The following was posted to the comp.sys.acorn.games newsgroup as a followup to several news postings speculating on the impact Phoebe will have on games:

From: R-Comp Interactive (RCI) <rci@arsvcs.demon.co.uk>
Newsgroups: comp.sys.acorn.games
Subject: Phoebe and 3DFX Voodoo 2 cards
Date: Wed, 12 Aug 1998 15:36:50 +0100

There seems to have been quite a discussion about 3D cards whilst I've been away, so I thought I'd post and bring you up to date and perhaps quash a few rumours etc.

Several members of RCI met with Acorn about 2 months ago to discuss the process of bringing 3D acceleration to Phoebe architecture and RiscOS. The main purpose of this was to bash out concerns regarding APIs and public access to support libraries, and other commercial issues.

The key points that came out from the meeting were as follows:

1. 2D/3D combo cards are out, due to the difficulties of moving the RiscOS video stream from the VIDC20 to a hardware board
2. This leaves PowerVR SG and Voodoo2. PowerVR 1 versions were heavily processor dependent (more so than Voodoo cards) and lacked some essential functions. Even the revised version caused major performance drop off at more than 640x480.
3. PowerVR SG isn't out yet, and hasn't got a foothold on the market barring the Sega Dreamcast. Voodoo 2, on the other hand, is the benchmark 3D accelerator. When you talk about 3D acceleration, you talk 3DFX - no other card has the same credability.
4. The hardware passthrough for Voodoo based cards is ideal given the on-board VIDC20. Software overlay (as in PVR) would require significantly more driver work.
5. A geometry setup engine may be necessary.

This led to the inevitable conclusion that we should work towards Voodoo 2 support, with the possibility of Voodoo 1 for the low end market.

The conversation then turned to APIs...

1. The current technology options are Direct3D, OpenGL, Glide
2. A RiscOS implementation would only have to support a "compatible" API for coders, so the needs of the coders are paramount
3. Direct3D was a joke, BTW. ;->
4. OpenGL is a colossal API, and would mean that the full 32Mb of a base Phoebe would be necessary for all 3D games. Also OpenGL is non-hardware optimized, so would only provide a "generic" layer, which would never give the full perforance required. It would, however, please 3D purists, but Acorn coders would prefer to work with something smaller and more efficient
5. This left Glide. Almost every 3D card accelerated game in the PC world has a Glide specific mode, tied to Voodoo (2) cards. Provided that the RiscOS implementation had a suitably effective API, then it would not be absurd to expect other cards to be "possible", if not easy.
6. Glide has a proven track record with programmers. Even OpenGL-only games (eg. Quake) have been "Glided" by their coders for test purposes without too much effort (compare this with Direct3D!!!) A Glide implementation for RiscOS would be a major incentive for outside developers, too.
7. OpenGL support could always be added later if necessary.

Thus we arrived at a Voodoo2/Glide solution.

The plan would be to offer the API for free on the net, including libraries and specs. The drivers, however, would only be available on a commercial basis.

At the moment, we haven't got any further on the driver side, as Phoebe hardware isn't available yet to developers. The PCI system has yet to be properly covered, and AFAIK even simple PCI projects (such as a reference network card) have yet to be done. Until such reference material is availble things can't be pushed forward too far.

Clearly, as and when developer hardware becomes available, we can begin work on the project in earnest.

Finally, I know a few of you will think "why RCI?", to which all I will say is - look at what's happened and what experience and contacts have been gained in the last 8 months. Look at the funding and bundle variables in the equation, and there was little option for us but to work with Acorn on this project.

OK, so I said "finally" above, but I'd like to add that whilst we want to make this happen, until Phoebe is alive and kicking, we still don't know feasibility of the project, so please don't get your hopes up.

--
R-Comp Interactive
22 Robert Moffat, High Legh, Knutsford, Cheshire WA16 6PS
Tel: (+44) 01925 755043            Fax: (+44) 01925 757377
http://www.rcomp.co.uk/

Richard Goodwin (13/8/98)

Back to News...