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

[bmw]

Out of curiosity I did a vertice count in the Facility Quad MP area (since I knew that this particular area suffers from slowdown in 4p mode) and came up with a count almost exactly at 3,000. Now this area is composed of only 4 rooms and generally speaking all 4 rooms are loaded and visible at all times, making the vertice count in 4p mode around 12,000. And that is bare with nothing else loaded into the stage.

Is there a general rule for total number of vertices visible at once before framerates begin to suffer? I'm guessing maybe 2500 total (or 10,000 in 4p mode) ?

[SubDrag]

I came to a number of around 3000 triangles total in my testing for bgfile. So in 4P, 3000 triangles / 4. It is difficult to get an exact number though. I guess in terms of vertices, it's 2-3x that number approximately.

[bmw]

Ok that's approximately where I was at. That tool you put in the editor for vertice count certainly is handy.

[Wreck]

I actually heavily optimized the Quad in my unreleased map pack update. Got rid of the pipes, extended some railings, removed decals, and even straightened the angled walls (which fixes some grenade travel issues). It does seem to run better, but it still isn't the best performance you'll ever encounter. I really do need to finish up that map pack...
_________________

YOUTUBE | TWITTER/X | FACEBOOK | VAULT | MOD DB | RHDN

[bmw]

Wreck wrote:

I actually heavily optimized the Quad in my unreleased map pack update. Got rid of the pipes, extended some railings, removed decals, and even straightened the angled walls (which fixes some grenade travel issues). It does seem to run better, but it still isn't the best performance you'll ever encounter. I really do need to finish up that map pack...

http://www.shootersforever.com/forums_message_boards/viewtopic.php?t=4102&postdays=0&postorder=asc&highlight=blackouts&start=0

Wow my memory must be really bad. I just went through and read every single post in that thread. We apparently had that very discussion, and I don't remember having it. Anyways, LOTS of valuable information in that thread. And now that the editor has features which replace the need for hex editing, some of those maps may be worth revisiting.

EDIT - out of curiosity I did a quick count on the number of polygons in the streets without any objects or object-type rooms loaded. Total number of primaries - 5435 Then throw in several hundred secondaries and this map has close to 6000 triangles. Its no wonder this thing bogs down in MP mode - with 4 players you could have upwards of 25,000+ triangles drawn at once with weapons and vehicles loaded.

[zoinkity]

There's a nuance too as to how they're loaded. Some combiner modes require more time for computation, for instance. Doing alpha converge on, say, 3000 tris might be worse than as-is with 5000. That said, the RSP is really freaking fast.

Then there's time just to load a room. The more texture calls that are used the more data has to be copied to expand the commands, and the smaller that data is the faster that goes. Sometimes it just pays to split rooms.

This is part of the reason why the basement room across the hall from where you start in Archives is split into so many little sections. It minimizes slow copying just to get the room loaded. There also aren't too many overlays to blend. That said, you're probably dropping more frames down there than anywhere.
_________________
(\_/) Beware
(O.o) ze
(> <) Hoppentruppen!

[bmw]

zoinkity - quick question for you about portals (I'm not sure if this specific question has ever been asked before or not) - Does the size of a portal have any effect on framerates? In other words, lets say you have a room with a lot of polygons but it is only visibly connected to another room through a very small portal - will this scenario run faster than if the portal were larger? Or is simply having any part of the room visible all that matters?

[zoinkity]

Good question.
That depends entirely on the selection method used to clip what's behind the portal, so minimum it requires poly calculation and the math to determine if it's within the clip area, but no texturing etc.
It would be somewhat faster then regardless, simply because you're not rendering something. That could be a marginal difference for all I know, but that ratio could make a difference with a large area.

Best bet to determine actual times is to look at Count when the RSP starts, then look at Count when it sends the complete message. Polling would be the only absolute method.
_________________
(\_/) Beware
(O.o) ze
(> <) Hoppentruppen!