|
|
|
|
|
Decoding MPEG Decoders
All rights reserved © Doceo Publishing, Inc. 1995
Think MPEG is a monolithic standard where all products work together in peace and harmony? HAH! Think again. This article presents a decision tree for purchasing MPEG decoding products, considering issues such as intended use, hardware vs. software decoders, hardware designs like VESA Feature Connector, VESA Media Channel, VESA Advanced Feature Connector and other alternatives like MPEG daughtercards and PCI MPEG decoders. If you're about to purchase 1 MPEG decoder card or 1000, read this article. To be published in CD-ROM Professional, November 1995.
"With freedom of choice comes confusion."
If one of our founding fathers didn't say this back in 1776, he surely would have after confronting the bedazzling array of MPEG playback options. In the eighteen months since the debut of the first commercially marketed MPEG board, Sigma Designs' REALmagic, the industry has seen new MPEG hardware and new MPEG software. MPEG on a graphics board, MPEG on a motherboard (in software and hardware), and MPEG on a daughterboard. MPEG overlay, MPEG inlay, and even an external MPEG adapter.
The inherent "standardness" -- one of MPEG's main attractions -- is a consistent thread throughout all these products. All CD-ROM drives can decode MPEG-1 encoded video. However, after installing only a few options, most users will quickly realize that MPEG is not the monolithic standard we expected. Whatever standards they share, all MPEG playback options are not created equal. They don't all play the same titles, they don't all integrate into the computer with equal smoothness, and they don't all perform equally well.
One way to explore MPEG playback options is through a decision tree, asking questions that help various types of buyers navigate towards the ideal MPEG option for a system's unique configuration. Even today, standing at the dawn of the MPEG market, we can't address every possible configuration, but a good discussion of the strengths and weaknesses of the respective approaches should be useful to all facing an MPEG playback decision and to all trying to understand how important MPEG will be for CD-ROM. Of course, the best place to start is by identifying the various components of MPEG playback devices.
But, at present, the larger question -- how will MPEG impact the CD-ROM title market and the title development scene -- can only be answered by inference. The number of MPEG solutions available today help tell publishers something about the role MPEG may play; the price, ease of installation, and quality of performance all contribute to a basis for prognostication. But the answers lie both in a closer inspection of the hardware and software products themselves and how they serve the multiple and various needs of developers and serious users of the technology.
The MPEG-1 compression standard comprehends both compressed audio and video. All MPEG playback devices -- hardware and software -- decode MPEG-1 video back to its original 352x240 resolution. MPEG decode chips decompress at 30 fps, with most of the key subjective differences relating to the quality of video scaled to higher resolutions. In contrast, there is a wide discrepancy between the playback rates of the various software playback devices. Hardware MPEG players decode the MPEG audio stream back to full CD-ROM quality (44 kHZ, 16-bit, stereo). While software players decode the same stream, they do so with varying levels of efficiency.
All MPEG playback devices have one or more "Application Programming Interfaces," or APIs, which define how programs incorporate MPEG playback maintain control of functions like starting and stopping MPEG playback. Virtually all MPEG playback devices conform to the new Open MPEG-1 (OM-1) Windows specification, a Microsoft specification under the Media Control Interface (MCI) family of devices, which includes laserdiscs and Video for Windows codecs. This OM-1 interface is well documented and familiar to anyone who has worked with other MCI devices.
However, because OM-1 is a relatively new specification, few MPEG titles to date conform to this API, which makes them incompatible with many new MPEG playback devices. In addition, very few MPEG decoders support DOS playback, making them not only unsuitable for many in-house uses, but also incompatible with all DOS MPEG titles. For those with immediate need for playback and development of titles with MPEG-compressed video, API is probably your most critical short term feature.
Depending upon the application, MPEG players should also decode the various higher level MPEG formats in addition to the lowest common denominator ISO 9660 format, which all decoders handle. These include CD-i (Green Book), Video CD (White Book) and karaoke. While most decoders support these formats, they do so with various levels of sophistication.
For example, as MCI devices, all MPEG decoders are accessible through Media Player, the basic Windows Media Playback application and OLE engine. However, not all decoders support track selection, a valuable feature for flipping through videos on a CD-i title. Decoders without track support treat all videos on the CD-ROM as one video, making it difficult to find precise starting points for the individual video tracks.
If CD-i or Video CD playback is a big part of the MPEG purchase, the buyer will want a decoder with a sophisticated display application. While consumer-oriented products like REALmagic Rave and Jazz Jakarta provide well designed, comprehensive display programs with features like volume control and multiple track queueing, most "industrial grade" products provide a low-end player, or point the user towards Media Player for MPEG playback.
Last but not least are bundled titles, especially for MPEG decoders purchased for home use. Today, most MPEG titles are bundled, and retail distribution is minimal. While this will change over the next eighteen months, the key players in the entertainment market, at least for now, are bundled titles, and the trend remains to tailor decoder board choices to the requirements of the desired bundles.
But the issues are more complicated than that, and the choices are always driven by individual circumstances and needs. Following a well wrought decision tree remains the best way to decode the question of what MPEG decoder to choose and how to choose it.
MPEG developers have different requirements than MPEG consumers, resulting in two tiers of MPEG playback products targeted towards the respective markets. Those encoding MPEG will appreciate advanced features like the ability to play back compressed video during encoding, and simultaneous output to the computer screen and an NTSC device like a television. An obvious rule of thumb for MPEG creation stations is to buy a decoder from the encoder manufacturer.
If you're rolling out a thousand kiosks, features like DOS and Windows developers kits, multiple video inputs and outputs, then multiple decoders in one computer and a history of OEM/VAR support will be more important than price. Conversely, these professional-strength systems offer little of the creature comforts of more retail-oriented decoder boards and are not well-suited for consumer systems.
The question of long-term needs versus present needs can be a determining factor in selecting one MPEG board or another, and is a question that goes directly to the issue of API. Although the OM1 specification was launched in June 1994, the few CD-ROM publishers who created MPEG titles since then wrote to Sigma Designs' REALmagic API, primarily for bundling arrangements with Sigma and its licensees. Even as late as August 1995, informal surveys for OM1-compatible titles turned up a grand total of one -- Compton New Media's Multimedia Encyclopedia -- and that was in beta testing. In addition, few decoders offer DOS playback, much less REALmagic-compatible DOS playback.
By now there are certainly more than a few OM-1 Windows titles, and DOS drivers will obviously become less important over time. However, to tap immediately into the widest installed base of MPEG titles, today's consumer would do well to purchase a product from Sigma or from one of their licensees, which includes Orchid, Genoa, and Digital Equipment Corporation. While other companies claim REALmagic compatibility, these are the only guaranteed compatible products.
On the other hand, if you're buying for title development or for in-house corporate projects, you'll likely develop for the OM1 API, supported by virtually all MPEG devices. While Sigma and its licensees are still in the running, they'll have to compete on features alone, rather than the installed base of compatible titles.
Whether or not one has opted for immediate gratification, the decision track becomes identical to more patient folks choosing generic systems, although the choices will be more limited. In either instance, your next decision is whether to upgrade or add-on.
Users can equip their systems with MPEG in two basic ways: "upgrading" with a combo card that supplies both graphics and MPEG playback and replaces the system's graphics card; and "add-on" boards that work with the system's current graphics adapter. In August 1995, the price differential between the two approaches was less than $50 in some instances, and will probably average less than $100 by first quarter 1996.
The best solution for users configuring a new system is clearly the combo route, which is cheaper, more stable, and saves an expansion slot. But if the issue is adding to an existing system, several factors must be considered in choosing between a combo and add-on approach.
First are the capabilities of the system's current graphics card. Graphics technology advances very rapidly, and the last twelve months have brought the debut of 64-bit architectures and video acceleration features like the Display Control Interface (DCI) and video scaling. If the system's graphics card doesn't offer these features, the solution may well be an upgrade. On the other hand, users who have just purchased the latest graphics screamer may be loath to discard it.
From an operational standpoint, combo cards tend to be more stable and are better suited for adding MPEG capability, especially if the work involves image editing or other 24-bit applications. Historically, separate MPEG and graphics cards have worked together uncomfortably at best, and often not at all. Early MPEG add-on cards were notorious for limiting base color depth of the graphics card and ignoring common Windows display protocols like yielding video display when switching between applications. While recent advancements may improve this situation, it's tough to recommend the add-on approach for general purpose computers. However, add-on boards work well for kiosks and similar applications or home computers used primarily for games.
The third factor is future expansion. As a general rule, combo boards offer more opportunities for additional features like image capture and compatibility with game formats like SEGA and 3DO. If you're converting your computer into a digital home entertainment center, combo boards save a slot and provide greater flexibility.
The heart of the decision for prospective MPEG purchasers choosing to follow the "upgrade" track is focusing on the combo card selection process. Several factors to consider step to the fore in a combo card purchasing decision, among them expansion options and playback vs. publishing orientation, as well as performance, price, and compatibility issues.
The first question buyers must ask themselves as they burst from the upgrade track blocks is, "Am I buying a graphics card with MPEG capabilities or an MPEG player that does graphics?" On PCI-bus computers, for example, the two foremost options may be the Jazz Jakarta and Matrox Millennium; one represents the graphics card strength and the other, the MPEG strength.
Jakarta comes with several sophisticated MPEG, CD-i, and Video CD players, the package often features a bundled title and a range of expansion options as well. While the graphics control software is outstanding, graphics performance, built around Tseng Lab's 32-bit ET4000 chipset, is comparatively modest.
In the other corner is Matrox Millennium, equipped with the optional MPEG daughtercard. Matrox's 64-bit MGA architecture is a perennial benchmark leader, producing faster scores at higher resolutions and refresh rates than Jakarta. Matrox's graphics control software, honed for the demands of the desktop publishing market, is also very sophisticated. However, Matrox's MPEG offering is very slight: no player, no bundled titles.
Graphics cards with MPEG are the best choice for users and developers who work at extremely high resolutions like 1280x1024, or perform graphics intensive functions like image or video editing. On the other hand, where MPEG playback is paramount, the MPEG playback device is the better option. Most of these devices, like Jakarta, are designed from the start to provide the ultimate in home entertainment.
For buyers choosing between boards in either category, several additional considerations apply. First is pure MPEG quality, which is usually determined by the MPEG chip and graphics chips. While most MPEG decoders produce similar quality video at 352x240, quality varies widely when the MPEG is scaled to higher resolutions. For example, one key difference between C-Cube's CL450 and CL480 is that the latter filters video during scaling. All else being equal, the CL480 will produce higher quality, full screen video.
On combo boards, other chips also contribute to scaled quality. For example, REALmagic Rave uses Sigma's own WM7010A chip to scale video produced by C-Cube's CL450 MPEG decoder. In comparison, Jakarta uses Tseng Labs ET4000 graphics engine and Viper video coprocessor with Zoran's ZR36100MPEG decoder. While quality is identical at lower resolutions, the Tseng chipset filters during scaling to produce much clearer video than Sigma's during scaling.
The next step in evaluating the product's pure graphics capabilities -- even if the buyer's primary concern is MPEG playback -- is to identify the graphics chipset type of video RAM appropriate to the particular case. In most instances, 64-bit chips will outperform 32-bit chips, while VRAM, SRAM, and the new Windows RAM are faster than the cheaper DRAM. Twice a year, in May and November, PC Magazine reviews video cards, usually ranking graphics processors as well. Keeping these issues on file will make it easy to determine whether a combo board is a graphics tortoise or hare.
Check for additional capabilities like acceleration and scaling of formats like Video for Windows (AVI) and QuickTime for Windows (MOV), extremely useful given that most video on CD-ROMs is compressed in these formats. While Millennium and Jakarta scale these formats, REALmagic Rave, Orchid Kelvin MPEG, and Genoa GVision DX don't, making them less useful in a general purpose multimedia setting.
Also important are expansion options. Jakarta offers the widest range of options, including TV tuning, still image and video capture, compatibility with Sega, Nintendo and 3DO games, and VGA to NTSC capabilities for viewing computer games on television and encoding video to tape. Most other combo boards focus on simple graphics performance and MPEG playback and offer few, if any options.
Not all boards are available for VLB and PCI configurations. Most notably, Sigma has neither released nor announced a PCI version of Rave, while most mainstream graphics developers like Matrox have eschewed the VLB platform, choosing to debut their MPEG options as PCI products. But MPEG capability seekers are ill-advised to select a particular product before checking its compatibility with the bus configuration of the system being upgraded.
Users looking to add MPEG options to their existing graphics subsystem, rather than upgrading their entire graphics subsystem, should follow the add-on track and consider the options it presents. For those whose current graphics card offers an MPEG hardware expansion option, the choice is simple: buy the daughtercard. For example, Diamond Multimedia offers the Motion Video Player MPEG expansion option for its Stealth video family, available for both VLB and PCI computers. Similarly, VideoLogic offers an MPEG Player option for their Movie 928 cards. If such a daughtercard isn't available, the question becomes more complicated; in such cases, video-upgrade-minded users must weigh several factors in their decision.
The first question users contemplating add-on options must answer is whether they are likely to work in 24-bit graphics mode or an 8 or 16-bit alternative is sufficient. The answer to this question relates to the type of connection used between the MPEG and graphics card.
Most popular MPEG playback boards like Sigma's REALmagic use the VESA Feature Connector (VFC) to communicate to the graphics card, connecting via an internal ribbon cable attached to both boards. In most instances, graphics are routed through the VFC to the MPEG card, which connects directly to the monitor. Unfortunately, because the VFC is an 8-bit serial connection, it can't handle the data flow necessary to support in high color modes at the fast refresh rates necessary for flicker free display. Consequently, the graphics card the user purchased for the unit's rock solid 1280x1024 wide screen display or blazing 24-bit graphics speed will likely be forced into low color, low resolution or low refresh mode, just to load the MPEG software. Because all graphics are pumped through the video card, users taking this tack are forced into the lower performance modes even when not running MPEG applications.
For example, back in June 1994, the original REALmagic software was incompatible with an ATI Mach 32 graphics adapter in all modes higher than 4-bit graphics, useless for playing back AVI files or image editing. By June 1995, Sigma's new drivers enabled operation in 8-bit mode, which was better, but still unacceptable for those working with graphics-intensive applications.
While Sigma has served its public well by compiling a small booklet of known incompatibilities with graphics cards, the mere existence of such a booklet highlights the architectural problem: the VFC is a 1991 technology that's simply inadequate for merging MPEG with high color graphics. If a particular project requires working with image-intensive 24-bit applications, or using a wide screen, flicker-free display for spreadsheet manipulation, the best bet is to avoid overlay. On the other hand, if adding MPEG capability means simply upgrading an older computer with maximum 8-bit graphics, these problems likely won't appear.
Another common architecture involves the VGA loopback connector, an external cable connecting the graphics card's monitor port to a port on the MPEG card. Once again, the monitor connects directly to the MPEG card. Since the external loopback cable isn't bandwidth-limited like the VESA Feature Connector, this approach doesn't place artificial limits on graphics performance and is better suited for users requiring access to high performance graphics.
Unfortunately, the VGA loopback connector isn't without its limitations. Unlike the VFC approach, which transmits digital information to the MPEG decoder, the loopback connector sends video already converted to analog format, which is more complicated than the VFC approach and more difficult to stabilize.
In addition, like the VFC approach, MPEG video is overlaid on top of the analog graphics during transmission to the monitor. Unfortunately, the lack of video overlay standard APIs makes it difficult for MPEG applications to conform to common Windows display protocols like yielding foreground screen display when boxes or other applications are placed over the MPEG video window.
While this isn't a problem for dedicated gamers, it could prove disconcerting for business users. Imagine ALT-TABing from your multimedia presentation to your word processor to review a proposal with a customer, only to have the MPEG video blissfully play on in the foreground. This is common behavior among VFC and VGA loopback-oriented boards.
In their recently announced REALmagic PLUS board, Sigma takes the VGA loopback approach one step further via a custom ASIC that mixes the MPEG and graphics streams together in analog format, promoting stability, according to Sigma. As the first non-VFC Sigma add-on product, the PLUS card provides REALmagic compatibility without limiting graphics performance. For ISA and VLB computers that don't meet Display Control Interface (DCI) or Graphics Device Interface (GDI) specifications, REALmagic PLUS could be the add-on decoder of choice for those working with 24-bit applications.
Two key players surging to the forefront of the connector question are the VESA Media Channel (VMC) and VESA Advanced Feature Connector (VAFC), two graphics port specifications recently defined by the VESA standards committee. Both let MPEG add-on cards work with graphics cards at any color depth, resolution and refresh rate.
Unfortunately, neither specification uses the popular VESA Feature Connector, since both require new connectors and on-board silicon. So while VMC and VAFC proponents vehemently promoted the technical merits of their approaches, graphics vendors ignored both camps in droves, looking towards the high bandwidth PCI bus to solve both the compatibility and bandwidth problems discussed above.
For the developer or user adding MPEG to a PCI bus computer, two viable options present themselves, both of which require DCI-enabled graphics on the system. DCI, which stands for Display Control Interface, is a relatively new graphics specification that lets codecs write directly to the frame buffer of the graphics card rather than working through Windows Graphics Device Interface (GDI). Essentially, DCI removes some of the overhead associated with decompression under Windows, boosting video display rates by up to thirty percent. DCI also provides a protocol for letting video cards accelerate video playback by performing higher level functions like color space conversion and video scaling. Both advanced PCI options require DCI.
Not all graphics systems use DCI, of course, and whether or not they do is also not always readily apparent; however, most mainstream computers and graphics cards purchased after March 31, 1995, implemented the new specification. The words DCI will always appear in the system's "system.ini" if the capability is installed. If it's not, the VFC and loopback approaches presented above are the only add-on options the system will allow.
The newest wrinkle in the MPEG add-on decoder market is advent of PCI-based MPEG decoders, like ELSA's new ELSAMotion, which transfer decompressed MPEG video to the video card via the PCI bus rather than a feature connector or loopback cable. Since the MPEG is sent to the graphics card, rather than vice versa, there are no limitations on graphics performance.
Under this approach, MPEG playback is architecturally similar to playing other formats like Video for Windows or QuickTime. Rather than maintaining two distinct video streams -- one graphics and one MPEG -- combined via overlay, the decompressed MPEG is transferred to the video card just like other decompressed video files, making it easier to conform to Windows' display protocols. Using the PCI bus rather than a direct MPEG-graphics connection also avoids most compatibility issues between MPEG decoder and graphics card.
The result is a well integrated, architecturally sound approach that should prove both more compatible and more stable in operation. It's definitely the best option for DCI-equipped, Pentium 60 PCI-bus computers -- in all cases except where immediate gratification is first and foremost. While Sigma has privately stated that they will ship an OEM version of a PCI MPEG card, no retail version is planned.
However, users who own Pentium 90 or above face one final question: What's more important, performance or price?
| CompCore | Mediamatics | ||
|---|---|---|---|
| Video-CD Sampler (Big Hand Productions) | |||
| Viper Pro - 2X | 20 fps | 20 fps | 10 fps |
| Matrox Millennium - 2X | crash | crash | crash |
| Motion 771 - 2X | 14.3 fps | 13.6 fps | 4.7 fps |
| Dragon's Lair |
| ||
| Viper Pro - 2X | 16 fps | 15.5 fps | 3.6 fps |
| Matrox Millennium - 2X | 10 fps | 9.4 fps | 3.6 fps |
| Motion 771 - 2X | 7 fps | 9.8 fps | 3.2 fps |
The question of whether performance or price should drive MPEG choices is itself driven by the ever-complicated hardware vs. software decision. Software MPEG will soon be a free option on most new graphics cards, and on Pentium 90 and above class computers, software-only MPEG can play back at 22 to 24 frames per second. Thirty frame per second performance has been reported on Pentium 133 systems equipped with high powered graphics cards.
Software MPEG players are actively marketed by the following three companies:
All three of these players are primarily concerned with bundling with graphics cards and titles, although it appears that Xing and CompCore may distribute a retail consumer version in the near future. All three players support the OM1 Windows specification; none supports the Sigma REALmagic API or offers a DOS version. IBM recently announced its own entry into the software MPEG market which will be shipped with Aptiva computers starting this fall.
Software MPEG performance varies by codec and graphics card. Mediamatics results were the most surprising, given that the software runs successfully on a Pentium 90 at 22 frames per second. Unfortunately, as with any test performed on only one computer, as in the case of Table 1, the results should be considered informative, but not conclusive.
Early tests also appear to indicate that Mediamatics makes certain quality-related trade-offs during decompression, not creating quite as detailed an image as either of the other software players. However, this was not apparent during playback of normal sequences. Again, the preliminary nature of these tests must be conceded, and the resulting limitations apply.
To achieve this impressive performance, all software MPEG players sacrifice audio quality, producing only 22 kHz, 16-bit stereo rather than the true CD-ROM quality 44 kHz, 16-bit stereo. However, this difference is apparent only to audiophiles working on high fidelity computer speakers. In informal tests comparing audio output from a fairly high performance computer audio system (SoundBlaster AWE32, Yamaha's YST-MSW10 speakers), the difference between hardware and software is not easily discerned.
In terms of comparative audio quality, Mediamatics led the pack. On several sequences Xing produced a warbling effect, and stuttered occasionally when played back on a SoundBlaster. CompCore was stable but didn't sound quite as good as Mediamatics.
Overall, the choice between hardware and software is determined by intended use. For users purchasing MPEG to play videos downloaded from the Internet or from sedate multimedia encyclopedias, software MPEG should suffice on Pentium 90 computers and above. In fact, the latest MPC-3 specification requires only software MPEG and includes systems as low as a Pentium 75. On the corporate side, software MPEG is probably also sufficient for most day-to-day playback of MPEG video.
On the other hand, software MPEG demands virtually all system resources during decompression, making the system slow and unresponsive. Hardware MPEG will be much more desirable for ardent gamers seeking to maximize their multimedia experience.
Similarly, in the corporate market, hardware MPEG will pay for itself with more dynamic presentations, and is a must for multi-tasking users needing to watch videos while performing common background activities like checking electronic mail. Overall, since software MPEG will be showing up free in more and more places, perhaps many will try software and then upgrade to hardware if performance is inadequate. Indeed, most developers of hardware MPEG devices claim that software MPEG will be their best advertisement -- give users a taste of MPEG video quality, and they'll want maximum performance.
Users planning imminent upgrades to Windows 95 are probably well aware that Microsoft has licensed Mediamatics software MPEG decoder for the new operating system. Rather than shipping the Mediamatics player in the main Windows 95 build, at least initially, Microsoft will distribute the player without charge to all developers, mimicking the successful strategy used to distribute Video for Windows runtimes. Expect the player to show up as the default player in many multimedia titles shipped for Windows 95.
But some initial testing suggests that Mediamatics may not be the best performance option, even if the price is right. Both Xing and CompCore will release Windows 95 versions that will be worth looking into for even casual MPEG playback. At the very least, however, Microsoft's move assures title developers that if they release OM1-compliant MPEG titles, there will be plenty of computers available to play them. This should boost the OM1 title development, which bodes well for OM1 software and hardware players.
For those buying a new computer and seeking integrated MPEG, two new options are staking a claim, one hardware, the other software. In June 1995, Compaq announced that they were including MPEG on the motherboard as a $250 option on new Presario computers. While the move is a nice design win for S3 Inc., the chip company supplying the MPEG decoder chips, because MPEG is an option -- and rather expensive at that -- it probably won't trigger a stampede towards MPEG on the motherboard. However, expect other computer vendors to offer a PCI MPEG decoder like ELSA's in the short term to counter the Compaq announcement.
On the software side, IBM recently announced that software MPEG will be shipped with each Aptiva computer starting this fall. Aptiva's software MPEG will have one key advantage over the other three players -- the programmable MWave audio chip on Aptiva's motherboard will decode the MPEG audio stream, not the host CPU. This should boost IBM's performance beyond that achieved by other software decoders using the host CPU for both video and audio.
And what about Windows 95, which has promised to deliver a significant boon for multimedia? In the short term, the shift from DCI to Direct Draw in the new operating system could make the transition from Windows 3.1 to Windows 95 far from seamless for many of the playback devices currently available. [See Larry Gussin, "Windows 95 and the Multimedia Initiative," CD-ROM Professional, Volume 8, Number 8 -- ed.]
After eighteen months of ardently pushing DCI, Microsoft reversed course last April, announcing Direct Draw, a new multimedia architecture that replaces DCI. Formulated in response to requests from game and multimedia developers, Direct Draw offers additional direct access to graphics hardware functions such as texture mapping, overlaying, rotating, and mirroring.
Direct Draw suffers from one key flaw, which is that it's not 100 percent backwards compatible. This means that all DCI-enabled devices will probably have to write Direct Draw drivers to regain total functionality. Specifically, while DCI-compatible devices will still have direct access to the frame buffer of the graphics card, the APIs for performing color space conversion and scaling have changed. Devices that use DCI to perform these functions, like the ELSAmotion, all software MPEG players and certain combo cards, won't perform them under Windows 95 without a Direct Draw driver. This virtually assures a short term performance drop, especially when scaling to full screen.
Unfortunately, Direct Draw support reportedly didn't make it into the first retail build of Windows 95, but should start shipping with the second build sometime in the early fall. In the interim, Direct Draw will be widely available from graphics card vendors, software MPEG vendors and games developers.
The bottom line, however, is this: If users have DCI in their system, and use video for mission-critical purposes (whether for business or pleasure), they should not upgrade to Windows 95 without first having a Direct Draw driver for their DCI devices. On the other hand, the video-hungry would be ill-advised to buy any MPEG decoder or graphics products that don't ship with a Direct Draw driver.
Emerging markets follow a similar pattern. Initial product entries come from small companies hoping to take a dominant share in what they've identified as a hot new market. If the market shows promise, clone manufacturers flood the channel with low cost products, especially if the market evolves around a standard like MPEG. Large established players sniff at an emerging segment until it becomes mainstream, then introduce products that ultimately gain significant share due to market factors like name recognition, shelf space and volume purchase agreements.
The small companies that kick-started the market often succeed, especially if they're lucky, plucky, and have plenty of capital. Clone manufacturers drift into new markets, as the heavyweights drive down prices and fill available channels. This scenario should be considered in early purchasing decisions.
Sigma and Jazz invested heavily in the MPEG decoder market, and can claim significant credit for its success. Their market positioning and early success should hold them in good stead as the market unfolds. The same, however, can't be said for many clone developers.
Over the next twelve months, virtually every major graphics manufacturer will enter the market, including ATI Technologies, Diamond Multimedia, Number 9 Visual Technologies, and Matrox. Over the same period, the general market will undergo substantial adjustment, from Windows 3.11 to Windows 95 and from DCI to Direct Draw, undoubtedly requiring significant driver updates. In this dynamic market, the safest course is to purchase from either an early market leader or established player.
From a CD-ROM title developer perspective, everything's coming up MPEG. An API standard has coalesced, software players are becoming real, hardware add-on and combo alternatives abound, and MPEG is budding on some pretty toney motherboards. All this bodes well for increased sales of MPEG-based CD-ROM titles.
Now, never mind that decision tree -- it may well be time to start figuring out how to plug that MPEG card in. . .
Jan Ozer is president of Doceo Publishing, a multimedia title developer and publisher of the Video Compression Sampler series. A CD-ROM Professional Contributing Editor, Ozer is author of Video Compression for Multimedia, published by AP Professional in late 1994. Ozer can be reached at Doceo Publishing, Inc., 568 14th Street, NW, Suite 200, Atlanta, GA 30318, at (404) 892-2889, or jan@doceo.com.
One of the most intriguing but least talked about issues arising from MPEG's recent ascent is royalties. Many companies, including Philips, IBM, and Sony, contributed to the pool of patents protecting the technology that makes MPEG possible. At some point, one assumes, they would like to be paid for their work.
In a press release dated March 27, 1995, Cable Television Laboratories, Inc. (CableLabs), announced the results of an MPEG Intellectual Property Rights (IPR) committee meeting held in Switzerland the previous week. The press release states, in part, as follows:
The group arrived at an initial conclusion on a model for paying royalties on MPEG-related products such as digital encoders; digital decoders, including settop boxes; digital videodisk (DVD) players; and prerecorded storage media such as video CDs, DVDs, and prerecorded magnetic media.
The group has discussed and has reached consensus on targeting a $3 to $4 (U.S.) royalty on each digital decoder, including MPEG-2 settop boxes, digital videodisk (DVD) players, and decode-for-general-purpose processors. Using that target and based upon the model discussed below, that would result in a $0.03 to $0.04 (U.S.) royalty on each video CD or DVD that would be purchased. Also, there would be a $0.30 to $0.40 (U.S.) royalty on each video CD or DVD that was for distribution use (i.e. rental market).
The press release did not state whether the royalty would be applied retroactively or prospectively, or when the MPEG-IRP group would start calling upon MPEG encoder/decoder developers and title developers. However, since the $3-$4 (U.S.) royalty roughly doubles or triples the licensee fee paid by graphics card developers for software MPEG decoders, it threatens to change the economics of including such decoders with every graphics card. One graphics card developer flatly stated that they were delaying shipment of MPEG software until the royalty situation was resolved. Several others expressed concern over potential MPEG royalty liabilities for previously shipped product.
Industry watchers focused on Microsoft's licensing of Mediamatics player for free distribution as a bellwether of the MPEG-IPR group's intent. Early rumors spread that Microsoft wouldn't distribute the Mediamatics player until an agreement was reached with MPEG licensing group. Microsoft officials, busy with the Windows 95 launch and its immediate aftermath, were not available for comment.
Sean O'Toole, marketing director for Xing Technologies, who stated "The MPEG IPR group sought input from the computer industry and got an earful. Apparently they listened, and as a result, will announce a patent plan by November calling for a non-retroactive, $0.10 (U.S.) royalty per decoder. Since the royalty won't be retroactive, it's not a major concern for us right now."
O'Toole had no information on royalties for titles or encoders. Calls made to CableLabs were unable to scare up any verification of O'Toole's information.
One thing is already certain, however. There will be lots of justifiably nervous chief financial officers worrying about potential royalty obligations on encoders, decoders and CD-ROM titles until this situation is resolved.
-- Jan Ozer
ATI Technologies 3761 Victoria Park Avenue Carborough, Ontario M1W 3S2 Canada 416/756-0718 Fax 416/756-0720 C-Cube Microsystems 1778 McCarthy Boulevard Milpitas, CA 95035 408/944-6357 408/944-6314 CompCore Multimedia, Inc. 1270 Oakmead Parkway, Suite 214 Sunnyvale, CA 94086 408/773-8310 Diamond Multimedia Systems, Inc. 2880 Junction Avenue San Jose, CA 95134-1922 408/325-7204 Fax 408/325-7411 Digital Equipment Corporation 1 Digital Drive Westminster, MA 01473-0471 508/874-4454 ELSA, Inc. 2041 Mission College Boulevard Santa Clara, CA 95054 Jazz Multimedia, Inc. 1040 Richard Avenue Santa Clara, CA 95050 408/727-8900 Matrox Electronic Systems, Ltd. 1055 St. Regis Boulevard Dorval, QC H9P 2T4, Canada 514/685-7230 Mediamatics, Inc. 4633 Old Ironsides Drive Santa Clara, CA 95054 408/496-6360 Fax 408/496-6634 Number Nine Visual Technology 1B Hartwell Avenue Lexington, MA 02173-3103 617/674-8513 800/GET-NINE Fax 617/674-2919 Orchid Technology 45365 Northport Loop WestFremont, CA 94538 510/683-0300 Fax 510/490-9312 Sigma Designs, Inc. 46501 Landing Parkway Fremont, CA 94538 510/770-2602 510/770-0100 Fax 510/770-2640 S3 Inc. 2770 San Tomas Expressway Santa Clara, CA 95051-0968 408/980-5400 Fax 408/980-5444 VideoLogic 245 First Street Cambridge, MA 02142 617/494-0530 Fax 617/494-0534 Xing Technology Corporation 1540 West Branch Street Arroyo Grande, CA 93420 805/473-0145 Fax 805/473-0147