Is nVidia Support for Older 3D Games Fading?

BrendaEM writes "A thread on Through the Looking Glass depicts the plight of fans of the original Thief Series and System Shock 2, who are asking nVidia fix rendering issues these 3D 16-bit games on their newer video cards and drivers. In the case of the original Thief series, in which the games build tension by their use of light and shadow, the rendering has been badly degraded from that which was originally intended. In another Slashdot article, the author asked the question whether or not video games were art. If one of the greatest video games of all time, with a growing wealth of hundreds of fan produced missions, as well as an entire full-sized expansion, does not play well because legacy support diminishes, then what will happen to lesser 3D video games?"

Comments

Re:Pure gaming bliss.

[Anonymous Coward]

As opposed to 10 years from now in the past?

nVidia not to blame

[zokier]

I'm quite sure that game programmers are using every undocumented feature and bug they can find to achieve better performance and to allow creating better games, and so they inheritely are fragile. I assume also that situation is getting better as API's are getting better and performance of hardware has increased giving programmers more freedom to produce good code vs fast code.

Re:nVidia not to blame

[cliffski]

I'm a game dev who uses a very out of date API for my games (DirectX 7). I do 2D turn based games, so there is zero incentive for me to use a newer API, which would mean rewriting engine code.
The problem is not so much that newer cards screw up rendering of old games (although this *can* happen), but that no card driver seems to stick 100% to the directx standard. DirectX is HEAVEN for game devs, because in theory it means we can write to a single standard for the windows platform, and have our games work on any card.
The problem is, there are so many minor quirks, differences and tweaks in the way each card implements the same directx calls, that in practice you will *always* encounter people who have rendering issues just on their PC. I wish ATI, intel (worst offenders) and nVdia would take more time to ensure that their cards actually come closer to the supplied reference rasterizer in terms of results. The entire point of directx is to allow devs to be free of individual card woes. shoddy drivers can undo all of that work.

Re:nVidia not to blame

[Tim C]

While I don't doubt that, nVidia at least has been caught cheating in their drivers before now to get better scores in benchmarking software, and I certainly remember them releasing new drivers to improve performance in popular games that reportedly broke other, less popular ones.

Even now, the latest nVidia drivers (which the Bioshock demo recommends you install) has caused a few minor glitches in Oblivion (for me at least), and that's hardly an old game.

Re:nVidia not to blame

[JNighthawk]

Don't assume. There are two big issues, aside from what you stated, here:

1. Drivers not sticking to spec. Many drivers have many bugs that game devs have to work around.
2. Legacy support leads to sub-optimal performance. When driver devs need to choose whether they'll devote their time to legacy support or a new whiz-bang feature, sometimes they choose the whiz-bang feature.

Virtualization to the rescue

[davidwr]

Those older video cards that have open specs will find themselves virtualized or emulated. Of course, there may be a 5-10 year or longer gap before an emulated video card can perform as well as the original. But in 2015, playing games from 2000 should be no problem, provided at least some vintage video cards are open-spec.

Re:Old games in general

[BobPaul]

This is probably a WinAPI issue. This has nothing to do with nVidia, ATI, Video Codecs (which would only affect cutscenes anyway...) or anything like that. This is almost certainly 100% caused by upgrades with the WinAPI and/or DirectX. Unfortunately for you, the best thing MS ever did was drop full compatibility with Win9x. Check the game vendor for updates or try it in DosBox or similar.

Dithering is supported but doesn't work

[Kegetys]

The problem with thief/sshock2 is that the 8800 series cards do not seem to do any dithering which leads to those ugly colors when using a 16bpp mode. The interesting thing however is that the cards claim they support dithering in D3D (D3DPRASTERCAPS_DITHER caps bit is set, which means "Device can dither to improve color resolution.") but they still just do not do it.

Makes me wonder if it is just something that's not implemented yet on the drivers or is it a hardware limitation. Either way the driver should not say it supports dithering if it doesn't.

Re:16-bit ?

[NSParadox]

16-bit as in 16-bit integer color per pixel, as opposed to 32-bit floating point color per pixel that all video cards have supported since the notorious NVIDIA GeForce FX/ATI Radeon 9700 series. This has nothing to do with 16-bit memory utilization/integer size vs. 32-bit.

Not merely a driver issue

[cpt kangarooski]

There's also a copyright issue involved; even the developer will cease publishing and supporting the game over time, and it's likely that it will stop being compatible with modern hardware and software due to underlying changes in APIs and such.

Part of the solution to this from a legal angle (in the US at least) would be: to mandate registration for all works for which a US copyright is sought; to mandate the deposit of a full, unprotected/unencrypted copy of the software and source, plus additional comments and information, so as to enable a programmer of ordinary skill (cf. PHOSITA in the patent field) to understand and make use of it freely; and to have a very short maximum copyright term -- perhaps five years -- in recognition of the especially short commercial lifetime of software.

As much as it would be great for the original parties -- the creators of the game, the OS, the hardware, etc. -- to provide long-term support, ultimately, it's safer to not put all of our eggs in that basket. Instead we should make sure that the resources are available so that even if they're not interested, but some third party is, that the software can be kept running in one way or another.

Re:Pure gaming bliss.

[rhyder128k]

I think that a new generation of emulators is probably a better idea for older games. I wonder what it would take to software emulate a 3dfx voodoo 1 card? Any such attempt could offload some of the work onto the real GPU, of course.

Re:There won't be an open source driver

[Randle_Revar]

ATI hasn't announced anything new. They've simply brought attention to the fact that they will support open source efforts, as they always have.

They never supported open source efforts in the past. This is the first time they have provided documentation and been willing to answer questions without an NDA (and sometime they were reluctant to talk even with an NDA).

 

As always, there will be 3d drivers for paleolithic versions of their cards, and 2d for everything else. If you actually want to use up to date cards, you'll have to use the closed drivers.

Actually, this should be very helpful to the efforts to create open drivers for all ATI cards.

http://airlied.livejournal.com/50613.html
http://www.phoronix.com/scan.php?page=article&item=838&num=1

Re:My next card will be Ati...

[Randle_Revar]

They already released two large documents, and it would seem pretty silly to release that much and then not follow up on the rest of the promise. I mean, if you aren't going to keep your promise, why not go all the way and and not keep any of it?

Re:16-bit ?

[Solra Bizna]

...You're... not quite right.

16-bit as in 16 bits per pixel, with 5 (or 6) bits per color component, as opposed to 32 bits per pixel with 8 bits per color component.

-:sigma.SB

Authenticode price is prohibitive for hobbyists

[tepples]

Technically by the way, the specs would allow a open source Windows driver to be written aswell instead of the one supplied by ati for windows, right? Unless the driver is one of those specific kinds of drivers that can run entirely in user space, you can't use drivers you compiled yourself because they are not digitally signed. Workarounds are to 1. press F8 every time you start Windows Vista or 2. pay $499 per year for a VeriSign code signing certificate and sign your code.

The damn thing generates a tick at the refresh rate of my monitor Perhaps this has something to do with the ability to block on the start of vertical blanking, which is crucial for smooth animation in PC games.

Re:Pure gaming bliss.

[gravos]

Exactly. One of the biggest reasons to be a PC gamer is because you can still run most of yesterday's games today, unlike on consoles.

If nVidia is taking that ability away, then PCs start to look a whole lot more like consoles...

Re:Old games in general

[Splab]

Find the exe file, right click, properties, pick the compatibility pane, check the run this program in compatibility mode for and pick a suitable version of windows. Also you might want to disable advanced services and visual themes.

Hope it works.

You're in violation of the Berne Convention

[tepples]

Part of the solution to this from a legal angle (in the US at least) would be: to mandate registration for all works for which a US copyright is sought; to mandate the deposit of a full, unprotected/unencrypted copy of the software and source, plus additional comments and information, so as to enable a programmer of ordinary skill (cf. PHOSITA in the patent field) to understand and make use of it freely; and to have a very short maximum copyright term -- perhaps five years -- in recognition of the especially short commercial lifetime of software. Such limitations on copyright, if applied to works from all countries, would violate international copyright treaties. If applied only to works first published in the United States, such limitations would just drive publishers to Canada or the UK; the United States has to honor Canadian and UK copyrights for at least the Berne minimum term.

Re:Games in Glide

[neocrono]

With a Glide wrapper. A proxy library that makes available the Glide API functions, but internally does some voodoo and maps them to another 3D API like OpenGL or Direct3D.

Half-Life, as you may remember, had a miniGL driver designed for better performance on Voodoo cards. There was also the third-party WickedGL "drivers" that did the same kind of OpenGL-to-Glide translation for other applications, using a drop-in opengl32.dll replacement. In both cases, the application treated the card like OpenGL, but the card ultimately received Glide calls, which performed better.

A Glide wrapper is the reverse: take an application expecting Glide, and help it make OpenGL or Direct3D calls.

ATi no better

[BlueCollarCamel]

I use the latest version of the Catalyst drivers. Day Of Defeat and all other HL1 mods were unplayable in OpenGL, and barely tolerable in Direct3D. I proceeded to switch to the Omega alternative, and all is well now.