3dfx Voodoo Graphic Card Emulation Coming To DOSBox 156
Posted
by
timothy
from the but-1996-feels-like-yesterday dept.
from the but-1996-feels-like-yesterday dept.
KingofGnG writes with this excerpt from King Arthur's Den: "One of the forthcoming versions of the best PC-with-DOS emulator out there should include a very important architectural novelty, ie the software implementation of the historical Voodoo Graphics chipset created by 3dfx Interactive in the Nineties. "Kekko", the programmer working on the project with the aid of the DOSBox crew and the coding-capable VOGONS users, says that his aim is the complete and faithful emulation of SST-1, the first Voodoo chipset marketed in 1996 inside the first 3D graphics accelerated cards on the PC."
Interesting but it looks slow (Score:3, Informative)
So this seems to be very different from something like, say, GliDos.
Voodoo emulation originally written by Aaron Giles (Score:5, Informative)
As per TFA [kingofgng.com], the Voodoo emulator is basically lifted from MAME [mamedev.org]. Granted, integrating it into DOSBox is important work and all, but I would judge the original code to be worth more than 90% of the effort. Yet Aaron gets no credit in the summary.
Re:Carmageddon (Score:3, Informative)
Re:Nice (Score:3, Informative)
Re:Technically, the Rendition Verite cards came fi (Score:5, Informative)
The S3 ViRGE was the "decelerator" of its time. Had they been used as glorified software renderers expected to do little besides push point-sampled, perspective corrected textures onto polygons, with all geometry calculations handled by the host CPU, they would have been better, but the competition was too steep for anyone to bother writing what would amount to an enhanced software renderer. Visual quality would have been shown up badly using such a scheme, so the native titles for the ViRGE were pretty but terribly slow. From what I recall the Descent II port was a pretty heroic effort.
The Rendition cards were really very solid by comparison, but the V1000 series took a noticeable speed hit when they were expected to handle on-chip z-buffering. Their fillrate was also around half that of the Voodoo1, but they would still have been price-competitive if RAM prices hadn't fallen through the floor and made the Voodoo Graphics board realistically obtainable.
Re:Glide and Matrox Support in DOSBox (Score:1, Informative)
They're very hard to find now but Tomb Raider screenshots of the Matrox version looked fantastic in underwater scenes. I owned a Matrox card for a short time and the 3Dfx Voodoo Graphics card I later got didn't come close to looking so good. The colour and shading had an ethereal quality to it when the Voodoo looked a bit rough. I really, really wish DOSBox emulated the PowerVR API so I could play that version of Tomb Raider again.
Re:So... (Score:3, Informative)
Re:Interesting but it looks slow (Score:3, Informative)
First, the patch is fully crossplatform (at least the dosbox part), but you require Glide support (real card or wrapper) on the host system. Patches have been submitted to OpenGlide that make it work in Linux and OS X. The full setup (DOSBox + OpenGlide) has been tested to work on Windows, Linux and OS X including using Glide (and OpenGL through the 3dfx minidriver) in guest Windows9x (yes, I've played Half-Life inside Dosbox, fully accelerated
Kekko's patch obviously offers true 3Dfx emulation and also works with games that cannot be emulated with a wrapper. But at the moment only has (single threaded) software rendering which is great for testing but unfortunately little use for playing the games.
Re:Huh? (Score:3, Informative)
Re:Interesting but it looks slow (Score:3, Informative)
This seems like a good place to point out DOS32/a. It's a drop in replacement for DOS/4GW. It was independently developed from Dosbox as a modern alternative to the old dos extenders, but it works quite well with Dosbox. It works with just about everything, and it makes most games better. I pre-emptively swap DOS32/a in when I install anything on dosbox.
Re:Technically, the Rendition Verite cards came fi (Score:3, Informative)
S3 had nothing to do with it. Intel bought Real3D, took the entirely competent architecture of the Starfighter, and pulled a stupid trick by forcing it to fetch out to AGP memory for texture storage, leaving the user with a large framebuffer and excruciatingly slow texture swapping over the bus. PCI versions of the card obviously couldn't do this, and were frequently better performers because this was properly compensated for in the drivers.
Subsequent integrated video chips continued along this shambling path, occasionally receiving updates - multitexturing here, S3TC support there, before receiving a minor overhaul somewhere around the i865G, which was allegedly DirectX 7-capable, but too slow to take advantage of most of that featureset. The i915 through GMA 3100 were native DirectX 9 parts, but pokey and prone to driver glitches; the x3100* onward are different DirectX 10+ parts, and then there's Sandy Bridge's integrated video which finally might not cause whimpering pain to all who behold it in operation. I'm STILL not confident in its OpenGL driver for anything besides desktop compositing.
* Yes, there's a huge friggin' gap between the 3100 and x3100, to the point that you could argue they aren't the same chip in any meaningful way. Alarmingly, this is actually less confusing than the naming schemes for their CPUs now...