3D --- Media Creation --- Shared Space
Published by 4th WAVE, Inc.
Issue #711 5/9/97
CONTENTS
Microsoft buys Dimension X [May 7]
Microsoft Corp. will acquire Dimension X, a start-up company devoted to the Java multimedia market. Dimension X began in March 1995 and had the first Java-based multimedia authoring tool, Liquid Motion, and the first Java-based VRML browser and runtime technology, Liquid Reality. The Dimension X Liquid products and technology will be integrated into the Microsoft DirectX multimedia system services. Java multimedia developers will also have access to hardware acceleration using DirectX. Liquid Motion and Liquid Reality are expected to be shipped with the DirectX media Software Development Kit (SDK). Most of Dimension X's 40 employees, now based in San Francisco, will join Microsoft in Redmond, including Dimension X founders Karl Jacob, Chris Laurel and Ryan Watkins and key engineers Patrick Schmitz, Charles Loop, Mark Halstead and Alan Greenblatt
711.2 Opinion: The Difficulties of 3D Benchmarking by Roy Latham
Graphics pros would like board and chip vendors to succinctly state the vertex capacity, the rendering capacity for polygons having different attributes, the significant bottlenecks of the architecture, and which software has been tuned. It would then be fairly easy to figure out which applications are well-suited to the particular hardware. There are two problems with this approach: (1) the detailed information would not be helpful to the average consumer, and (2) only the top performing vendors benefit from making performance attributes clear, so the information is rarely made available. In fact, since there probably is no overall "best" hardware, with all the variations in price, performance, bottlenecks, and tuning, many, many products will be "best" at some application. So instead of making the overall performance attributes clear, each vendor finds it advantageous to either cite the benchmarks at which they excel or just publish unqualified mostly-meaningless numbers.
As a practical matter, while putting together the graphics accelerator survey for our professional newsletter "Real Time Graphics" it was difficult to even get information as to whether polygon performance was for individual triangles, strips, or meshes. This was not usually deliberate obfuscation. Often, the marketing department did not understand what we are taking about, and engineering assumed that whatever it was they did was obvious and the standard way of doing it for the industry. Engineers may assume that the bottlenecks in their design are the "normal" and "necessary" ones, and hence not worth commenting upon. In fact, they vary considerably from product line.
The bottom line: use benchmark data that is as close as possible to a particular application, understand what the benchmark means, and if it is really important, do you own tests or have an evaluation done by a pro.
Roy Latham is the editor and publisher of the Real Time Graphics newsletter. It is a newsletter for professionals working in the fields of simulation and virtual reality. It covers all aspects of the technologies of virtual environments, with special emphasis on image generators and displays, and associated software for real time operation.
711.3 Conference Report - Computer Game Developers' Conference by John Latta and David Lohse
The 11th Computer Game Developers' Conference was an outstanding event. 3D was the buzz in the sessions, on the floor and in the halls. This is where the technology and product advocates (chip companies, API developers, authoring tool companies, platform companies and OEMs) court the content developers. Game developers are a Wild West bunch who work all hours, party and there are many who dress and appear outside of the main stream - even in Silicon Valley. 5,000 attendees crowded into an over burdened hotel and Santa Clara Convention Center. Next year it moves to Long Beach. The plan is to alternate between the SFO and LA in alternate years.
The OpenGL vs. Direct3D API for games debate still rages. We went to several sessions on OpenGL and they were all packed. In the informal sessions OpenGL was a much debated topic. Yet, we could not determine if the game community is serious about creating titles for OpenGL. It seems the bigger issue among developers - should they do a title which runs only with 3D accelerators for 1997? We saw a small but increasing level of commitment to "acceleration only" titles.
In a number of sessions another hot topic was, will 1997 be the year of shake out in online games? Dogged by uncertain business models companies are trying to find a niche which will be self sustaining. A proposition which looks all the more difficult.
Presentation - Mark Pierce - Coin-Op Arcade Games
Mark Pierce is Senior VP & Executive Producer for Coin-Op Product Development at Atari Games. He has many years experience in developing and managing the development of coin-op games. He is working on a book on the coin-op industry and many of his insights came from his research and experience.
The domestic sales of coin-op equipment are approximately $150M. The coin drop is approximately $2.5B in the U.S. and $3.5B worldwide.
He sees coin-op as the most difficult interactive entertainment venues where the players must have satisfaction in the play in 90 seconds with no manual or tutorial. A game "lives or dies based on the attraction of the game and its initial play." Game development costs run from $500,000 to $4M.
In the U.S. the number of leading distributors of games is approximately 40 and one of their most important roles is to provide financing to the operators. The typical distributor mark-up is 10% to 18%. A system which costs a distributor $3,000 will sell to the operator from $3,400 to $3,600. Kits, an upgrade of the electronics for an existing game, will sell based on the title. For example, a leading title by Atari, Area 51, sold new for $3,200 and will go in a kit for $2,100.
Many of today's operators have been in business for 10 - 20 years. A typical standup game will cost an operator $4,000 and that amount should be recovered in 6 - 8 weeks in coin-drop if the game is to be successful. This is not what the operator receives because most machines are on an income split basis with the owner of the location. Thus, in terms of actual operator income they would like for the machine to cover its costs in 8 - 20 weeks.
A hit will sell 10,000 units. To a manufacturer who sells to a distributor the gross income is $30M and of that 1/3 is profit. Atari's Crusin did very well and sold 23,000 units. The best selling game of all time was Ms Pac Man which sold 126,000 units and it is felt that there were 100,000 illegal copies also sold. Mark compared this to games for the home where a title, across all SKUs and platforms (PC and home video game), could sell up to 500,000 units. The consumer sales can reach from $25M to $100M per title. With a successful consumer title the income is "mostly profit" after the development costs have been written off. In contrast he described "...coin-op as not being far from a loss-leader."
The number of street locations is estimated at 60,000 - 80,000 in the US. A good game will make $100/week. There are 9,000 to 13,000 arcades in the U.S. and each have from 20 - 40 games which average in income from $95 - $200/week. A top ten title will gross $300 - $900/week.
The peak year for the arcade business was from 1979 to 1981 when the cash box was $6B and this was higher than movies.
FEC's, Family Entertainment Centers, are the breeding grounds for new interactive entertainment. In 1996 there were approximately 2,700 of them and they are located near large population centers. Most offer large, expensive attractions and expensive prices.
Mark was negative about VR and stated that "VR never made money in arcades - don't ask me about goggles."
Arcade game design was described as a special talent which is only passed on as an oral tradition. He stated that "...maybe only a dozen individuals can make a good arcade game." The game design rules he cited include:
Mark classified all games into seven genres:
Presentation - Lee Isgur - Status of the Interactive Entertainment Software Industry
Lee Isgur, Managing Director of Corporate Counselors in New York, gave interesting statistics:
1997 Hardware Units - Video games (>= 32bit) 47%, PC 53% 1997 Hardware $ - Video games 8%, PC 92% 1997 Software Units - Video games 60%, PC 40% 1997 Software $ - Video games 30%, PC 16% and Coin-Op Drop 56%
He described 1996 as a good and bad news year for PC game software. While a number of titles were able to sell 100,000 units at full price (unbundled) the critical industry issue is that there are too many titles - estimated to be 4,000+. He believes that less than 50 sold above 100,000 units. Thus, even a weak platform such as Sega Saturn with an installed base of 1.6M units could offer a better potential market (unit sales/title) than the PC in spite of the fact that there are more than 10X the number of PCs over Saturn. The PC market continued to be difficult to create winners when not a single PC title in 1996 achieved as much as 1/2 the first year success of Myst or Doom the prior year. In spite of the prospect of the PC, home video game unit and $ sales for the N64, PlayStation and Saturn, are substantially higher than those for the PC.
In spite of a confusing market Lee was optimistic on its growth. He described the market today as being larger than it was 5 years ago for the following reasons:
There are a greater number of people capable of using interactive entertainment. There are a greater number of consumers who are purchasing interactive entertainment software. There is a greater number of households with one or more interactive platforms in active use. The geographic spread of consumer interactive software as a form of mass media. The number of consumer interactive software titles that are being published annually has increased. The number of units of consumer interactive software being bought annually is increasing.
Lee was most bullish on the increasing number of individuals who he described as Interactive Literate. This could create up to 80M+ consumers by the year 2020.
Panel - Jon Peddie and Others - 3D Games Aren't Really 3D Until They Use 3D Hardware
Jon Peddie was the moderator of a panel which included Microsoft, Rendition, 3Dfx and others. He began by giving a short presentation on the market. Jon cited his statistics that there were 14M 3D chips shipped in 1996 and he expects that 40M will be shipped in 1997 (We regard these numbers as high.) Jon then chastised the audience of game developers for not targeting this installed base and asked that they "get off the dime." Some developers considered this a cheap shot. One shouted back from the floor that those chips were poor, they did not provide enough advantage over software, and that the API they had to use was also poor.
A spirited panel discussion ensued along with shouting from the audience. A number of key points were also made by the panel members.
3Dfx reiterated its position that the minimum frame rate for games will be 30f/s and the coming standard will be 60f/s. (This observation is consistent with the experience in flight simulator development where 60f/s is essential. We heard a number of comments during the conference that there is a demonstrable difference in the play quality as one goes from 30f/s to 60f/s.)
Microsoft disclosed that a number of companies had licensed Talisman technology. They also disclosed that 3D Baseball was originally targeted for Talisman but now that "this has shifted" it will not immediately support Talisman.
3Dfx stated that they can "achieve Talisman (performance) without a substantial change in present technology." While Rendition believes that "we can meet Talisman performance without Talisman technology and at OEM price points."
The prospect of Intel's 740 chip (the Auburn 3D and 2D accelerator) came up when Jon Peddie asked if "...it is a threat to the industry? Is it good enough?" In response the audience gave its negative assessment.
When asked where the feature performance priorities are the panel responded that fill rate is the highest followed by on-chip set up.
Presentation - Murali Sundaresan (3Dfx) - High Performance 3D Graphics on a PC
Using a deadpan delivery with humorous drop-offs Murali gave one of the best presentations at the conference. He began by stating that he was asked to show at 3Dfx how 1M p/s could be achieved on a Pentium chip. The paper began with an overview of the analysis of the number of instructions to do geometry and lighting on a Pentium - 55 instructions. The tools, all the way to a logic analyzer, were reviewed to allow for measurements that would be made on the system. Murali then proceeded to review a series of tests that showed that independent of what he did the answer was always 600,000 p/s, yet, there was no apparent reason why the 1M p/s rate should not be achievable. The last test found the answer - the Pentium L1 cache would fill up with vertices and when full this data would be written back to either the L2 cache or system memory. During write back the processor is locked and this is wasted time with no productive processing being done. One way to address this problem is to partition the geometry data so that vertex information only occupies 1/2 the cache and static data the other 1/2. It was also concluded that the API could manage the data flow into the L1 cache to help alleviate the problem. When it was stated from the audience that AGP has no impact on this issue Murali confirmed it.
We begin first with our comments on the presentation by Murali Sundaresan.
3Dlabs has often spoken of the limitations in geometry processing performance imposed by the host processor and 3Dfx has shown explicitly why. Game developers have been pushing that the host processor be used for game logic, dynamics, collision detection and media processing and they now have all the more reason to champion the placement of geometry, lighting and rasterization on the 3D chip. This frees the microprocessor to do game functions. It is clear the industry is moving to eliminate as much 3D image generation from the processor as possible. AGP is an excellent path to move vertices from system memory to the geometry front end of the 3D chip and, in fact, this is what 3Dlabs is doing with Gamma. Intel has spoken much about the need for a balanced platform design, however, we have not seen Intel speak to how they expect a balanced design to be accomplished. For example, to have the API shoulder the responsibility of L1 cache management so that the Pentium can do functions it was not designed for is ridiculous. High performance 3D is a systems issue where architecture and integration of all the components is a critical factor. This has been known for many years by companies building flight simulators. The PC industry is beginning to realize the complexity of 3D system performance.
On the show floor we spoke with all the chip companies to collect an overview of the current state of 3D. We zeroed in on the preparations for the 1997 Holiday buying season. OEMs will lock down their purchases for Q4 by the end of May. Basically, if a chip was not on the floor at CGDC it would not have an impact in Q4. The primary competitors are: S3, ATI, Matrox, Cirrus, 3Dfx and nvidia. The RIVA 128 has impressive early results and it is posed as a spoiler - that is, OEMs are holding off decisions as they evaluate the chip. Given that the chip was only 20 days from first silicon achieving this level of attention was impressive. Absent from the list of Q4 contenders, in all of our interviews, was the Intel 740 chip. A number of Intel personnel were wearing 740 caps on the show floor and Intel had a separate invitation only location. Intel would not respond to a request to discuss its plans for the roll out of the 740 at CGDC.
In spite of Intel's silence much attention is being directed to the 740. The game developers are certainly sensitive to Intel's entry into this market as was evident in the panel discussion. The announcement this week of the Pentium II only increases the attention directed towards the 740. AGP and the 740 are closely coupled. For AGP to work there must also be the 440LX chip set which is now rumored to be delayed to Q3 1997. Given all the emphasis Intel has placed on the use of system memory for texture storage the 740 will take maximum advantage of this. We see that all of the following need to fall in place for the 740 to have an impact: Pentium II, AGP, Memphis, and motherboards populated with 740 chips.
At CGDC we sought the opinions of many in the industry and our best construction of the 1997 road map is the following. Intel is expected to go public with the 740 at SIGGRAPH, however, this is likely to be demos and not shipping silicon. The rollout of products will move to Fall COMDEX and be too late to have much, if any, impact on Q4. It remains uncertain if any units will ship in Q4. At the WAVE Report we have spoken before about operating system support for AGP and the critical gate this represents. At CGDC we picked up a number of rumors that Intel may have an update to Windows 95 to support AGP. This could make AGP and the 740 more realistic in Q4 especially if Memphis is slipped into 1998, which continues to appear likely. Yet, we discount the prospects for a Windows 95 version to support AGP. First, Microsoft made it very clear at the Windows Driver Conference that they are not doing this. In spite of Intel's desire to have AGP systems in 1997 we question if Intel is willing to take on the customer support burden for the Windows O/S in this area. Further, is Intel willing to test and certify the AGP version of Windows 95 which supports all versions of AGP capable chip sets, including those from IBM, for example? Thus, we continue to reach the conclusion that AGP and 740 are Q1 1998 prospects.
Why all the attention being directed at Intel's plans? If Intel raises the public expectation for the hottest thing in PCs since MMX and does not deliver in volume the industry ripple effect could be significant. We have heard many express concerns that if consumers see better PCs just over into the next quarter they will hold off buying at Christmas. The prospects of a poor Christmas with shelves full of non-AGP computers is not a pleasant thought to many.
Even with this attention being directed to hardware we also came away disappointed with developments on the software front. Where are tools for real time game development and delivery? The only authoring tool shown for real time 3D was Catalyst by NewFire. Nichimen Graphics also announced Fast Track a modeling package targeted for game developers but this is based on existing software.
Lee Isgur captured the competitive landscape well. The PC industry is in direct competition with the home game console - both for consumer titles and game developer commitment. Only in 1997 do we have the potential of seeing 3D performance rival that of Nintendo 64. However, it is unlikely that the number of PCs at this performance level will be able to match the installed base of Nintendo 64 by the holiday buying season. The PC is still playing catch-up and Q4 1997 is a very important indicator of its ability to go against the game consoles.
Q4 1997 is looking shaky as a 3D Christmas. Must we wait until 1998 to see polygons fall?
The uncertainty in 3D accelerators is simple: who will deliver what, at sufficient performance levels, and in quantity to make the PC a competitive 3D platform by Q4 1997. All the 3D companies are working on 3D AGP chips but they could be sorely disappointed if Intel does not deliver - Cirrus Logic is the first company to have gone public with the impact that Intel's delays have had. Yet, this is much more than a chip issue. There must be compelling games and just not home video game ports. From the brief sample at CGDC there are early indications that high quality titles will surface. It looks like we will have to wait for E3 in June to make another assessment of the status of a 3D Christmas.
SHOW FLOOR, ANNOUNCEMENTS AND PRODUCTS
Rendition - New Technology Shown
It seems like everyone knows about the next generation chip - 2000, yet, Rendition, remains mum. But in the booth were screens running "New Technology." Although its performance was scaled back the results were interesting and we await the announcement.
Rendition did announce the availability of Rredline its Windows 95 version of their proprietary API. This is available now for developers. An MCD is also available for OpenGL.
Showing the first title running on the RIVA 128 was G. Police from Psygnosis. Many agreed that the combination of the game content and chip performance was significant for a new chip.
Oak Technology - Tape Out Is Near
Oak Technology expects to tape out in the next few weeks. Chips will be shown at E3. Expected performance is 33M p/s with all features turned on including tri-linear texture mapping.
Newfire was showing its new authoring tool for creating commercial quality games for the Internet. Developers can use art created in 3D Studio MAX and other packages for input. Included in Catalyst are the following:
*Project Manager, which allows for the management of project output including its assembly for playback on CD-ROM or the Internet; *Real Time Playback and Analysis which enables developers to set the results rapidly; *Scene Composition to assemble scenes into game elements; *Interactivity Editor for the assignment of behavior to game components, and *Performance optimizer to tailor 3D scene geometry and textures for fast playback in the playback engine - Torch
The base product is $1995 and plug in modules cost from $250 to $1,000 each. Beta release is in June with shipment in August.
Qsound Labs - Provides 2 Solutions for 3D Sound
QSound Labs announced that they support both the Microsoft DirectSound3D API and their own QMixer API. When developers use the Microsoft API they can bundle the QSoft3D mixer and rendering engine. QMixer is an audio mixer which provides interactive placement of individual sounds in multi- dimensional space. No special hardware is required. The API works on both Windows 95 and Power Macintosh. Intel MMX is also supported. Sample rates include 11, 22 and 44.1 kHz at 8 and 16 bit resolution.
Immersion - Force Feedback Primer
At their booth Immersion Corp. was passing out copies of "A Force Feedback Programming Primer" a 177 page book on development of games which support PC peripherals which use I-Force 2.0 and DirectX 5.
Immersion claims that the following companies support its I-FORCE: Logitech, CH Products, SC&T International, ACT Labs, Interactive IO, Advanced Gravis, InterAct Accessories, and Nuby Manufacturing.
http:///www.force-feedback.com
Sonic Foundry - New Audio Editing & CD-burning Software
Sonic Foundry introduced CD Architect to their line of Windows-based audio editing products. It provides waveform editing and audio CD burning capabilities, allowing users to create professional audio CDs on their PCs. Audio can be imported from compact discs, recorded from digital audio tape, or digitized through any Windows-compatible sound card. Sound editing capabilities include dozens of effects and tools to process sound files. CD Architect will be available for $395.
Duck - New Version of Video Codec
Duck Corporation has released version 2.0 of their TrueMotion video codec, offering 24-bit RGB at 30 fps.
Microsoft - Licenses Duck TrueMotion 2.0
Microsoft licensed Duck Corp's codec technology called TrueMotion 2.0. DirectX media with TrueMotion will allow developers to deliver video on a wide range of Windows platforms. Features include: data rate control, scalability and optimization for multiple device platforms and lower computational complexity for decoding video.
DirectX media is now in beta 1.0 and it is scheduled for release in DirectX 5.0 this summer.
http://www.microsoft.com/directX/
SciTech - Graphics Library Now Supports OpenGL
SciTech Software has released a public beta version of SciTech MGL 3.1, their professional graphics library for DOS and Windows platforms. New features in version 3.1 include full-screen OpenGL support, Borland Delphi support, and a new framework for game development, along with other enhancements. SciTech MGL supports three different OpenGL implementations: Microsoft's OpenGL, SGI's CosmoGL and Mesa (a freeware OpenGL implementation).
Floating Images - New 3D Imaging Technology
Floating Images, Inc. debuted Real-Depth, a new 3D imaging technology that brings depth to computer and video games using a multi-planar imaging technique. Their new (patent pending) "floating 3-D, off-the- screen-experience" technology produces images at different depths on any display. In addition to providing traditional depth cues such as perspective and background image occlusion, it also provides both horizontal and vertical binocular parallax, which coincides with convergence. Implementation of Real-Depth on a TV or computer monitor requires a Floating Images adapter, which can be manufactured in quantity for less than $25 for up to a 20" display.
Fuseworks - Fuseworks Server Debuts
Fuseworks, Inc. unveiled Fuseworks Server 1.0 and SDK. They allow developers to create multi-user Java, Shockwave or VRML applications within a Web browser. They have developed Battlefield.net, the "first real-time multi-player web based game" with the technology (available from their Web site).
Lightscape - Plug-Ins for 3DS MAX and LightWave
Lightscape Technologies announced the availability of import/export plug- ins for the direct translation of data between 3D Studio MAX and Lightscape 3.0. This will allow content developed in 3DS to make use of Lightscape's advanced radiosity rendering engine. Beta versions of the plug-ins are available now online, with the final versions expected to ship in June. Similar plug-ins were also announced for NewTek's LightWave 3D; these are expected to ship in Q3.
Lightscape also announced that they have selected Nichimen Graphics as the Master Distributor for Lightscape products in Japan. They are also working with Nichimen Graphics to integrate Lightscape 3.0 with Nichimen's N-World product line, which is in the process of being ported to Windows NT. These applications will be marketed for the Windows NT platform under the brand name "Fast Track."
Aureal Semiconductor - 3D Sound SDK
Aureal Semiconductor introduced the A3D SDK, which enables positional 3D audio hardware acceleration for Microsoft's DirectSound3D API.
3D Construction Company - Builder Pro 2.0
3D Construction Company announced version 2.0 of 3D Builder Pro and PowerLite, new software for building 3D models from photographs.
ImagiNation Network - Agreement with Microsoft Announced
A wholly owned business unit of AOL Studios, the ImagiNatiion Network, announced that it will support Microsoft's DirectPlay. In turn, Microsoft will distribute with its DirectX 5.0 SDK unique elements of the ImagiNation's SDK. Thus, there will be a standard set of APIs that developers can use which will also support content delivery on the ImagiNation Netowork. Some of these elements include: tournament server, service providers invite mechanisms, game watching and chatting.
Nichimen Graphics - Fast Track Announced for 3D Real Time Modeling
Nichimen Graphics rolled out Fast Track which includes components from its existing game development software N*World. Included components are N-Geometry, N-Paint 3D and N-Exchanger Lite. This software will run under Windows NT and SGI. A special promotion was announced at CGDC where two licenses for $6,495 one each on NT and SGI. The normal individual price is $9,995. Delivery is Q3 1997.
ST Labs - Opening of Regional Game Testing Facility Announced
ST Labs, Inc., the nation's largest outsourced software testing service provider announced the opening of its newest lab in Los Angeles. There are three laboratories. The Game Test Lab includes the latest gaming input/output devices such as "force feedback" joysticks, 3-D hardware components, and popular machines, including Pentium 166s, Pentium Pro 200s, and Pentium MMX 200s. The Static Lab is outfitted with the leading OEM-bundled machines, covering the most popular end user configurations. There are more than 90 PC and Macintosh configurations in the Static Lab. The Component Test Lab provides a way to test component and technology compatibility which use video, sound, and CD-ROM capabilities The lab has more than 30 PCs and Macs of various speeds, 65 video cards, over 40 sound cards, and 20 CD-ROM drives of various speeds.
Kinetix - Company Background Provided
Kinetix, a division of Autodesk, released interesting corporate data at CGDC. It provides nearly $30M in 1997 to its parent company of which 54% is international and 46% domestic. They claim that 41% of all PC game developers use 3D Studio MAX. Their share of the entire professional 3D animation market and authoring market - both PC and workstations - is 32%. The market share for 3D Studio MAX in the 3D PC market is 46% and the installed base if 66,000 for 3D Studio and 25,000 for MAX.
The following Spotlight Awards were presented at the CGDC. A complete list of winners can be found on their Web site.
711.4 Intel - Visual Computing for Arcade Games by John Latta
At the DirectX conference after the CGDC Intel announced its initiative for arcade games. The specification is on the Intel web site and it is entitled the "Open Architecture Coin-Op Reference Platform" - OAA CORP. This specification is based on the Pentium II using either the Windows 95 or Windows NT operating system. System memory is 32MB. 3D acceleration is at 3Dfx Voodoo graphics or equivalent. During the presentation Albert Teng stated that this is a scaleable platform and that a 4 SMP system could be used to support theme park requirements.
It is expected that systems will ship by October 1997.
We came away from the presentation and specification wondering:
How is this specification sufficiently different from a home PC to allow the arcade machine to have unique and compelling out-of-home games?
Are machines based on this standard cost effective in the arcade market?
Why will 4 Pentium Pro processors make a system suitable for theme parks when the real bottleneck is system bandwidth not processing power?
711.5 3Dfx - Developer's Conference before CGDC by John Latta
For the second year running 3Dfx had its developer's conference in advance of CGDC. Last year we estimate there may have been 50 but this year it was 350+.
3Dfx tested 6 games which used D3D. We found some of the results surprising. For example, of the six none used the Z-buffer capability and only one used RGB Texture with Fog and one other RGB Texture with Alpha. The observation made by 3Dfx is that the game development community has yet to leverage the capabilities being made available in hardware.
Some of the game developers described their experiences with both Glide and D3D. The developer of Formula 1 for Psygnosis stated that they wrote one version for Glide to maximize performance. However, for D3D they had to create 4 versions for each card that they were to run on. The observation was that optimizing for accelerators under D3D is less general than card specific API. To them D3D has not yet isolated the developer from being card specific.
Copyright 1997 4th WAVE Inc.
May be redistributed in full for individual readership and posted to newsgroups, Web, and FTP sites. May not be reprinted or redistributed for profit. Short quotes are permitted but must be attributed to the WAVE Report on Digital Media.