Tom’s Look Back at Comdex 1998 – THG.RU

Introduction

Comdex is now over for more than a week and looking back at it doesn’t leave many exciting memories. It certainly wasn’t the big show it used to be and there wasn’t much outstanding, particularly in terms of PC hardware. Companies like IBM, Intel, Compaq or Dell could not even be bothered to have a booth at Comdex this year and so you could hear the question ‘Is Comdex going down?’ all over the place. The other comment you could hear very often was ‘well, the hardware has obviously outran the software’. I don’t quite agree with that, but it seems a matter of fact that there have to be new ways of bringing computer to the people. This also requires new software, and I could imagine that this challenge could be a great chance for new software companies winning ground versus Microsoft.

3Dfx

The week of Comdex 1998 started with 3Dfx press conference in the ‘Dive’, a bar/café in submarine look. The presentation could have been that bit better if the projector would have worked properly, but unfortunately it took one hour until the projector could at least display a blurry and almost illegible presentation. I personally couldn’t care less about presentation slides, so I could concentrate even better on what Scott Sellers, Senior Vice President 3Dfx, had to say.

Ex-quarterback Scott is telling us why 3Dfx will soon win the 2D/3D accelerator Superbowl

The upcoming 3Dfx Voodoo3 is actually nothing much different from what we thought would be called ‘Banshee 2’. 3Dfx decided against this name since Banshee doesn’t really have the very best reputation and because the new chip is meant to win the crown of the 2D/3D accelerator chips, so that it should well deserve the promising name ‘Voodoo3’.

The features of Voodoo3 read like this:

• Dual 32-bit texture rendering architecture
  32 bit internally, but 16 bit externally, then interpolated by the special RAMDAC, leading to a color depth of around 22 bit
• True multi-texturing: 2 textures per pixel per clock
  means either one pixel per clock that’s rendered using two texture maps or 2 pixels per clock using normal bi-linear filtering of one texture
• Full hardware setup of triangle parameters
  nothing special, that’s common these days
• Support for multi-triangle strips and fans
  also common now
• Single Pass, Single-cycle bump mapping
  using multi (dual) texturing, no special hardware routine, this means one bump mapped pixel can be rendered per clock
• Single Pass, Single-cycle tri-linnear mip-mapping
  same as above, means one tri-linearly filtered pixel can be rendered per clock
• Alpha Blending on source and destination pixels
• Sub-pixel and sub texel correction to 0.4×0.4 resolution
• Per-pixel atmospheric fog with programmable fog zones
• Full-scene polygon-based edge anti-aliasing
• Floating point Z buffer (W buffer)
  this Z-buffer is only 16 bit deep though
• True per-pixel, LOD MIP mapping with biasing and clamping
• Texture composting for multi-texture special effects
• 8-tap anisotropic filtering
• Support for 14 texture map formats
• 8-bit palletized textures with full bilinear filtering
• Texture compression through narrow-channel YAB format
  this is an interesting feature, proving that texture compression will be the future for all 3D chip makers
• Texel Fill Rates 366 Mtexels/sec (Voodoo3 3000), 250 Mtexels/sec (Voodoo3 2000)
  Voodoo3 3000 will thus have a tiny bit higher fill rate than Voodoo2 SLI, but it can use it up to 1600×1200 and above
• Triangle Rate 7 million triangles/sec (Voodoo3 3000), 4 million triangles/sec (Voodoo3 2000)
  this rate is fine, but so far there are hardly any games out that would ask for only 1 million triangles/sec. It only states that the 8 million triangles/sec of TNT-2 won’t make this chip really faster. Fill rate is more important right now.
• RAMDAC 350 MHz(Voodoo3 3000), 300 MHz RAMDAC (Voodoo3 2000), allowing resolutions of up to 2048×1536 at 75 Hz refresh
  this feature really rocks the boat. There is no chip announced that would have such a fast RAMDAC
• Core Graphics Clock 183 MHz (Voodoo3 3000), 125 MHz (Voodoo3 2000)
  using .25 micron technology
• 32 MB onboard memory support (was only 16 MB at first)
  ATI will make 32 MB local memory a common thing with Rage 128 soon, thus 3Dfx changed the specs and will support 32 MB also.
• DVD acceleration
  this doesn’t mean hardware MPEG2 decoding! It seems a matter of fact that Voodoo3 will also not have ‘motion compensation’, which would be helpful for MPEG2 with weaker CPUs
• Digital Video Output (NTSC, PAL, SECAM)
  pretty common these days, but nice
• Digital flat panel screen support (LCDfx)
  This will soon be a very important feature, but there is no common interface for didgital output for flat panels, so we’ll have to see how important LCDfx will be
• Banshee’s 2D engine
  has proven to be very fast, certainly fast enough for everyone
• AGP 2x support
  well, this feature is not quite clear to me, since 3Dfx still votes against AGP texturing. It could be that Voodoo3 will still not be able to do AGP texturing, which could be a serious problem. At the time that Voodoo3 ships there will AGP 4x come out. That will only be supported with a future Voodoo3 chip.

More on 3Dfx

This altogether sounds pretty cool, but there are also a few missing features. Voodoo3 will be the only new chip without real 32 bit rendering capabilities and without a 24 or 32 bit deep Z-buffer support. 3Dfx says that those two are frame rate killers and that a frame rate of 60 fps should become the standard for 3D gaming. I still don’t quite agree with this, although 3Dfx showed a special demo, internally called the ‘Tom Demo’, which is meant to prove that 60 fps is a lot better than 30 fps. At first I was impressed too, but after talking to several specialists, I again want to express my doubts that 60 fps is what it takes for high class 3D gaming. It really depends on the scene that’s shown, otherwise you could never watch MPEG2 encoded movies (as DVD). MPEG2 takes so much information out of a picture and displays it at 25 or 30 fps, but we still think that the quality is excellent. This is due to several ways of ‘cheating’ the human eye. In dark scenes with few contrasts, as e.g. in Quake or Quake 2, the difference between 30 and 60 fps is becoming very little, so boasting only the frame rate is a questionable approach. When you raise the image quality or even use some eye-cheating features also, instead of increasing the frame rate to much over 30 fps, you can get the very same result. Anyway, 3Dfx thinks Voodoo3 can do without 32 bit coloring in 3D and 60 fps at 1280×1024 are supposed to equalize that. You 3D gamers out there will have to decide if they are right.

Another missing feature is the ‘motion compensation’ for MPEG2 decoding. This feature is really only needed in systems with CPUs of less power than a Pentium II 266. 3Dfx told us that a Celeron 333 is used by 80% when displaying MPEG2 (DVD) at 30 fps. Slower CPUs will have problems with software MPEG2 decoding using Voodoo3.

The benchmarks we’ve seen with a Voodoo3 chip running at only 160 MHz are pretty promising, although not earth shattering. 32 fps in Quake2’s timedemo demo2 at 1600×1200 resolution and 16 bit color depth are certainly impressive, because it’s almost double of what nVIDIA’s RIVA TNT is currently able to offer. 54 fps in massive1 at 1152×864 are also about double of what TNT is scoring right now. At the same time the quality looked good, however certainly not better than TNT or G200, but better than Voodoo2 (thank God!). At 183 MHz Voodoo3 will score about 15% better and with mature drivers and a few optimizations here and there Voodoo3 3000 will most likely be the fastest 2D/3D chip once it comes out. Voodoo3 2000 will probably perform pretty much as well as TNT-2, but Voodoo3 2000 is meant for the OEM business and supposed to be included on the motherboard. I personally don’t think much of motherboards with included 2D/3D chips, since it kills the upgrade path, but it may become an interesting solution somewhere in between high end systems and systems with Intel’s Whitney chipset.

My personal view of Voodoo3 is that it’s certainly the most promising 2D/3D chip that’s announced for Q2 next year, but 3Dfx will have to come up with something faster pretty soon. NVIDIA’s NV10 is due for next summer and it sounds as if it will be way ahead of Voodoo3. The other chip makers are also working on chips with 4 texture pipelines and highest fill rates, but those guys won’t dare to release their chips without 32 bit rendering abilities and 32 bit Z-buffers.

NVIDIA

The visit at NVIDIA’s booth was a bit of a disappointment to me. I expected to see TNT-2 and some real high performance 3D, instead of this I could look at a 3D rendered bicycle by the name of ‘Voodoo’, how tasteful! NVIDIA showed their new business chip ‘VANTA’. Vanta is supposed to be targeted to the enterprise market. ‘There is no killer application for 3D in corporate America’ are the words of NVIDIA’s CEO Jen-Hsun Huang, saying that corporate America doesn’t require a high-end 3D chip, but a chip that is able to display 3D well enough. Vanta is supposed to be the right chip for those needs, being a stripped down version of TNT with e.g. a 64 bit instead of 128 bit memory interface and overall lower 3D performance than TNT. At the same time Vanta offers the same high 2D performance of TNT.

TNT-2 was announced later on, but it’s nothing but the current RIVA TNT, this time running at 125 MHz in .25 micron.

All in all there’s not too much to say about this. I hope that NVIDIA will have to show something exciting on the next computer show. I keep my fingers crossed, NV10 with included geometry engine and many other cool features will rock the boat next year, but don’t expect NVIDIA to tell you much about it.

AMD

The one company I looked forward visiting most was certainly AMD. I knew they would have K7 and Sharptooth running there, the chips that scare the pants off Intel. When I entered the booth I had the nerve of interrupting Dana Krelle, AMD’s vice president marketing, who was busy playing Viper Racing. Well, I guess Dana has forgiven me that I screwed up his race, as far as I know he was only training anyway.

Dana and the K7 500 in AMD’s first Slot A board with AMD chipset

AMD had two demo rooms. In the first room I could inspect the first Slot A motherboard with AMD’s own chipset and K7 itself. The K7 was equipped with 512 kB of L2 cache, running at 1/3 of the CPU core speed, thus 500 MHz / 3 = 166 Mhz.

This is it, the mysterious K7, the possible Intel killer chip.

The front of the K7-PCB, after removing the cartridge.

And the back of the K7-PCB, you can see the huge amounts of pins that K7 has compared with Pentium II.

There was of course a working K7-system in that room too, which ran a mixture of Winstone98 and 3D games at really high speed. I could not see any benchmark numbers, but the benchmark part ran faster than anything I’ve seen yet, definitely faster than a Pentium II 504.

The second room showed two systems next to each other. The left one was a system equipped with AMD’s Sharptooth, the upcoming K6-3 (if the name won’t be changed), at 400 MHz. Sharptooth is nothing but K6-2’s new CXT core with an on die L2 cache of 256 kB running at CPU core clock. The system to the right was a Pentium II 450 MHz system, equipped with the same peripherals. Again a mixture of business and high-end Winstone98 was ran, starting both systems at the same time. The Sharptooth finished those benchmarks significantly earlier than the Pentium II 450, proving that Sharptooth or K6-3 will be extremely fast in business applications, even faster than Pentium II at 50 MHz higher clock rate.

All in all I wasn’t really too surprised, since the demonstrations only showed that I was right with my estimate published around Micropocessor Forum time 5 weeks before Comdex. AMD has some really strong products lined up for the future, as you can see in this comparison.

AMD and Intel CPU Roadmap Comparision

When you read this evaluation you will probably understand why I’d recommend selling Intel stocks and get some AMD stocks before they skyrocket. AMD is in better shape than ever to take on Intel. Intel has gotten a bit lazy these days, new technology is taking too long, KNI alone won’t save Katmai, neither will the largest marketing campaign in the history of computers. I also doubt that Celeron and PGA370 will be attractive enough vs. K6-2 and K6-3 on the Socket 7 platform. I understand why Intel only talks of AMD’s possible inability to ship product, there is not much more what Intel can hope for. But this is not the right answer for a company with the slogan ‘only the paranoid will survive’. I advise Intel to become paranoid right now, otherwise they may not survive, at least not as the super successful company they are today.

Intel

The news I have to tell is not really from Comdex, but it’s most important to many of you. The future Intel chips will put a final end to overclocking. New Intel processors won’t only come with the already common multiplier-locking, but making an Intel CPU faster by increasing the bus speed will also not be possible anymore. Some new circuits on the chips will make sure that the system remains dead if you raise the bus speed from e.g. 66 to 75 MHz or from 100 to 112 MHz. It is rumored that AMD will do the same thing.

Motherboard Manufacturers

There isn’t much news from the motherboard front. Most manufacturers sell BX boards in all kinds of flavors as well as Super7 platforms, some offer Slot2 solutions for the Xeon CPU as well. The only new thing that’s coming up early next year is PGA370, the new socket for the socketed version of Celeron. As you already know, the PGA370 will offer Intel’s Celeron at a more attractive price point than with Slot1 as it’s sold now. PGA370 looks almost the same as Socket7, but it uses the Intel P6 interface instead. Intel will ship the socketed Celeron at up to 366 MHz in January 1998, this chip will include the on-die L2 cache known from Celeron 300A and 333. The motherboards for PGA370 will either use the well known Intel LX chipset, or the new ZX66 chipset, a stripped down version of BX, officially only supposed to run at 66 MHz FSB. Alternatively there will be non-Intel chipsets available for PGA370.

PGA370 looks almost identical to Socket7, except for one corner.

Cyrix WebPAD

For me the product that should have won the price for the most innovation and user friendliness is the Cyrix WebPAD. Following Brian Hala’s speech at Microprocessor Forum, Cyrix is following new ideas to bring computers to people who can’t be bothered to buy a clumsy, super user unfriendly Windows PC. WebPAD is a reference design for a product that will make Internet browsing real fun.

Instead of sitting at a desk in front of a more or less clumsy monitor, instead of using a mouse, which is still a problem for computer beginners, instead of listening to the monotony of your PC fans and your hard drives spinning, sitting in an uncomfortable office, … you can now make Internet browsing as enjoyable as reading your favorite car magazine. WebPAD can be taken anywhere inside the house, it’s small and it connects wirelessly to the Internet . You use it with a highly intuitive little pen.

I can see that many of you power users are already falling asleep reading this, but just try and make yourself free from the old fashioned and highly inconvenient PC platform idea. I am a power user myself. However, I’m not playing 3D games all of the time, neither do I always create spread sheets or presentations. Sometimes I’m just browsing the web for information, without any stress or hurry on my mind. With WebPAD I can now do this on my super comfortable leather couch, watching TV at the same time, I can do this even in the bath tab or at this well known place, where so many people spend hours for to many other people’s unknown reasons. My mother who is certainly no PC-crack could surf the web without the slightest problems, kids can use it, without having to know how to boot up a PC and the super easy to use and easy to crash Microsoft operating systems.

WebPAD is much easier to use than WebTV and it will be available at an attractive price too. It is the best way of bringing the Internet to pretty much every citizen of a developed country, particularly all the people who disapprove or simply don’t like computers.

You can see that I am really in love with the idea of WebPAD and I guess Cyrix will be surprised about that too. I haven’t had many positive words for Cyrix products in the past, but something like WebPAD is what it takes to take the development of computers to the next step. We don’t need even bigger or even hotter computer systems, as a matter of fact we don’t even need that one super PC system at all anymore. What I want is an easy to use but versatile 3D gaming computer that fits in seamlessly into the recreation area of my house. It should not be in my office where I don’t want to be after hours anymore. I want an easy to use, small and mobile thing that lets me surf the Internet where ever I am. I want the ability to dictate my emails to this thing as well, so that I won’t need a keyboard anymore. And then there may also be my office system, which is where I write or dictate my articles and where I prepare my spread sheets and do other office work. This system doesn’t need to be a great 3D gaming platform, since I prefer playing in my living room or elsewhere. This is how I see the future of computing, the PC is meant to die sooner or later, I don’t give it more than 2 years anymore.

WebPAD is a step into the right direction. I’ll make sure to get several of them as soon as there’s a manufacturer that picked up Cyrix’ reference design. I will use one, the rest will be given to my family, my parents and my brother. None of them are Internet cracks, but I bet WebPAD will change that.

Palim Palim … 🙂

What else happened at Comdex? Well, 3Dfx showed how the perfect party can be!!!

This party had all that it takes, a cool place, no waiting for drinks at all, good music, super atmosphere, especially when Cool&The Gang was playing .. and the right people as well as the right amount of people! This party was all that the Eidos party at E3 could have been. Yes, I take it back, Americans do know how to throw a party as well!

On Tuesday Ben from KillerApp, Pete from Gamers Extreme, famous Dennis ‘Thresh’ Fong, Alex ‘Sharky’ Ross from ‘Sharky Extreme’ and me went to a special event for the Taiwanese computer industry in the Mirage. Each of us had to give some kind of speech, Dennis, Alex and I were told about 20 minutes before it took place. After the speeches I did a short Q&A-session as well. It was a lot of fun doing this, although the auditorium wasn’t very big and loud music came in from a room next door. After the Q&A I had a quick interview with a Taiwanese TV station and then we went over to dinner. I’d just like to show a few photos from there. Unfortunately I haven’t got pictures of Ben and Dennis (Thresh). Hope you’ll enjoy the photos of Alex and me.

Alex – concentrating … ?

Blah blah -. 🙂 Tom in a world of his own

Later on we all went to my favorite place in Las Vegas, the Voodoo Lounge and Michael was almost too greedy to pay a nice limo for us all. The guys who were at the Eidos party know that this time it wasn’t me to sort out the limo.

Wednesday was kinda relaxed, at 1 PM I did the ZDTV interview with Leo Laporte. I received a lot of mails where people complained that Leo cut me off several times. You bet that I don’t like this either, but in a TV show you’ve got to stick to the time frame, so Leo did what he had to do, and I’d say he did it pretty nice.

In the evening we went to Mystere, Cirque de Soleil in the Mirage, which was kinda cool, although I heard it’s not quite as cool as O, Cirque de Soleil in the Bellaggio. Afterwards it was Voodoo Lounge time again, but on that Wednesday night the life music kinda sucked a bit.

I am not going to write about the flight back to San Francisco on Thursday with my favorite nightmare airline United Airlines, but I can say that I really like Las Vegas McCarran Intl. Airport, you can even get quite good food there.

Some of you may know that I dislocated my left shoulder even twice in the week before Comdex, giving me a completely new definition of the word pain. I first thought that I wouldn’t be able to fly over to Las Vegas at all, since my left arm was pretty unusable. Thanks to my brother Rainer I could go. He canceled all his appointments for that week of Comdex and flew over together with me, helping and body guarding me for the whole Comdex time. Thanks Rainer, you know I love you man!