NVIDIA GeForce FX 5900 XT

The NVIDIA GeForce FX 5900 XT is an end-of-life graphics card built on the NV35 chip, employing the Rankine architecture. Fabricated by TSMC on a 130 nm process, the die contains 135 million transistors across a 207 mm² area, resulting in a transistor density of 652.2K per mm². The card ships with 128 MB of DDR memory on a 256-bit bus, delivering 22.40 GB/s of bandwidth. The memory operates at 350 MHz with an effective rate of 700 Mbps. The GPU integrates 8 texture mapping units and 4 render output units, achieving a pixel rate of 1.600 GPixel/s and a texture rate of 3.200 GTexel/s. With a TDP of just 35 W, it requires a single Molex power connector and a suggested 200 W power supply. It connects via AGP 8x and provides one DVI, one VGA, and one S-Video output. Released on 2003-05-11, this card sits at the 50th percentile of all GPUs in the database, with an average benchmark score of 0, though no specific benchmark entries are listed.

Who Should Consider It

The FX 5900 XT fits a narrow niche defined by its 128 MB frame buffer. With only 128 MB of DDR memory, high-resolution textures will quickly exhaust the available space, forcing the card to rely on slower system memory over the AGP 8x bus. The 256-bit bus width is generous, but the 22.40 GB/s bandwidth remains the ceiling for data transfer. For users targeting moderate resolutions and reduced texture detail, the card can function adequately. The 1.600 GPixel/s pixel rate and 3.200 GTexel/s texture rate indicate that basic 3D geometry and texture mapping are handled without severe bottlenecks, but scenes with heavy overdraw or complex shaders will push the 4 ROPs and 8 TMUs to their limits. The 50th percentile ranking against all GPUs in the database suggests a middle-of-the-road standing, though the absence of benchmark entries means this placement is aggregate, not test-derived. Given the 35 W TDP, the card is also a candidate for systems with limited power budgets, provided the motherboard has an AGP 8x slot. The single-slot design and low power draw make installation straightforward, and the 200 W PSU recommendation is easy to meet. However, the lack of dedicated ray tracing or tensor cores means any such workload is unsupported, so users requiring those features should look elsewhere. This card is not suited for modern high-detail gaming or professional 3D rendering, but for legacy software and applications that respect the 128 MB memory limit, it remains a functional option.

Ray Tracing and Feature Set

The FX 5900 XT does not include any dedicated ray tracing cores or tensor cores; the FACT PACK lists neither, which means hardware-accelerated ray tracing and AI-based features are entirely absent. The card's API support is limited to DirectX 9.0a, OpenGL 1.5 (full), and OpenGL 2.0 (partial). There is no Vulkan support. This places the card firmly in the early DirectX 9 era, where shader model support is basic compared to later iterations. The partial OpenGL 2.0 support means some advanced OpenGL features may not function correctly, though the full 1.5 support covers the majority of legacy titles. For users relying on DirectX 9.0a applications, the card is compatible, but it will not run software that requires later DirectX versions, as only 9.0a is listed. The lack of tensor cores also means no DLSS or similar upscaling, and no AI denoising. In summary, the feature set is strictly functional for its time, with no forward-looking hardware acceleration. Users who prioritize modern rendering features will find the card lacking, but for pure rasterization at the listed API levels, it provides the expected baseline.

Power and Cooling

Power consumption is a notable strength of the FX 5900 XT. The TDP is rated at 35 W, which is remarkably low for a graphics card of this generation. This low power draw allows a single-slot cooling solution, keeping the card compact and easy to install in tight cases. The power connector requirement is a single Molex, a legacy connector commonly found on power supplies of the era. The suggested power supply rating is 200 W, which is modest and should be readily available in most systems. Given the 35 W TDP, the card does not generate significant heat, so a basic single-slot heatsink and fan are sufficient. The AGP 8x interface provides power through the slot, but the additional Molex connector is still required to meet the card's needs. For users building or upgrading a system with a 200 W PSU, this card is a safe choice. The low TDP also means less strain on the system's thermal envelope, which is beneficial in small form-factor cases. However, users should ensure their power supply has a free Molex connector, as the card cannot operate without it. The single-slot design is a practical advantage over multi-slot cards, allowing for additional expansion cards in adjacent slots.

FAQ

Q: What is the memory configuration of the NVIDIA GeForce FX 5900 XT?

A: The card features 128 MB of DDR memory on a 256-bit bus, providing 22.40 GB/s of bandwidth. The memory clock is 350 MHz, with an effective data rate of 700 Mbps.

Q: What is the power consumption and PSU requirement?

A: The TDP is rated at 35 W. The card requires a single Molex power connector, and the suggested power supply is 200 W.

Q: Does the FX 5900 XT support ray tracing?

A: No. The FACT PACK lists no ray tracing cores or tensor cores, so hardware-accelerated ray tracing is not available.

Q: What graphics APIs are supported?

A: The card supports DirectX 9.0a, OpenGL 1.5 (full), and OpenGL 2.0 (partial). It does not support Vulkan.

Q: What is the bus interface and what display outputs are available?

A: The card uses an AGP 8x interface. Display outputs include one DVI, one VGA, and one S-Video connector.

Q: What is the manufacturing process and die size?

A: The GPU is fabricated on a 130 nm process by TSMC, with 135 million transistors on a 207 mm² die, resulting in a transistor density of 652.2K per mm².

How It Compares

The FACT PACK for the FX 5900 XT does not include any nearest rivals, meaning there are no benchmark scores or percentile deltas to reference for direct comparison. Consequently, a numeric comparison against competing cards is impossible within the provided data. The card is positioned in the product stack between the GeForce 4 Ti, which is listed as its predecessor, and the GeForce 6 AGP, its successor. The GeForce 4 Ti represents an older generation, likely with lower performance and fewer features, though no specific numbers are given. The GeForce 6 AGP is the next generation, which typically brings architectural improvements and increased performance, but again, no specs are provided. Without benchmark data, users must rely on the card's own specifications: the 128 MB memory, 256-bit bus, and 22.40 GB/s bandwidth, along with the 1.600 GPixel/s pixel rate and 3.200 GTexel/s texture rate. The 50th percentile ranking against all GPUs in the database indicates it is a median performer, but the lack of individual benchmark scores means this is a general placement, not a specific comparison. For users upgrading from a GeForce 4 Ti, the FX 5900 XT offers the Rankine architecture and DirectX 9.0a support, which is a step forward. For those considering a GeForce 6 AGP, the FX 5900 XT will likely be slower, but the exact margin is unknown from the available data.

Memory Subsystem

The memory subsystem of the FX 5900 XT is defined by 128 MB of DDR memory, a 256-bit bus width, and 22.40 GB/s of bandwidth. The memory clock runs at 350 MHz, with an effective rate of 700 Mbps. The 256-bit bus is a wide interface, which helps mitigate the relatively low clock speed, but the overall bandwidth of 22.40 GB/s is still a limiting factor for high-resolution workloads. At high resolutions, the 128 MB frame buffer becomes a critical constraint. Textures and frame buffers must be stored within this 128 MB limit, and exceeding it forces the card to rely on system memory over the AGP 8x bus, which is significantly slower. The 22.40 GB/s bandwidth is sufficient for moderate resolutions and lower texture detail, but it will bottleneck when handling large textures or multi-sampled anti-aliasing, which consumes additional memory bandwidth. The 4 ROPs also limit the fill rate, with a pixel rate of 1.600 GPixel/s, meaning that high-resolution output will be constrained by both memory capacity and bandwidth. For users planning to run at high resolutions, the 128 MB capacity is the first barrier, followed by the bandwidth. The 256-bit bus does provide a solid foundation, but the DDR memory type and 350 MHz clock are not high-speed by later standards. In practice, the memory subsystem is adequate for its intended era, but it will not scale to modern high-resolution textures or heavy post-processing effects.