About this machine

NVIDIA sold the GeForce 256 as the first GPU, on the strength of moving transform and lighting off the processor and onto the card. The marketing word outlasted the argument it was coined for, but the shift underneath it held: geometry work that had belonged to the CPU now ran beside rasterization.

Configuration

SDR 32 MiB reference card

Component Role Count Clock Capacity
NVIDIA GeForce 256 (NVIDIA GeForce 256) GPU 1 120 MHz Nominal clock
NVIDIA GeForce 256 SDR 32 MiB frame buffer Video memory 1 32 MiB

DDR 32 MiB reference card

Component Role Count Clock Capacity
NVIDIA GeForce 256 (NVIDIA GeForce 256) GPU 1 120 MHz Nominal clock
NVIDIA GeForce 256 DDR 32 MiB frame buffer Video memory 1 32 MiB

Figures for the whole machine

NVIDIA GeForce 256: system figures

provisional A figure marked provisional is recorded but not accepted: its evidence is unsettled. It may rely on one source, an unconfirmed announcement, or a digit read from a damaged scan. The reason appears under the figure. The catalog shows the figure but never uses it in a multiplier. How records are judged.

Metric Value Evidence Basis Method Source
Memory bandwidth memory provisional 4.800 GB/s Rumored Theoretical Theoretical peak From community documentation Tier C table figure for the DDR frame-buffer interface, not a measured application result.
Memory bandwidth memory provisional 2.656 GB/s Rumored Theoretical Theoretical peak From community documentation Tier C table figure for the SDR frame-buffer interface, not a measured application result.
Memory capacity whole-system provisional 32 MiB Rumored Vendor-rated Design capacity From community documentation Tier C table figure for the DDR reference frame buffer, not host memory.
Memory capacity whole-system provisional 32 MiB Rumored Vendor-rated Design capacity From community documentation Tier C table figure for the SDR reference frame buffer, not host memory.
Memory bandwidth whole-system provisional 4.800 GB/s Rumored Theoretical Theoretical peak From community documentation The DDR configuration has one fitted frame buffer, so its whole-system figure repeats the cited interface bandwidth.
Memory bandwidth whole-system provisional 2.656 GB/s Rumored Theoretical Theoretical peak From community documentation The SDR configuration has one fitted frame buffer, so its whole-system figure repeats the cited interface bandwidth.

Figures for its components

NVIDIA GeForce 256 (NVIDIA GeForce 256)

provisional A figure marked provisional is recorded but not accepted: its evidence is unsettled. It may rely on one source, an unconfirmed announcement, or a digit read from a damaged scan. The reason appears under the figure. The catalog shows the figure but never uses it in a multiplier. How records are judged.

Metric Value Evidence Basis Method Source
Clock frequency gpu provisional 120 MHz Confirmed Vendor-rated Nominal clock Published independently Two independent technical reviews agree on the NV10 core-clock figure.
Peak FP32 rate gpu provisional 960 MFLOP/s Rumored Theoretical Theoretical peak From community documentation Tier C table figure for the NV10's theoretical fixed-function FP32 rate.
Pixel fill rate gpu provisional 480 Mpixel/s Rumored Theoretical Theoretical peak From community documentation Tier C table figure for the NV10's theoretical pixel fill rate.
Texel fill rate gpu provisional 480 Mtexel/s Rumored Theoretical Theoretical peak From community documentation Tier C table figure for the NV10's theoretical texel fill rate.
1 documented absence
NVIDIA GeForce 256 (NVIDIA GeForce 256): documented absences
Metric Value Evidence Basis Method Source
Rated power consumption gpu provisional unknown No checked source, including the VisionTek product brief, states a board-power rating for either reference configuration. Not stated Vendor-rated Rated power From community documentation No checked source, including the VisionTek product brief, states a board-power rating for either reference configuration.

Editorial notes

NVIDIA released SDR and DDR reference configurations. Later partner cards with more frame-buffer memory are outside this record.

Sources

Back to top