Market Overview ? Video Cards: Rainbow Warriors
Next time you?re saving the planet from the Empire, spare some thought for the humble graphics card. Chris Long reports on this booming market sector
You can tell just how far a certain area in the computer business has developed when its name changes and no one notices. Several years ago a video card was, well, just a video card. It sat inside a PC, took up a single slot and provided the pictures for the screen.
These days when you mention video cards you are surrounded by anxious marketing men desperate to tell you about video conferencing. Now, it would seem, video cards are graphics cards. Of course, they used to be graphics cards in the old days too, but then a graphics card was a new invention, because the early PC didn?t have graphics. So they were called video cards ? but we are in danger of repeating ourselves here.
When the PC first hit the street in August 1981 it had a simple ? and rather small ? boxy monitor and you could have only one screen colour: green.
IBM was just about to embark on what will probably be remembered as its TLA (three letter acronym) period and this would certainly be seen as the highpoint of the genre. The TLAs were needed as IBM (itself a TLA) desperately tried to keep up with what it thought the users wanted and make buckets of money. In the end it did neither, but who cares ? it laid the foundation of the video/graphics card industry of today (and made a lot of TLAs).
Ironically, the first graphics adaptor wasn?t graphic at all. The monochrome display adaptor (MDA) was a text screen driver ? it supplied the standard computer 80 x 25 character layout, each character made up in a 9 x 14 pixel matrix. Although, because one pixel at each side of the box, two at the top and one at the bottom are used for character spacing, the actual matrix size is 7 x 11.
Unfortunately this wasn?t very colourful. A combination of IBM noticing this and the company getting flak for not having anything to produce colour on its screens gave us the first colour offering, the colour graphics adaptor (CGA). At last the PC business world had something to draw pie charts and bar graphs with. CGA was by no means the answer to everyone?s prayers, though: 320 x 200 pixel resolution with four colours and 640 x 200 with two colours could hardly be called colourful.
All the same, CGA allowed graphics software that had been waiting in the wings a chance to do its stuff. Lotus 123 had been around for a while, but with CGA a rudimentary graphics standard emerged and finally it was possible to use colour on screen.
Around this time, Hercules released its monochrome graphics board. Realising that colour wasn?t all, it produced the Hercules graphics adaptor (HGA), which used ordinary mono screens and drove them at 720 x 348 pixels to produce high-definition graphics.
Enter IBM (again), with the enhanced graphics adaptor (EGA), part of a joint release EGA and PGA (professional graphics adaptor). PGA, with its 640 x 480 pixel screen and 256 colours out of 4,096, was aimed at professional graphics and Cad applications. PGA died without much fanfare ? its mid 80s price of $4,300, including screen, saw to that.
Some critics thought EGA was lacking, but this did not stop it taking off. It offered 640 x 350 pixel resolution and 64 colours. A couple of years later came the video graphics array (VGA) with its resolution of 640 x 480 pixels. After that Super VGA, at 800 x 600, took over at the bottom level and the graphics market became a bit of a free for all.
The price and availability of memory and monitors, plus the proof that when the chip manufacturers put their mind to it they can produce some very highly integrated silicon, all conspired to take graphics adaptors to a different universe from the one they started in.
As memory prices fell, it was possible to move on from the 256Kb of many mid-range cards to the 2Mb to 4Mb on today?s cards. Now a 17in monitor running 1,024 x 768 at 65,000 colours is not regarded as anything special and we no longer talk of standards and modes. Windows 95 has effectively seen to that with its introduction of a flat playing field ? albeit one that is full of pot-holes. All you need to do is produce a card and write a driver for it.
All of which brings us to today. Today we are confronted by developments in graphics cards that would have been almost unbelievable five or six years ago. Developers, with far too much spare time on their hands, have gone beyond the simple graphics card and are now intent on building more sophisticated equipment into their products.
The first stop was to develop Windows accelerators, which take graphics processing off the motherboard and processor, bringing it directly on to the card. In effect these are standard equipment on most motherboards these days.
Despite all this, the way the card works hasn?t really changed. Its job is to get information from the inside of the PC on to the monitor as quickly as possible ? no mean task, although the number of chips to achieve it has drastically reduced. For example Chips & Technologies saved EGA when it came up with a four-chip replacement for IBM?s 19 chips.
There are about four separate sections on a graphics card: the connection between it and the rest of the computer; a graphics processor that does the graphical number crunching; some video memory (VRam); and a Ram digital analogue converter (Ramdac) that converts the digital signal from the memory into electronic signals that drive the monitor.
The VRam is special because it is dual ported, which means it allows simultaneous access through two routes. So, a CPU or video graphics chip can manipulate the contents of video memory at the same time as other circuitry, usually a Ramdac chip, can read memory to send the picture to a monitor.
Other developments are found in the video memory bus size. Three years ago, everybody was amazed by the new 32-bit video cards which have a 32-bit data path between video chipset, video memory and Ramdac. Later there came the 64-bit video cards which can shift eight bytes in one go ? these are now more or less standard. Recently, chipsets were announced which have a 128-bit data path ? 16 bytes in one go.
The pressure of performance is spread around, though. The higher the screen resolution and the higher the colour resolution, the more data has to be transferred from the video chipset to the video memory, the faster the data has to be read by the Ramdac to be sent to the monitor. But improving the Ramdac is a good thing: the faster the Ramdac, the faster the screen refresh ? which means less flicker.
The growing number of PCI PCs means that there are few standard ISA graphics cards on the market too. Most of them are PCI now ? in fact Video Logic, for one, doesn?t sell any ISA graphics cards at all, they are all PCI.
So, having made cards that are so quick that most people scroll their wordprocessors or spreadsheets too fast to read, the manufacturers looked around for more to do, some sort of added functionality. Some manufacturers have built video capture systems into their cards; others have added TV tuners or Mpeg compatibility.
But perhaps the shift in the market is summed up by Video Logic announcing a #40 card that can produce 1,024 x 768 graphics. With that kind of price for that kind of performance, companies have been looking around for more to supply just to interest the users.
And for the past year or so the new technology has been arriving: 3D is here amid a flurry of competing standards and price slashing.
Sounding like something created at a marketing seminar, 3D graphics is, for the time being at least, the way of the future. Adding to the marketing feel, the label has also, with a resultant sinking of hearts, allowed companies to rename their current products as 2D.
3D takes the accelerator card philosophy to another level. Where the accelerator card takes actions and processes from the main processor, the 3D card does the same thing, but in the area of perspective and texture mapping. Thus having supplied lots of raw processing power for the creating of pictures, these new systems are aimed directly at computer games players.
In effect, 3D brings two general functions to the graphics board: rendering and geometry. Geometry involves lots of number crunching to work out things like perspective, direction of lighting and scaling the scene. Rendering is the painting of the image taking the geometry into account, so, say, if a picture had two light sources, the rendered picture would have to light splashes from the sources plus any shadows. Both these processes take serious processing power and it is here the 3D systems do their stuff.
As ever, this is not as simple as it sounds. Rather than supply the user with a perfect solution to the growing problem of underpowered computer games, the industry has supplied a number of choices that will serve to undermine the market for at least the next year.
The problem is how to talk to the 3D hardware. Not for the first time manufacturers have gone their own route and left the users behind and developed different APIs for their kit. As you?d imagine, Microsoft is in there somewhere.
Direct 3D, Microsoft?s API for Windows 95, enables applications to quickly produce high-quality pictures. This is what is needed for any sort of animation work, especially in games, where the user?s point of view changes quickly.
This is thought by some as a bit clunky and slow, which isn?t particularity a reflection on Microsoft?s code producing abilities, but more that it has attempted to produce a system that is all things to all users. The software is for use by everyone, but not everyone is using 3D ? currently it is mainly games people.
But games developers are past masters of the hyper quick shortcut, and are loath to use anything that might be slower (and by Microsoft). Thus while most of them are still developing for the Direct 3D API, some are actually porting games directly to specific hardware ? like the Creative Labs cards ? to give an extra speed boost. This, while obviously helping the user?s game play, just serves to confuse an emerging market.
One of the other APIs is Silicon Graphics? Open GL, descended from the company?s motion picture-oriented systems. While this is a popular and well established high-end API, it doesn?t quiet seem what games manufacturers want, which suggests that Microsoft may well win again.
It doesn?t stop there. While all the video hardware companies are producing faster and more powerful video cards, Intel has been doing what it is best at: upsetting the market.
The recent MMX chip with its 57 extra instructions is likely to change the game somewhat too: the new instructions are all intended to speed up multimedia operations, including sound and graphics processing. The Pentium II (Klamath) is the next iteration of the MMX design, and while the MMX-ness hasn?t actually changed, the processor will have the benefit of the new accelerated graphics port (AGP).
This is a dedicated high-speed bus which runs at 66MHz as opposed to the 33MHz of the PCI bus, setting a maximum data transfer rate of 533Mb per second between the processor and graphics system. At these speeds the AGP motherboard and processor (the Intel 440LX chipset will provide AGP support in Pentium II systems) can use main system memory rather than VRam for holding graphics data.
There will be products soon too ? the AGP group has over 80 companies readying AGP compliant products, all set to upturn the graphics market once again. There aren?t any products on the market yet, but the beginning of 1998 is likely to see them arriving.
With all this movement in the graphics market, it isn?t surprising that there is finally some in the channel. Some firms have been talking about the past 18 months being almost static. Users have, they say, not been upgrading because the onboard graphics have been of a standard that made upgrading unnecessary. Even the professional market is shrinking as the bog-standard accelerators that occupy the middle ground of the market do what a lot of the users want, although there is still a need for serious high-end graphics systems.
All the same, the big change in the upgrade market is in the games arena, where hardware and software are coming together to beat off the Japanese console game market. One vendor said its 3D graphics card was four times more powerful than the Sony Playstation ? all it needs now are some games. And even then there have been a couple of announcements from companies active in the games market to concentrate on the PC as a games platform.
Card manufacturers are surprisingly coy about the lack of ISA cards. Obviously the performance enhancements offered by the PCI bus are necessary for super-fast graphics, but there seems to be no real move to mop up the older generation of users with bog standard ISA machines ? although some still make VL-Bus cards available.
Dataquest projects that a worldwide market for devices with 3D acceleration in addition to 2D and video acceleration will emerge, reaching $582 million in 1997 and almost $1.4 billion in 1998. This suggests that we are seeing a new beginning in the video graphics arena. If the games become available, and if the MMX and AGP grouping doesn?t happen too soon and if the API issue is solved ? there may be a chance that the revolution to switch to 3D will succeed.
But if one ?if? in this business is bad enough, more than one is decidedly dodgy. Anyone looking to use any of these new technologies in solutions needs to keep a weather eye out for the storms that are likely to be on their way. A lot of people ? dealers and users ? are going to end up with obsolete kit if they aren?t careful.