Articles to write ? (Feedback)

[Kristof]

What are the articles we have to write to bring Beyond3d back to its old self ?

I was thinking about :

1. FSAA : the complete overview
-> Starting from the 3dfx document write a new updated version that includes all the latest techniques, sample patterns, efficiency. Remove the fluff from the 3dfx version and add-in things that were "accidentially" not mentioned.

2. Bandwidth saving techniques
-> Write a whole document about bandwidth saving techniques. Try to compare where possible. Identify all the possible bottlenecks in the system.

Other idea's are welcome. It will probably be next month before I can get to actually writing any of this. Holidays and GDC coming up.

Another Question :

How would you guys feel if all these bigger articles would appear online in a regular html form with banners and would also be available in PDF but for a small price (since PDFs have no advertising revenue). A system like this is being used on AcesHardware (http://www.aceshardware.com). Would you find this acceptable ?

_________________
Kristof

Developer Relations Engineer PowerVR Technologies (http://www.PowerVR.com)
Technical Editor Beyond3D (http://www.Beyond3D.com)

<font size=-1>[ This Message was edited by: Kristof on 2002-02-10 10:07 ]</font>

[Pete]

I agree with all three. I think a full GF4 vs. 8500 128MB review, with your in-depth review of minimum FPS and tech bottlenecks would also be a big traffic generator. Anything that brings the focus to consistent framerates over absurd max FPs numbers will be welcomed.

[mat]

I'd like to see a 8500 vs GF4 too, but it would be great if there is a 8500LE vs GF4 TI 4200 review. (everyone has a high end compare, but these cards are too expensive for me)

As for the PDF... i dont really need downloadable PDFs of any article (even if its free)... i mean i can just watch it as HTML and if i want it, i can save the whole HTML with pics on my harddrive.

I'd like too see a PS 1.1 vs 1.4 (NV vs ATI) too... whats possible with the new one that the old cant do? from what i'Ve read PS 1.4 has different instructions. Are the compatible with the old (1.1) ones or does ATI has to "emulate" PS 1.1?

<font size=-1>[ This Message was edited by: mat on 2002-02-10 10:46 ]</font>

[ram]

I'd rather prefer theoretical articles than reviews. Information in reviews is more focused on one particular product, while theoretical articles can be interessting for several products, even several generations of products.

[nggalai]

I second ram's notion.

I'd love to see an article about pixel shaders, btw. :smile:

ta,
.rb

________
Ducati GT1000 (http://www.cyclechaos.com/wiki/Ducati_GT1000)

[mboeller]

An article about the different FSAA-methods would be great ( OGSS;RGSS;Z3;Sparse Supersampling etc.. )

But how about an article about Rip-Mapping, Anisotropic filtering (different methods and taps) FAST; Footprint assembly etc. IMHO the filtering methods now in use will not be the end of it, but only the beginning, so an article about all the new more advanced forms of Texel-filtering would be great.


[edit] : Or how about older but golder FSAA and Filtering algorithm like the ZZ-buffer I just found minutes ago (improved form of A-buffer which approaches ray tracing in quality but is much faster according to the authors )

Manfred

<font size=-1>[ This Message was edited by: mboeller on 2002-02-10 15:15 ]</font>

[JF_Aidan_Pryde]

PDF idea is acceptable. I think the Full FSAA coverage is a great idea, looking foward to it.

[Brimstone]

How about an article on FC RAM? How might the reduction of latency effect graphic card performance? Lower latency should produce more effective use of clock cycles, so what kind of performance boost would be seen? Is Quad Data Rate FC RAM a possibility in future graphic cards?

Here is a link on FC RAM

http://www.fma.fujitsu.com/mem/memoryFAQ.asp?grOut=FAQ

[nAo]

On 2002-02-10 13:33, Brimstone wrote:
How about an article on FC RAM? How might the reduction of latency effect graphic card performance? Lower latency should produce more effective use of clock cycles, so what kind of performance boost would be seen?

Almost none. Pipelines on a GPU are so deep that most of memory latencies can be hidden trough prefetching. Low latencies make sense only with caches in a tipical GPU, imho.

ciao,
Marco

[Ailuros]

I'll second ram's notion too.

I'd just like to add a request for a theoretical "Multisampling AA on Tilers" article.

[Yannis]

About the articles-suggestions: Both seem fine to me. I would also like an article which would explain the higher color precision..what do we need 64bit color?

A nice article would be one that would explain the way a gfx card works. What are the registers etc...How do the new features work..

For the pdf: Personally I wouldn't pay not because I don't want to support B3d, but because I don't trust the internet security yet (credit card fraud) :smile:.
I would prefer the printable version of Anandtech...

Keep up the good work!

[nAo]

A question..can some one outside to the Beyond3d's staff provide informations or articles that will be published on the site ?(if it'll be a good work, of course :smile:)

ciao,
marco

[tEd]

I would like to see an article about lighting and shadowing. There are lot of words coming from developers and hardware makers.


realtime lighting,per pixel lighting,vertex lighting,dynamic lighting,phong shading,diffuse,specular,anistropic lighting

what does it mean exactly and what different approaches we looking at in games today, whats todays hardware capable of, stuff like that.

about the pfd files depends of how much it will cost.

[Dave Baumann]

A question..can some one outside to the Beyond3d's staff provide informations or articles that will be published on the site ?

We'd be gald to provide a forum for good articles to be published if they suit the site. Peter Greenhalgh (idris5) has had the Transmeta article here (http://216.12.218.25/domain/www.beyond3d.com/articles/transmeta/index1.php) for some time.

I believe Rev had a few others lined up prior to our downtime.
_________________
'Wavey' Dave
Beyond3D (http://www.beyond3d.com)

<font size=-1>[ This Message was edited by: DaveBaumann on 2002-02-10 16:04 ]</font>

[Doomtrooper]

No offence guys but whats the big deal about FSAA technology. With all the other neat features out there, pixel and vertex shaders, HSR , truform, vertex shaders, EBMM...etc... why is all this attention being placed on BLUR technology.
Does anyone here actually think it should take precedence over games engines like Codecult CODECREATURES Demo running at playable frames ??
A great example of what I'm talking about is the Aquanox review 4X Accuview shot on Chip Online.

http://www.chip.de/artikelbilder/1604243_80e76d4da9.jpg

With games becoming more complex and realistic looking modern video cards will not be able to use FSAA at playable frame rates..example above 24 fps..Yehhhh. As game continue to put more polys on the screen (UT 2) FSAA is not a option.

[Doomtrooper]

Does anyone actually think FSAA is a option with a engine like this:

http://www.codecreatures.com/screens/c-5.jpg
http://www.codecreatures.com/screens/c-1.jpg
http://www.codecreatures.com/screens/c-7.jpg

I know personally If I had a choice I'd take the above game with a few jaggies at 1280 x 1024 and play on a engine like that.

[Galilee]

Well there is a point with FSAA. Even if I don't use it a lot myself.
Even at 1600*1200 you have jaggies, and only FSAA can remove them.

So an article about FSAA would be great :smile: Love reading about that stuff. Also an article about anisotrophic filtering would be fun to read.

Why are you (Doom) always posting that picture from Aquamark? To show how damn fast GF4 is? :smile:

<font size=-1>[ This Message was edited by: Galilee on 2002-02-10 16:21 ]</font>

[Doomtrooper]

To show a example of how the fastest FSAA is card is brought to its knees with a modern game engine with lots of eye candy...simple to understand Galilee.
FSAA is not useable here, now apply that to the same engine from Codecult with rockets and planes flying overhead..explosions..high poly models running around.
Is FSAA a Feature gonna be used..I high doubt it.

[nggalai]

This is slightly OT, but still . . .

Doomtrooper, pray do explain how high/complex polygon load and FSAA are mutually exclusive as you imply with your posts in this thread.

ta,
.rb

________
vaporizers (http://vaporizers.net/vaporizers)

[Doomtrooper]

Nggalai,

Not sure if I understand your question but FSAA and high polycount games don't mix. Try running a game like Medal of Honor: Armored Assault with FSAA on and see what happens. Not even a Ti4600 could get playable frame rates with 4x FSAA on the OMHA beach map.
So my original request was to see more reviews done on ADVANCED features like PS, VS etc..vs the same old FSAA. I personally believe far too much time is spent on getting rid of a jaggies..but that is my opinion.

FSAA must basically draw the scene a couple of time depending on the sampling to my understanding, saying that... cards today can't give the performance required using a game engine similar to the one above and use FSAA too.

So IMO I'd rather see more reviews done on OTHER advanced features....as FSAA should be taking a backseat until we can get game engines like the one above playing with respectable frames @ 1280 x 1024 32 bit with 8 high poly models on the screen.

[Galilee]

Doom:
You would be surpriced how great 1024*768*32 and 4X FSAA works in MOHAA :smile:
Omaha drops to the 20+ :smile: but it's not that high without FSAA either :grin:

http://www.stud.ntnu.no/~vidaralm/Img/omaha1.jpg
http://www.stud.ntnu.no/~vidaralm/Img/omaha2.jpg
http://www.stud.ntnu.no/~vidaralm/Img/omaha3.jpg

<font size=-1>[ This Message was edited by: Galilee on 2002-02-10 18:51 ]</font>

[Doomtrooper]

Galilee,

There is NO way your getting playable frame rates on OMAHA beach with the Geometry set to this:

ftp://stuff:stuff@65.93.72.141/shot0000.jpg

ftp://stuff:stuff@65.93.72.141/shot0011.jpg
ftp://stuff:stuff@65.93.72.141/shot0008.jpg
ftp://stuff:stuff@65.93.72.141/omaha3.jpg





Tested and confirmed on a Radeon 8500, GTS, Geforce 3 classic, Geforce Ti 500.





<font size=-1>[ This Message was edited by: Doomtrooper on 2002-02-10 18:54 ]</font>

[Galilee]

No I had bilinear and only High on model detail, and shadow standard. Same on the other settings. I'll try yours setting.

Are those screens with 4X? Looks playable (30+)

Edit: I tried your setting: min fps 17. About 5-6frames lower than my previous setting. But you should not use Complex on the shadows. Meaningless.

http://www.stud.ntnu.no/~vidaralm/Img/omaha_new_1.jpg
http://www.stud.ntnu.no/~vidaralm/Img/omaha_new_2.jpg
http://www.stud.ntnu.no/~vidaralm/Img/omaha_new_3.jpg

<font size=-1>[ This Message was edited by: Galilee on 2002-02-10 19:12 ]</font>

[Doomtrooper]

Those screens are without FSAA, I took a shot of your screen shots and I see no difference in the jaggies vs. mine, especially on the cover debris in the water..still lots of jaggies ?? You sure you got FSAA on ?

Example:

ftp://stuff:stuff@65.93.72.141/omaha7.jpg


17 FPS is not realistic for playing online is it. I use eveything cranked for eyecandy, the shadows look great in the city levels.



<font size=-1>[ This Message was edited by: Doomtrooper on 2002-02-10 19:16 ]</font>

[tEd]

i'm playing mohaa right know with all to max except shadows on simple. I don't know the fps but it's very playable for me at 1024*768@32 with max anistropic and 2xFSAA.

i think anti-alisaing gets more and more important with higher polycount games.What i noticed in games lately (mohaa,rtcw,ss2) is that these games using a lot of trees created with alpha-textures and without proper anti-aliasing you get this annoying flickering what really bothers me.

[Galilee]

I think you are making the mistake to compare performance using 4X FSAA on a GF3, and 4X on a Radeon. 4X on GF3 is actually pretty playable.

Yes those pictures are with 4X FSAA, it's easy to see, but you had to find the only object on six screens that showed jaggies hehe. It's all about angles and stuff.

No 17fps online is not playable, but I didn't know Omaha was a online map.

but this is entirely off topic :wink:

<font size=-1>[ This Message was edited by: Galilee on 2002-02-10 19:28 ]</font>

[Doomtrooper]

Galilee,

Actually there is jaggies all over the place , just look at your photos. I just took the first one I saw. As for comparisons we tested on the same platform, 1.33 tbird on a Kt133 board..none of the cards gave playable framerates..so they all play with FSAA off. Try playing a city map like France with a 32 player server with as many as 20 models running around and watch FSAA enabled cards drop down to single digit frames.

Cards that were tested were:
Asus Geforce 2 GTS
Geforce 3 Classic
Geforce 3 Ti 500
Radeon 8500

None of these cards gave playable frame rates online with FSAA enabled plus Anistropic set to MAX.



<font size=-1>[ This Message was edited by: Doomtrooper on 2002-02-10 19:38 ]</font>

[Galilee]

On 2002-02-10 19:37, Doomtrooper wrote:
Galilee,

Actually there is jaggies all over the place , just look at your photos. I just took the first one I saw. As for comparisons we tested on the same platform, 1.33 tbird on a Kt133 board..none of the cards gave playable framerates..so they all play with FSAA off. Try playing a city map like France with a 32 player server with as many as 20 models running around and watch FSAA enabled cards drop down to single digit frames.

Cards that were tested were:
Asus Geforce 2 GTS
Geforce 3 Classic
Geforce 3 Ti 500
Radeon 8500

None of these cards gave playable frame rates online with FSAA enabled plus Anistropic set to MAX.



Aniso Max? When did this option get enabled? I used it off on my test. Max aniso on GF3 is killing the framerate, not so much together with FSAA, but it's definately lowering the framerate.

No there are not jaggies on my pictures (atleast not very visible). You used the third one I posted in the first post. It's the only place on all six pictures where jaggies are visible.

noFSAA:
http://www.stud.ntnu.no/~vidaralm/Img/omaha_noFSAA.jpg
2X:
http://www.stud.ntnu.no/~vidaralm/Img/omaha_2XFSAA.jpg
4X:
http://www.stud.ntnu.no/~vidaralm/Img/omaha_4XFSAA.jpg

<font size=-1>[ This Message was edited by: Galilee on 2002-02-10 19:57 ]</font>

[Doomtrooper]

Still jaggies on the walk off platform that falls down when landing on your 4X shot :smile:

Anyways try this, Anistropic MAX, set your 4X FSAA..join the France map and try and get a server with 30 players or more..take a screen shot then of your frames.
You can turn ANISTROPIC off if you wish, either way your looking at 8-20 fps.

This reason is why I'm on the FSAA kick is ' NOBODY I KNOW USES IT REALISTICALLY ' online.
Games like UT is a online game, so how realistic is it....

Race games etc...I see FSAA usefull as the games are not as demanding but advanced games and engines like Codegen shots are not realistic.


BTW Ted if you want to see your frames, enable the console in the advanced options-->tilde key--> type toggle fps

<font size=-1>[ This Message was edited by: Doomtrooper on 2002-02-10 21:17 ]</font>

[Doomtrooper]

Aniso Max? When did this option get enabled?


My shots are with ansitropic on as its on by default with the 6018 drivers. I always have anistropic on.

[MikeC]

On 2002-02-10 16:10, Doomtrooper wrote:
No offence guys but whats the big deal about FSAA technology. With all the other neat features out there, pixel and vertex shaders, HSR , truform, vertex shaders, EBMM...etc... why is all this attention being placed on BLUR technology.

Hang out a the forums at SimHQ for a while. They will tell you why AA is an important feature. Download and play the demos of IL2-Sturmovik and Nascar Racing 2002. Try Microsoft's Flight Simulator and you'll see why AA is necessary.

[MikeC]

On 2002-02-10 16:20, Galilee wrote:
Even at 1600*1200 you have jaggies, and only FSAA can remove them.

Even at 2048x1536 I can see jagged edges in Quake 3...

http://www.nvnews.net/reviews/prolink_gf3_ti200/page_6.shtml

[Doomtrooper]

On 2002-02-10 21:37, MikeC wrote:

On 2002-02-10 16:10, Doomtrooper wrote:
No offence guys but whats the big deal about FSAA technology. With all the other neat features out there, pixel and vertex shaders, HSR , truform, vertex shaders, EBMM...etc... why is all this attention being placed on BLUR technology.

Hang out a the forums at SimHQ for a while. They will tell you why AA is an important feature. Download and play the demos of IL2-Sturmovik and Nascar Racing 2002. Try Microsoft's Flight Simulator and you'll see why AA is necessary.



Sorry Mike,

For FSAA I'd rather have games looking like the codecreatures shots than a few less jaggies..to each their own.
I have Nascar 4 with Opengl Anistropic on it almost looks real at 1280 x 1024....
I asked on Gamespy last night in the MOHAA game room where there was at least 2000 people online and only two people replied that were using FSAA, and even then they were running 800 x 600.

[Althornin]

Doom, you and Galilee have hijacked this thread! If you wish to debate the merits of FSAA in various games with various cards, please create your won thread.


Kristof: I like The ideas of an FSAA article, a Texture Filtering Article, and a Pixel Shader Article. Those three seem to me to be the most important.

As for the PDF thing, i would never buy one. I can deal with a few ad banners for this site :smile:
_________________
YHWH is here.....
Use Trillian! AOL Sucks.

<font size=-1>[ This Message was edited by: Althornin on 2002-02-10 22:09 ]</font>

[Geeforcer]

Actually, in the future ultra high-poly games when the system is going to be limited by geometry rather then fillrate/bandwith FSAA could become "free".

[MikeC]

On 2002-02-10 21:56, Doomtrooper wrote:

For FSAA I'd rather have games looking like the codecreatures shots than a few less jaggies..to each their own.

Agreed, to each their own. However, one cannot discount that the future of 3D graphics is heading in the direction of making AA mainstream. There's really no cost effective alternative as it would take one hell of a resolution to completely eliminate jagged edges. We're still not there yet, but it's exciting to see the technology get closer to that level.

[bystander]

Kristof.

I agree with the first two articles.

On the third point however, I really don't like PDFs, so I don't think I'd be inclined to download one even if it were free. HTML pages with banners are perfectly fine with me. :smile:

[Dave Baumann]

Actually, in the future ultra high-poly games when the system is going to be limited by geometry rather then fillrate/bandwith FSAA could become "free"/

Actually, as Ghost of Envy said on the old forums, with increased usage of per pixel operations for 'high detail geometry emulation', and the increase programmability of future processors then its likely to be more of a 'computational' bottleneck in the short to medium term.

[Doomtrooper]

On 2002-02-10 22:15, MikeC wrote:

On 2002-02-10 21:56, Doomtrooper wrote:

For FSAA I'd rather have games looking like the codecreatures shots than a few less jaggies..to each their own.

Agreed, to each their own. However, one cannot discount that the future of 3D graphics is heading in the direction of making AA mainstream. There's really no cost effective alternative as it would take one hell of a resolution to completely eliminate jagged edges. We're still not there yet, but it's exciting to see the technology get closer to that level.


Actually Mike that is my arguement, I think FSAA is light years away from being mainstream when coupled with games like Aquanox, a very NICE looking game with lots of effects, FSAA becomes useless.
Is FSAA usefull..yes..is it reality with new games, IMO NO.
We have skipped a step here, we still haven't got the high polygon games running well before moving on to getting rid of jaggies.

[John Reynolds]

We have skipped a step here, we still haven't got the high polygon games running well before moving on to getting rid of jaggies.

Why must one come before the other? And I play quite a few recent and/or graphically demanding games with both AA and anisotropic filtering enabled (Max Payne, Sacrifice, etc.), so I'm not sure why you feel FSAA is years away from being a viable feature.

But perhaps we should return this thread to its intended topic and continue the discussion in nggalai's new one.

[Reverend]

Why, Kristof should have asked me b4 he posted this. :smile:

There is only one article I'd like to see, and if I have the time and no other website writes such an article in a scope that is at any length or depth that I'm personally satisfied with, I will write it myself (but it'll be difficult to realize, since I have to depend on other individuals). and that is :

1) Exactly how the 2 predominant APIs, DX and OGL, evolve with the various IHVs around
2) Exactly if and how developer feedback influence a chip's design
3) Exactly how a developer "designs" his game around a certain level of hardware capabilities, taking into account the all important "majority market" conundrum

These are just 3 that came OTOH. I personally do not believe in this site spending a lot of time explaining how things such as PS or VS works, otr how such a thing as aniso filtering works, or how FSAA may evolve (as these can be found elsewhere already, and as for FSAA, there are so many ways this can be done that "limiting" an article to discussing the possiblities may appear, er, close-minded or unknowledgeable).

I suppose, what I'd really like to see, hopefully on our site, is one article explaining in detail the intricacies of the evolvement of DX/OGL vis-a-vis IHV interaction... that is after all how the current hard-core crowd judge a piece of hardware, regardless of knowledge of the undeniable way the gaming industry works when it comes to using announced technology.

[Livecoma]

I'm really curious to learn more about how ATI, NVIDIA, PowerVR, Matrox, etc. manage their driver development.

For instance, as far as I can see they recieve input from two major sources. customers and developers. You have customers wanting better performance, control panels, bugs fixed with current games, and frequent updates etc. Then you have developers working on unreleased games spoting really specific bugs, and needing them fixed so they can make their games better. Also you gotta consider the chipmakers vision for their drivers, and resource concerns. QA in making sure fixing customer complaints dosn't unleash a new bug that a developer is going to complain about etc...


Pretty much I would like a little more insight into how a chipmaker melds all these concerns into a solid driver hehe. I think it may shed alot of light on peoples questions and concerns about drivers.

I don't know if that would be beyond the scope of the site though...

<font size=-1>[ This Message was edited by: Livecoma on 2002-02-11 05:00 ]</font>

[sriracha]

There's lots of other stuff I'm curious about:
- You always hear about Nvidia executing on their roadmap. Now that we have a few releases out under the roadmap, how good has Nvidia been at actually hitting it? What's been held back? How much has it evolved as they've learned more or had to beat competitors?
- With the programmability of the NV30 being hinted at now, a clear analysis of what aspects would be programmable and very different from recent releases and what would not be. E.g., obviously the memory subsys is not programmable, but shaders would be. What are their choices? Where are there possible 'abstractions' that we hear hinted at by Carmack and others, but not deeply discussed or explained? It would be interesting to hear developers' perspective in a coherent form, as opposed to just some Q&amp;A's.
- Along those lines, what are developers the most frustrated with about programming graphics engines under DX8, OGL, Nvidia or ATI's choices. Maybe related to above question.
- What's really keeping shading/bumpmaps out of games? Is it just the time lags in including new hardware features in a 2 year game dev cycle or is it too time consuming, are ther skill set probs or some other barrier to implement? With a new shader engine make a difference?

Overall, I'm more interested in the "take a step back and tell us the big things that we don't know because we're not in the industry" kind of questions than the "RGSS vs. Supersampling" minutae.

As for pdf, I personally prefer html.

[Kristof]

About Driver development and hardware development procedures... its pretty unlikely that any company will hand out details, they might give you some hyped up PR statement but I doubt you'll get any real in-depth answers on it.

[Nick[FM]]

I would like to read something on lighting and shadows too. I mean, 8 lights (HW) with realistic reflections and shadows is something you still don't find anywhere in games.

Also DirectX vs OpenGL (thorough) in raw performance, could be fun to read. I mean, why certain things might be faster in DX and some in OpenGL. Also why some manufacturers drivers are performing better in OpenGL but worse in DX (&amp; vice versa), and how they optimize stuff for certain engines etc. How they do it, etc. You get the point! :grin:

<font size=-1>[ This Message was edited by: worm[MadOnion.com] on 2002-02-11 11:59 ]</font>

[Nite_Hawk]

Hi Kristof,

The FSAA article could be pretty cool, though I feel like the last FSAA article here at b3d already gave me a pretty good understanding of how it generally works. (is that/can it be posted?)

I think another article that could be interesting is a comparison of the opengl extensions written by nvidia/ati, and the differences in directx revisions for each card. What and how do ati and nvidia expose features through through opengl/directx for their cards? Do nvidia and ati implement them similarly? How do developers and competitors react to each implementation, and how will the current implementations be affected by the release of DX9/opengl 2.0?

Nite_Hawk