PhysX Accelerators

[Final8y]

Post 58 was spot on.

 

[Marc Fraser]

Im sorry, but a properly coded game with a quad core is more than capable of running a game with a decent amount of physics in.


CPUs are capable of millions and millions of calculations a second, im sure they are quite capable of calculating physics when its coded properly.

The argument that something is poorly coded is getting to be the most overused argument that gets posted on these forums. Most of the time it is used by people who have no idea how to program a computer to do even the simplest of things or have any idea of the level of calculations required to produce the graphical and physical effects they see in games.

Can you let us know what programming experience you have and how you are qualifed to comment on whether something is coded properly or not?

 

[FreeStream]

The argument that something is poorly coded is getting to be the most overused argument that gets posted on these forums. Most of the time it is used by people who have no idea how to program a computer to do even the simplest of things or have any idea of the level of calculations required to produce the graphical and physical effects they see in games.

Can you let us know what programming experience you have and how you are qualifed to comment on whether something is coded properly or not?

I have none at all.

But tell me why Supreme Commander runs so poorly compared to total annihilation and when my GPU is hardly loaded (55C instead of 60+ in either game?)

You cannot say that most games that come out have been properly tested and well coded, you simply CANNOT. Bugs, performance issues everywhere...

I don't need any knowledge about coding to work out that more time spent on tweaking the code the better it will perform.

I could write a program if I was bothered to learn in a day, it probably wouldn't work that well or very fast but after a month I'd bet it work a hell of a lot better.


However that task force video is impressive, but I want to see how it does on a dual core.

But still aint enough for me to part with £150...

 

[james.miller]

The argument that something is poorly coded is getting to be the most overused argument that gets posted on these forums. Most of the time it is used by people who have no idea how to program a computer to do even the simplest of things or have any idea of the level of calculations required to produce the graphical and physical effects they see in games.

Can you let us know what programming experience you have and how you are qualifed to comment on whether something is coded properly or not?

rubbish, you dont need to know how to program in c++ just to know when i game is badly coded. Its blindingly obvious in most cases. stalker - generally poor outdoor areas but still murders performance even with the 8800's. just about EVERY ubisoft pc game in existence - low res, poorly optimised textures and no end of performance/stability problems with them. straight console ports arent good enough, and they are perfect examples of poorly coded games.

 

[Marc Fraser]

I have none at all.

But tell me why Supreme Commander runs so poorly compared to total annihilation and when my GPU is hardly loaded (55C instead of 60+ in either game?)

You cannot say that most games that come out have been properly tested and well coded, you simply CANNOT. Bugs, performance issues everywhere...

I don't need any knowledge about coding to work out that more time spent on tweaking the code the better it will perform.

I could write a program if I was bothered to learn in a day, it probably wouldn't work that well or very fast but after a month I'd bet it work a hell of a lot better.


However that task force video is impressive, but I want to see how it does on a dual core.

But still aint enough for me to part with £150...

I haven't played Supreme Comander or Total Annihilation so can't really comment but I need to ask why you are only talking about the GPU load. You do realise that some games are CPU limited not GPU limited?

Also are you talking about the Total Annihilation that came out in 1997 and comparing it to Supreme Commander?

When you say you could write a program in a day if you could be bothered to learn what kind of program are you talking about?

 

[Marc Fraser]

rubbish, you dont need to know how to program in c++ just to know when i game is badly coded. Its blindingly obvious in most cases. stalker - generally poor outdoor areas but still murders performance even with the 8800's. just about EVERY ubisoft pc game in existence - low res, poorly optimised textures and no end of performance/stability problems with them. straight console ports arent good enough, and they are perfect examples of poorly coded games.

If you don't know how to program how can you possibly say something is badly coded. You will have no idea what kind of issues there are when implementing certain techniques such as the deferred renderer that Stalker uses for example.

 

[sja360]

If you don't know how to program how can you possibly say something is badly coded. You will have no idea what kind of issues there are when implementing certain techniques such as the deferred renderer that Stalker uses for example.

you can tell its poor coding because it wont run smooth on the latest and greatest hardware. how else would you explain the slowdown on occasion? or the need to drop resolutions or HDR affects? i mean we're talking about hardware thats powerful these days and yet the games seem to have performance issues.

 

[Marc Fraser]

you can tell its poor coding because it wont run smooth on the latest and greatest hardware. how else would you explain the slowdown on occasion? or the need to drop resolutions or HDR affects? i mean we're talking about hardware thats powerful these days and yet the games seem to have performance issues.

But how do you know it is possible to run all of the effects that are going on at that time on current hardware smoothly without any stuttering?

You may be able to point to similar games that you think are producing the same effects but unless you understand what is going on you do not really know if that is the case.

Modern games are designed to stress the hardware to its limits, especially on the highest settings and hence it is not suprising that you experience slow down. If I recall correctly Crytek have said that they will have certain settings that will not be able to be run on current hardware but it is intended that they will only be able to be run by future hardware. Are you going to accuse them of poor coding?

Of course the hardware is a lot more powerful but the effects being produced are also much more complicated. From experience I have found it is very easy to cripple the latest hardware without getting anyhwere near the quality of todays games.

 

[sja360]

Of course the hardware is a lot more powerful but the effects being produced are also much more complicated.

true, but then i presume the developers are at a stage where they dont know which technique to code with is best perhaps?

 

[FreeStream]

I haven't played Supreme Comander or Total Annihilation so can't really comment but I need to ask why you are only talking about the GPU load. You do realise that some games are CPU limited not GPU limited?

Also are you talking about the Total Annihilation that came out in 1997 and comparing it to Supreme Commander?

When you say you could write a program in a day if you could be bothered to learn what kind of program are you talking about?

Im well aware of CPU bottlenecks, FS9 and 10 both have huge CPU bottlenecks...

and yes, I am. Because a 2 vs 2 500 unit map AI runs > 60 FPS yet the same in Supreme Commander runs < 10, why? Same amount of units, same map size, whats the problem?


As for a program, any program! Im well aware programs are compiled and called builds, and the first builds are very basic and very slow and very buggy, but as the build count increases the bugs and fixed and the performance is increased as the code is refined. Same thing applies to games, they are being pushed out far far far too early, another year should have been added to FS-X atleast, aswell as Sup Com, TDU, BF2, infact any game thats come out in the last year or two...

 

[Gashman]

Im well aware of CPU bottlenecks, FS9 and 10 both have huge CPU bottlenecks...

and yes, I am. Because a 2 vs 2 500 unit map AI runs > 60 FPS yet the same in Supreme Commander runs < 10, why? Same amount of units, same map size, whats the problem?


As for a program, any program! Im well aware programs are compiled and called builds, and the first builds are very basic and very slow and very buggy, but as the build count increases the bugs and fixed and the performance is increased as the code is refined. Same thing applies to games, they are being pushed out far far far too early, another year should have been added to FS-X atleast, aswell as Sup Com, TDU, BF2, infact any game thats come out in the last year or two...

yeah you can't really compair total annihilation to supreme commander, for one everything is much more complicated in supreme commander, i mean all the hits are newtonian calculated, there you go huge hit on your CPU right there, plus the way the units move is proberbly more complicated, the way the AI acts is proberbly more complicated, it all adds up really, though your right and there are a lot of badly coded games knocking around where you get performance hits you just shouldn't get. heres an example, far cry looks pretty stunning still with high AA and AF yet runs fluidly as hell (well coded, brilliantly infact) yet some games that look similar have terrible performance and non of the huge view ranges in far cry

 

[Durzel]

I agree with Marc, unless you know what goes into programming these engines then you can't call it "bad coding". You can certainly say it performs poorly on your equipment, but that's it.

Also it's quite easy to be deceived by high quality textures...

 

[FreeStream]

I agree with Marc, unless you know what goes into programming these engines then you can't call it "bad coding". You can certainly say it performs poorly on your equipment, but that's it.

Also it's quite easy to be deceived by high quality textures...

No, you can. As i've said before if its rushed out then its nowhere near what it could be if the time had been spent tweaking it. There is NO argument for rushing it out of the door, and rushing out means rubbish tweaking of the code. END OF.

We could have restored our Concorde in 1/4 the time if we didn't bother sand, seal, prim and 2 coat it. But we did take the time to do it properly. Same should be done with games.

 

[Participant]

It's not possible to code all games equally well. Therefore some games must be coded better than others, that's a fact. What other conclusion can be drawn from the fact that we can take two games that look equally pretty, run them on the same system, and one runs great and the other runs like a dog? Therefore I'm with concorde on this one.

 

[Marc Fraser]

No, you can. As i've said before if its rushed out then its nowhere near what it could be if the time had been spent tweaking it. There is NO argument for rushing it out of the door, and rushing out means rubbish tweaking of the code. END OF.

We could have restored our Concorde in 1/4 the time if we didn't bother sand, seal, prim and 2 coat it. But we did take the time to do it properly. Same should be done with games.

If you have no idea how to program then how do you know that it can definitley be tweaked to perform better.

When you were comparing the performance of Total Annihilation and Supreme Commander earlier were you comparing their performance based on the hardware available at their respective launch dates or were you talking about how they both perform on current hardware?

I really hope you were talking about the hardware available at launch.

 

[Marc Fraser]

It's not possible to code all games equally well. Therefore some games must be coded better than others, that's a fact. What other conclusion can be drawn from the fact that we can take two games that look equally pretty, run them on the same system, and one runs great and the other runs like a dog? Therefore I'm with concorde on this one.

As I said earlier they may look equally pretty but may be using completely different techniques to get to the same visuals effect. For example there are various techniques used to generate shadows in games which don't all perform the same and have different limitations. Therefore it may not always be appropriate to use the fastest performing algorithm in a certain game. Another example of this is the deferred renderer used in Stalker. As far as I know most game engines prior to Stalker used a more traditional forward renderer.

There are also numerous ways of optimising a graphics engine to speed it up. However, depending on the content of the game it may not always be possible to use these or you may not achieve the same performance increase as you would if you implemented the optimisation in another engine. Some of the most costly operations in DX9 are the draw calls needed to render primitives or changing the state of the device. e.g. switching textures or shaders. One way around this is to batch objects to be drawn so they use one draw call to render them all. However the amount of benefit you get from this will not always be the same as depending on the objects to be rendered it may not be possible to create large enough batches and reduce the number of draw calls.

You also need to realise that there is a lot more going on than just graphics. Two games that look equally pretty but perform differently may be using completely different amounts and types of Physics and AI.

In summary you cannot just look at two games, see they look similar and then accuse the one that runs slower of being badly coded.

 

[Participant]

they may look equally pretty but may be using completely different techniques to get to the same visuals effect

And, if one result yields better performance than the other, this is known as bad coding.

 

[Durzel]

And, if one result yields better performance than the other, this is known as bad coding.

So by the same token if you compare a Golf GTi to a Ferrari then the Golf is "rubbish" simply because it's not as fast?

You might as well call 3DMark06 "bad coding" because it doesn't look a whole lot better than HL2 and yet still manages to run much slower?

As already said HL2 (as an example) is very clever with textures. As an engine Source isn't hugely groundbreaking, Havok provides a lot of the "oooh" factor together with the very immersive level design. That's not in any way a diss on HL2/Source - just saying that you can't call other game engines "bad coding" just because they don't perform as well on your equipment, and you especially can't call it "bad coding" when you have no idea what coding is.

 

[Supercow]

Fad peice of crap IMO.
Look at the Alan Wake vids. In stead of using PhysX, they use cores of c2d or x2 chips to do the same job... or am I being a nub ?

 

[Energize]

I can't believe ageia are still in business, developers must be insane to code for a ppu that nobody has just to add a few extra effects while crippling framerates.