Digital Foundry Article Technical Discussion Archive [2011]

Status
Not open for further replies.
The big enemies of high resolution triangle mesh geometry are memory and aliasing.
There's also the pixel shader quad utilization problem and numerous authoring and pipeline problems too.
These factors are not even close to being fully solved even for next-gen.
The memory can be addressed via tessellation, but to drive that you need hightmaps or high order surfaces (B-splines, sub-D surfaces etc). I don't know of any game studios using these in their production workflow.
The aliasing and quad utilization can be addressed by moving away from the current rasterization methods, but that is also a long way from being viable for production games.
 
I never said anything about eliminating normals I was implying they need to scale back the game design so where they aren't stretching memory requirements and getting results like the image I included.

I don't get what you mean here at all.
Are you suggesting more simple models for the source artwork or what? Have less characters on screen, smaller levels??

The only place where anything looks a bit problematic is in-engine cinematics, otherwise most of the characters aren't even using the highest MIP levels most of the time.


To be more clear on my last statement I meant creating a normal for a ultra high resolution mesh that wont even come close to the in-game mesh wont give you good results so making the original mesh closer to what you will get in game will yield better results.

Absolutely not true. You want the highest possible quality source geometry because the normal map exctraction can use supersampling. A 2K map has about 4 million texels' worth of info and you want a little more surface detail than a single poly per texel. Which is why most game character models nowadays are in the 10-25 million polygons range altogether.
 
I am not picking on Gears, just using it as an example because many games do the same thing. It just makes me wonder why devs keep pushing normal maps + color maps when they cannot keep the quality up. Just focus on better color maps and be done with it.
Many games have per pixel specularity, glossiness, displacement (parallax) and ambient (self occlusion) maps as well, and per pixel control for many other more specialized things such as emissivity (glowing surfaces). Sophisticated dynamic lighting needs much more than simple color texture to look realistic. Materials are fat, and thus we need to stream the textures if we want to have perfect quality everywhere (player might want a closeup of any object). With the most sophisticated texture streaming systems, the console RAM isn't the bottleneck anymore, it's the disc/HDD size. You only have limited amount of pixels at the screen once (720p = 921k pixels), you don't need to keep considerably more texels in memory if the streaming system is able to fetch the missing ones in a few milliseconds.

A new game could have 4k (4096*4096) textures on every object, but you couldn't have much different objects in the game, as you would run out of DVD space. 4k texture takes 16 times the storage space of 1k texture.
 
It will be intersting to see the PC version of Batman: Akham City which will apaprently be the first UE3 game with DX11 support. RockSteady is apprently adding Tesselation support to it among other additions (then again it's a NVIDIA TWIMTBP game so it may very well be a shity implemetation ala Crysis 2 DX11 patch)..

Meh... Batman is getting NVIDIA-Style Tesselation http://www.pcgameshardware.com/aid,...e-PC-version-DX11-tessellation-and-more/News/
Let's see if it's as bad and Crysis 2 and if it's simply done to make Radeon's look bad..
 
"Manually added" at least offers a little hope that it'll be worth the effort.

I sure hopes so. They also noted that all the PhysX/APEX effects can run on the CPU and not be artificially restricted to NVIDIA GPU (although they are using the old sucky none CPU freindly PhysX 2.8.4 so we can expect relatively crappy performance even on super high end CPUs)
 
http://www.eurogamer.net/articles/digitalfoundry-vs-console-battlefield-3

Interesting bits...
As you've seen in the performance analyses dotted throughout this piece, both games run at a capped 30 frames per second and in order to preserve visual fluidity and controller response, v-sync is disengaged when rendering runs over budget, resulting in screen-tear - though by and large, frame-rates are maintained fairly well on both consoles. Neither version acquits itself fantastically well in terms of the tearing, but here it appears as though the Xbox 360 game commands a small advantage. In the final urban section of the level, after players emerge from the Paris subway, Xbox 360 generally seems to offer a smoother experience while PS3 can drop frames significantly.

General framebuffer set-up is exactly as DICE claims - on PS3 at least. Here, resolution is definitely 1280x704, with tiny black borders top and bottom. We were expecting the same on Xbox 360 after a tweet from DICE's rendering architect, Johan Andersson, but the borders are absent on the Xbox 360 version. Perhaps it's full res, perhaps it's been upscaled - in combination with the anti-aliasing method chosen for the Microsoft console, it's very difficult to do any pixel-counting. Post-process AA is active on both console builds: Andersson has previously confirmed MLAA on the PS3, while NVIDIA's FXAA would seem to be a good fit for the Xbox 360 (there's been no word from DICE yet on this). Both seem to do the job as well as you would expect and in this regard, it's a significant upgrade from the old Frostbite titles.

BTW, remember guys when Joker said RSX was sloppy on geometry side back in 06? And then he said something pretty funny in FB2 thread?


If you want to discuss that just out of academic reasons then sure, but it's totally unrealistic. Having to wait 6 years for well funded rockstar developers or whoever to finally show the benefits of a design choice all but demonstrates how it was the wrong design choice all along. In this case it still hasn't proven itself even after 6 years, we'll have to wait and see if Frostbite 2 finally shows some advantage. I wonder what people will do if it turns out in the end to still have no advantage overall, even with Frostbite 2 engine. It will be funny to see if the Frostbite guys get hit with the "lazy dev" tag if after it's all said and done the 360 version ends up keeping pace with the ps3 version.
 
BTW, remember guys when Joker said RSX was sloppy on geometry side back in 06? And then he said something pretty funny in FB2 thread?

"If you want to discuss that just out of academic reasons then sure, but it's totally unrealistic. Having to wait 6 years for well funded rockstar developers or whoever to finally show the benefits of a design choice all but demonstrates how it was the wrong design choice all along. In this case it still hasn't proven itself even after 6 years, we'll have to wait and see if Frostbite 2 finally shows some advantage. I wonder what people will do if it turns out in the end to still have no advantage overall, even with Frostbite 2 engine. It will be funny to see if the Frostbite guys get hit with the "lazy dev" tag if after it's all said and done the 360 version ends up keeping pace with the ps3 version."

Crickets...
 
http://www.eurogamer.net/articles/digitalfoundry-vs-console-battlefield-3

Interesting bits...




BTW, remember guys when Joker said RSX was sloppy on geometry side back in 06? And then he said something pretty funny in FB2 thread?

If you want to discuss that just out of academic reasons then sure, but it's totally unrealistic. Having to wait 6 years for well funded rockstar developers or whoever to finally show the benefits of a design choice all but demonstrates how it was the wrong design choice all along. In this case it still hasn't proven itself even after 6 years, we'll have to wait and see if Frostbite 2 finally shows some advantage. I wonder what people will do if it turns out in the end to still have no advantage overall, even with Frostbite 2 engine. It will be funny to see if the Frostbite guys get hit with the "lazy dev" tag if after it's all said and done the 360 version ends up keeping pace with the ps3 version.

Whether it's matter of tech discussion I have to said a lot of things about the last comments...
 
Last edited by a moderator:
All I know is the 360 beta is quite satisfactory. Other than audio drops and some texture pop, it looks and runs great.
 
kind of weird that the result is other way around where the PS3 version slow down on physics and 360 version slows down on the fog. Shouldn't it be other way around?
 
kind of weird that the result is other way around where the PS3 version slow down on physics and 360 version slows down on the fog. Shouldn't it be other way around?

Its Japanese developer, You will never know what they will cook :p
 
kind of weird that the result is other way around where the PS3 version slow down on physics and 360 version slows down on the fog. Shouldn't it be other way around?

Even more odd, is the fact that both titles are apparently implemented with Sony's PhyreEngine middleware.
 
the middleware is available for free since day 1 and some multiplatform games like gripshift already use it ages ago. I don't think its a full engine like UE3, but should get dev started quickly. Kind of sad that audio always get overlooked in all the DF comparison, if a game is very similar on both platform, PS3 version always have the audio advantage, especially DTS sound that is supported by many 5.1 receiver. And 7.1 PCM is the best!
 
the middleware is available for free since day 1 and some multiplatform games like gripshift already use it ages ago. I don't think its a full engine like UE3, but should get dev started quickly. Kind of sad that audio always get overlooked in all the DF comparison, if a game is very similar on both platform, PS3 version always have the audio advantage, especially DTS sound that is supported by many 5.1 receiver. And 7.1 PCM is the best!

Huh? They just said in Dark Souls choose it for the PS3 audio, or friends...

I disagree though, a trivial audio difference sounds like really stretching it as a reason to recommend a platform, DF probably only does it cause of all the backlash they get from PS3 fans. I'd say even something like controller preference, or cross game chat, etc rates a much bigger factor than that.
 
Status
Not open for further replies.
Back
Top