Reverend at the Pulpit #1

[Reverend]

(Ignoring the current favourite war topic) I'm sure nobody likes to read my ramblings... but so what!?

Rev's World
As America/Bush appears to have won a war they claim is essential for "human rights and world-wide peace-of-mind (as opposed to simply "peace")", I still fail to believe any human being is capable of being so completely righteous. Maybe Bush is such a righteous person, who knows, in detail as provided by his intelligence employees, the terror of Saddam (and I, for one, do not believe Saddam is a "good" man) but IMO I am sure this is also about 911, his father and money in addition to him being possibly absolutely more concerned about the world than his own country. I just hope the rich countries will contribute the money to help the Iraqi people -- while it is undeniably sad to see Iraqi kids without arms/legs due to solely to America's bombs, I just hope such kids will live to either see another day or are, however impossible it is to be the case due to their age/immaturity, aware that they are sacrificial lambs, determined by Bush/America to be so, that will benefit their future Iraqi generations.

Rev's PC Games
Ah, the "It's what it's all about" stuff that appears to have lost its apparent priority based on its forum description at our site'o'love. Anyway, I am embarassed to say that my current fav game is FIFA 2002 -- with the current English Premier League (EPL) scenario and being a EPL fanatic, it should be understandable, no? I've been a Gunners fan for as long as I can remember, and if Arsenal does not win the EPL title, I sure hope they comprehensively beat ManU tomorrow. Other than this, I had been helping Ubisoft with their v1.2 Splinter Cell, specifically wrt its benchmarking feature, and then I'd been talking to various devs about their perception about the current video card/video card company scenario... as usual, their opinions are different in certain aspects from the usual "hardware-centric" crowd here at B3D (vs software-centric) and it is certainly a little trying at times to differentiate the differences and what's important and what's not when it comes to my software-based knowledge and B3D's hardware-based concentration.

Rev's B3D involvement
Not having a regular vendor supplier means not having much to do/say officially at B3D. Of course, Dave appears to have all things covered and even though he may complain he's been extremely busy wrt site-stuff, I'm sure both he and you guys enjoy hearing him say that . Of course, I'm trying to gain new good-relationships with a few vendors but this is completely governed by B3D/Dave -- hey, I can't review new products if Dave hasn't already! Anyways, I'm currently in the midst of coming up with synthetic benchmarking programs that will hopefully help B3D in its bid to be the premier 3D investigation site (if it isn't already ).

Rev's Health
I have the flu -- so what else is new... I have this at least twice a year. The SARS scare is real and while I haven't been to a doc for this flu of mine, I don't think I need to -- it's just the bloody nose, minus any SARS-type symptoms (dry cough, fever, tough-to-breath). More important to me is how I can cope with restricting myself from playing too often with my 4+ kid (since I don't want him to get the flu) as that is one of the most enjoyable things I look forward to everyday.

Rev's things to-look-forward to
DOOM3 (tried to get B3D to be its beta-test member, with no apparent success forthcoming), 3DMark04, Half-Life2, MOHAA2 and for Dave to send me some hardware-love (hint, hint ).

PS. I anticipate this to be part of a "Rev at the Pulpit" series at this forum -- I will have more to say about many other stuff in the future... so sue me if this isn't interesting... I'm part of this site, dammit!

 

[Randell]

Other than this, I had been helping Ubisoft with their v1.2 Splinter Cell, specifically wrt its benchmarking feature

what is your take on the different code paths used by the benchmark for different cards - how can it be a fair benchmark if different shadow and lighting engines are used? What are the realistic differences between the Class 1 and 2 code paths and why does the R3xx use the Class 1 path, isnt it capable of everything in the class 2 path?

 

[Reverend]

Other than this, I had been helping Ubisoft with their v1.2 Splinter Cell, specifically wrt its benchmarking feature

what is your take on the different code paths used by the benchmark for different cards - how can it be a fair benchmark if different shadow and lighting engines are used?

We can specify shadow projection to be used by all video cards for apples-to-apples/fair comparative benchmarking.

What are the realistic differences between the Class 1 and 2 code paths

The v1.2 patch (which, btw, looks like Ubi's QA dept didn't do its job properly -- they forgot to include the three benchmarking demos although they have batch files to benchmark these three demos! Use B3D's own Splinter Cell benchmark demo instead, that should work fine) has a patch-info readme that explains the differences. As Ubi says, shadow buffer algorithm is basically easier to implement than shadow projection algorithm. The main differences between the two modes are :

shadow resolution
shadow buffer mode = max 2048x1024
shadow projection mode = 512x512

rendering
shadow buffer mode = Z only
shadow projection mode = intensity/color

shadow quality
shadow buffer mode = sharp shadows edges
shadow projection mode = soft shadows edges

and why does the R3xx use the Class 1 path, isnt it capable of everything in the class 2 path?

The difference is the shadow mode used -- The GF3/4Ti/FX/Xbox shadow buffer algorithm is a straight forward implementation of a shadow buffer algorithm using the NV capability to render into depth textures. This feature is not available, under DX8.1, with the R3xx core.

The main thing is that we can force apples-to-apples shadow settings, using shadow projection algorithm. NV cards will default to shadow buffer, so video card reviewers should be careful to set NV cards to use shadow projection if they intend to do shootouts between NV and ATi cards.

The following is also useful info in case I never mentioned this before :

Ubisoft's re-worked Unreal engine (the Unreal build they use is way, way old, nothing like UT) never goes beyond 6 passes per triangle for it's main rendering loop. This means that the R2xx are able to do everything in a single pass while the GF4Ti sometimes takes 2 passes. The main rendering loop is not using pixel shaders. On a R3xx card you need 1 pass for general texturing, 1 pass per shadow (single texturing) + screen post processing (glow, special effect filters). Overall, 2-4 passes per triangle (if we forget about the post processing that is not polygon base). When rendering as a class 0 adaptors, several textures are removed and not rendered to speed up rendering and avoid spending something like 4 passes per triangle for general texturing. Details textures and most of the specular reflection is gone with those cards.

 

[Randell]

Rev,

Thanks for the detailed reply. Are Gf4's faster using shadow volumes or shadow projections? I guess shadow volumes or there's no point

 

[Reverend]

I haven't tested SC on an NV card so I wouldn't know the difference (if any) in performance between buffer-mode and projection-mode. Maybe Dave can fill us in when he has time to test all those NV and ATi cards he has.

However, I'm more than a little depressed that ATi cards display this while NV cards display this as a result of the visual difference between buffer-mode and projection-mode. The difference, as you can see, is pretty stark and I hear that SC fans with ATi cards are complaining to no end about being "cheated", hehe. If I were ever a NVIDIA-hater, this game alone may convert me to buying a NV card!!

 

[Randell]

yes, I'm aware of that effect. So no ATI card can do that, if Ubi-soft had bothered to try? Is that also why shadows kill performance on ATI cards in NWN, Bioware used shadow buffers or is there another reason?
Can I get the benchmark to work on the demo version?

[H]OCP seemed to indicate you could.

 

[Reverend]

So no ATI card can do that, if Ubi-soft had bothered to try?

Ubisoft can't even try, because there is no hardware support in R2xx/R3xx.

Is that also why shadows kill performance on ATI cards in NWN, Bioware used shadow buffers or is there another reason?

Don't know enough about NWN to make any comments I'm afraid.

Can I get the benchmark to work on the demo version?


[H]OCP seemed to indicate you could.

Well if HOCP can do it, I doubt that you can't but I think you'll need to get the relevant files (recorded demo files, maybe some game dlls or even maybe a special executable, dunno really) from someone (I don't have the files related to getting this to work on the demo version). There was supposed to be a new official demo with benchmarking feature but I don't know what's happened to that.

 

[Randell]

There was supposed to be a new official demo with benchmarking feature but I don't know what's happened to that.

That's what I was thinking off. I have a new demo with 3 maps in it, but no mention of a benchmarking facility.

 

[Reverend]

Okay, sorry... now I remember. Ubisoft did give me files for benchmarking to work on SC demo but I don't think I am allowed to distribute that.

 

[Randell]

fair enough, I'll wait and see when it comes out properly. Thanks Rev.

hmm I might get Splinter Cell on X-Box then, that way I get proper 'nVidia' shadows