-
We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.
New Xbox info
Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
I should go look up the quote where I believe it was Blackley who said they have gotten in game upto 95% efficiency of the claimed geometry performance for the GTS. Somewhere around 22M polys.
Funny how some people will defend to the hilt some cards when supposedly developers haven't taken advantage of it yet and it gets a bad rap, but vice-versa for another companies card.
Funny how some people will defend to the hilt some cards when supposedly developers haven't taken advantage of it yet and it gets a bad rap, but vice-versa for another companies card.
BenSkywalker
Diamond Member
Dave-
"There is no such thing as using polygons, it is always vertices."
I work at the poly level over 90% of the time. Different POV between hardware engineer and end user on this one🙂 I understand that the hardware always deals with vertices, however I don't.
"And Ben, I think we both know that is a crazy assumption for any real world situation. Typically your polygon count is 1/3-1/2 your vertex count."
Not for me, not by a long shot. I think with the current insanely low poly games that is very likely true, when you increase geometric complexity you tend to run much closer in the poly/vertice ratio.
Most of the time you are not going to be increasing the amount of objects on screen, simply the LOD in those objects. When you have many objects with low levels of complexity, you will likely see the ratio of polys to vertices at 1:2 or 1:3, but if you significantly increase the poly count per model the amount of vertices increases only slightly.
10K polys between 20 models bad ratio, probably in between 1:2 or 1:3
200K polys between 20 models very good ratio, possibly 7:8 or better
Thorin-
"The proof is in the linkage........so where is it?"
1600x1200@100FPS = 192MPixels per second(no link needed on this one🙂). All of the current boards(save Kyro) are horrible at rasterization efficiency.
"There is no such thing as using polygons, it is always vertices."
I work at the poly level over 90% of the time. Different POV between hardware engineer and end user on this one🙂 I understand that the hardware always deals with vertices, however I don't.
"And Ben, I think we both know that is a crazy assumption for any real world situation. Typically your polygon count is 1/3-1/2 your vertex count."
Not for me, not by a long shot. I think with the current insanely low poly games that is very likely true, when you increase geometric complexity you tend to run much closer in the poly/vertice ratio.
Most of the time you are not going to be increasing the amount of objects on screen, simply the LOD in those objects. When you have many objects with low levels of complexity, you will likely see the ratio of polys to vertices at 1:2 or 1:3, but if you significantly increase the poly count per model the amount of vertices increases only slightly.
10K polys between 20 models bad ratio, probably in between 1:2 or 1:3
200K polys between 20 models very good ratio, possibly 7:8 or better
Thorin-
"The proof is in the linkage........so where is it?"
1600x1200@100FPS = 192MPixels per second(no link needed on this one🙂). All of the current boards(save Kyro) are horrible at rasterization efficiency.
actually
if you take 3 vertices to make 1 triangle
15Mvertices = 5M triangles
actually this is actually pretty close to the real life output of the geforces
my GTS although can chuck out 22M, is giving out 4-5M triangles/s in the demo of X-Isle
which is pretty close to the theoritical amount
if you take 3 vertices to make 1 triangle
15Mvertices = 5M triangles
actually this is actually pretty close to the real life output of the geforces
my GTS although can chuck out 22M, is giving out 4-5M triangles/s in the demo of X-Isle
which is pretty close to the theoritical amount
BenSkywalker
Diamond Member
"actually
if you take 3 vertices to make 1 triangle"
You don't need to use three seperate vertices for each triangle, in fact it would be very difficult even if you tried to. Draw a square and an X through it, corner to corner. There you have four polygons and only five vertices, this is how models are built(oversimplification, but you get the idea).
"With a single model perhaps, but not with multiple models running through out an enviroment."
Give me a scenario🙂 I'll slap together some primitive models and see how it turns out in polys:vertices.
if you take 3 vertices to make 1 triangle"
You don't need to use three seperate vertices for each triangle, in fact it would be very difficult even if you tried to. Draw a square and an X through it, corner to corner. There you have four polygons and only five vertices, this is how models are built(oversimplification, but you get the idea).
"With a single model perhaps, but not with multiple models running through out an enviroment."
Give me a scenario🙂 I'll slap together some primitive models and see how it turns out in polys:vertices.
blurredvision
Lifer
You guys get into too much of an hysteria over theoretical numbers and the such. Who cares if what they say is true?? I agree with what BurnedNIU said above. If they say 8x, I'm expecting AROUND 8x. Everyone of you talk so much of numbers that you should be smart enough to realize that these numbers they put out are purely theoretical, and stand to give you an idea of what the machine can do. Dave3BD, you seem to be biased against Microsoft and Nvidia, why I don't know.
<< i might wait in line the first night cuz the first revision will be the most hackable and you know linux will be on it within two weeks >>
That shows you are ignorant about consoles in the first place because that has to be the most stupid remark I've heard yet against the XBox. For those who know what I'm talking about, you guys are the ones who should be making comments. For those who don't know what I'm talking about, i.e. Dave3BD, I suggest you get informed.
All IMO of course, so don't bitch at me!! (as you so elegantly put it)
<< i might wait in line the first night cuz the first revision will be the most hackable and you know linux will be on it within two weeks >>
That shows you are ignorant about consoles in the first place because that has to be the most stupid remark I've heard yet against the XBox. For those who know what I'm talking about, you guys are the ones who should be making comments. For those who don't know what I'm talking about, i.e. Dave3BD, I suggest you get informed.
All IMO of course, so don't bitch at me!! (as you so elegantly put it)
BenSkywalker
Diamond Member
"Dave3BD, you seem to be biased against Microsoft and Nvidia, why I don't know."
Dave is an engineer for 3dfx, if he wasn't biased against nVidia then I would have some serious doubts about their future😀
Dave is an engineer for 3dfx, if he wasn't biased against nVidia then I would have some serious doubts about their future😀
I found it:
<< Nonetheless, it does make sense to talk about theoretical polygons and gigapixels, because these set the limits on what's possible. Also, in a balanced system, performance does tend to scale roughly with theoretical specs, so it's a useful indicator. And finally, it gives programmers something to shoot at. On the GeForce 2 GTS, initially we were getting only a few million triangles a second, but we knew the theoretical limit was 25 million, so we kept after it, and now we've come within about 10% of theoretical maximum throughput (actually more like 96% of max if you don't worry about little details like not having degenerate triangles 🙂. >>
I was pretty close, but its Abrash not Blackley. 🙂
myurl
<< Nonetheless, it does make sense to talk about theoretical polygons and gigapixels, because these set the limits on what's possible. Also, in a balanced system, performance does tend to scale roughly with theoretical specs, so it's a useful indicator. And finally, it gives programmers something to shoot at. On the GeForce 2 GTS, initially we were getting only a few million triangles a second, but we knew the theoretical limit was 25 million, so we kept after it, and now we've come within about 10% of theoretical maximum throughput (actually more like 96% of max if you don't worry about little details like not having degenerate triangles 🙂. >>
I was pretty close, but its Abrash not Blackley. 🙂
myurl
blurredvision
Lifer
I'm sorry thorin, but I don't follow what you are saying. My comment was toward the fact that it would stupid, if not impossible to put linux on the XBox. Let me explain why I think this.
First, you can't have an 'operating system' on a console. Some of you might say, "Well, the Dreamcast has Windows CE." It has the code for it, so making games is easier. You can't actually go in and do anything with the version of CE, it's there as a base for the programmers to make games. The same will hold true with the XBox, so I've read. Whichever base code it uses will not be able to be modified, simply because its not an operating system to be used by normal users. How in the hell would you put Linux on the Xbox? It's not even conceivable, let alone possible. Yeah, the Xbox will have a hard drive, but not to install stuff on. Even if there was a way to put Linux on that console, you wouldn't be able to run any games because all games for a console are built around only a couple of variables. Games for computers are built around a seemlingly infinite combinations of variables.
First, you can't have an 'operating system' on a console. Some of you might say, "Well, the Dreamcast has Windows CE." It has the code for it, so making games is easier. You can't actually go in and do anything with the version of CE, it's there as a base for the programmers to make games. The same will hold true with the XBox, so I've read. Whichever base code it uses will not be able to be modified, simply because its not an operating system to be used by normal users. How in the hell would you put Linux on the Xbox? It's not even conceivable, let alone possible. Yeah, the Xbox will have a hard drive, but not to install stuff on. Even if there was a way to put Linux on that console, you wouldn't be able to run any games because all games for a console are built around only a couple of variables. Games for computers are built around a seemlingly infinite combinations of variables.
BenSkywalker
Diamond Member
"I dunno Ben.. just a scene from Q3 with some guys running around.."
Alright, just letting you know now it will probably take me a day or so🙂
I'll try and use NV15 as a base map to keep it fairly accurate(already existing map) and then just "fix" the player models to acceptable levels. I don't want to come up with something too heavily otpimized for this particular scenario🙂 Should be in the 100K range or so when done(per frame), but I'll try and up it to ~250K also just to get some comparitive numbers(also I'll reduce as much as possible if I can).
Alright, just letting you know now it will probably take me a day or so🙂
I'll try and use NV15 as a base map to keep it fairly accurate(already existing map) and then just "fix" the player models to acceptable levels. I don't want to come up with something too heavily otpimized for this particular scenario🙂 Should be in the 100K range or so when done(per frame), but I'll try and up it to ~250K also just to get some comparitive numbers(also I'll reduce as much as possible if I can).
blurredvision
Lifer
I didn't put up the quote as coming from you, Dave. If you'll read it, I never once said that you said it. As being a quote, they can reference it back to who said it.
I didn't put anyone else's name in my reply because I was attacking you at the time. Thanks 🙂
I didn't put anyone else's name in my reply because I was attacking you at the time. Thanks 🙂
ok.. then.. So you are trying to make it look like I'm saying something I'm not. That is really stupid and VERY immature.
Well i'm pretty convinced that I'm already a good deal more informed than you.. so any time buddy... Don't antagonize me.. I don't take it well and I dish it out a lot harder.
Well i'm pretty convinced that I'm already a good deal more informed than you.. so any time buddy... Don't antagonize me.. I don't take it well and I dish it out a lot harder.
BenSkywalker
Diamond Member
D@mn Dave, I don't think I've seen you so riled up as you have been around here😉
You get used to people flaming you that don't agree with you after a while, even if they don't understand what you are talking about.
You get used to people flaming you that don't agree with you after a while, even if they don't understand what you are talking about.
blurredvision
Lifer
In response to your comment above that I was trying to make it sound like it was you who said it, I didn't do this intentionally. I suppose your way of 'dishing it out a lot harder' is twisting words around to make you look good???
CAD tests don't count. Nothing close to a game.
Actually they do. They show how much real-world potential a certain piece of hardware has when creating a new game engine. With optimal utilization of strips, just as Ben demonstrated, you only get number of polygons + 2 vertices per object. When number of polygons approaches infinity while number of objects is just a large constant, number of vertices per triangle approaches 1.
It's true that GeForce's T&L was never really useful for the vast majority of gamers, unlike nVidia claimed. However, for developers and 3D-modellers a cheap and very powerful hardware T&L solution was a blessing.
But... What I'm saying that no one should claim that GeForce T&L unit is only capable of 4-5 million polys per second in real-world situation, because this is just not true. However, it's true in all current games and engines, which are not optimized for hardware T&L. Claiming current hardware T&L to be unuseful for enhancing gaming experience based on current T&L games is like claiming 3D hardware accelerators are not useful by using Diablo II as an example.
All my opinion.. don't bitch about it.
Please don't take this the wrong way, but if you don't want people to bitch about what you post here, what's the point in posting here? 😀
Actually they do. They show how much real-world potential a certain piece of hardware has when creating a new game engine. With optimal utilization of strips, just as Ben demonstrated, you only get number of polygons + 2 vertices per object. When number of polygons approaches infinity while number of objects is just a large constant, number of vertices per triangle approaches 1.
It's true that GeForce's T&L was never really useful for the vast majority of gamers, unlike nVidia claimed. However, for developers and 3D-modellers a cheap and very powerful hardware T&L solution was a blessing.
But... What I'm saying that no one should claim that GeForce T&L unit is only capable of 4-5 million polys per second in real-world situation, because this is just not true. However, it's true in all current games and engines, which are not optimized for hardware T&L. Claiming current hardware T&L to be unuseful for enhancing gaming experience based on current T&L games is like claiming 3D hardware accelerators are not useful by using Diablo II as an example.
All my opinion.. don't bitch about it.
Please don't take this the wrong way, but if you don't want people to bitch about what you post here, what's the point in posting here? 😀
blurredvision
Lifer
Dave, dude, calm down. I'm not here to flame, I just like a healthy argument. It's not a matter of who can flame better and harder. It's not my fault you're mad because Microsoft chose Nvidia over you 3dfx guys......
Just kidding!! Bet you got pretty riled up over that one, eh?? Oh well, since you so politely asked, this will be my last post in this thread. I'm not looking to win an argument like you are, I just wanted to put in my opinion, and cause a little ruckus at the same time. Being a 3dfx engineer, I would assume you are used to being bent over by other people all the time anyways. Have a good day 🙂 You may now have the last word......
Just kidding!! Bet you got pretty riled up over that one, eh?? Oh well, since you so politely asked, this will be my last post in this thread. I'm not looking to win an argument like you are, I just wanted to put in my opinion, and cause a little ruckus at the same time. Being a 3dfx engineer, I would assume you are used to being bent over by other people all the time anyways. Have a good day 🙂 You may now have the last word......
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 25K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 24K
-
-
