3dfx and The Graphics Game
Author: Patric Grinsell
It's been a long time since my last update on 3dfx and much has happened since then. Recently 3dfx has announced Voodoo3, a slew of OEM's including Gateway, and a record time to fab Voodoo3. I'll take each piece bit by bit and break down what I think are the effects.
Voodoo3:
I'll give it to you short and sweet. Voodoo3 is Voodoo2 SLI and Banshee crunched into one card. (There are other minor differences, but I want to keep it simple.) This raw performance will almost certainly seat 3dfx on top of the performance heap, but let me get to the performance thing in a second. This chip was actually intended to be part of the Banshee family and wasn't intended to be their next generation. Sources are saying that the Voodoo3 name was actually switched from Avenger in the last few days before the product announcement. Needless to say, many 3dfx fans were disappointed that this “Voodoo 3” didn't come with a whole new slew of visual enhancements as well as performance that is only on par with Voodoo2 SLI. I have been aware of this product for some time and 3dfx originally billed it as their “Banshee follow-on product.” It was very clear that the major reason for this second version of Banshee was to include AGP 2x, a marketing mistake that they regretted the first time around. You can imagine my surprise when 3dfx announced something as powerful as Voodoo3, a product that seemed originally conceived only as an afterthought.
The key here is that Voodoo3's primary target is the OEM and not the consumer, just like with Banshee. With the exception of performance increases and 2d, Voodoo3 is essentially the same product as Voodoo1. While the net has been full of technophiles bemoaning the fall of 3dfx for not supporting bleeding edge visual enhancements, these are the types of things that OEMs hate. Why? Because OEMs put reliability as a top concern, even over wiz-bang function. The average consumer won't put up with downloading the latest drivers to get their brand new video card working in the brand new machine. If the apps don't work, they might just return it. Secondly, a feature is only important if it is named in the advertising of the machine. AGP 2x is a prime example. Even though AGP 2x yields minimal performance results at best, the average consumer thinks of 2x being better than 1x. Features like hardware bumpmapping will never make it to the graphics features list in an add and are therefore unimportant.
So this kind of begs the question, “What do OEMs want?” Well for now here is the list of the important items:
1) AGP 2x (or better)
2) 16 MB framebuffer (or better)
3) 3d functionality
4) Cheap price
5) Powerful performance*
What's that little (*) doing on #5? It isn't real performance that matters to the OEMs but benchmark performance such as 3d Winbench. The graphics technophiles will note there is a huge difference between benchmark performance and real world performance. ATI has managed to stay on top of the OEM heap because they are pros at fudging the benchmarking stuff. The OEMs know this, but I'll let you in on a little secret: They don't care! As long as their machine is being benchmarked as the fastest graphics performer then they can dupe many a consumer into buying their overrated piece of garbage. Don't get me wrong, they wouldn't mind if those benchmarking numbers were real, which will be the case with Voodoo3, but I suspect it would be impossible to throw in the kitchen sink in terms of visual quality, and performance, and still make it as cheap. Because 3dfx is still working on their next-generation 3d (Rampage, it's officially trademarked now), they must make a conscious decision about what will and what will not be included in Voodoo3. Every added feature costs more money and delays shipment of the OEM product and subtracts from the uniqueness of the next-generation product. 3dfx has stated regret for only one drawback of Voodoo3 and that's the limit of a 16MB framebuffer. You'll notice that it happens to be #2 on the OEM checklist which is probably why 3dfx is thinking twice.
The choice of Voodoo3 for the name was a rather interesting one considering that it is really only a Banshee2. The on-line community hates this because they know better. They feel that calling the product Voodoo3 is somehow cheating. The bottom line is that this more knowledgeable crowd can get as up in arms as they like because it doesn't matter what they call it. They could call it the SST6000 and it would still be Banshee2 to these guys. Voodoo3, on the other hand, implies a whole new generation of brand new technology to the unwashed masses and it just so happens that it's the unwashed masses that Voodoo3 is shooting for. It's a great marketing move in my opinion.
So given the above analysis of the OEM market, how does Voodoo3 stack up? The chip will be just slightly more expensive than Banshee, be cheaper to produce (.25u process), and twice as powerful. Oddly enough Nvidia seems to be just figuring out that throwing everything but the kitchen sink on their chip may not have been such a great idea for the TNT. The chip is pretty expensive and Nvidia has decided to create a scaled down version called the Vanta to approach the mainstream and business markets. The OEMs they have currently won are only in super high-end machines, or as options. TNT is showing nowhere near the promise that the Riva128 did on its release as the Riva128 hit just the right combination of price/performance. In contrast, Banshee is being included in the mainstream machines at Gateway (read high volume) and seems to be sold out into Q1. I suspect they'd have even more OEM's on-line if they had the volume to support them.
Nvidia is planning a TNT2 and the Vanta but from what I hear both will be too little too late. ATI, the current graphics leader, could pose a problem. They have what looks to be an all-around compelling chip in the Rage128 and for the first time might have competitive technology as well as marketing. Either way, Voodoo3 should put 3dfx easily in the top 3 OEMs for 1999.
3dfx has publicly stated that Voodoo3 will be shipping retail in Q2 of 99, but I suspect that's a bunch of bunk. Why? 3dfx announced Voodoo2 at Comdex last year and had the product actually shipping by February. Voodoo3 is now breaking records at the fab plant (TSMC) by being delivered in 19 days. Add to that the fact that Alex “Sharky” Ross got to play around with this bad boy for several hours without one glitch or crash and the silicon was only 3 days old. This type of reliability is exactly what OEMs are looking for and is probably the result of sticking with known technology. I've heard rumors that we'll be seeing Voodoo3 in stores in January. I don't know if this is true, but I wouldn't be surprised.
Valuation:
I've been doing a lot of soul searching on my stock picking recently, especially with 3dfx. It would seem that from its IPO ($11 per share) the stock price has had only minor appreciation in the long run. If a stock grows its revenue by ten times, such as 3dfx, it isn't unusual to find the stock price has followed suit. I've started asking myself what has gone wrong. Did the market have higher expectation than what 3dfx has actually achieved? I think not. At the IPO the only thing 3dfx had going for them was that they had the Sega deal lined up. After the Sega deal collapsed, nobody was even sure if a retail-only, 3d-only, fabless chipmaker had a chance at making money. I don't think anyone had any idea they would change the face of the 3d roadmap.
Looking at the Internet stocks, I think I've finally figured it out. If 3dfx keeps executing the business plan and displaying strong growth it will be considered one of the steady growth stocks. It is this change in attitude by institutional investors which will eventually transform the stock. I've seen it happen many times before, but I've rarely been able to time it. What I can do is say that this company is ten times more valuable then when I bought way back during the IPO and has that much more of a chance at making the transition.
I'm going to stay away from making specific guesses on actual earnings, but I believe that 3dfx will soundly beat this quarter's earnings. The recent announcement about one million Banshees is nothing new but is comforting. It's essentially one million Banshee in three months, which is exactly what 3dfx expected.
The recent uptick in the stock has me raising an eyebrow. Today we just broke through 16, a level I haven't seen in quite a while and we were unable to attain these heights on even the Comdex hype. There are enough block trades and volume is sufficiently high to lead me to think that this is more than just good earnings. Experienced players in this stock know that betting on a good earnings announcement to bring the stock price up is not necessarily a sure thing. Compaq is due to announce that they have an OEM with 3dfx, but given the lack of enthusiasm over the Gateway announcement I doubt that's it either.
So what is to come?
3dfx announced at Comdex several new directions they will be heading in that may be only now coming to light. Set-tops, hand-helds, and laptops are just three of the additional markets 3dfx will attempt to penetrate. It is well acknowledged in the industry (even if 3dfx won't) that 3dfx will have their true next-generation part for the avid gamer in late 1999 too. I also don't think it out of the question that 3dfx might partner with an independent hardware vendor. It just recently came to light that 3dfx and Diamond were in merger talks but the merger failed because of Diamond's lousy balance sheet. I also heard a nasty rumor that Diamond withheld a large purchase of Voodoo2 orders to put pressure on 3dfx into making the deal. This isn't the first time I've heard of Diamond playing nasty. If you wan't a good idea on how a merger can effect the financials of the company just check some of my previous analysis. Add to all this the fact that 3dfx and AMD have been closeted away working on some secret technology and it certainly looks like 3dfx has plenty of ways to make money come 1999.
…oh yeh, an interview with Scott Sellers gets him on record as saying there's going to be a new product for Christmas. Hey Scott, I wouldn't mind a 3dfx Palm III.
|
