Geometry's threat to Intel in the CPU race

For the CPU, Sega selected Hitachi's latest generation of its SuperH RISC architecture, a 200MHz SH-4 CPU, which delivers 128-bit performance in 3D calculations, resulting in highest resolution and faster graphics. In fact, the SH-4 processor is optimized to perform floating-point operations -- the key to super-fast 3D calculations four times faster than the Pentium(R) II.

If that wasn't bad enough for Intel, now there is now a conflict brewing between Intel and the graphics accelerator companies that goes beyond the fact the Intel is getting into the graphics business.  Intel continues to look for more reasons for people to buy the latest and greatest processors and as all us gamers know a slow processor is more important to us than just about anyone.  But the high CPU costs are getting rather prohibitive.

In early May graphics companies at the Computer Game Developers Conference (CGDC) repeatedly said they want to expand the image processing and set-up on their chips.  This desire is in response to the growing bottleneck in the CPU.  The goal is to churn out more triangles than the competition, but the reliance on the CPU makes things much more complicated for the game developers.

Why? PC's use the CPU to calculate geometrical transforms and lighting effects.  A dedicated hardware geometry engine, as compared to a CPU, is more efficient at handling the exceptions that happen when setting up a geometric objects.

Game developers have three goals:

1. develop games for the masses
2. develop games that are easy to play
3. develop, cool games with the latest bells and whistles to capture the users attention.

The first two goals do not conflict, but in the race to develop the coolest full featured, polygon rich, games the developers are running into a direct conflict with their first goal- develop games for the masses.  If a developer produces a game with too many polygons in a scene, then they run the risk of alienating most of their audience.  For instance Quake II uses higher polygon rates and is thus much more CPU dependant than most games, this causes the user not only to worry about the graphics chip they have, but also the CPU.  If you don't have at least a Pentium 200mhz to go along with your 3Dfx Voodoo board then you probably will not be happy playing the game.

Most hard core gamers will upgrade to play the latest games, but the price tag becomes daunting when they have to buy CPUs at $400-500 a shot- not to mention buying a new motherboard and memory.  When you consider the continuing costs of upgrading the CPU you start looking for some other solution.

The good news is that 3DLabs, 3Dfx, S3, SiS, TriTech, and Nvidia are going to build integrated processors with geometry engines. Microsoft is going to allow for geometry hardware in DirectX 7 which should be out in 1999 and will be smart enough to use the features that are supported by the new chips.  This will allow the geometry to be a standard on-chip feature in 1999.  This will allow a new market to open up- the sub $1000 high end gaming machine.  All the gaming performance of a high end Pentium II or Katamai system, but a $1,000 box price.  The only way to achieve this is to put the geometry in the graphic chip's engine.  The sub $1000 market looks like the sweet spot right now.

While sales are dropping in the traditionally dominant (usually 60+% of the market) $1000-1500 range, the sub $1000 computer has eaten away profit margins and gained 10% of the sales in just the first month of the price wars (January to February 1998.) Dropping margin are causing the big boys to be concerned about their bottom lines.

The bad news is that Intel and Microsoft, along with some of the PC makers, are mounting an effort to prevent this from happening.  Their goal is to increase the need for multimedia in the work place to keep people looking towards their CPU for help.  These business apps would require more than just geometry setup.  If they are successful, they may be able to prevent the graphic chip makers from creating this low cost gaming machine.

Intel and the large PC makers will lose suffer as they see their margins drop.  Intel, in particular, stands to lose the most since they are geared to always be ahead of the technology curve and constantly move the end user to the next CPU to keep sales going.

Intel is committed to stopping the erosion of their market share and the establishment of low end alternatives.  Last year they established a 75 person team, the Business Development Relations Group, whose goal is to help ISVs develop multimedia software for Pentium II PCs.  The group is working hard with intensify Intel's own multimedia efforts and to grant early access to system/CPU prototypes to the ISVs to enable them to provide the business applications as close to the launch of new CPUs as possible.  Of course Intel's plan is challenged when you are able to run Quake II at 60fps on a 133Mhz processor and a graphics card that does the geometry work.  Intel wants us to believe that a "balanced" system is better, and that their next-gen Katmai processor will erase any doubt that the CPU can handle the geometry.

On the other side of the argument the graphics chip makers argue that none of the new processors from AMD, Cyrix, IDT, and Intel, and the extra instructions that they will bring, will still lag behind what the graphics chip can bring to the table.

Conclusion: Great multimedia is nice, fancy spreadsheets and presentations are cool for business, but how many of us home users are going to need that? In the end the PC makers will need to build the systems that users want.

With sub $1000 prices, the PC is becoming a commodity that more and more average people will want.

The question is: What will the average user want to use the computer for? I admit, I'm a speed hound myself, but it is becoming increasingly obvious that my home use of my computer does not even come close to challenging my Penitum II 266Mhz.

In the end lets hope that the graphic chip makers can stand up to Intel's might, and provide us with lower cost solution to upgrade our systems.

[Previous Page]

[Top of Page]  [Reload Page]

[Forward]