Hi all. I am currently developing an open-sourced MMORPG using the Ogre3D engine. We are still early in development so it will be quite some time until we arrive at Beta, and release. This is where you input is needed. Our current polygon estimate for game models is around 3000 (for charadcters). Since this will be a MMO, it will have mainly outdoor scenes with many trees, plants, buildings, etc. Since it will be a long time until we will be at the release stage, I feel that I want to up my poly limit, since by then (2-3 years), the low, standard, and high end gfx cards will be able to handle so much more. I have been looking all over the place (Gamasutra, CgTalk, Google) for an idea of how I am going to make my estimate and have not found anything conclusive yet. I figure with all the developers around, someone might want to point me in the right direction. My plans are to get the main tri-count that a good card can handle in one scene (with normal mapping and shaders), then estimate the amount of items that we will have in any given scene. This is assuming that a good card from today will be the low end card in 2-3 years. Current trends would say I am very safe in this assumption. From there, I will prioritize the types of objects and give the highest priority the highest limits. My main concerns are that we have limited resources, so I do not want to have to redo things too many times. I do not want to go too high, but too low might not be good as well. At least with too low, we can do simple poly dividing (I have ZBrush) to up it if it comes to that. Removing detail seems to be more counter-productive than adding, IMO. Anyone have any input to add or subtract? All types are welcome. Thanks ahead of time.

It really depends on what you forecast your audience to be.
If you're aiming for a release in 2-3 years, aim high due to obvious hardware improvements.

You have to remember that since this is an MMORPG, hundreds of people will probably be viewing the same scene all at once. If you where able to divide maps up into sectors, then that would simplify this issue.

My plan is to aim a little higher, but how high is my real question. To be specific, I guess my question is about how to actually calculate my budget based on the hardware. Where can I find documented polygon counts for cards? Or, where can I find documented budgets of other games?

This will be a seamless world, so dividing is not really my concern. Draw distance is something that can be adjusted, but overall performance is a concern. I do not want to alienate users due to having high minimum specs, however, I do not want to scare them away by having the game look dated. Everyone can remember a game that they tried and then didn't like because the graphics were not up to par.

What is your idea of dividing maps up into sectors? Maybe I am misunderstanding this. Thanks for the input so far!

Just separating maps into viewing areas so that only entities within a certain radius, area, view, etc. are even processed. Once you get to the border of a sector, they are both processed until you are within the area of a single one.

I would look towards benchmark tables for references. Google it.

A tip that I think could offer you some insight is to load as many models with textures into your engine and monitor the frame rates. If you put them in large clumps in the area (30k poly models),40k, 50k and 60k, you could spin around and watch how quickly it loads the (cheaters way out emulation of a cell) Basically what you are wishing to know is what amount of polygons can be displayed with the engine you are using as well as hardware players will have on your screen at any given time with 0 latency. This is a simple way to get a general idea, but do keep in mind that it will be further altered by the other processes the computer will make that will consume your cpu resources (all the other game mechanics). If you start to lag in any way, you probally should reduce the poly count of your character models or scenes down 5% and see if that helps. Right now my project is using ogre also and we are set at 60k scenes, it actually can be raised, but we are keeping it a little lower to leave us room to improve it later. Our FPS are high. I dont know the exact numbers off hand though. Also high scene poly counts can be great, but always try to figure the areas that will most likely be congested with players and try to reduce the scene polys in that area if at all possible.

Having a high poly count is not always the answer to a great game. The model itself being made with extremely low polys can look extremely nice if the proper texture is applied with a neutral shadowing method used on the texture. Most polygons are spent on the head, which is typically the main focus of the players when looking at another player.

To end I would also like to remind you that all model poly counts can be reduced using LOD as well as reducers in many modeling programs to get you right at the limits you seek. We currently have ours set at 3k and it reduces in stages as they player gets farther way.

I hope I helped in some way. If you have further questions, send me an email and I will see what aid I can offer. Good luck.