In 2000, NVIDIA released the GeForce 2 series of dedicated graphics cards. In 2001, the GeForce 3 series was released. In this article, I'll discuss why these graphics cards are excellent for a hybrid DOS/Windows machine, and go over some tips when buying these cards.

Advances in Graphics Processing

Let's go back to the year 2000. This is the year we would see the debut of Intel's Pentium 4 and AMD's Thunderbird Athlon processors.

GeForce 2 (2000)

NVidia's GeForce 256 was already making waves on the scene, having been released the previous year in October. They already had a capable graphics processor that boasted hardware T&L, which was a steadily emerging demand from game developers and hardware enthusiasts alike. Their graphics processors were also very capable in DOS. Prior to that, the TNT2 and Riva TNT were just as capable.

Other companies were racing just as fast to dominate their preferred market segments. S3 released the Savage 2000 the previous November, in 1999. Nvidia released the GeForce 2 GTS in the middle of 2000, ATi released the first Radeon, 3dfx released the Voodoo 5, Matrox released the G450. All the big names seemed to have found ways to drastically increase the performance of their graphics processors this year.

The GeForce 2 would go on to win the hearts of many. We would eventually receive four models based on NV11: GeForce 2 GTS, GeForce 2 Pro, GeForce 2 Ti and GeForce 2 Ultra.

Despite the advancements of NVidia's competition, it was too difficult for many to keep up with NVidia and ATi. Even though the Voodoo 5 was met with critical acclaim, 3dfx went bankrupt and was sold to NVidia. After the disaster that was the Savage 2000, S3's graphics division was sold to Via. It's just as well that S3, a very popular company for OEMs to get their low cost graphics solutions from, sold out: NVidia would release the GeForce 2 MX later in 2000 to scoop up the remainder of the budget segment. So much came and went in the blink of an eye.

NVidia would follow up the next year with the Quadro2 line for workstations. The Quadro2 Pro remains one of the most performant cards of this generation. Its overclocking potential was demonstrated by PhilsComputerLab on YouTube.

GeForce 3 (2001)

The next big item on NVidia's list was to ability to utilize programmable shaders. The GeForce 3, based on the NV20 GPU, was released in 2001 with support for Shader Model 1.1 and DirectX 8. These capabilities were very important to rapidly advance the fidelity of realtime computer graphics. Doom 3 is one such game that makes heavy use of this technology.

The DOS speed and compatibility persisted into this generation of NVidia GPUs as well. We eventually received three versions of the GeForce 3: the standard model (simply named GeForce 3), the budget GeForce 3 Ti 200, and the enthusiast GeForce 3 Ti 500.

These cards also tend to have a DVI port in addition to the standard VGA port. DVI was considered an emerging technology at this point and wasn't found on any NVidia cards before it, nor many other manufacturers' products for that matter.

For workstations, the Quadro DCC would be released.

Onward (2002+)

NVidia would go on to release the GeForce 4 line, based on the NV25 GPU. However, in implementing a new memory architecture, DOS performance greatly suffered. It's possible that this was necessary in order to boost its other more modern capabilities, which would be understandable. After all, there wasn't much DOS software released in 2002.

DOS Performance

There are only a few chips from NVidia that should be considered for use in DOS machines for the best performance:

• NV3 (Riva128)
• NV4 (Riva TNT)
• NV5 (Rive TNT2)
• NV10 (GeForce 256)
• NV11 (GeForce 2 MX)
• NV15 (GeForce 2)
• NV16 (GeForce 2 Ultra)
• NV20 (GeForce 3)

The NV1 is the Diamond Edge and has rather unfortunate DOS compatibility. The NV17 and NV18, despite coming in between the GeForce 2 Ultra and the GeForce 3 when considering only NV numbering, have roots in the GeForce 4 as far as memory access goes, and suffer from the same greatly reduced performance as the GeForce 4 and later cards do.

P-Model List

P-model numbers have been on Nvidia PCBs for a while. This practice continued well into the thousands, but starts low. Sometimes looking for these part numbers nets you more results because people don't often know what these cards are.

The model and PCB codes refer to the PCB layout itself. The GPU that is mounted to the board may differ depending on binning. That's why multiple cards can be listed. A Quadro DCC will have the same PCB code as a GeForce 3, and the amount of RAM mounted to the card can also be different.

OEM Model List

There's a number of models out there that may or may not be labeled as 'obvious' NVidia GeForce cards. Sometimes searching for the OEM model/part number yields results on auction sites that can't be found otherwise. Here's a list of cards I've encountered with their OEM model/part numbers.

More to Come

There's more information to be added here! However, I've opened up this post so that others who are looking to build a powerful hybrid DOS/WIN98 machine will have more resources to work with.

Drivers and benchmarks will follow at a later date. Thanks for your patience!