame) 1 Cycle EnvMap 2048 = 20FPS - 3 (2.2) Frames (930 tris/frame) 5056 = 12 fps - 5 (4.3) Frames (1,175 Tris/Frame) 11,200 = 6FPS - 10 (9.1) Frames (1,230 Tris/frame) 2 Cycle 2048 = 20FPS - 3 (2.5) Frames (819 Tris/Frame) 5056 = 12 fps - 5 (4.5) Frames (1,123 Tris/Frame) 11,200 = 6FPS - 10 (9.2) Frames (1,217 Tris/Frame) 1,200*60 = 72,000 Triangles per second

Ah now that's interesting!

So the optimum approach would be:

1. Find an enemy model with the fewest polygons.

2. Construct a test level that has the fewest polygons possible as well. Basically an empty 'box'.

That could be an interesting test to run IMO.

[GoldeneyeGuru]

Wreck wrote:

Besides the polygons currently being rendered, there are other processes going on with the AI. Enemies can easily start to bog down the performance when a cluster of them are actively on screen. In my personal opinion, I'd rather buff (within reason) the stats on a handful of guards, rather than flood the map with tons of them. I see a lot of enemy spam in mods, in a wide range of games.

Yes absolutely - I suspect the AI overhead is probably as taxing on the system if not moreso than worrying about small differences in polygon count.

I'm trying to think what is the maximum number of enemies I see on screen at once time anywhere in the 'normal' game...Ouroumov's guards at the end of Facility and/or the endless spawning at the end of Facility can result in a higher than normal number of guards on screen but it's probably never even as high as 10 at once.

I seem to remember the 'defend Natalya' section in Control grinds to a halt at times but that's quite a sophisticated level to render I guess once the glass has smashed and you've to render what's behind it as well as the guards.

I suspect that under 'perfect' conditions the N64 would still struggle to render 20 guards never mind 100. I'm sure with emulation we could get some crazy number being rendered.

[Lazlo52]

The calculations involved with guards being alive is definitely the biggest hit to performance, at least as far as I've seen personally. The difference between 18 walking around on screen and 18 dead on screen is a difference of an unplayable framerate and a decently playable framerate.

I think the highest amount of guards you can get vanilla-wise comes from turning on invincibility and trying to corral everyone into the same few rooms. The game tries to prevent that, though, by forcing some to lose track of you.
_________________

Quote:

22 not happening nerds. forget about it. 23 til the day i die.

[zoinkity]

There is a physical limit; the table of guard data (0x1DC each) is a fixed size and includes the player, pretty sure it's on a per-stage basis. I think Facility was 75, with several reserved off the top and bottom. There should be a cap as well based on ID so wouldn't expect over ~250 in ideal (aka brutally hacked) conditions.

Without a path network, one guard in a level with one player will crawl so slowly you'll think the game crashed.


Subdrag accidentally did something like this once by (probably) corrupting the setup during decompression, the infamous TREV ARMY. It flooded Facility with Trevs. It was somewhat recreated based on a savestate near the end of it.
To quote a 10-year-old private email:

Quote:

Should correspond to these codes: 811CFE78 0009 811CFE7A 0413 811CFF58 0009 811CFF5A 0413 811CFF74 0009 811CFF76 0413 811CFFAC 0009 811CFFAE 0413 811CFFC8 0009 811CFFCA 0413 811D0000 0009 811D0002 0413 811D001C 0009 811D001E 0413 811D0054 0009 811D0056 0413 811D0070 0009 811D0072 0413 811D00A8 0009 811D00AA 0413 811D00C4 0009 811D00C6 0413 811D00FC 0009 811D00FE 0413 811D0118 0009 811D011A 0413 811D0150 0009 811D0152 0413 811D016C 0009 811D016E 0413 811D0268 0009 811D026A 0408 811D0284 0009 811D0286 0408 Those last two Trevs will be agro just after the first security door, but the rest of them will be walking toward the bottling room. They're all over the place, so you should run into quite a few along the way. Since they're in the middle of pathing they don't want to interact with you or anyone else, but it's still pretty cool and they do open all the doors for you ;*)

Those codes must be on when the stage loads. They're overwriting the setup file.