3dfx. An engineering blunder.
Jun 03 '00
Here's my take on the NVIDIA vs. 3DFX front.
First of all, 3dfx has been in trouble for quite some time now. They've been stuck behind the 8 ball, and they seem to botch everything they can on their way to a new product release.
There was a time where 3dfx ruled the roost and was simply the only game in town. They delivered 30fps performance in 3d with their original Voodoo Graphics chipset, and at the time it was revolutionary. To think that you could get that kind of performance in a card that cost $150 or less. It was great.
They had the market right through their Voodoo2 product line. That's when they developed a neat technology that allowed you to use 2 cards instead of one to increase performance. While this was a great idea, 3dfx seemed to like this technology more than the industry did. While I owned 2 Voodoo2 cards, and loved the performance at 1024x768, it was an added expense that really wasn't necessary.
It seems 3dfx never got past this kind of technology.
NVIDIA developed their Riva series of cards, and officially caught up with 3dfx's performance when they released the TNT2 Ultra series of cards. At that point they were on par (or faster) than 3dfx's Voodoo3 3000 line of cards, *and* offered 32 bit rendering at appreciable speeds. 3dfx kept bantering about how they didn't feel 32 bit rendering was important, and continued to develop 16 bit products.
In my opinion, 3dfx's stance on what they feel the public needs rather than what the public would rather see is troubling to say the least. I feel 3dfx's stance on this will hold back the market, rather than allow it to move forward. Here's why.
NVIDIA keeps developing technology for FUTURE games. When they developed the TNT line of cards with 32 bit rendering, no games took advantage of it. The most popular game I remember in that time was Quake 2, which had all 16 bit source art, according to the lead developer John Carmack. It's obvious that NVIDIA was looking to create a higher standard for gaming, while 3dfx remained complacent with current technology. A bad move in my opinion.
If NVIDIA had not developed 32-bit capable accelerators and just stuck with 16 bit, would any of our games look better today? There's no denying that a game that takes advantage of 32 bit rendering looks better than in 16 bit rendering. If NVIDIA had not developed a card that was capable, the software development industry would not have followed. After all, why would John Carmack make Quake 3 a gorgeous 32 bit game if no one had the hardware to use it?
The unfortunate side to this is that so many people STILL bought 3dfx cards because of the brand name. With all the voodoo cards still in existence out there, software developers could still code a 16 bit rendered game, and know it would sell because they were covered on both fronts. If half the people out there had a 16 bit card, and the other half had cards that could do both 16 bit and 32 bit, create a game that works for the lowest common denominator, a 16 bit card.
Now, how about Transformation and Lighting? 3dfx has decided to ignore it as well. Why? They feel that it's a "FUTURE" enhancement and does nothing for today's games. Same mentality. So if the Voodoo5 series of cards sells like hotcakes, what incentive will game developers have to promote this technology in their games, freeing up CPU resources for AI?
A lot of 3dfx folks will argue that T&L is pointless because today's CPU's handle T&L better. This misses the point of T&L though. If you can offset *ANY* functions to your video subsystem, that still leaves more CPU resources for other game apsects, like AI. That to me IS important. Two heads are better than one.
Now let's take a look at the Voodoo5's design. The Voodoo5 is based on the VSA-100 chip, their scalable architecture graphics processor. My first question to the industry: Would you want a card based on a SINGLE VSA-100? If you ask me, the VSA-100 on its own seems no faster than the GPU on the Voodoo 3 3000. I don't see the VSA-100 as being a significant engineering feat by any stretch of the imagination.
In an age where die sizes are shrinking, chip fabrications are reverting back to a 2" by 2" square rather than a 6" by 3" cartridge, 3dfx has decided to place more on a circuit board in an attempt to keep up with the competition. While NVIDIA is still packing as much punch into a single chip as they can, 3dfx is trying to counter their moves by just adding more chips to their boards. That's HARDLY a step in the right direction towards better engineering. My only hope here is that NVIDIA doesn't decide to do the same thing and offer a Geforce 2 GTS card with more than one GPU. That's not a step in the right direction. While it would be viable, and would likely sell like crazy, I'd much rather see engineers do "more with less".
With the huge trend towards overclocking in this industry (a trend I *don't* like), just how well-received is the Voodoo5 going to be anyway? When the Voodoo5 6000 arrives with its gargantuan-sized PCB, and 4 VSA-100 GPU's, what are overclockers going to do? You now have FOUR more fans blowing hot air into your case, something overclockers hate. It's a scary thought if you ask me.
About the only thing 3dfx did right was including full-scene antialiasing. While performance-mongers will find this feature useless, it does improve the looks of a game by straightening out jagged lines. While NVIDIA has countered this in their newest Detonator 2 drivers, they don't do it as effectively as 3dfx, since 3dfx's FSAA is a complete hardware solution. FSAA is hardly as major an improvement to games however when you compare it to the Geforce 2 GTS' T&L processing (adding 8 hardware lights and transformation offloaded to the GPU), not to mention the new "GTS" portion, the GigaTexel Shader. Per-pixel effects is something that was reserved for very high-end openGL cards in the past, and is now on the consumer market. While I hardly consider NVIDIA's accomplishments in the graphics industry "INNOVATIONS", I consider them significant because all of the additions they've made to their GPU's came from high-end graphics boards.
The bottom line is this: if you want the industry to move forward, you shouldn't support 3dfx. They concentrate on the here and now, rather than what you could have in the future. The 3d industry shouldn't stand still just because 3dfx thinks it should. I *WANT* to see per-pixel effects in my games. I *WANT* to see more hardware-based lighting in my games. I don't want to just blur what's there so lines look straight. Give me more polygons. Give me more realistic environments. Don't just slap more CPU's on your circuit board in an attempt to keep up.
And that's my 2 cents.
 |
|
|
|
|
|
|
Epinions.com ID:
kweckstrom
|
 - Top 500 |
|
Member: Karl Weckstrom
Location: Emerson, NJ
Reviews written: 100
Trusted by: 99 members
About Me:
Voted "Most Likely to be Photographed as a Bigfoot Sighting" by his senior class.
|
|
|
© 1999-2009 Shopping.com, Inc.