mpswaim:

You're ignoring where I said that they had to go to the software raterizer to get a measurable difference in performance. If you ran both versions with the OpenGL engine, the performance was essentially identical.

So then that makes the platform (native vs. managed) even less important. My point was simply that perception is what is important. If your players are happy with the speed, you should be happy with the speed.

Regardless, there is a performance cost to managed code, just like there is a performance cost to not using hand tweaked assembly. It's a question of trade offs. Is the benefits in the tools worth the performance cost for the app? If you're not pushing the envelope, then the answer is probably, "yes."

Agreed.

Nick Gravelyn:

Either way I still hold that if testers don't notice a difference and you don't release the native version for public comparison, that 15% hit is negligable. Even Vertigo Software said that the garbage collector and move to .NET in general (at least from my perspective while reading the white paper) was a good move. And as Andy mentioned above, this port was done years ago so who knows how much (if any) of the data is still relevant.

It really depends on what you're doing. If you're doing a lot of stuff with lots of polygons, and modeling object interactions, you'll almost certainly get better performance with native code and a really good C++ compiler. For most games, it shouldn't matter, though. (In other words, it won't matter until it does. Then it'll really matter. :) )

I have an exciting story about this.

I was on the train recently, reading a book about shaders, and a man asked me if I worked in the game industry. We chatted a while.

I explained that I used XNA, and had no C++ experiance, we talked about managed code, he mentioned that his studio was looking for developers.

He gave me his email address, and I later on sent him a link to my game....

... Later on he sent me an email saying I should send my CV and possibly come for an interview.

I will be pleasently suprised if I come out of the interview alive :)

Anyway, I just thought I would share the story as I think it shows one of the benefits of XNA.

It allows you to focus on making a game, you get to do all the bits you want to learn or explore(3D, UI, physics, AI, gameplay, Networking), for other bits you can just use a sample, or library.

All of those bits are transferable knowledge.

I am not saying that learning XNA is a better route in, just that it appears that it may help you poke your nose round a corner.

(If you are a student, you probably have the time to do both anyhoo.)

J.

XNA let me learn my 3D programming very very quickly compared to C++.

A game studio here in Colorado told a group of 10 students at our school (I was one of them), that we had 3 months until we had a date to demo a game to them. Noone in the group had even really started on 3D programming, but we came up with this. We had no money, and no free engine we could think of would work for us for free, so we decided to try XNA. Thank god we did, there is no way we could've come up with a game like this, with no money. They were impressed we did this from scratch with only the XNA API.

I now have a job in the industry. XNA FTW. nuff said.

Joopsy:

I was on the train recently, reading a book about shaders, and a man asked me if I worked in the game industry. We chatted a while.

Finally decided to learn shaders? :P

Anyway, on the discusison at hand, I'll also add one great thing about XNA: It made my life learning shaders very very easy. Compared to the amount of code it takes to set up a shader in C++/DX, XNA is a breeze.

mpswaim:

Nick Gravelyn:

See that quote wasn't accurately represented above. The fact that the performance of the port, even in 2003, was not noticed by the testers means that the performance hit was negligable. If it runs 15% slower and nobody cares, does it really matter that it runs slower?

You're ignoring where I said that they had to go to the software raterizer to get a measurable difference in performance. If you ran both versions with the OpenGL engine, the performance was essentially identical.

That does not mean much of anything when you're looking at a game whose purpose in life is to be multi-player eye candy.  What id game has not been completely bottlenecked by graphics performance (either hardware or software)?  When you say the OpenGL versions of both run identically, you're not measuring the relative performance of general C# code.  Instead, you're measuring the speed difference between native C calls and P/Invoke calls from C# to C.  The bulk of the work is being done on the GPU.

mpswaim:

Regardless, there is a performance cost to managed code, just like there is a performance cost to not using hand tweaked assembly. It's a question of trade offs. Is the benefits in the tools worth the performance cost for the app? If you're not pushing the envelope, then the answer is probably, "yes."

I wish more people agreed with you.  This is probably the most useful statement made in this thread.  People can look up and skew data in any way they like to make their point, but at the end of the day most are just stating opinions and misinterpretations.

I've read almost all posts in this thread,
even if like some of you say xna can not be a tool used for making industry games, and some other say "coding with xna gives a valuable experience which can be applicable for your job later", i believe xna is playing a great role bringing more game coders to industry. just think about if using cpp and directx was the only starting point for making games, how many coders/hobbyist would ignore their desire.
and i really appreciate teaching games to high school students which there is a thread posted  by mr. brian scarbeu about this here.
can you ever think without such a tool like xna, could it be possible teaching such stuff to 16 year old teenagers?
with xna young can start making simple games just a few months after starting writing simple prorams.
I'm just wondering can microsoft make a programming language and a framework even much simpler than C# and XNA one day, so future gamers can start their journey when they're 12.
if simplicity brings less power or make things slower, it gives more people and younger people a welcome which returns higher talent and capacity

comboy:

I've read almost all posts in this thread,
even if like some of you say xna can not be a tool used for making industry games, and some other say "coding with xna gives a valuable experience which can be applicable for your job later", i believe xna is playing a great role bringing more game coders to industry. just think about if using cpp and directx was the only starting point for making games, how many coders/hobbyist would ignore their desire.
and i really appreciate teaching games to high school students which there is a thread posted  by mr. brian scarbeu about this here.
can you ever think without such a tool like xna, could it be possible teaching such stuff to 16 year old teenagers?
with xna young can start making simple games just a few months after starting writing simple prorams.

I mostly agree, but to be fair that's not the point we're arguing in this thread.  I do not think anyone is arguing that XNA GS has no place in game development, especially academic learning.  The OP was asking why XNA GS development has not been brought into mainstream, professional game development.  That is a completely different topic.

comboy:

I'm just wondering can microsoft make a programming language and a framework even much simpler than C# and XNA one day, so future gamers can start their journey when they're 12.

Why does it have to be Microsoft?  And why does it need to be a new programming language?  There are already packages out there that allow you to just drag and drop objects into 3D worlds and build quasi-games with no programming required.

comboy:

if simplicity brings less power or make things slower, it gives more people and younger people a welcome which returns higher talent and capacity

Again, this is a good point but no one is arguing against such a point.  Easy-to-use educational tools will always be great to have, even if they sacrifice speed and power in favor of simplicity.  The problem is that current compiler technology is not at a level where we can have the simplicity without losing performance, and that's a large part of the reason why tools such as XNA GS have not penetrated into mainstream game programming.

comboy:

I'm just wondering can microsoft make a programming language and a framework even much simpler than C# and XNA one day, so future gamers can start their journey when they're 12.

Well, they're looking into it. :)

Clicky1
Clicky2
Clicky3

A lot of people say C# is slow.

Vista desktop (WPF C#) is running 10* faster than XP desktop (pure C), my practise XNA program ran 2* faster than an open source C++ engine, ODE engine (C++) run 5* slower than BulletX (C#), see the singularity C# OS etc.

C# is only a language with easier/faster development, not more. C# compiler is very intelligent, but it can’t be thinking instead of you.

Uhm... what?!

Robsoft:

A lot of people say C# is slow.

I never said C# was slow, just that it limits what you can do in some cases.  However, that's more the fault of the runtime than the language itself.  Though if you feel the need to push the point, I will say this:  there is a reason no-one uses C#/Java for high-performance scientific computing on clusters.

Robsoft:

Vista desktop (WPF C#) is running 10* faster than XP desktop (pure C)

State your source.  The entire Vista desktop composition engine is most definitely not a WPF app, nor is it written in C#.  Look at the explorer.exe file in Vista.  Does it have a CLI header?  Nope.  Not to mention XP blows Vista away in desktop responsiveness on my machine (Q6600, 4 GB RAM, 8800 GTX).

Robsoft:

my practise XNA program ran 2* faster than an open source C++ engine

Are you comparing apples to apples here?  Or are you comparing a spinning triangle to a full game?  Is this a graphics-only comparison?  If so, the difference is probably in the shaders, not the CPU code.  You have to support your statements with facts, or at least context, not just make blind claims.

Robsoft:

ODE engine (C++) run 5* slower than BulletX (C#)

Please tell me this is a joke.  Do you have to pick the open-source physics engine with the least active support?  ODE has never been at the top of the list of high-performance physics engines, and it hasn't been updated in what?  2 years?  Use Bullet (the original native version) or Newton for a real comparison.  Or if you feel up to, do the comparison against PhysX or Havok.

Robsoft:

see the singularity C# OS

What does Singularity have to do with anything?  It's a Microsoft Research project to explore operating system concepts.  Have you actually looked at the source?  A lot of the boot code is still in C/ASM, and they're using a special in-house C# compiler (Bartok) for the kernel code that produces native code (an AOT native C# compiler).  How can you possibly make a speed comparison based on this?  What are you even comparing it to?




I know I've come off sounding harsh, but if you're going to make such strong claims, then please do your research and provide supporting evidence.  It helps no-one when people come around spewing statistics without basis.  Even if these statements were to be true, I would still call you out on this.