ThanX to Michael Birke to help me out with the translation !

While performing a Hardware-test I asked ELSA and Diamond 10 questions;
for ELSA Ingo Nadler, productmanager Multimedia answered:

RS: ELSA uses SGRAM for the Erazor II in contrast to the other manufacturers using SDRAM. What were the reasons for this?

Ingo Nadler: If the TNT would have been running at the original 150Mhz Memclock specification the difference would have been much higher. SGRAM would have a considerable advantage at that clockspeed. But nVidia had lowered the chip specifications, so that the differences have less impact. The board is nevertheless built for 150Mhz memory clock. And even if SGRAM is more expensive than SDRAM, we stick to it to offer our customers maximum performance!


RS: With which clock frequency will the memory of the Erazor II run? The PC-Professionell claimed 110 Mhz.

Ingo Nadler: Correct and nVidia specifications are: 90Mhz chip clock and 110Mhz cemory clock.


RS: The small difference to Diamond's Viper V550 is interesting. nVidia's reference board is using SDRAM, too. Is it possible that the reference drivers favour SDRAM?

Ingo Nadler: No, the differences within the drivers are marginal. Even a higher clock frequency would only take advantage of fast host systems (PII/400+).


RS: What role does the BIOS play on a modern 3D graphics card?

Ingo Nadler: 90% of the BIOS maintains compatibility with the old VGA standard. It also contains the VESA modes and handles the "old" INT10-calls. About 10% are used for the initialization of the graphics chip. Immediately after booting the Windows driver is in charge!


RS: ELSA has bought Hercules. What will happen with the TNT boards Hercules announced? Will they be available in Germany?

Ingo Nadler: According to latest information available on AdHoc and Business Wire the takeover has been canceled. Therefore I cannot comment on the availability of Hercules boards in Germany.


RS: ELSA decided not to build a Erazor II with Video I/O. What were the reasons.

Ingo Nadler: It might sound profane, but it was simple economics. The market for
cards without Video is just bigger and most of the customers are (unfortunately) not
willing to pay more for video I/O. Now we have a fast 3D-Board that attracts mainly
through its software bundle. The new 2.0 drivers are speaking for themselves.
However, if we have the bandwith we might do another version with video in future.


RS: What function do the 17- and 5-pair connectors on the card have? At the end
of the card there is space for additional parts. What functions do they have.

Ingo Nadler: These connectors are reserved for an optional VIP-based add-on module.
At the moment ELSA does not offer such a module. You could think of a DVD-Decoder or Video-In that could plug into the conector.But since that would only work on the Erazor II and the demand is very low we decided not to offer any solution like this.


RS: There were rumors from time to time of a Video I/O card. Are there any informations about this?

Ingo Nadler: In spite of the rumors there will be no Video Add-On for the Erazor II currently. With new drivers being released soon you can put an Erazor PCI beside it and use its Video-In - no joke, that works! And you even get a larger desktop under Windows98 - just hook up a second monitor...


RS: The latest drivers of the Riva 128 had a significant performance advantage over the first ones. Will this be the same with the TNT?

Ingo Nadler: Certainly will there be an increase in performance, but not as much as with the Riva 128. The main increase will come through faster CPUs because the Riva TNT scales very well, it becomes faster with the system. A much lesser increase will be done through driver optimization. The first Riva128 drivers had a high optimization potential. The TNT drivers instead are already close to their optimum.


RS: What will be the next big step in the developement? Geometry-processors? Texture Comprssion? 3D-Now!?

Ingo Nadler: A bit of theory to start with: There are two limiting factors for the ultimate 3D-experience: The perfomance of the CPU limits the number of polygons possible and the texture space is too limited, (16MB onboard and 32 MB adressed by AGP), to produce photorealistic textures. The first problem can be solved by giving the graphics card the task to calculate the geometry, freeing the CPU -> geometry processor and by supporting faster 3D-commands in the CPU (AMD 3D!Now, or Intel Katmai KNI). The 2nd problem can be solved with more texture memory onboard (e.g. 32MB). A good start is texture compression, which needs at least 16MB, so that the compressed textures don't have to be swapped permanently. And all these features need to be supported by the game software. I do not expect this in a larger scale before christmas 99 ... although the idea makes sense! The next steps will be to increase the video memory to 32MB and AGP4x to allow more polygons to be passed over the bus and let the graphics board use system memory faster. This will solve the problems with large texture amounts. The next step will be multiple multitexturing in one pass (light, bump, shadow, reflection) and a geometry engine in the graphics processor. A look in the crystal ball tells us of hardware based light and shadow processing with multiple light sources. In less than 5 years we will finally experience photorealistic 3D on Home PCs.


RS: Thanks a lot for answering my questions!

RIVA Station 11/1998