DX10 - How Far Have We Come?

MojoKid writes "When DirectX 10 was first introduced to the market by graphics manufacturers and subsequently supported by Windows Vista, it was generally understood that adoption by game developers was going to be more of a slow migration than a quick flip of a switch. That said, nearly a year later, the question is how far have we come? An article at the HotHardware site showcases many of the most popular DX10-capable game engines, like Bioshock , World In Conflict , Call of Juarez, Lost Planet, and Company of Heroes, and features current image quality comparisons versus DX9 modes with each. The article also details performance levels across many of the more popular graphics cards, from both the mid-range and high-end." PC Perspective has a similar look at DX10 performance.

Comments

DX9 looks better?

[Chemisor]

Am I the only one who find the DX9 version of the pictures more appealing? With the exception of the Bioshock fog examples (which had sharp boundaries in DX9) they just look more "natural" to me.

Re:DX9 looks better?

[b100dian]

If you look at "Company of Heroes - Image quality", the first "grass effects" comparison shows am octogonal wheel.
I mean, 2007! and we still have octogonal circles!!

I think that the "realism" isn't worth it. Go out and create DX7 games that are fun :P !! (or openGL games that don't require much extensions;)

Re:DX9 looks better?

[fbjon]

Yes, a lot of the differences also seemed to be unrelated to DX10/DX9 differences. The soft fog and some lighting effects is really the only feature of significance I could see.

Re:DX9 looks better?

[Poromenos1]

TF2 is the best looking game I've seen in a while, and it's not realistic. That's a big part of why it's so much fun. I've died numerous times from standing around looking at stuff :(

Re:DX9 looks better?

[Presence2]

I second that, the positive differences between DX9 and 10 are not significant enough to warrant the negative differences between XP and Vista.

I'd much rather see game developers expend their man-hours on making PC games creative and better to play (and in a perfect world, not restrictively ready-to-port-to-console), rather then focus on making them graphically unique to DX10.

Re:DX9 looks better?

[QuietYou]

When you consider the FPS differences of the tests, it's not a fair comparison of the abilities of each each API.

On one test the DX9 version was running at 110fps and the DX10 version running at 30fps. The DX10 version damn well better haver higher image quality if it takes nearly 4 times as long to render a scene. Push the DX9 version futher by throwing more polygons and more complex shaders at it until you reach the performance of the DX10 version, THEN do a comparison. You'll find that there is precious little image quality difference between the two when you're pushing BOTH versions to their limits. The DX9 version might even look better.

Re:DX9 looks better?

[suv4x4]

Am I the only one who find the DX9 version of the pictures more appealing? With the exception of the Bioshock fog examples (which had sharp boundaries in DX9) they just look more "natural" to me.

Some did, some didn't.

You gotta understand that DX10 can do absolutely everything DX9 can, so if the DX10 image looks less natural, it's more of a human flaw than technological: it's a new area and people are only starting to discover what works best, both devs and designers.

Also I imagine that fine-tuned the DX9 version more since the majority of people out there have DX9 cards. DX10 are barely out there, they probably don't even have a good selection of DX10 cards yet to test everything thoroughly.

The only thing that worries me is that DX10 shows up slower on the benchmarks. DX10 was promised to have better performance than DX9, but don't forget all of the reviewed game use different code paths for DX10, thus load more effects and use higher precision buffers/processes in the DX10 versions. So while DX10 may be faster, it's not a fair comparison when DX10 is loaded with twice the texture sizes and effects of the DX9 version.

We'll need a more objective test written to use the same elements in DX9 and 10 and compare that.

One way or the other DX10 is the future. Even if the first few generation suck, the new features show lots of promise that will come to fruit in the coming years. DX10 has no choice but to become great. If you don't want to burn, just don't buy DX10 card YET, it's the worst moment to do so.

Wait at least until there's a DX 10.1 card out there with good price and review (DX 10.1 will come with Vista SP1). I don't expect this to be before Q3-4 2008 (which is great since Microsoft would have fixed lots of things in Vista by then, and 3rd parties would have better drivers and hardware for Vista).

Re:DX9 looks better?

[DAldredge]

What is so bad about vista?

Re:DX9 looks better?

[jez9999]

I think that XP+DX9 would have made useful supplements to the results they gave, but their goal was to measure DX9 vs DX10, and you don't do that by changing two variables.

Yup. That's when I tested the speed of my car vs a train, I ran the car on the tracks. I was testing the speed of a car vs train, and you don't do that by changing two variables.

DX10 still Windows Vista only?

[shawnmchorse]

Then the answer is going to have to be "not very far". I can't see game developers getting that excited about something supported only on a version of the operating system that people are specifically NOT migrating to in droves.

Wow DX10

[Dusty00]

These numbers to me validate my suspicion that DX10 was nothing more than a cheep angle to sell Vista. The performance isn't a tremendous improvement and the resulting graphics are enough of an improvement that I'm going to let Vista suck down that much of my hardware.

How far have WE come?

[VeteranNoob]

What's this "we" business? DX10 is only available with Vista, and Vista sales are abysmal. And with this being a *nix-oriented site, it's falling on deaf ears.

The summary states that DirectX 10 was "introduced" to by the hardware manufacturers and Windows adopted it. I have always understood it to be the other way around. If it is the hardware makers, then why are they actively supporting two different 3D APIs (DX, OpenGL)? Does this mean that DirectX could be adopted by another OS, say Linux? Only for a fee?

I urge everyone to vote with their wallet and try games that support OpenGL and Linux. Sure, you may never get to play Halo 3 in Linux, but if the game developers see the market growing there, I'm sure we'll start to see more big names soon. It will be to their ultimate benefit, too, since they can take advantage of the advanced technologies that Linux has to offer (mostly for free). Can you imagine games or applications that worked with Beryl to create actual 3D desktop objects instead of just 2D windows. Or how about a Linux LiveCD that did nothing but boot a kernel with drivers and ran a dedicated game; every single CPU cycle would be dedicated to giving you frames per second.

Personally, I've been addicted to this great 3D bridge-building game, Bridge Construction Set. Of course, it supports Linux

The real joke

[Moryath]

The real joke is that neither DX9 nor DX10 are inherently "better" any more than the original Glide API was inherently "better" than DirectX or OpenGL. Hardware has been changed constantly, to give "better" responses to this call or that call, but inevitably you have to write a driver that converts the OpenGL or DX9/DX10 or whatever into something your card understands.

In the really old days, you had people actually coding for the card on hand. This is why there's a gazillion different releases of Mechwarrior 2, each of which varies greatly in image quality and features - each had to be hand tuned to the card.

If Bioshock had been intended for DX9, it would probably look the same as that DX10 shot on DX9. They'd have figured out what they needed to do, perhaps coded a few "If ATI, do this, if NVidia, do this, if Intel Extreme fail 'your video card is too crappy to play this game'" decisions for specific hardware, and that would have been that. Since it was backported (and MS would have thrown a fit to have "no difference") they had to just do a more slappy job of it.

Then again, if not for the emphasis on ridiculous graphics, think about how many games would be able to use their processing power for some seriously wicked AI. Even Bioshock only has half-decent AI that can be twigged to and predicted fairly easily - you know that a wrench guy is going to rush you, you know that the spider slicers will hang from the ceiling and lob stuff all day till you burn or freeze them, you know where the houdinis are going to land long before the animation starts merely because you can figure out what the AI tree says for them to do in what radius... it's sad.

Hell, you can predict the precise spot on the health bar where they'll run for the health station, and if you're smart you trapped that thing half an hour ago. Now you get to watch as four of them all kill themselves on the same damn one, never paying attention to the 3 dead bodies on the floor that obviously just gassed themselve using a booby-trapped station.

But nevermind. I know the reason they want graphics over AI - the same fucking morons that could never defeat a decently programmed AI (hell, they have trouble getting through Halo on NORMAL), drool over thinking that they can see the shoelaces on Madden's boot.

Re:The real joke

[Mikachu]

Hold on for a second here. The graphics are usually very GPU intensive, but the CPU is generally not overworked by them at all. If they wanted to write good AI, they could do so without sacrificing graphics quality at all.

Re:The real joke

[pyrbrand]

I don't think you get it. There's a reason people aren't writing assembly any more and there's also a reason they wrote in assembly instead of 1's and 0's. Yes, it's technically possible to write everything you write in C++ in binary, heck, the compiler and linker pump that out for you, but the point is to be more productive and make less errors so that you can get more done in the limited time you have. In that sense, it's way better to have modern programming tools since you can finish up the graphics and have time left over for AI too.

Further, your comment that we haven't come further on the technology train since glide is just wrong. New APIs provide access to new hardware capabilities. The advent of shaders and a breakaway from the fixed function pipeline is what makes games look different from each other these days instead of seeing the same plastic-y models in the same hyperreal lighting only with the dials turned up or down. DirectX10 takes this even further and instead of just being able to custom program the geometry shader part of the pipeline which thin has to pass things off to the pixel shader part of the pipeline which you can custom program, the whole pipeline becomes programmable giving a developer a lot more flexibility.

Lastly, the point of AI isn't necessarily to make things brilliant, and even in games where people think the AI is brilliant it can sometimes be brain dead simple. If I recall correctly, FEAR only had 3 states to its AI, but the system was so brilliant, people thought the AI was far more complex and assigned it intentions based on why they assumed it was doing what it was doing. Also, an actually intelligent AI isn't any fun to play. Most of the time, you will get shot in the head before you see the bad guys (since you're field of view is smaller than your field of, uh, not-view). That's why in games like Half Life 2, the AI guys were smart and did things like make sure the AI always misses the first shot. There's a great article on these sorts of design choices by one of the guys who did the AI for HL2 - I think it was on Gamasutra although I can't find it now, so it might have been in one of the AI programming gems books. Having predictable behavior is also one of the fun parts of games - learning how to outsmart the system can be entertaining itself (and gives you that smug feeling you're emoting when you talk about how you can predict the behavior of all those Bioshock baddies). Not saying it's the only way, just that it can be fun in itself. Completely erratic an unpredictable AI wouldn't be any fun either. That said, more time spent tuning and play testing and redesigning AI is never a bad thing, I'm just questioning your assumption that the AI should be smarter - I prefer more fun.

Re:The real joke

[Anonymous Coward]

If Bioshock had been intended for DX9, it would probably look the same as that DX10 shot on DX9.

I can tell you for a fact that Bioshock was coded for DX9 and forward ported for DX10 in the last months of development. Features weren't sacrificed to go to DX9, they were simply added when the DX10 rendering path was made. Do check your facts before making disparaging comments...

obvious parallel

[fred fleenblat]

I don't think DirectX 10 will achieve any kind of market acceptance until DirectX 11 is released. Then everyone will bitch about DirectX 11's high-end hardware requirements, DRM lockdowns, and poor performance and they'll start clamoring for the good old days of Direct X 10.

Re:Shadows are wrong!

[TheNetAvenger]

That's great, except for the fact that shadows don't have crisp edges in the real world. Unless it's illuminated by a point-source (which immediately excludes the sun, lamps, flashlights, and pretty much every other light source you're likely to encounter), there will be a penumbra. The DX9 image here: http://www.hothardware.com/articleimages/item1031/big_stateofdx10_wic_shad.jpg is more realistic.

Not sure how this got confused by either bioshock or the reviewers...

DirectX 10 allows for both 'crisp' or 'soft' shadowing, as some games demonstrate, the DirectX 10 shadows are 'softer' and more realistic.

The 'difference' with DirectX 10 is that shadows are done on the GPU, in DirectX9 shadows are done on the CPU. This is the 'main' difference between DX9 and DX10.

The 'crisp' choice by bioshock is NOT what DX10 is about, this is a game developer choice. PERIOD.

I know reviews like this can lead people down wrong paths, but it doesn't hurt to look up this type of information before making fun of a fact that is incorrect in the first place.

It is strange that any site 'reviewing' DX10 in comparison to DX9 would not even know the basic 'consumer' terminology for the differences, so they would know what they were looking at... Maybe someday we can get a review posted on SlashDot that is actually done by gaming professionals... (gasp)

Here is a quick list from the MS Consumer Info site on DirectX10, notice the reference to shadows specifically.
-----------------------
Summary

In summary, DirectX10 provides the following benefits to gamers:

More life-like materials and characters with:
Animated fur & vegetation
Softer/sharper shadows
Richer scenes; complex environments
Thicker forests, larger armies!
Dynamic and ever-changing in-game scenarios
Realistic motion blurring
Volumetric effects
Thicker, more realistic smoke/clouds

Other
Realistic reflections/refractions on water/cars/glass
Reduced load on CPU
    -Re-routes bulk of graphics processing to GPU
    -Avoids glitching & system hangs during game play

Re: ai

[Anonymous Coward]

As someone who writes AI for text-based games, let me clear you of some misconceptions.

First, the goal of "AI" isn't always to be as smart as possible. Often, the goal is to make something believable and/or of the appropriate difficulty level. It's possible that Bioshock missed the mark there, but I haven't played Bioshock yet, so I don't know.

I can write "AI" that will kick your ass every time, even without cheating. (Mobs have the advantage of being on home turf, and they outnumber you.) But that's not fun for the player, so I don't do it. Instead, I'll write something with a pattern you have to figure out. Once you learn one of the ways to beat it, the mob will be easy for you, and it's time to move on to the next area. Very few mobs get the full "try to survive at all cost" treatment, and even fewer are programmed to actually learn from your behavior.

You're describing the classic "I wish this mob would keep getting harder" remorse, but think about it: would it really make sense for those mobs to learn from your new tactics? Are they supposed to be smart, or are they just supposed to be an obstacle?

As for your dead bodies example: would you really prefer to have an infinite standoff as the mobs decide it's not worth getting killed, so they go hide somewhere with their own traps and wait for you to attack? Right... so get over it. If games were realistic, you would realdie on level 1.

Re: ai

[Chemisor]

> I can write "AI" that will kick your ass every time, even without cheating.
> (Mobs have the advantage of being on home turf, and they outnumber you.)

You are assuming that the mob would just sit there and wait for the player, like it usually does in pretty much every game. In reality, a "level" would not necessarily know that Gordon Freeman is on his way. Neither will they have the patience to sit in their assigned ambush places, waiting for him all day long. A better AI would actually "live" in the environment where it is placed, so that it would react to the player instead of waiting for him. It would also be fun to watch. In Half-Life I really enjoyed watching those occasional scenes where monsters are wondering around doing things; like when the bullsquids feed on the headcrabs. I wish there were more things like that, things worth watching.

> would it really make sense for those mobs to learn from your new tactics?
> Are they supposed to be smart, or are they just supposed to be an obstacle?

If the AI was smart, you wouldn't need a mob. You would only need a few individuals. It would be like a multiplayer deathmatch, and, judging from the popularity of those, would likely be more fun than the current mob situation.

> As for your dead bodies example: would you really prefer to have an infinite standoff
> as the mobs decide it's not worth getting killed, so they go hide somewhere with their
> own traps and wait for you to attack?

An infinite standoff will only happen if the game designer makes you kill off the entire mob before setting off some stupid trigger to open some stupid door. Don't program artificial obstacles and the player will be able to ignore the hiding mob and go on, just like in real life.

Re:Just as far as it needs to to displace OpenGL.

[Tim C]

Thinking for yourself rather than letting Slashdot do it for you.