The Technology of GTA IV/RDR *Rage Engine*

[MazingerDUDE]

I think I finally understand why some would say the PS3 version looks better because artistically the blur of the PS3 image blends in nicely with the blur background/silhouette in the distance like a natural DOF, where the 360 image is clear and crisp then suddenly jarringling blurred when creeping into the distant background/silhouette

While dithered image of the 360 is more easy on your eyes, it's the blurred PS3 image that actually provides with more details, more color. You might not notice this at a first glance, but when you spend 10+ hours playing, things like these would add up to impress you, and I think that's why people who have spent considerable amount of time on both versions thinks that the PS3 version actually looks better.

 

[Mintmaster]

Interesting. I guess one of two things was going on then. Either we had extreme overdraw (at the time), or there was a transparent pass going the fp16 route.

I recall you mentioning a fairly high polygon count, so heavy opaque overdraw is quite possible now that I think about it.

If you have 2M triangles on screen, that could easily mean 5-10M excess pixels from quad-level rasterization, and that level of overdraw (5-10x) would definately hurt you with FP16. I can certainly see nAo's point about quad-level inefficiency now.

 

[nAo]

That's why I love z pre passes
Modern GPUs are super fast at that!

 

[rapso]

Just take a look at the sign on the center 'DENTAL CENTER' which is clearly readable on PS3, but NOT readable at all on 360.

Also check out the phone number on the right side 555-3263


* added highlights, and increased gamma, so those with faded monitors can read as well

Maybe it's some noise/dithering to hide streaming latency, as long as you don't try to read text, it probably looks better then interpolated textures.

 

[Shifty Geezer]

No wonder people like versus threads and comparing PS3 and 360 screenshots - it's what we've been primed to do in kindergarten.

And then some spoilsports ruins it all by highlighting the differences in pen!

 

[Fafalada]

I've yet to experience tearing on PS3.
It happened about 3~4 times on 360 that lasted until weather/lighting changes.
So, I suppose it happens only in certain circumstances (it makes cut scenes tear as well) and I'm pretty sure it's got something to do with weather/time.

Thing is, the few impressions that mention fps differences all sound exactly like what you see when you take a generally sub 30fps VSynced game and run it with VSync off.

With the supposed lack of tearing on PS3, I'm wondering if they just left VSync enabled. If that's indeed the case this could serve as nice insight into what people actually prefer - tearing or a touch more juddery fps.

 

[grandmaster]

If you're that interested, check out TimeShift on PS3 and 360. The 360 version has a v-lock option (although it's still prone to a bit of tear even with it enabled) while the PS3 version has no such option and looks bloody awful.

The difference with GTA - unlike TimeShift - is that the screen tear is barely noticeable whatsoever, even when I'm scanning through captures on a frame by frame basis. The fact that it is there at all on 360 is a curiosity, but certainly nothing anywhere near the level of the other differences between the two versions.

 

[nAo]

Not many games do it, but it's obviously possible on PS3 as well to vsync if the game runs at 30 (60) fps or more and hsync if it's slower than that.

 

[Fafalada]

grandmaster, I'm not talking about tearing being noticeable or not - my last project I've gone with no VSync if below 30 and people that have seen both were overwhelmingly in favour of that over always VSync.
And yea - in the said case tearing was barely noticeable as well.

I was merely bringing this up because comments from a few people like "360 seems to have higher max fps" are rather symptomatic of no-VSync above 30 (or if game is mostly under 30, VSync off below 30).
I think we need someone that does FPS counting in games like Quaz does pixels

 

[nAo]

I think we need someone that does FPS counting in games like Quaz does pixels

This is going to be even easier than counting pixels!

 

[AlNom]

oh lordy.

 

[MazingerDUDE]

did some more 200% shots

http://www3.telus.net/public/dhwag/E.jpg
http://www3.telus.net/public/dhwag/F.jpg

http://www3.telus.net/public/dhwag/I.jpg
http://www3.telus.net/public/dhwag/J.jpg

http://www3.telus.net/public/dhwag/L.jpg
http://www3.telus.net/public/dhwag/M.jpg

http://www3.telus.net/public/dhwag/P.jpg
http://www3.telus.net/public/dhwag/Q.jpg

 

[Scott_Arm]

PS3 is definitely the cleaner of the two. The speckles on the 360 almost remind me of artifacts I've seen on the PC when I overclocked my video card. It definitely doesn't look intentional. If it's a bug related to software, you'd think they would have caught it. I don't notice it when I'm playing, but in these screenshots it's very apparent.

 

[grandmaster]

I've got a tool that allows me to count the number of unique frames in a video capture and gives an average frame rate. As my captures of GTA are 720p60 24-bit RGB lossless HDMI there's no scope for analogue capture or frame rate decimation affecting the result. The trick is to make sure that your sample clips are the same and have no areas of zero movement.

As the whole game is rendered using the same engine, measuring cut-scenes should suffice. I'm on a crushing deadline at the moment (only posting while After Effects renders!) but I'll give it a go later in the week. Gut feeling based on running clips side by side is that 360 has the advantage but right now that's entirely subjective.

One thing to note about the screen tear on 360. The reason it's basically unnoticeable is that an affected frame typically only stays on screen for one frame before going - that's 0.016 seconds. There's no continuous tearing (the Ubisoft effect if you will).

 

[pjbliverpool]

Lol, you tell the internet you have the ability to directly compare the framerate of GTA4 on PS3 and Xbox 360 and then say you can't do it until later in the week!

What a tease!

Looking forward to your results!

 

[grandmaster]

One thing I did observe is that the game seems to have a fluctuating frame rate… a 30fps game should be:

AABBCCDDEEFF etc

However, on a cut-scene I looked at, the frames were:

AABBCCCCDEEEFGGHHIJJKKLLMN

 

[Fafalada]

One thing I did observe is that the game seems to have a fluctuating frame rate…

I don't think that was ever in doubt - it's a T2 game after all. Anyway I take it you can also scan for tearing in your test (to confirm if it happens in PS3 version or not).

 

[grandmaster]

PS3 doesn't tear at all based on five days working on the captures - all 500gb of them (at 3:1 lossless compression).

The scanner will detect a torn frame as a new frame - not much I can do about that. This might be seen to artificially boost 360's frame rate, but it's going to come down to choosing clips where the tear isn't evident, which shouldn't be too difficult. And if a clip is, say, 1000 frames, 2-3 rogue frames won't make much difference to the average.

 

[Zassk]

Looking at those latest comparison shots... is that temporal AA on the Xbox 360? Temporal AA would look speckled when frozen in a screenshot, wouldn't it?

 

[AlphaWolf]

Looking at those latest comparison shots... is that temporal AA on the Xbox 360? Temporal AA would look speckled when frozen in a screenshot, wouldn't it?

No. The speckling is some sort of dithering that becomes more pronounced with the zoomed images. It's pretty easy to see the AA on the railing in the zoomed 360 shots, although with the scaling it just looks wrong.