[Response to Flaconfly]

I never really though of it like you pointed out -- but I have kept up my reading on Nvidia, and one of the PC mags I read (don't remember which one, was a couple months ago) was touting 256Bit color and were breaking apart how the current color schemes work, and how 256bit will bring better color.....blah,blah,blah... so knowing Nvidia as we do, it will mean they'll use it as an actual color mode. Nvidia hasn't ever been known to use the technology like that as 3Dfx did.

[Response to Boui Andrei]

AGP brought a great promise, but I think it ends there, most of what it promised remained on paper, it did improve things, but not like promised - I have read many documents like Intel's Document Explaining AGP and it all looks great and makes sense why AGP -- but I from what I have read on the subject -- the bus speed its self doesn't even get anywhere =near= as high as any AGP specification design says it will.

With the new video cards w/128MB -- that is ridiculous -- why you'd need =that= much frame buffer since there aren't that many games that will utilize it. I think that's great that someday our video cards will have as much memory as there is main memory. But they haven't made the boards work hard enough to use it; they should find a way to utilize this memory. Otherwise you've just bought an expensive video card w/memory it won't ever touch.

Too Much Of A Good Thing? The Lowdown On 128MB GeForce3 Ti200 Cards

High-Precision Color

That and considering when (with the new blazing fast types or memory/motherboards out there) it can steal from main memory.

There have been tests done, and it was found out that those boards don't even make use of more than 1/2, more if you are lucky or can push it hard enough.

I think all these ideas are great ones, and I would LOVE nothing more than to see them deliver what they promise, if not more. But sadly, we may not ever see that day for a long time - if at all.

NV18 & NV28:
NVIDIA Chips with AGP-8x Flavor

But they admit - silently, that AGP didn't do what is was supposed to so now they ditch that and jump on the PCI Express bandwagon. Rather than improve/re-design the current technology and get it to work as promised. Then just keep redesigning it rather than spend gobs of $ for new design plans, just to scrap it later.

This link, prooves that they are going to trash AGP and go to PCI Express. Meaning AGP was a BIG mistake?

That all makes sense but one that I still wonder.

Have you ever gotten the =true= 1.2 Gig/sec data trnafers from AGP 4X, etc. or whatever the teoretcial tansfer rate is, for any AGP spec. for that matter?

The reason I call it a failure is because it cannot live up to that standards it was supposed to.

Last I knew 8X was the thing I was not aware they have gone abouve that -- or even implemented it in cards that are alread in retail.....I must be on drugs   because last I saw, the current cards you can go out and buy are 8X AGP max....

I keep reading - anywhere AGP is tested in a mutlitiude of ways (via the web, magazines) that it doesn't even reach more than 1/2 of its promised bandwidth -- it doesn't even go to the qouted Gigahertz range for the current and past specifications by a land slide. Some features are great some don't matter worth jack squat.

And I don't know if you read that Intel link about AGP8X but is specifcally said they see the future technology as PCI Express, in the desciption of AGP 8x! That clearly states, they are switching to that standard as a replacement -- for whatever reason.

I have come to that anlyzartion because if it weren't true PCI Express wouldn't be mentioned on that page and just it's own page.

sounds kinda fishy readin that and actually thinking they'd keep AGP....just me but I think they are trashing AGP (good/bad/etc.) and going to PCI Express, they have gotten the press to hype it up to no end and you see it on every web site, mention it on the AGP homepage.....it looks like it will be phased out in the near future, improved to some extent but ultimately phased out.

@Boui :

lol

The Textures are already compressed.
(what were you thinking ?)

Uncompressed, they would be somewhere around 2-4 GigaBytes, wouldn't fit anywhere
Without Texture Compression, nothing works anymore since a long time now.

And DXTC is basically S3TC (DXTC being a relabeled DirectX variant, but they are quite similar, as they do about the same job).

16bit Color Palette ?
Well, last time I saw that was on Quake 1 I think...
Most Games now are fully optimized towards 32bit; anything else is only for fallback Options (hence really not optimal).

UT's original 256x256 Textures don't cut it by today's standards; they looked "as good as it gets" though, since it was really optizimed for 3dfx Glide with some really nice tricks.
But now, just compare it to the S3TC Compressed counterpart in OpenGL (if available) and the difference is simply gigantic.

I do agree that some of the Textures of UT2003 didn't turn out to be as good as expected, but it rather looks like they rushed the Release, than the Texture's own fault. When I look at DM-TokaraForest at Max. Detail, though, it becomes apparent what High-Res Textures really can do

What card you choose to have fun with, is of course your own decision. I play Unreal Tournament and UT2003 "competitive", meaning I cannot accept even 0.1 Seconds delay, still I want the Option of the best Graphics existing at the same time.
And for that, there aren't many Options left that can still produce solid 100+fps at 1152x864 32bit plus Aniso and FSAA.

@ Black_out

Yes, of course I do...
However, those figures really are theoretical, always loses something to latency and overhead.

But no matter if it's 1.2GB/s or 900MB/s, the Result remains the same : progressive performance loss, increasing with the Proportion of Textures not in local VRAM, but stuck in AGP Texturing.

And you're right, 8x AGP is the current standard (doesn't look like they try to raise it again, since the classic AGP Architecture doesn't handle it anymore).

But big Problem with AGP really is, that is has to compete with other Components for the precious RAM Bandwitdh, and the simple fact that even High-End RAM can hardly deliver sustained 8x AGP Transfer rates.
All that usually results in strictly limited AGP performance, far from the maximum possible.

PCI Express will indeed replace AGP quite shortly, but I'm not sure whether this is yet another Marketing gag, or of real benefit (which I doubt to some extend).

Ehm, yes... "maybe"...
(for a second, I imagined UT2003 shipping with 12 CD's , of which 10 would be needed for Textures alone, hence my "lol" )

So you're saying a 16x Aniso Image rendered by the current "Aniso King" of Visual Quality (ATI) looks worse than an S3 Savage rendered Frame...

Well...

Personal preference I'd call that

About your S3TC vs. uncompressed Statement in UT :
Sorry, but either you made a significant mistake, or you'd really need heavy glasses :

One of the most frequently posted Screenshots to demonstrate the superior quality of the S3TC Textures was the Planet Texture rotating around CTF-Face.

The difference is nothing short of dramatic, so I think this is beyond personal preference; it's an obvious leap in Quality.
To stay with this Map, same applies to the Wall and floor Textures.

If you wish, I'll post 2 unprocessed, uncompressed 24bit Bitmaps to demonstrate both.

Well, in all honesty, running a Game at 158fps on the CPU limit, or "only" at 150fps on Max. Quality, really doesn't make a big difference.

I'll go and make some Screenshots, so you can see the difference yourself (I can only reckon that you've actually never seen a true, High-Res Texture image)

If Texture compression for High-Res Textures didn't bring any advantage, nobody would use it *hint*

http://www.falconfly-central.de/cgi-bin/yabb/YaBB.pl?board=games;action=display;...

I already know what compaines have done this before and that it has been done - but I think you took it out of context - I was referring to Nvidia, ATi and the up and coming XGI.

But like you said the reason they problaby don't is becuase they wouldn't have people buying their new video cards every 6 months, rather they'd upgrade the memory as a cheaper alternative.

Alright, I'm almost out of this Thread...

Some clues :

- Manually upgradeable Memory is currently hardly possible anymore.

Why ?

The cuicuitry to connect to the VRAM must be extremely short (the faster the clock, the shorter they need to be).

Eletrical Resistance for an external Module connector would be completely inappropriate to accomodate Signal Lines, that need specific Fall/Rise times and Amplitudes over half a clock clyle at 300MHz DDR and beyond.

In the past these things simply were not such a big factor, and Upgrade Modules were possible.

- any curreny 3D Video Card can usually render a simple Game Menu faster than the Game-Side framecounter goes (e.g. beyond 1000fps)

Untextured Objects are rendered nearly as fast as Textured objects nowadays, provided the number of Pipelines can still render in single-pass.
The only notable effect can be observed, if the Amount of Polygons per scene takes sufficient, additional toll on the GPU, or the amount of Textures used really requires the Card's Memory bandwidth, or eventually, by general Overhead induced by really heavy Texture usage.

But then, who would want untextured Objects (?)

@Boiu_Andrei

This the last time I "clear up" your weird statements.
Refrain from making them in the future, or face consequences.
Read up on those basics > then post, and not the other way around...