Tomb Raider: Underworld

[ShootMyMonkey]

If you believe that those things mean TRU must necessarily be sub-HD then I personally do not agree.

The most fundamental flaw in the average gamer's technical analysis of any game is that it never goes further than skin-deep. It never really matters how advanced anything is -- it only matters how apparent you make it. You want a list of things that go further than what you can plainly see on just about any game you can imagine, I don't think I can fit that list in under 1 MB of text.

Although, if you don't think that draw distance affects render performance, you're not at all fit to complain about any game being under 720p at any point in time for any reason. Though on stuff you can't see, the fact that TR:U is actually a platforming puzzle game means that the search trees for motion path predictors is at least an order of magnitude above the average. Conay's guess about IK feet is a little short of the reality -- it's a full-body IK and that applies not only to Lara, but every soft-body object in the game right down to the foliage. Having a lot more variance in the per-instance illumination model... Having a million things from traps and puzzles to HUD and menus all running in a threaded VM... All this stuff eats up time. Just having the per-instance irradiance estimation alone eats up a quarter to half a millisecond in some levels. Sounds like nothing by itself, but when you add everything up and realize that you have very little time to share for everything including rendering, you're going to take ways of giving you that time which don't interfere with your priorities. And in the end, that's the real difference. What is it that you prioritize?

Now of course, I know the standard response is ultimately going to amount to "well, somebody else did it and therefore it's universally solvable." That's a complete load of crap. Why doesn't LBP have the same priorities as FarCry? Why doesn't Haze have the same priorities as TR:U? Why doesn't Spore have the same priorities as Halo 3? They're not the same games! That's all there is to it. It doesn't matter if you're comparing superficially similar games (which is what the TR<->Uncharted comparison is) or completely unrelated games. The reality will always be the same. You can disbelieve all you like and pretend that there everything must work exactly as superficially as you think, but that doesn't make it a valid position.

 

[deepbrown]

The most fundamental flaw in the average gamer's technical analysis of any game is that it never goes further than skin-deep. It never really matters how advanced anything is -- it only matters how apparent you make it. You want a list of things that go further than what you can plainly see on just about any game you can imagine, I don't think I can fit that list in under 1 MB of text.

Although, if you don't think that draw distance affects render performance, you're not at all fit to complain about any game being under 720p at any point in time for any reason. Though on stuff you can't see, the fact that TR:U is actually a platforming puzzle game means that the search trees for motion path predictors is at least an order of magnitude above the average. Conay's guess about IK feet is a little short of the reality -- it's a full-body IK and that applies not only to Lara, but every soft-body object in the game right down to the foliage. Having a lot more variance in the per-instance illumination model... Having a million things from traps and puzzles to HUD and menus all running in a threaded VM... All this stuff eats up time. Just having the per-instance irradiance estimation alone eats up a quarter to half a millisecond in some levels. Sounds like nothing by itself, but when you add everything up and realize that you have very little time to share for everything including rendering, you're going to take ways of giving you that time which don't interfere with your priorities. And in the end, that's the real difference. What is it that you prioritize?

Now of course, I know the standard response is ultimately going to amount to "well, somebody else did it and therefore it's universally solvable." That's a complete load of crap. Why doesn't LBP have the same priorities as FarCry? Why doesn't Haze have the same priorities as TR:U? Why doesn't Spore have the same priorities as Halo 3? They're not the same games! That's all there is to it. It doesn't matter if you're comparing superficially similar games (which is what the TR<->Uncharted comparison is) or completely unrelated games. The reality will always be the same. You can disbelieve all you like and pretend that there everything must work exactly as superficially as you think, but that doesn't make it a valid position.

I personally don't think reducing resolution is the only solution. It is A solution. From what I've seen from other developers (apart from COD4, where the developers had a lower resolution in mind from the begining in order to target 60fps) it's a last resort towards the games deadline. You'll look at how much memory you're missing towards the end of development, and you don't have the time to further tweak the game, or even the knowledge to find a way to use the resources more efficiently, and you'll work out how much you would gain by reducing resolution. You drop it, and everything fits within the memory and works perfectly. It's the easy, but often necessary, option.

All developers have different problems and decide on different solutions - but some developers simply have better tech, better programmers, and more indepth research of the hardware. It's ND's talents and their ability to work very closely with the makers of the hardware and all the 1st party developers who are working to find solutions to the same problems, that means they can achieve the results they do. And despite Uncharted being described as a "corridor shooter" - there are 720p games with greater draw distances than TRU...and games with much more going on - Resistance 2 comes to mind. Insomniac are another developer who clearly have the knowledge. It's nice of them to share that knowledge with other developers.

 

[FirewalkR]

Everyone continue to argue with SMM because he'll answer with great posts like the last one. Please SMM, talk even more about implementation details next time.

 

[deepbrown]

Everyone continue to argue with SMM because he'll answer with great posts like the last one. Please SMM, talk even more about implementation details next time.

Debatng always brings out the gems

 

[ShootMyMonkey]

I personally don't think reducing resolution is the only solution. It is A solution. From what I've seen from other developers (apart from COD4, where the developers had a lower resolution in mind from the begining in order to target 60fps) it's a last resort towards the games deadline. You'll look at how much memory you're missing towards the end of development, and you don't have the time to further tweak the game, or even the knowledge to find a way to use the resources more efficiently, and you'll work out how much you would gain by reducing resolution. You drop it, and everything fits within the memory and works perfectly. It's the easy, but often necessary, option.

Well, that much, at least we agree. The only thing I'd add is that the kinds of issues that lead to that decision can really come up at any point. It's entirely possible that late in the development, one might be left with the possibility to reconsider, but it almost never actually works out towards a change back to higher res. Again, capacity is usually not as big a concern as bandwidth and having the bandwidth to spare. It is the easy solution, and I doubt anybody would do it if they didn't feel it was necessary (mainly because it's just extra work). At the same time, it's much more often that people would not consider it important compared to a thousand other things.

It's one of the reasons why the whole "720p w/ 2x MSAA or equivalent" TRC/TCR is by far the one console manufacturers are least picky about. They get picky about things like error messages being expressed in passive or active voice or even having the word "error" in them... They'll complain that adjusting the volume from 0 to max requires too many or too few gradations... They'll tell you to fix a crash that involves duct-taping a button down for 24 hours during a full moon while you do the hokey-pokey and you turn yourself around... But resolution, the TCR/TRC compliance testing somehow ends up "Eh. Whatever." There are times where they do complain about it when it's things like aspect ratio mismatches between buffer size and target resolution (e.g. having a buffer that's close to 16:9 when the display mode is SD and so on)... but even things like that, the 1st and 2nd party developers get concessions on them.

Insomniac are another developer who clearly have the knowledge. It's nice of them to share that knowledge with other developers.

As much as I think that Insomniac's "code-reuse-is-bulls**t" approach to development is maniacal, they do engage in militaristic enforcement of development rules and explicit boundaries for every single snippet of code that anybody writes, and that seems to be a powerful thing in the end. Based on what they've openly been willing to say at conferences and several informal talks they've given at developer day meetings and such, I wouldn't dare call their development approach sane or scalable to large teams (and definitely not to studios which develop more than one game on the same tech simultaneously), but insane works sometimes.

 

[makattack]

I for one can appreciate what's going on in the game engine-wise, and also appreciate the motion-capture mapping done on the main character, particularly with the rock-climbing animations! They're a little too quick for my taste, but I still love them -- right down to the outrageously dynamic, dynamic moves! Never have a bought a game based on a single technical aspect, but when I first controlled the character in the climbing bit, I was sold.

 

[deepbrown]

Well, that much, at least we agree. The only thing I'd add is that the kinds of issues that lead to that decision can really come up at any point. It's entirely possible that late in the development, one might be left with the possibility to reconsider, but it almost never actually works out towards a change back to higher res. Again, capacity is usually not as big a concern as bandwidth and having the bandwidth to spare. It is the easy solution, and I doubt anybody would do it if they didn't feel it was necessary (mainly because it's just extra work). At the same time, it's much more often that people would not consider it important compared to a thousand other things.

It's one of the reasons why the whole "720p w/ 2x MSAA or equivalent" TRC/TCR is by far the one console manufacturers are least picky about. They get picky about things like error messages being expressed in passive or active voice or even having the word "error" in them... They'll complain that adjusting the volume from 0 to max requires too many or too few gradations... They'll tell you to fix a crash that involves duct-taping a button down for 24 hours during a full moon while you do the hokey-pokey and you turn yourself around... But resolution, the TCR/TRC compliance testing somehow ends up "Eh. Whatever." There are times where they do complain about it when it's things like aspect ratio mismatches between buffer size and target resolution (e.g. having a buffer that's close to 16:9 when the display mode is SD and so on)... but even things like that, the 1st and 2nd party developers get concessions on them.


As much as I think that Insomniac's "code-reuse-is-bulls**t" approach to development is maniacal, they do engage in militaristic enforcement of development rules and explicit boundaries for every single snippet of code that anybody writes, and that seems to be a powerful thing in the end. Based on what they've openly been willing to say at conferences and several informal talks they've given at developer day meetings and such, I wouldn't dare call their development approach sane or scalable to large teams (and definitely not to studios which develop more than one game on the same tech simultaneously), but insane works sometimes.

Well I guess you have to measure the publicity the game might get if it's sub-hd. Perhaps there's no backlash at all. But that's not what we're debating about - you're saying that lowing resolution is the only solution to the problems you face - and that knowledge of other ways to find memory are not relevant in your case.

I didn't mention Insomniac's development process, I mentioned their tech and their tech research. They've done a lot of the work for you - they know the PS3 so well and release most if not all their documentation, that a development team working on the PS3 would be mad to ignore it. The pride I found at Free Radical saying they didn't need Sony technicians and just "knew PlayStation" and how talented their engineers were - it obviously didn't pay off.

 

[ShootMyMonkey]

Well I guess you have to measure the publicity the game might get if it's sub-hd. Perhaps there's no backlash at all. But that's not what we're debating about - you're saying that lowing resolution is the only solution to the problems you face - and that knowledge of other ways to find memory are not relevant in your case.

That is quite far from what I said. You're taking two completely different points and putting them together. I mentioned memory capacity constraints in regards to texture arrangements and generic resource management. I mentioned resolution in regards to memory bandwidth and/or fillrate, and how not having as much framebuffer bandwidth hit (which, let's face it, is the single biggest consumer) can potentially mean room to do other stuff. And not once in any instance, did I ever say that it was the "one and only" solution. What happens is in reality is that it can become the only remaining solution.

Perhaps if you're talking about not wanting to tile while still using hardware MSAA, then shrinking resolution is invariably the only solution because that's simply the way it works.

I never really got into specifics about tech that's out there nor do I really plan to, but there are two basic ways in which a technique can be rendered irrelevant. An example of the former would be a case that joker454 mentioned when talking about his last project. He mentioned, for instance, that the use of the visibility and backface culling aspects of EDGE were useless for his title... why? because it was a game that involved a pretty limited set of viewpoints, so the geometry could safely be built with most all of those surfaces already non-existent. That's actually quite a common case for any and all context-sensitive performance techniques out there when you take it from one context to another.

The other case is when the overall performance picture is bottlenecked by something that is completely unrelated to the lack of performance in the areas a technique attacks. You have to do something about that bottleneck. If you are filling too many pixels per frame, then optimizing the hell out of everything in your codebase even until it's infinitely faster than it was before will still have you running slow because you're still filling too many pixels and you only sped up that which was not limiting you. That's what I meant when I said that no amount of time, money, or genius will save you from what you're working with.

It's also in the nature of hardware limitations that when you approach ever closer to them, everything looks just fine, and then you just suddenly go down the toilet even for going 0.1% beyond the line. It's really a pass-fail story.

 

[AlNom]

With 360, this is less of a concern because the framebuffer isn't going to share any of that main memory bandwidth, but tiling as implemented on the 360 (that is to say, highly intrusive and non-transparent) can never be viewed in a positive light.

hm... Do the TCRs not allow you to render at a higher resolution without MSAA? It sounds like the pixel work wasn't a problem, but I suppose free AA is free (within the eDRAM).

 

[ShootMyMonkey]

hm... Do the TCRs not allow you to render at a higher resolution without MSAA? It sounds like the pixel work wasn't a problem, but I suppose free AA is free (within the eDRAM).

They most likely do... That is the TCR which is really quite easily the most loose. I think there was an issue there, that if we did that, we became limited on texel fill. It's quite common throughout TR:U that you have some materials with something like 10-12 texture layers or something. Most all of the environment geometry has a minimum of 6, I think. I'd have to look back on that. 360 having 16 TMUs makes sense that it would have some trouble keeping up if you oversampled a little bit.

I'm also a little wary about just how good the downsampling would be, though. But then, I'm far more negative about just about everything.

 

[DrJay24]

I finally tried the demo. I like the way Laura moves and swims, her climbing is smooth. I climbed up to the cave and picked up some things, but I could not figure out where to go after that, so I just swam around. I might go back to it when I have more time, I'm sure I missed 90% of it.

Laura tended to stick to things and warp a bit too. One part I would climb up and she would look like she fell through the geometry and them warp to the correct level.

The graphics are alright, about average this day and age. Nothing near Uncharted, but that is the way with most multi-platform games. The sub-HD just gives it a soft sterile look.

I loved blowing a shark up with a grenade.

 

[Daozang]

Hey, SMM, in the 360 demo, lara has a motion blur effect when doing jumps and cartweels that really ROCKS!. But it's missing from the pc version... Hope it's there in the full game. It was the only thing that I found better for the 360... Everything else was just a matter of adjusting the brightness-contrast of the game 'cause on default settings, it kind of looked flat on the pc.
One more game that I'll buy on release day. This Christmass, I'm going bankrupt!

 

[pjbliverpool]

Cool, I hadn't realiseed this game was out already, i'm downloading the demo now :smile:

This one will almost certainly go onto my buy list as I love all the TR games. Nothing too intense, just easy going adventure to play on a Frriday night with a beer or 6

I'd be interested to know what the reason is behind the missing feature in the PC version though (the version I'll be picking up), care to comment SMM?

EDIT: just read over the last 2 pages and it seems this was already explained. I'm a bit miffed tbh, every other game out there in this day and age manages to look as good or better on the PC and most of the time thats using DX9 so I see no reason for TR to come off worse. Especially when you consider what games like Warhead are doing shader wise under DX9 which goes way beyond what the consoles are capable of. And now that DX10 GPU's are cheaply available right down to entry level I don't see why this game couldn't have been developed under DX10 to get around shader limitations that DX9 might impose.

I paid more for PC hardware so that I could get a better gaming experience, not a second rate one. I'm afraid this just fell off my buy list.

 

[pjbliverpool]

Tried the demo and the game certainly looks good. I'll download the 360 version at some point to see if the differences are really as big as some are making out (better lighting, better textures etc..) I guessing not but we'll see.

Performance wise my GTS640 seems to handle 1360x768 + 16xAA +16xAF pretty well. Its usually somewhere between 30 and 60 fps but very occasionally falls below 30 when your looking out over the water from certain angles. Nothing to signifcantly damage gameplay though.

COntrols also seem ok with a 360 control pad, I'll be comparing that to the 360 version too once i've downloaded.

All in all, i'm impressed. If the game had been given a DX10 path to get around DX9 shader length limitations (its not like the hardware doesn;t have multiple times the shading capability of consoles) then it would be a definate buy. As it stands, its back on the possible list depending on how different the 360 version is.

 

[Phil]

BTW: is there a way to change the controls to Resident Evil style (left-right rotates Lara instead of running left/right). I never really got around to liking the Uncharted/MGS style of controls.. just didn't seem right after the playing so many good Tomb Raiders with the original control scheme...

 

[Daozang]

Tried the demo and the game certainly looks good. I'll download the 360 version at some point to see if the differences are really as big as some are making out (better lighting, better textures etc..) I guessing not but we'll see.

Either my eyes are biased (!!) or there is something wierd going on... To me it's the pc version that has better lighting, shadows and textures. The only think that isn't there is motion blur. What the 360 has, is good contrast and brightness with default settings...

 

[pjbliverpool]

I played a little of both for comparisons sake on my TV this morning and I honestly can't see any graphical differences between the two apart from the very subtle motion blur on Lara's jumps and the water being more foggy in the 360 version (not sure if thats supposed to be a good or bad thing).

The one big obvious difference though was image quality. The 360 version seems to suffer from quite a bit of aliasing while on my PC at least, that games looks as smooth as a CG movie.

Based on that alone I would prefer the PC version so I guess I will be picking this up eventually.

 

[Farid]

First the moderation stuff: Please let's keep the game comparison stuff out of this thread. It doesn't read as a sane discussion at all and only it bring noise to the mix. So, let's keep that detail comparisons betwen engines to the general game tech threads.

Second an actual question about the PC game: SMM, can you tell us whether CD handles the port in house or is it the Dutch folks from Nixxes that handle this version? I'd guess the latter, given the great relationship between CD and Nixxes, but a confirmation would be even better than a guess.

 

[Neb]

I paid more for PC hardware so that I could get a better gaming experience, not a second rate one. I'm afraid this just fell off my buy list.

Went straight off my list, "lazy" or bad optimistion work is a no no. Like with PS3 a different code approach is in need to achieve same effect suited to API and/or hardware. And as said everything in TR:U has been done before + waaay more all under DX9 at the same time.

I gave it a go under DX9 and although quite nice the texture detail, res and shadow res is average. The 2D photo backdrop also rears its ugly face when you get near. Would also be nicer with vegetation that looks more volumetric with more geometry since they look rather flat and "edgy". Lighting seems so and so, IMO doesnt give the scene uniform depth.

 

[pjbliverpool]

Just finished the demo, I actually thoguht it looked pretty damn good, one of the best looking console ports I've played in any case.

I'm not impressed that a technically more advanced and massively more powerful system isn't pulling off the same effects based purely on a software limitation which must have had possible to work around but I'm pretty sure i'll pick it up eventually.