I'm all in And have faith you all will provide us something nice that we can keep building upon onwards into the future.
I't is of course ridiculous to put urt hd as it is like in quake 3 into unreal 4, I guess i'm wondering what kind of performance increase you would see on the same hardware if that would be done.
From what i can make out unreal engine 4 is basicly what qauke 3 would become if you where to implement the same possibilities you see in unreal 4 so it's the right engine for this game, I did found out playing with the fov looks a little bit weird especially on 110 or above.
Let's hope when it comes out everybody forget about that and just upgrade your pc if you have lag/bad fps and all we be fine i'm sure.
Thanks for the clear answers.
Greetings,
This post has been edited by killspree: 10 October 2015 - 03:01 AM
I't is of course ridiculous to put urt hd as it is like in quake 3 into unreal 4, I guess i'm wondering what kind of performance increase you would see on the same hardware if that would be done.
I think the point of my previous post didn't come across very well. What I meant to say is that there is no direct correlation. On hardware which was barely sufficient to run Quake 3 well, the performance would be much worse. Quake 3 makes certain assumptions about the content which lets it run faster on weaker hardware (but limits what you can do with it). UE4 doesn't make those assumptions so it will run slower given equal hardware and details.
However, UE4 takes much better advantage of modern hardware features. Using a modern GPU on Quake 3 doesn't do all that much, because the engine makes poor use of the capabilities. So if you own a modern piece of hardware, the UE4 counterpart will likely run much better. But in that case, you can also throw a lot more polygons at it without it having any noticeable affect on the framerates whatsoever.
So you can't really say that one is more optimised than the other. Both simply take full advantage of the hardware capabilities that were/are common at the time, and that's what they run best on.
I think the point of my previous post didn't come across very well. What I meant to say is that there is no direct correlation. On hardware which was barely sufficient to run Quake 3 well, the performance would be much worse. Quake 3 makes certain assumptions about the content which lets it run faster on weaker hardware (but limits what you can do with it). UE4 doesn't make those assumptions so it will run slower given equal hardware and details.
However, UE4 takes much better advantage of modern hardware features. Using a modern GPU on Quake 3 doesn't do all that much, because the engine makes poor use of the capabilities. So if you own a modern piece of hardware, the UE4 counterpart will likely run much better. But in that case, you can also throw a lot more polygons at it without it having any noticeable affect on the framerates whatsoever.
So you can't really say that one is more optimised than the other. Both simply take full advantage of the hardware capabilities that were/are common at the time, and that's what they run best on.
Yes You are right I seen on my gpu monitor no gpu used on qauke3 urt unreal does use it.
Let's keep moving forward taking with us what has been learned.
Thank's and love to see more updates and work done in unreal 4.
Where the performance curve as to the difference between idtech3 and UE4 is noticeable is in the GPU rendering of the player models.
Animations is being authored at a frame rate of 30 FPS so streaming wise is feed to the GPU at a rate of 30 FPS which in turn outputs the rendered result at a rate of 30 FPS.
As a test using idtec3 I made a player set in MD3, which is point cache data (think digital flip book), at 10k poly (20k tris) and can not be hardware rendered. With two bots my frame rate stayed at a reasonable 110-125 but with additional bots added the FPS started to drop that by the time I reached 10 bots the FPS had dropped to 60 FPS. Proof that poly counts does effect performance.
Working on the fstech1 engine design I performance the same test but used MD5 which has the same animation pipeline as UE4 but this time I used a 60k (120 tris) player model and with 10 characters the frame rate stayed the same.
But
With each additional character add the "feel" started to get sluggish but still playable.
What I'm assuming is since the animation was still being outputted at 30 FPS with a frame rate of 120 the bottleneck was not GPU speed but rather the lack of bandwidth that was needed to push out 10 characters at the same time.
Keep in mind that this was done using the a limited feature set of OpenGL v2 (2004) but as a comparison fstech1 outperformed idtech3 using more or less the same frame work.
With UE4 everything changes as to numbers that may or may not indicate a performance issue, which is much easier to track down, and frame rates is no longer a reliable means as an indicator of performance, as it is in idtech3, as compared to being able to establish a reliable way of being able to track down an issue of "feel" due to bandwidth limitations over rendering speeds.
As a personal observation disregarding performance indicators the "feel" is almost instant to the point that it might as well be instant as to player input and would be willing to say that it would be possible to increase game speeds to the point that it would become uncontrollable and that $1000 dollar high end video card someone just bought would only increase the to much problem where in idtech3 it did not matter if you had a cutting edged card or you brothers hand me down. ;)
This is about as close as I can get as to how fast. :D
And of course when ever I do play UrT/idtech3 I "feel" like I'm wearing a suit made of mud.
Where the performance curve as to the difference between idtech3 and UE4 is noticeable is in the GPU rendering of the player models.
Animations is being authored at a frame rate of 30 FPS so streaming wise is feed to the GPU at a rate of 30 FPS which in turn outputs the rendered result at a rate of 30 FPS.
As a test using idtec3 I made a player set in MD3, which is point cache data (think digital flip book), at 10k poly (20k tris) and can not be hardware rendered. With two bots my frame rate stayed at a reasonable 110-125 but with additional bots added the FPS started to drop that by the time I reached 10 bots the FPS had dropped to 60 FPS. Proof that poly counts does effect performance.
Working on the fstech1 engine design I performance the same test but used MD5 which has the same animation pipeline as UE4 but this time I used a 60k (120 tris) player model and with 10 characters the frame rate stayed the same.
But
With each additional character add the "feel" started to get sluggish but still playable.
What I'm assuming is since the animation was still being outputted at 30 FPS with a frame rate of 120 the bottleneck was not GPU speed but rather the lack of bandwidth that was needed to push out 10 characters at the same time.
Keep in mind that this was done using the a limited feature set of OpenGL v2 (2004) but as a comparison fstech1 outperformed idtech3 using more or less the same frame work.
With UE4 everything changes as to numbers that may or may not indicate a performance issue, which is much easier to track down, and frame rates is no longer a reliable means as an indicator of performance, as it is in idtech3, as compared to being able to establish a reliable way of being able to track down an issue of "feel" due to bandwidth limitations over rendering speeds.
As a personal observation disregarding performance indicators the "feel" is almost instant to the point that it might as well be instant as to player input and would be willing to say that it would be possible to increase game speeds to the point that it would become uncontrollable and that $1000 dollar high end video card someone just bought would only increase the to much problem where in idtech3 it did not matter if you had a cutting edged card or you brothers hand me down. ;)
This is about as close as I can get as to how fast. :D
And of course when ever I do play UrT/idtech3 I "feel" like I'm wearing a suit made of mud.
Frankie I agree with you on that a faster gpu for urt is pointless because it almost only used the cpu.
From a p4 to a quad core is a noticeable difference but from a hd 5670 to a gtx 560ti not so much, also indeed we are now talking about rendering relatively high poly models compared to current version like you say we make a model about 10k or 20k its not so much but indeed if you where to take 2 of those you already have more than most maps are currently.
And also I understand where you guy's coming from and doing this for sooo long and me I'm still learning and keeping up with you guy's is not just a walk in the park.
Also I noticed indeed that with a map let's say about 500k and some nice effects the fps stay's stable even when moving very fast I read a lot about draw calls and other optimization like billboards.
Still learning the engine right now but feeling more at ease also by the friendly responses, there is a lot of scepticism but please don't mistake when I point something negative out that i'm just being negative I like to solve problems when I need to get something done but idk how I keep at it until it's done, that's what I like.
About TP now Is Wu working with you guy's?
I hope it's not because he feels like his map is taken over I mean a mapper like wu clearly deserves to be let in on the fold right?
All mappers deserves to be in the fold. ;)
Overall though I feel it would be better to ramp up CMM as the Unreal 4 mapping community where things are already set up to accommodate the <cough cough> management requirements.
The only reason I can see to join the blue shirts is/was so that contributions to idtech3/UrT came along with a Contributors Agreement.
With UE4 contributions can be made, including maps, either by CA or Creative Commons, MIT, and even under Epic's fair us license. This would mean that we could use the asset any way we wish but at the same time the contributor maintains copyrights.
Overall though I feel it would be better to ramp up CMM as the Unreal 4 mapping community where things are already set up to accommodate the <cough cough> management requirements.
The only reason I can see to join the blue shirts is/was so that contributions to idtech3/UrT came along with a Contributors Agreement.
With UE4 contributions can be made, including maps, either by CA or Creative Commons, MIT, and even under Epic's fair us license. This would mean that we could use the asset any way we wish but at the same time the contributor maintains copyrights.
Do you mean setting it up as a repository for assets maps materials and such?
Do you mean setting it up as a repository for assets maps materials and such?
Yea CMM could do that. After a year there is now a TON of free assists but most of it is not catalogued.
Thinking to the future though CMM really is the only mapping portal that I know of, except for maybe Polycount, and by the looks of things game moding is making a comeback.
Unreal Tournament 4 is free and what I understand mods can be made for it and then there is the game called ARK that you can buy and down load an entire mod package, 42 Gigs I believe, that includes all the art asset you need for that game.
CMM has a huge advantage though as there number one product is community members that knows what makes for a good or bad map. ;)