Linux vrs Windows - FPS

#1 Jason, that's not a very fair compare since Windows XP isn't designed for 64bit processors unless you're using Microsoft's beta of XP64 then yeah Microsoft sucks then. :)

#2 You cant really present results using 2 different system setups... c'mon. Do same system, with different OS...

#3 It's not really meaningful to benchmark graphical apps across different OSes. You're mostly comparing nVidia's linux driver versus nVidia's windows driver.

It'd be more interesting to compare software renderers, I think.

#4 @Nathan both were 32bit OS's
Mepis doesn't support 64 bit yet.
Nor am I using the 64 bit beta software from Microsoft
However I have it, might be neat to compare them all

@Mark I think it is meaningful btw.
Shows you can get better performance on Linux
with lesser hardware. The difference in fps and generation of hardware shows alot for linux. Whether its better drivers or whatnot(Nvidia wrote both drivers), the performance is better.

@Craig, I'll take ya up on that, will post results soon.

#5 Try running a win32 version of Torque compiled with VC++ 6.0, VC++ 7.0 Standard, VC++ 7.0 Pro and GCC

See which one gets the best FPS

#6 Also make sure you're running it with exactly the same pref settings for the visuals...

#7 Actually, the Legends autobuilder builds with GCC, and I build mine wiht VC++ 7.0 standard, and the GCC builds get much better performance. I also get nearly twice the framerate on Linux as opposed to Windows on the same system.

#8 Chris, is this with one of the fixed function pipeline impaired nvidia drivers ? (ie the whole released Forceware line so far)
As Mark said, this points at drivers performance discrepancies, rather than anything else, and Chris's experience seems to confirm that.

#9 It also points at -> COMPILER <- differences

#10 I used to own tribes 2 for Linux, and my brother owned a copy for windows. Well, just for kicks one day, we installed the linux version on my gentoo partition, and the windows version on my xp partition (yes, I dual boot when I need to). I don't remember the exact results, but I know that on the same settings (1280x1024), I used to get a 10-15 fps improvement on linux over windows.

For what ever that info is worth...

#11 Hobbs, I'd also like to point out that I'm using VC++ 7.1 Enterprise Architect, and it builds faster Release builds than the builds from the bot and GCC.

#12 Actually this proves nothing. It is a complete apples to monkeys comparison, there is no empirical or scientific conclusions that derived from your results.

there are way to many variables like compilers, video drivers etc.
to derive anything meaningful from your numbers.

40 frames per second could be just because of the drivers or video cards alone. It could be caused by some incorrect bios setting that introduces wait states on the memory bus, it could be 1 or more of dozens of things. It could even be motherboard related. 333mhz motherboards with 333mzh were out performing the first generation 400mhz motherboards with 400mhz ram because of bio issues.

The GeForceFX cards can potentially score lower on older non-optomized for FX games because of the drivers and hardware architecture.

#13 Apples to Monkeys ya say?
Here's an expensive ferrari, its way out of tune, doesn't accelerate fast, and causes you to crash alot.

Here's a free volkswagon beetle, all suped up, runs way fast, hardly ever crashes.

They both are made up of different parts but use the same gas,
which would you pick?

Hrm wonder is there is a Linux distro called Beetle.....kinda catchy

#14 I agree with Jarrod. Back "in the day" I wrote a little wire frame 3D "engine" and compiled a demo with it. It ran at 60FPS on my machine. I took the exe to another machine from the same manufacture, but was x2 my speed with a better vid card and the same OS. For some reason the code ran at 1/2 the speed on my friends machine (less than 30fps)

In short as Jarrod states "there are way to many variables... to derive anything meaningful from your numbers." So I can't tell from your data that Microsoft is the Ferrari you speak of or the Volkswagon. A more scientific approach would need to be done to proof this wasn't just a fluke with the users setup.

Although some may already "know" the answer and I always question results from those who are sure of the outcome before even looking at the data or even before designing the tests. Its like asking a parent if they think their child is smart. The answer is almost always "Yes the smartest and of course I have the proof!"

I would like to see this sort of testing though. Would be interesting. Wonder if any of the GG games that go across platforms have data on "frame rate" or other running complants and the machine type and OS.

#15 Well Jason you don't understand anything about how computers and software works and how benchmarking in general is rarely scientific or even useful. if you really think this way, actually, your example with the cars proves my point.

If you REALLY want to understand benchmarking go read the articles on Tom's Hardware on how benchmarking is pretty much apples to monkeys.

#16

Quote:Well Jason you don't understand anything about how computers and software works

@Jarrod LOL, I find it funny you need to insult someone to get try and get your point across. You should really use some more tact, it tends to invalidate anything else you say.

I have a sound understanding of computers regardless of what you believe.

I'm just saying you walk into my computer room and there is a linux box and a windows box, playing on the linux box, there is better performance. It has lesser hardware.

Yes you can continue to argue well if this driver was better and if that compiler did this and so on, well they don't, the specs on the windows box hardware should blow away the linux box's hardware. It doesn't.

I'm in the process of wiping and installing os's and rerunning the tests on the same machine, maybe that will be more scientific for you. You should have seen that above unless you just posted to argue and didn't bother reading.

@Dan, Both demo's were compiled for the platforms they were run on. You should have been able to see the the difference in video cards. With all things equal(which they are not) the faster video card should have edge over the older video card)
Linux isn't my child, I actually use windows alot more because the types of games I play are more numerous on windows. So my "child" in this case is windows. And yes I use to think my child was smarter, but damn was I suprised.

I'll have those tests done later today, and will post them shortly after.

I invite others to do the test as well. It will be interesting for sure.

#17 I downloaded the demo for Windows and Linux on the same machine (set up as dual boot) and I saw interesting results as well. I have a P4 2GHz, 1GB DDR RAM, GeForce 3 128MB DDR. I use Windows XP and Red Hat distro 9. I found a FPS difference of about 25 fps, linux having the upper hand. So, as you can see, using the EXACT SAME HARDWARE, there was a definate difference between the OSes.

-Barzahd

#18 Jason, I didn't say anything insulting, I simpley stated an observation that it is obvious that you don't understand the fundemental concepts behind what you are claiming therefore you don't understand why you are wrong.

You even quoted me incompletely to make my statement out of context.

You "comparison" is EMPERICALLY flawed. The fact that you insist that running against different hardware is some how a controlled enviornment proves that even further.

Travis, on the other hand has more controlled comparision.
But non scientfic, it still does not scientifically PROVE that the Operation System is the cause of the descrepancy. The driver alone could be causing the problem. And there is absolutely no way to run the windows driver on linux or the linux driver on windows.

If you have more than one variable you can't actually do any benchmarking that actually means anything.

You using completely differnt hardware and all the other proves nothing about anything other.

Here is an example.

Me and a friend have the EXACT same MODEL ATI Radeon 9500 non-pro graphics cards.

With the same exact drivers installed and the same exact 3DMark appliation installed. He gets about 10$ - 15% better scores in the multi-texture test.

Here is the kicker he is running an old ass 1G Athlon with a 100mhz front side bus and PC150 memory.

I am running an Athlon 2500 333 fsb with 1GB of DDR400 Dual Channel ram.

According to your testing methodology you would derive that his machine is better.

All it proves is the 3DMark multi-texture test is not affected by the CPU or main MEMORY his PARTICULAR card executes that particular code faster.

#19 Jeez. Jason didn't say it was some incredible, scientific benchmarking experiment. It was a casual observation that a Linux box with much lower specification hardware appeared to be significantly kicking a WIN machine's butt.

This is cool material for a website though. It would be interesting to see how different compilers stack up against each other, or how the different drivers affect performance. This type of benchmarking would be great for indie developers and is much different than the typical benchmarking that takes place on HardOCP or Tom's Hardware.

Some Torque specific banchmarks would help the community find performance holes in the engine. For instance, what type of machine and card on a Mac would be needed to equal the performance of a basic Windows machine? Is there really a difference in preformance of identical systems using LIN or WIN?

Maybe the benchmark is how many Orcs can be displayed at full LOD and still maintain 30 FPS in the sales demo on each of the systems. That could help answer the continual question of how many polys should I put in my character? It could help find performance differences in the terrain system across platforms, etc.

-Jeff Tunnell GG

#20 K Jarrod, no harm no foul.

Lets do some testing and compare.

Some have already and the trend is following what I experienced.
This is gonna be fun!