Software rendering?
by Tomas Dahle · in Torque Game Builder · 03/25/2005 (11:20 am) · 18 replies
I guess this question goes out to melv.
Seeing that I am already plunging head-first into development with this spectacular engine, I am wondering when software rendering will be available. (may I say when, not if? ;))
With great libraries such as swShader (not to mention the commercial pixomatic) I also wonder whether or not T2D will use one of these or roll it's own optimized renderer.
Anyway, software rendering would be absolutely utterly fantastically useful and bring in tons of extra cash for any game I put out so I'm excited to know when this will be possible. :)
Seeing that I am already plunging head-first into development with this spectacular engine, I am wondering when software rendering will be available. (may I say when, not if? ;))
With great libraries such as swShader (not to mention the commercial pixomatic) I also wonder whether or not T2D will use one of these or roll it's own optimized renderer.
Anyway, software rendering would be absolutely utterly fantastically useful and bring in tons of extra cash for any game I put out so I'm excited to know when this will be possible. :)
About the author
#2
heres a pic
Melv had this to say
I'd agree completely.. I ran it at 150 rigid-body tiles, got to 5 fps... it still didnt crash, lol I was expecting it too since I had at least 8 other programs open
03/25/2005 (12:07 pm)
I ran my Physics Demo on a Intel Pentium III 600Mhz... with a silly 32mb intel integrated (peice of crap) video card... and only 256 mb of ramheres a pic
Melv had this to say
Quote:@Matthew: Download is fine. :) I find that pretty cool that a P3-600Mhz can handle 67 rigid-body tiles at 10+ fps. All smiles over here!
I'd agree completely.. I ran it at 150 rigid-body tiles, got to 5 fps... it still didnt crash, lol I was expecting it too since I had at least 8 other programs open
#3
03/25/2005 (1:08 pm)
It is not currently in our plans to have a software renderer. 
#4
03/25/2005 (1:40 pm)
You could, however, extend the renderer to swShader as long as you dynamically linked the library and didn't include any of the T2D source in the swShader modifications. But that's an exercise best left to the reader!
#5
That's becuase you're killing the CPU with rigid-body physics. The graphics card has nothign to do with it. How does it run if those objects are just rendered sprites?
You should be able to get hundreds of 2D sprites on-screen with your card/CPU. Granted, I've never used an Intel "accelerator", but it's got to be as good as a TNT2, right?
03/25/2005 (1:51 pm)
Quote:I'd agree completely.. I ran it at 150 rigid-body tiles, got to 5 fps... it still didnt crash, lol I was expecting it too since I had at least 8 other programs open
That's becuase you're killing the CPU with rigid-body physics. The graphics card has nothign to do with it. How does it run if those objects are just rendered sprites?
You should be able to get hundreds of 2D sprites on-screen with your card/CPU. Granted, I've never used an Intel "accelerator", but it's got to be as good as a TNT2, right?
#6
Altogether this definately supports your comment, I'm sure I could run a whole lot more sprites on the screen, though what I did would probably be a better representation of what a game might be considering the dynamics going on in a game (though not rigid body physics exactly it gives you a better guess) even that alone is impressive... will try with non-collision tiles, and no physics... when I get a chance I want to make a cycle that goes through and runs it at all of these and stores framerates to keep some sort of benchmark.
but your definately right, if anything the system is worse than what you think and I'm sure it could handle lots of more images...
I really don't think software rendering is needed
03/25/2005 (2:05 pm)
Actually I would doubt it compares to a TNT2... its a built in card from the manufacturer, probably a good 2-3 years old (if not more)... back then especially manufacturers put OEM and "extra" versions of cards in there (often cards that never made it to the shelves of the stores)... plus its merely a 600 mhz processor, keep in consideration it only has 256 mb ram, I had the printer queued up to 50+ documents (that right there is a killer), and about 10+ other windows open , roughly 8 of them being other programs, word, excel... (all draining on ram, processor, and by then a good deal of virtual memory). (EDIT: excuse the rant of a pc-tech lol)Altogether this definately supports your comment, I'm sure I could run a whole lot more sprites on the screen, though what I did would probably be a better representation of what a game might be considering the dynamics going on in a game (though not rigid body physics exactly it gives you a better guess) even that alone is impressive... will try with non-collision tiles, and no physics... when I get a chance I want to make a cycle that goes through and runs it at all of these and stores framerates to keep some sort of benchmark.
but your definately right, if anything the system is worse than what you think and I'm sure it could handle lots of more images...
I really don't think software rendering is needed
#7
03/25/2005 (2:33 pm)
@Tunnel: No plans for software? Hmm. That's actually enough for me to reconsider recommending T2D for serious development. I'm guessing at least a 25% market increase for each game I make if the engine has software rendering. I guess it would be possible without extreme difficulties to create a software renderer myself (by using some other library) when the time is right, but as I said, I'm not sure if I'm comfortable recommending a pure HW 2d engine.
#8
Doesn't DX take care of throwing things into software already? As for OGL... I have no clue.
03/25/2005 (2:40 pm)
%25? How do you arrive at the figure? I'd like to see software rendering too, but I'm not sure how critical it is.Doesn't DX take care of throwing things into software already? As for OGL... I have no clue.
#10
Are you sure you read my post correctly? I said a "3D software rasterizer". Plus, I promoted the TNT2 as being able to do "everythign that 2D games need."
03/25/2005 (6:59 pm)
Quote:True five years ago. These days, 3D hardware kills 2D renderers dead.
Are you sure you read my post correctly? I said a "3D software rasterizer". Plus, I promoted the TNT2 as being able to do "everythign that 2D games need."
#11
"To get an idea of how much Pixomatic might increase your potential market, consider that more than 40 percent of Half-Life players responding to a recent survey are using software rasterization." source
Now I am sure a few of those people have since gotten a better gfx card or just worked out their drivers, but the truth remains, software rendering always works, hardware rendering never _always_ work. At least not yet ;)
03/26/2005 (1:06 am)
To answer the question, will software rendering increase the market and how I came up with the 25% figure."To get an idea of how much Pixomatic might increase your potential market, consider that more than 40 percent of Half-Life players responding to a recent survey are using software rasterization." source
Now I am sure a few of those people have since gotten a better gfx card or just worked out their drivers, but the truth remains, software rendering always works, hardware rendering never _always_ work. At least not yet ;)
#12
It's not in our current plans as we are not planning that far ahead, we've already got lots of other stuff stacked up. Even if we had it on the list of things to do, it wouldn't be for a long time.
Not to put too fine a point on it but you don't have to wait for us to provide any particular feature and we wouldn't want you to. Writing a cross-platform software renderer is no trivial task and would distract from getting all the other core technologies out.
Your games shouldn't be limited by our vision. We give you the cake, you put the icing on. :)
- Melv.
03/26/2005 (1:41 am)
T2D provides a majority of the solution for you here and provides all the code. Go grab an existing software renderer and integrate it yourself, T2D doesn't make many rendering calls and they are all isolated within a single render function. The imagemaps encapsulate all the texture loading.It's not in our current plans as we are not planning that far ahead, we've already got lots of other stuff stacked up. Even if we had it on the list of things to do, it wouldn't be for a long time.
Not to put too fine a point on it but you don't have to wait for us to provide any particular feature and we wouldn't want you to. Writing a cross-platform software renderer is no trivial task and would distract from getting all the other core technologies out.
Your games shouldn't be limited by our vision. We give you the cake, you put the icing on. :)
- Melv.
#13
But I didn't say anything about hardware rendering (except pointing out that it the TNT2 will beat any software you try to use). 2D Software rendering will beat 3D software rendering attempting to be 2D software rendering. That's what I said.
How about this:
3D Software doing 2D (re: using Pixomatic for T2D) < 2D Software < 2D Hardware < 3D Hardware doing 2D
How recent is that survey? HL1 is a very old game; most of the people who initially played it originally probably did use software rendering. Since then, people have purchased new computers that have made this number decrease. If that survey is truly over the life of HL1, rather than a year-by-year survey, then its results are misleading.
I'd be interested to see what the Source version of that survey looks like.
03/26/2005 (9:24 am)
Quote:Smaug, still the same result. You are wrong in assuming software rasterization will be faster as it is not.
But I didn't say anything about hardware rendering (except pointing out that it the TNT2 will beat any software you try to use). 2D Software rendering will beat 3D software rendering attempting to be 2D software rendering. That's what I said.
How about this:
3D Software doing 2D (re: using Pixomatic for T2D) < 2D Software < 2D Hardware < 3D Hardware doing 2D
Quote:"To get an idea of how much Pixomatic might increase your potential market, consider that more than 40 percent of Half-Life players responding to a recent survey are using software rasterization."
How recent is that survey? HL1 is a very old game; most of the people who initially played it originally probably did use software rendering. Since then, people have purchased new computers that have made this number decrease. If that survey is truly over the life of HL1, rather than a year-by-year survey, then its results are misleading.
I'd be interested to see what the Source version of that survey looks like.
#14
The survey data that pixomatic points to is way, way, way out of date and totally worthless now. Here is the survey they used, it's old. Here is the updated version.
There are so many differences between the two. Only 2% of people report using the software renderer now. Look at the list of hardware as well... at least 95% of people have hardware accelerated cards in their machines that Torque works great with.
There are other big differences too. In the survey you pointed to almost 50% of people were using Windows 98. If you use that data the same way you're using the software renderer data, you'd say that it is of primary importance to support, and even optimize for, Windows 98. But look at the updated survey. 3% of people in the survey use Windows 98 now. This is another indicator of how out of date and worthless the old survey data is.
Even the survey I'm pointing to is a couple months old... and it'll take you a little while to finish your game anyhow... these numbers are not going to go up in the mean time. The return on investment for software rendering on the PC goes down and down, and software rendering is totally worthless on OS X. :)
Using the updated data, you might expect a 5% increase in sales of your game from including a software renderer. Given the change in data above, I'm not surprised Pixomatic has chosen not to update their sales and marketing info. :) You should be careful about trusting such material when it comes to business decisions about your game. In fact, you should be careful about using the half-life player base as your source of statistics altogether... I'm sure there's quite some overlap between this survey population and your target market, but you can't say for sure how much.
In the end though, if software rendering is personally important to you... please don't get me wrong. It'd be very cool to see somebody implement a software render for T2D. As Melv says, you could incorporate a software render yourself pretty easily. If you do so, please keep us up to date on your progress... it'd be fun to see a software renderer. I hope you can understand now why this wouldn't be a high official priority for us though, and I certainly hope you'll reconsider the remark above about recommending T2D to others based on these plans. :)
03/26/2005 (10:11 am)
Tomas,The survey data that pixomatic points to is way, way, way out of date and totally worthless now. Here is the survey they used, it's old. Here is the updated version.
There are so many differences between the two. Only 2% of people report using the software renderer now. Look at the list of hardware as well... at least 95% of people have hardware accelerated cards in their machines that Torque works great with.
There are other big differences too. In the survey you pointed to almost 50% of people were using Windows 98. If you use that data the same way you're using the software renderer data, you'd say that it is of primary importance to support, and even optimize for, Windows 98. But look at the updated survey. 3% of people in the survey use Windows 98 now. This is another indicator of how out of date and worthless the old survey data is.
Even the survey I'm pointing to is a couple months old... and it'll take you a little while to finish your game anyhow... these numbers are not going to go up in the mean time. The return on investment for software rendering on the PC goes down and down, and software rendering is totally worthless on OS X. :)
Using the updated data, you might expect a 5% increase in sales of your game from including a software renderer. Given the change in data above, I'm not surprised Pixomatic has chosen not to update their sales and marketing info. :) You should be careful about trusting such material when it comes to business decisions about your game. In fact, you should be careful about using the half-life player base as your source of statistics altogether... I'm sure there's quite some overlap between this survey population and your target market, but you can't say for sure how much.
In the end though, if software rendering is personally important to you... please don't get me wrong. It'd be very cool to see somebody implement a software render for T2D. As Melv says, you could incorporate a software render yourself pretty easily. If you do so, please keep us up to date on your progress... it'd be fun to see a software renderer. I hope you can understand now why this wouldn't be a high official priority for us though, and I certainly hope you'll reconsider the remark above about recommending T2D to others based on these plans. :)
#15
03/26/2005 (10:14 am)
And remember the old adage: 83.7% of all statistics are made up on the spot. :)
#16
I guess for more casual gamers that doesn't play Half-Life the hardware will be a bit more out of date and maybe bring the percentage up a tiny bit but maybe it's not worth implementing a software renderer at all.
In the end, everyone including myself will get first hand information about how many people can't play the game due to HW and it will be easier to consider adding a software renderer myself at that point.
@Josh: I will of course gladly keep recommending T2D to people. :)
03/26/2005 (10:49 am)
@Josh: That is an excellent answer. Surely this means the % of people that might play my game but can't because of HW problems is surely much lower than the 25% I guessed.I guess for more casual gamers that doesn't play Half-Life the hardware will be a bit more out of date and maybe bring the percentage up a tiny bit but maybe it's not worth implementing a software renderer at all.
In the end, everyone including myself will get first hand information about how many people can't play the game due to HW and it will be easier to consider adding a software renderer myself at that point.
@Josh: I will of course gladly keep recommending T2D to people. :)
#17
The most important thing to worry about is making sure your game is good and fun. Then you can worry about smaller technology concerns like whether implementing a software renderer is a good idea. :) Make a fun game, and you'll be in great shape!
And thanks to you and Rob (and everyone else) for the recommendations!
03/26/2005 (1:42 pm)
There ya go Tomas. :) The most important thing to worry about is making sure your game is good and fun. Then you can worry about smaller technology concerns like whether implementing a software renderer is a good idea. :) Make a fun game, and you'll be in great shape!
And thanks to you and Rob (and everyone else) for the recommendations!
#18
Especially if you want it to be a drop-in replacement for what is already supported by T2D via a 3D hardware backend. Sure, you can software blend, rotate and scale 32-bit images (with full alpha support) but then you start to push your minimal requirements for the software renderer up. So, it ends up your min specs lie between between a slow main CPU with a 3D accelerator, or a 1 GHz+ machine running a Trident 2D-only 4M video card. I'd say the former is more likely than the latter. :)
I've got a 300 MHz system sitting in the garage, with a Voodoo 3 video card waiting for it. I plan on trying T2D on some REAL base systems and honestly -- I don't expect too much trouble.
03/26/2005 (8:04 pm)
Josh, once again you dazzle us with numbers. I agree with the other posters -- losing sleep over a software renderer is really just not worth it.Especially if you want it to be a drop-in replacement for what is already supported by T2D via a 3D hardware backend. Sure, you can software blend, rotate and scale 32-bit images (with full alpha support) but then you start to push your minimal requirements for the software renderer up. So, it ends up your min specs lie between between a slow main CPU with a 3D accelerator, or a 1 GHz+ machine running a Trident 2D-only 4M video card. I'd say the former is more likely than the latter. :)
I've got a 300 MHz system sitting in the garage, with a Voodoo 3 video card waiting for it. I plan on trying T2D on some REAL base systems and honestly -- I don't expect too much trouble.
Torque Owner Smaug
Secondly, why would you need a 2D software rasterizer at all? Even a lowly TNT2 can do everythign that 2D games need. If people have any kind of 3D accelerator in their system, even those silly Intel integrated ones, they'll have everything they need for T2D.