Linux vrs Windows - FPS

#41 Windows XP 64bit edition
K8T Neo Mobo
1000 MB DDR-400 Ram
Nvidia Geforce FX 5950 Ultra

Nvidia driver(53.04)
1024x768 = 166 FPS(MAX) 90(Nominal)
1280x1024 = 166 FPS(MAX) 90(Nominal)

Latest Nvidia drivers(57.30)(March 22nd)
1024x768 = 297 FPS(MAX) 170 FPS(Nominal)
1280x1024 = 297 FPS(MAX) 170 FPS(Nominal)

Wasn't able to push it past 1280x1024.
20" LCD's still kinda expensive.

I have been having trouble installing the Nvidia
drivers for 64bit Suse Linux. So no comparison there.

The 32 bit Suse Linux OS on same hardware resulted in about the same
FPS as the Windows XP 64 bit did before the driver update.
I'll be looking for the new update for the Linux drivers.
Hopefully there will be a increase in performance like I saw
with the new driver for Windows XP 64bit.

#42 Quite a bit late to this discussion, but I would recommend a few things. When doing your benchmarks, if you are getting 100+ FPS on any of the platforms, you are probably not benchmarking enough. The difference between 150fps and 100fps really isnt very much. Load your levels with geometry and try it. When you get down towards the threshold of playability, say around 20-30fps, on one OS, then try it on the other OS on the same hardware. Testing on different hardware is certain to render the results irrelevent for the OS performance as Jarrod has mentioned, even if the older hardware is getting better performance. And if the video hardware has different capabilities, then the results are all screwed, because the game may not even be trying to use the capabilities on the older hardware, thus giving better performance but less visual quality. (I am not sure how much Torque relies on video caps for its pipeline, but id imagine it has some reliance)

Also keep in mind that many Linux distros come with very few services running by default. Windows XP, on the other hand, comes with scores of services running by default. Make sure you go into Admininstrative Tools and disable all of the services that you don't need. Disable your themes, screensavers, messengers, indexing services, etc... unless you have similar items running on your Linux install.

Also, compare the outputs from the profiler and see if some part of the code is using a greater percentage of processing time under one OS than under the other. Then check to see if that part of the code uses OS dependent mechanisms, such as pthreads vs winthreads, or some such.

And finally, keep in mind that the timing mechanisms in Linux are pretty crappy, particularly in multithreaded applications. If you are going to do a performance test, do it by recording the starting time of the first frame rendered, and the ending time of the last frame rendered, and divide the difference by the number of frames processed. Trying to gauge performance by looking at a realtime FPS counter under linux, and comparing it to a realtime FPS counter under Windows, is not going to be very useful.

Just some thoughts.

Peace

#43 What Gerald said.

Also, Jason - you point out a very large difference in FPS between Mandrake and Mepis. Only that proves what mr. Jarrod said even further, since Mandrake and Mepis both share the same core, only different drivers and services running.

#44 I hear ya's. But what I started out to do was casual observation. I used typical installs just to get an idea of what kind of performance I could see. I think disabling everything has some merit, but its not the typical play environment. As developers, most of us know that disabling im's and apps not needed by your game helps the game run better. Most people however aren't power users or developers and would never think to do that or they expect a certain performance level even with the other stuff running. I basically found based on the latest drivers of the time, Linux edged out windows. It had alot to do with drivers. But drivers are software, and each system type has its own software. Make your own conclusions.

I've kinda lost interest in pursuing this anymore really. So much negative feedback and it takes time to do installs and such.
Only 1 person did a comparison and posted it other then me. It is what it is. If you want to know more, then please, run some tests your way and show everyone how it is suppose to be done.

Peace

#45 All of the discussion about benchmarking aside, who cares?

111 fps vs 157 fps? Considering that monitor refresh rates don't go that high and the human eye couldn't tell the difference even if they did, there isn't even a real world difference in those two numbers, so the whole discussion is pretty moot.

A more interesting topic is the fact that there is no viable market for selling Linux games, indie or retail, whereas there is for Windows and MacOS.

#46 On George's note, I'm going to add to the rambling nature of this wonderful thread.

I'd say that as an indie game developer that in the Linux and Mac game market you can get a lot more bang for your development buck because generally the big AAA game developers/producers stay away from these markets so you've got more room to manuver in the market. This could be just because it is not a proven viable market at this time, like George said. You don't have the high noise factor of the big game producers PR dollars that you have in the Windows market. Plus the Linux community is seems to take kindly to indie developers. Probably because the Linux OS basically came from an indie venture.


CatB!

#47 Truly. We sell a lot of our games to mac and linux people, not least of all because no one else is out there making good games for those platforms...

#48 Ben,

I have heard that Linux is the smallest of the 3 markets from GG.

From Jeff T. on think tanks

Quote:
60% Mac sales

Quote:
Linux is still a little bit of a labor of love. It runs from 7-13% of sales. We do it because we think the OS needs more games in order to eventually be successful.

on this thread

That means the PC make at the ver least 100-60-13 or 27% of the think tank sales. Thats more than 2x the linix sales.

Have things changed since last nov or was your point that the linux market is abanded by AAA so thats how you get >10% of your sales?

Still Torque is a great deal! Who cares which sells more. You can ship for all three!

Edit - Now that I thought about it a bit more...
Being maybe (AKA my guess) at most 1 out of 20 home PC are running Linux 13% is a huge market penetration compared to 27% sales on windows.

#49 There is one thing that is being demanded of AAA developers that a lot of them are not able to produce expediently because they ignored Linux during game development. I submit Ubisoft's Ravenshield as a prime example. They promised a Linux dedicated server shortly after shipping, and thus far have not delivered, which if you read their community forums is a substantial cause of ire among their users. For multiplayer games that allow for dedicated servers, people want dedicated Linux servers because a dedicated Linux box can be leased for quite a bit less than the Windows counterpart, and is generally easier to maintain, without licensing restrictions. There are also several companies that offer solely game server hosting on Linux boxes for dirt cheap.

If you are developing a multiplayer game, and can sell enough game copies to the Linux market just to cover the cost of the added Linux effort, it should be worth it. It will give you Linux server support for practically free, which will most likely result in increased popularity, and increased sales in the Windows and Mac markets, even if you don't turn a profit from the Linux client side. That is assuming it is a good enough game that people want to spend $25-30 a month to host a server for it to begin with.

And while the rendering speed may not be of paramount importance (no rendering on a ded server), the physics, game logic, and netcode are pretty important on the server.

Peace

#50 I would guess 1 out of 50 home computers run Linux, so the market penetration is really much much higher. Anyone have real* numbers tho?

(*Numbers by surveys/analyzis corps or whatever)

Edit: And I agree completely with Gerald. I would not have selected Torque hadn't it been multiplatform.

#51 I second that. I dev on Linux and Mac and Windows. I like all three sometimes, I hate them too :) But in the end Torque let's me build game code on all three with very little extra effort.