Hubbry Logo
GeForce FX seriesGeForce FX seriesMain
Open search
GeForce FX series
Community hub
GeForce FX series
logo
8 pages, 0 posts
0 subscribers
Be the first to start a discussion here.
Be the first to start a discussion here.
GeForce FX series
GeForce FX series
GeForce FX series
View on Wikipedia
from Wikipedia

GeForce FX series
GeForce FX logo
Top: Logo of the series
Bottom: A GeForce FX 5950 Ultra released in 2003, the series' flagship model
Release dateJanuary 27, 2003; 23 years ago (January 27, 2003)
CodenameNV30, NV31, NV34, NV35, NV36, NV38
ArchitectureRankine
ModelsGeForce FX series
  • GeForce FX-VE series
  • GeForce FX-LE series
  • GeForce FX-ZT series
  • GeForce FX-XT series
  • GeForce FX-Ultra series
  • GeForce PCX series
Cards
Entry-levelFX 5100
FX 5200
FX 5200 LE
FX 5300
FX 5500
Mid-rangeFX 5600
FX 5700
PCX 5750
High-endFX 5800
FX 5900
PCX 5950
Enthusiast5800 Ultra, 5900 Ultra, 5950 Ultra
API support
Direct3DDirect3D 9.0a
Shader Model 2.0a
OpenGLOpenGL 2.1
History
PredecessorGeForce 4 series
SuccessorGeForce 6 series
Support status
Unsupported

The GeForce FX or "GeForce 5" series (codenamed NV30) is a line of graphics processing units from the manufacturer Nvidia.

Overview

[edit]

Nvidia's GeForce FX series is the fifth generation of the GeForce line. With GeForce 3, the company introduced programmable shader functionality into their 3D architecture, in line with the release of Microsoft's DirectX 8.0. The GeForce 4 Ti was an enhancement of the GeForce 3 technology. With real-time 3D graphics technology continually advancing, the release of DirectX 9.0 brought further refinement of programmable pipeline technology with the arrival of Shader Model 2.0. The GeForce FX series is Nvidia's first generation Direct3D 9-compliant hardware.

The series was manufactured on TSMC's 130 nm fabrication process.[1] It is compliant with Shader Model 2.0/2.0A, allowing more flexibility in complex shader/fragment programs and much higher arithmetic precision. It supports a number of new memory technologies, including DDR2, GDDR2 and GDDR3 and saw Nvidia's first implementation of a memory data bus wider than 128 bits.[2] The anisotropic filtering implementation has potentially higher quality than previous Nvidia designs.[1] Anti-aliasing methods have been enhanced and additional modes are available compared to GeForce 4.[1] Memory bandwidth and fill-rate optimization mechanisms have been improved.[1] Some members of the series offer double fill-rate in z-buffer/stencil-only passes.[2]

The series also brought improvements to the company's video processing hardware, in the form of the Video Processing Engine (VPE), which was first deployed in the GeForce 4 MX.[3] The primary addition, compared to previous Nvidia GPUs, was per-pixel video-deinterlacing.[3]

The initial version of the GeForce FX (the 5800) was one of the first cards to come equipped with a large dual-slot cooler. Called "Flow FX", the cooler was very large in comparison to ATI's small, single-slot cooler on the 9700 series.[4] It was jokingly referred to as the "Dustbuster", due to a high level of fan noise.[5]

The advertising campaign for the GeForce FX featured the Dawn, which was the work of several veterans from the computer animation Final Fantasy: The Spirits Within.[6] Nvidia touted it as "The Dawn of Cinematic Computing".[7]

Nvidia debuted a new campaign to motivate developers to optimize their titles for Nvidia hardware at the Game Developers Conference (GDC) in 2002. In exchange for prominently displaying the Nvidia logo on the outside of the game packaging, the company offered free access to a state-of-the-art test lab in Eastern Europe, that tested against 500 different PC configurations for compatibility. Developers also had extensive access to Nvidia engineers, who helped produce code optimized for the company's products.[8]

Hardware based on the NV30 project didn't launch until near the end of 2002, several months after ATI had released their competing DirectX 9 architecture.[9]

Overall performance

[edit]
GeForce FX 5200

GeForce FX is an architecture designed with DirectX 7, 8 and 9 software in mind. Its performance for DirectX 7 and 8 was generally equal to ATI's competing products with the mainstream versions of the chips, and somewhat faster in the case of the 5900 and 5950 models, but it is much less competitive across the entire range for software that primarily uses DirectX 9 features.[10]

Its weak performance in processing Shader Model 2 programs is caused by several factors. The NV3x design has less overall parallelism and calculation throughput than its competitors.[11] It is more difficult, compared to GeForce 6 and ATI Radeon R300 series, to achieve high efficiency with the architecture due to architectural weaknesses and a resulting heavy reliance on optimized pixel shader code. While the architecture was compliant overall with the DirectX 9 specification, it was optimized for performance with 16-bit shader code, which is less than the 24-bit minimum that the standard requires. When 32-bit shader code is used, the architecture's performance is severely hampered.[11] Proper instruction ordering and instruction composition of shader code is critical for making most of the available computational resources.[11]

Hardware refreshes and diversification

[edit]
See also: Comparison of Nvidia graphics processing units
Personal Cinema FX 5700
Personal Cinema FX 5900 Ultra

Nvidia's initial release, the GeForce FX 5800, was intended as a high-end part. At the time, there were no GeForce FX products for the other segments of the market. The GeForce 4 MX continued in its role as the budget video card and the older GeForce 4 Ti cards filled in the mid-range.

In April 2003, the company introduced the GeForce FX 5600 and the GeForce FX 5200 to address the other market segments. Each had an "Ultra" variant and a slower, budget-oriented variant and all used conventional single-slot cooling solutions. The 5600 Ultra had respectable performance overall but it was slower than the Radeon 9600 Pro and sometimes slower than the GeForce 4 Ti series.[12] The FX 5200 did not perform as well as the DirectX 7.0 generation GeForce 4 MX440 or Radeon 9000 Pro in some benchmarks.[13]

In May 2003, Nvidia launched the GeForce FX 5900 Ultra, a new high-end product to replace the low-volume and disappointing FX 5800. Based upon a revised GPU called NV35, which fixed some of the DirectX 9 shortcomings of the discontinued NV30, this product was more competitive with the Radeon 9700 and 9800.[14] In addition to redesigning parts of the GPU, the company moved to a 256-bit memory data bus, allowing for significantly higher memory bandwidth than the 5800 even when utilizing more common DDR SDRAM instead of DDR2.[14] The 5900 Ultra performed somewhat better than the Radeon 9800 Pro in games not heavily using shader model 2, and had a quieter cooling system than the 5800.[14]

In October 2003, Nvidia released the GeForce FX 5700 and GeForce FX 5950. The 5700 was a mid-range card using the NV36 GPU with technology from NV35 while the 5950 was a high-end card again using the NV35 GPU but with additional clock speed. The 5950 also featured a redesigned version of the 5800's FlowFX cooler, this time using a larger, slower fan and running much quieter as a result. The 5700 provided strong competition for the Radeon 9600 XT in games limited to light use of shader model 2.[15] The 5950 was competitive with the Radeon 9800 XT, again as long as pixel shaders were lightly used.[16]

In December 2003, the company launched the GeForce FX 5900XT, a graphics card intended for the mid-range segment. It was similar to the 5900 Ultra, but clocked slower and used slower memory. It more thoroughly competed with Radeon 9600 XT, but was still behind in a few shader-intense scenarios.[17]

The GeForce FX line moved to PCI Express in early 2004 with a number of models, including the PCX 5300, PCX 5750, PCX 5900, and PCX 5950. These cards were largely the same as their AGP predecessors with similar model numbers. To operate on the PCIe bus, an AGP-to-PCIe "HSI bridge" chip on the video card converted the PCIe signals into AGP signals for the GPU.[18]

Also in 2004, the GeForce FX 5200 / 5300 series that utilized the NV34 GPU received a new member with the FX 5500.[19]

GeForce FX model information

[edit]
Further information: Rankine (microarchitecture)
  • All models support OpenGL 1.5 (2.1 (software) with latest drivers)
  • The GeForce FX series runs vertex shaders in an array
Model Launch
Code name
Transistors (million)
Die size (mm2)
Core clock (MHz)
Memory clock (MHz)
Core config[a]
 Fillrate Memory
Performance (GFLOPS
FP32)
TDP (Watts)
MOperations/s
MPixels/s
MTexels/s
MVertices/s
Size (MB)
Bandwidth (GB/s)
Bus type
Bus width (bit)
GeForce FX 5100 Mar 2003 NV34 TSMC 150 nm 45[21] 124 AGP 8x 200 166 4:2:4:4 800 800 800 100.0 64
128 		2.6 DDR 64 12.0 ?
GeForce FX 5200 LE 250 1,000 1,000 1,000 125.0 64
128
256 		2.6
5.3 		64
128 		15.0 ?
GeForce FX 5200 AGP 8x
PCI 		200 3.2
6.4 		64
128 		21
GeForce FX 5200 Ultra 6 Mar 2003 AGP 8x 325 325 1,300 1,300 1,300 162.5 10.4 128 19.5 32
GeForce PCX 5300 17 Mar 2004 PCIe x16 250 166 1,000 1,000 1,000 125.0 128
256 		2.6 64 15.0 21
GeForce FX 5500 Mar 2004 NV34B 45[22] 91 AGP 8x
AGP 4x
PCI 		270 166
200 		1,080 1,080 1,080 135.0 64
128
256 		5.3
6.4 		128 16.2 ?
GeForce FX 5600 XT Oct 2003 NV31 TSMC 130 nm 80[23] 121 AGP 8x 235 200 940 940 940 117.5 64
128 		3.2
6.4 		64
128 		14.1 ?
GeForce FX 5600 Mar 2003 AGP 8x
PCI 		325 275 1,300 1,300 1,300 162.5 64
128
256[24] 		8.8 128 19.5 25
GeForce FX 5600 Ultra 6 Mar 2003 AGP 8x 350 350 1,400 1,400 1,400 175.0 64
128 		11.2 21.0 27
GeForce FX 5600 Ultra Rev.2 400 400 1,600 1,600 1,600 200.0 12.8 24.0 31
GeForce FX 5700 VE Sep 2004 NV36 82[25] 133 250 200 4:3:4:4 1000 1000 1000 187.5 128
256 		3.2
6.4 		64
128 		17.5 20
GeForce FX 5700 LE Mar 2004 AGP 8x
PCI 		21
GeForce FX 5700 2003 AGP 8x 425 250 1,700 1,700 1,700 318.7 8.0 128 29.7 20
GeForce PCX 5750 17 Mar 2004 PCIe x16 128 25
GeForce FX 5700 Ultra 23 Oct 2003 AGP 8x 475 453 1,900 1,900 1,900 356.2 128
256 		14.4 GDDR2 33.2 43
GeForce FX 5700 Ultra GDDR3 15 Mar 2004 475 15.2 GDDR3 38
GeForce FX 5800 27 Jan 2003 NV30 125[26] 199 400 400 4:2:8:4 1,600 1,600 3,200 300.0 128 12.8 GDDR2 24.0 55
GeForce FX 5800 Ultra 500 500 2,000 2,000 4,000 375.0 16.0 30.0 66
GeForce FX 5900 ZT 15 Dec 2003 NV35 135[27] 207 325 350 4:3:8:4 1,300 1,300 2,600 243.7 22.4 DDR 256 22.7 ?
GeForce FX 5900 XT 15 Dec 2003[28] 390 1,600 1,600 3,200 300.0 27.3 48
GeForce FX 5900 May 2003 400 425 27.2 28.0 55
GeForce FX 5900 Ultra 12 May 2003 450 1,800 1,800 3,600 337.5 128
256 		31.5 65
GeForce PCX 5900 17 Mar 2004 PCIe x16 350 275 1,400 1,400 2,800 262.5 17.6 24.5 49
GeForce FX 5950 Ultra 23 Oct 2003 NV38 135[29] 207 AGP 8x 475 475 1,900 1,900 3,800 356.2 256 30.4 33.2 83
GeForce PCX 5950 17 Feb 2004 PCIe x16 425 27.2 GDDR3 83
Model Launch
Transistors (million)
Die size (mm2)
Core clock (MHz)
Memory clock (MHz)
Core config[a]
 Fillrate Memory
Performance (GFLOPS
FP32)
TDP (Watts)
MOperations/s
MPixels/s
MTexels/s
MVertices/s
Size (MB)
Bandwidth (GB/s)
Bus type
Bus width (bit)

GeForce FX Go 5 (Go 5xxx) series

[edit]

The GeForce FX Go 5 series for notebooks architecture.

Model Launch
Fab (nm)
Core clock (MHz)
Memory clock (MHz)
Core config1
 Fillrate Memory Supported API version
TDP (Watts)
Pixel (GP/s)
Texture (GT/s)
Size (MB)
Bandwidth (GB/s)
Bus type
Bus width (bit)
 OpenGL
Hardware
Drivers (Software)
GeForce FX Go 5100* Mar 2003 NV34M 150 AGP 8x 200 400 4:2:4:4 0.8 0.8 64 3.2 DDR 64 9.0 1.5 2.1** Unknown
GeForce FX Go 5500* 300 600 1.2 1.2 32
64 		9.6 128 Unknown
GeForce FX Go 5600* NV31M 130 350 1.4 1.4 32 Unknown
GeForce FX Go 5650* 350 Unknown
GeForce FX Go 5700* 1 Feb 2005 NV36M 450 550 4:3:4:4 1.8 1.8 8.8 Unknown

Support

[edit]

NVIDIA has ceased driver support for GeForce FX series.

Final drivers

[edit]
  • Windows 9x & Windows Me: 81.98 released on December 21, 2005; Download;
Product Support List Windows 95/98/Me – 81.98.
  • Driver version 81.98 for Windows 9x/Me was the last driver version ever released by Nvidia for these systems; no new official releases were later made for these systems.
  • Windows 2000, 32-bit Windows XP & Media Center Edition: 175.19 released on June 23, 2008; Download.
    • Note that the 175.19 driver is known to break Windows Remote Desktop (RDP).[30] The last version before the problem is 174.74. This was apparently fixed in 177.83, however this version is not available for the GeForce FX series of graphic cards.[31] Also worthwhile to note is that 163.75 is the last known good driver that correctly handles the adjustment of the video overlay color properties for the GeForce FX series. Subsequent WHQL drivers do not handle the whole range of possible video overlay adjustments (169.21) or have no effect on those (175.xx).
  • Windows XP (32-bit): 175.40 released on August 1, 2008; Download.
  • Windows Vista (32-bit): 96.85 released on October 17, 2006; Download;
  • Windows Vista (64-bit): 97.34 released on November 21, 2006; Download.
  • Linux/BSD/Solaris: 169.12 released on February 26, 2008; Download.
    • Also available: 177.67 (beta) released on August 19, 2008; Download.

The drivers for Windows 2000/XP can also be installed on later versions of Windows such as Windows Vista and 7; however, they do not support desktop compositing or the Aero effects of these operating systems.

(Products supported list also on this page)

Windows 95/98/Me Driver Archive
Windows XP/2000 Driver Archive
Unix Driver Archive

See also

[edit]

References

[edit]
[edit]
Revisions and contributorsEdit on WikipediaRead on Wikipedia

GeForce FX series

View on Grokipedia
from Grokipedia
The GeForce FX series is a family of graphics processing units (GPUs) developed by Corporation, announced in late 2002 with releases spanning 2003, marking the company's first implementation of full hardware support for Microsoft's 9.0 application programming interface () and featuring the innovative CineFX shading architecture for advanced cinematic-quality visual effects in gaming and multimedia applications. Announced amid high expectations as a revolutionary leap in PC graphics, the series debuted with the high-end FX 5800 Ultra launching in March 2003, built on the NV30 core using a 130 nm manufacturing process, followed by revisions and mid-range variants throughout 2003 to address initial production challenges and . Key models included the flagship GeForce FX 5950 Ultra (NV38 core, launched October 2003) for top-tier performance, the GeForce FX 5900 Ultra and 5900 (NV35 core, May 2003) as balanced high-end options, the GeForce FX 5700 Ultra and 5700 (NV36 core, August 2003) targeting mainstream gamers, and entry-level cards like the GeForce FX 5600 (NV31 core) and GeForce FX 5200 (NV34 core, March 2003), all supporting AGP 8x interface and DDR or DDR-II memory configurations up to 256 MB. Mobile variants under the FX Go branding, such as the Go 5700 and Go 5200, extended the lineup to laptops with optimized power efficiency. Despite its advanced feature set—including and vertex shader model support with up to 1024 instructions for complex procedural , 16 simultaneous textures per pass, 128-bit floating-point precision for studio-quality color rendering, and technologies like Intellisample for efficient and —the series faced criticism for underperforming relative to competitors like ATI's 9700 series in 9 workloads, exacerbated by inefficient implementations that caused significant performance penalties in shader-heavy games. Additionally, early models suffered from poor thermal management, running excessively hot and requiring bulky cooling solutions, which contributed to reliability issues and higher power draw compared to prior generations. These shortcomings led NVIDIA to iterate rapidly with the NV40-based in 2004, but the FX lineup remains historically significant as a bold, if flawed, transition toward programmable and high-fidelity graphics that influenced subsequent GPU architectures.

Introduction and Development

Background and Announcement

The GeForce FX series represented NVIDIA's strategic response to ' Radeon 9700 series, which had launched in August 2002 as the first graphics card to fully support Microsoft 9's programmable shader model, moving away from fixed-function pipelines toward more flexible, developer-controlled rendering. To regain market leadership in high-end graphics, NVIDIA developed the NV3x architecture to deliver comparable programmable vertex and pixel shaders, enabling cinematic effects like advanced lighting and procedural textures that were previously limited by hardware constraints. This shift aimed to empower game developers with tools for more realistic visuals, directly challenging ATI's early dominance in the 9 era. NVIDIA officially announced the GeForce FX series on November 18, 2002, at the trade show in , positioning it as a breakthrough for "cinematic computing" in PC gaming. The event garnered support from major game studios including , , and , who praised the series' potential to elevate game visuals to Hollywood standards. NVIDIA executive Bill Rehbock emphasized that the GeForce FX would "unleash no-holds-barred creative talent" with its CineFX engine and Cg shading language, targeting enthusiast gamers seeking immersive, high-fidelity experiences. Development of the FX, codenamed NV30, faced significant challenges during the transition from the NV2x of the GeForce4 series, including issues with TSMC's 0.13-micron manufacturing process and the integration of low-k materials that were later abandoned for a standard process. Rumors of the chip circulated throughout 2002, with initially promising a fall launch to counter ATI's momentum, but persistent yield problems and architectural refinements delayed availability to early 2003. These setbacks stemmed from the ambitious goal of balancing high transistor counts—125 million in the flagship NV30—with power efficiency and thermal management in a competitive landscape. The initial pricing strategy focused on the premium segment, with the high-end GeForce FX 5800 targeted at around $400 to appeal to performance-oriented enthusiasts willing to invest in cutting-edge technology. This positioned the series as an aspirational upgrade for prioritizing features over immediate value, aligning with NVIDIA's emphasis on for the core PC gaming .

Launch Timeline and Initial Market Position

The GeForce FX series marked NVIDIA's entry into the DirectX 9 era, with the flagship GeForce FX 5800 Ultra launching on March 6, 2003, at a recommended price of $499. This initial high-end model was followed by the GeForce FX 5900 series in May 2003, priced at around $399 for the base variant, and lower-tier options such as the FX 5200 (March 2003) and FX 5600 (April 2003) to broaden market accessibility. These releases came nearly a year after ATI's Radeon 9700 (R300) debut in August 2002, positioning NVIDIA as a late challenger in the high-performance discrete GPU segment. Upon launch, the FX series faced intense competition from ATI's maturing R300 lineup, including the 9800 PRO released in March 2003, which often outperformed the FX 5800 Ultra in benchmarks while consuming less power. NVIDIA aimed to reclaim market leadership by emphasizing the FX's programmable capabilities for emerging 9 titles, but ATI captured a slight majority of the discrete GPU market share by Q4 2003. Meanwhile, Intel's push into integrated via chipsets like the 865 series with Extreme Graphics captured 67% of the desktop integrated segment by late 2003, pressuring discrete vendors at the low end by bundling basic visuals into affordable OEM systems. Early reception was marred by controversies, including reports of excessive heat generation and fan noise on the FX 5800 Ultra, which drew around 85W and required aggressive cooling that reviewers described as ""-like. Driver instability further compounded issues, with initial ForceWare releases exhibiting precision errors in pixel shaders, leading to suboptimal 9 performance and artifacts in games. NVIDIA responded by quickly revising the reference cooler and issuing driver updates, but these problems contributed to the FX 5800's short shelf life, with production halted by mid-2003 in favor of the improved FX 5900. Sales for the FX series in 2003-2004 reflected a mixed initial position, with reporting total revenue of $1.82 billion in fiscal 2004 (ending January 2004), driven partly by FX adoption but tempered by competitive losses. The series solidified 's role as a flagship provider for 9 gaming, notably supporting id Software's upon its 2004 release, where FX cards delivered playable frame rates with advanced shading effects that highlighted their pixel strengths. Despite early hurdles, the lineup helped maintain a strong presence in enthusiast segments, shipping millions of units amid the transition to shader-heavy titles.

Technical Architecture

Core Design and Specifications

The GeForce FX series utilizes NVIDIA's NV3x family, with the NV30 serving as the foundational core fabricated on TSMC's 0.13-micron ( technology. This chip integrates 125 million s across a die area of 199 mm², enabling a balance of performance and efficiency for 9-compliant rendering. Subsequent variants like the NV35 refined this design while maintaining core architectural principles, with later cores such as NV36 (for ) and NV38 (for high-end with 256-bit ) featuring similar counts around 130 million. Core clock speeds across the NV3x lineup range from approximately 250 MHz in entry-level implementations to 500 MHz in high-end configurations, such as the GeForce FX 5800 Ultra, allowing for scalable performance tailored to different market segments. Memory interfaces vary from 64-bit to 256-bit across models, paired with DDR or GDDR2 (DDR-II) SDRAM, supporting capacities up to 256 MB and delivering bandwidth ranging from about 3.2 GB/s to 30.4 GB/s based on memory clock rates up to 1000 MHz effective. The pipeline architecture consists of 4 pixel pipelines equipped with 8 texture mapping units and 4 render output units, capable of processing up to 8 pixels and 8 texture samples per clock cycle under optimal conditions, though complex operations like multi-texturing may limit effective throughput to 4 textures per cycle. Integrated shader hardware includes 4 pixel shader units and 3 vertex shader units, providing foundational support for programmable shading effects. High-end NV3x cores demand significant power, with around 74 W in top models, driving the adoption of advanced thermal management. NVIDIA's FlowFX cooling innovations, featuring copper-based heatsinks, integrated heat pipes, and assemblies, address these demands by enhancing airflow and heat dissipation across dual-slot cooler designs.

Shader Pipeline and Rendering Features

The GeForce FX series introduced NVIDIA's CinéFX 2.0 engine, which provided comprehensive support for 9's vertex 2.0 and pixel 2.0 models, enabling developers to create more sophisticated programmable effects. This allowed for up to 65,536 instructions in vertex shaders and 1,024 instructions in pixel shaders, facilitating complex transformations and per-pixel lighting calculations that were previously limited in earlier generations. A key innovation was the native support for floating-point precision, including FP16 for high-speed operations and FP32 for higher accuracy, which enabled cinematic effects such as (HDR) rendering and realistic material simulations without precision artifacts in most scenarios. The shader pipeline in the GeForce FX series incorporated dynamic branching capabilities, particularly in vertex shaders, allowing conditional execution and loops based on per-vertex data to optimize processing for varied . Pixel shaders supported conditional execution through predicates, though full dynamic branching was more constrained compared to subsequent architectures. These features extended to advanced texturing techniques, including support for via dot-product 3 (DOT3) operations and higher-order methods in programmable shaders, enhancing surface detail without additional polygon . Additionally, the series fully supported Microsoft's High-Level (HLSL) under 9, permitting developers to write shaders in a C-like syntax that compiled to the underlying assembly, streamlining development for effects like procedural textures and dynamic lighting. Rendering enhancements in the GeForce FX series included IntelliSample technology, NVIDIA's proprietary approach to and , which improved image quality by reducing jagged edges and aliasing in transparent elements. This implementation supported modes up to 8x, along with transparency antialiasing for alpha-blended objects like foliage and particles, delivering smoother visuals in real-time applications. While effective for its era, the GeForce FX's handling of complex pixel shaders proved inefficient relative to later architectures, as its scalar often processed instructions at full FP32 precision even when FP16 sufficed, leading to bottlenecks in shader-intensive workloads. This limitation stemmed from the design's emphasis on cinematic precision over optimized throughput for branching-heavy or long-instruction shaders.

Product Models

Desktop Variants

The GeForce FX series for desktop systems encompassed a range of graphics processing units (GPUs) targeted at high-end, , and entry-level markets, all built on derivatives of the NV30 architecture introduced in 2003. These models utilized advanced features like pixel shader 2.0 support and CineFX engine for 9 compatibility, with variations in core clock speeds, memory configurations, and manufacturing processes to address different performance segments.

High-End Models

The flagship desktop variant was the GeForce FX 5950 Ultra, released in October 2003 and based on the NV38 core fabricated at 130 nm. It featured a core clock of 475 MHz, 256 MB of DDR memory clocked at 475 MHz (effective 950 MHz) across a 256-bit interface, 8 pixel shaders, 3 vertex shaders, 8 units (TMUs), and 4 render output units (ROPs), with a of 135 million. This model was positioned as NVIDIA's top-tier offering for enthusiasts, emphasizing improved yields and thermal efficiency over earlier high-end FX cards. Earlier high-end models included the GeForce FX 5900 Ultra and FX 5900 XT, both using the NV35 core at 130 nm as a die-shrunk refresh from the initial 150 nm NV30 process. The FX 5900 Ultra, launched in May 2003, operated at a 450 MHz core clock with 128 MB or 256 MB DDR memory at 425 MHz (850 MHz effective) on a 256-bit bus, incorporating 8 shaders, 3 vertex shaders, 8 TMUs, and 4 ROPs with 135 million transistors. The FX 5900 XT, introduced in January 2004, ran at a slightly lower 400 MHz core clock with 128 MB DDR at 350 MHz (700 MHz effective) on the same 256-bit bus, incorporating 8 shaders, 3 vertex shaders, 8 TMUs, and 4 ROPs, targeting value-oriented high-end users while maintaining full FX feature parity. These refreshes improved production efficiency and reduced power draw compared to the original FX 5800 Ultra on the 150 nm NV30.

Mid-Range Models

The mid-range lineup centered on the GeForce FX 5700 series, powered by the NV36 core at 130 nm and released starting in August 2003. The FX 5700 Ultra variant featured a 475 MHz core clock, 128 MB GDDR2 memory at 450 MHz (900 MHz effective) via a 128-bit interface, 4 pixel shaders, 3 vertex shaders, 4 TMUs, and 4 ROPs, with 82 million transistors. The GeForce FX 5600, launched in March 2003 on the NV31 core at 130 nm, offered a 350 MHz core clock with 128 or 256 MB DDR at 500 MHz (1000 MHz effective) on a 128-bit bus, 4 pixel shaders, 1 vertex shader, 4 TMUs, and 4 ROPs, with 80 million transistors, serving as an upper entry-level option. Lower-tier options like the FX 5700 and FX 5700 LE adjusted clocks downward for cost efficiency—the standard FX 5700 at around 400 MHz core and the LE at 250-300 MHz core with DDR memory options up to 256 MB—while retaining the core's shader capabilities for mainstream gaming. This series balanced performance and affordability, often bundled with 128 MB or 256 MB configurations to support resolutions up to 2048x1536.

Entry-Level Models

Entry-level desktop variants included the FX 5200 and FX 5500, designed for budget systems and basic multimedia tasks. The FX 5200, launched in March 2003 on the NV34 core at 150 nm, had a 250 MHz core clock, 64 MB DDR at 200 MHz (400 MHz effective) on a 128-bit bus (with some 64-bit variants), 4 pixel , 1 vertex , 4 TMUs, and 4 ROPs, totaling 45 million transistors. The FX 5500, introduced in 2004 using the NV34 core also at 150 nm, offered a modest with a 270 MHz core clock, 64-128 MB DDR at 200 MHz on a 128-bit bus, and the same /TMU/ROP setup but with 45 million transistors for slightly better efficiency in low-end applications. These models prioritized compatibility with AGP interfaces and supported 9 features without the power demands of higher tiers.
ModelCoreProcessCore ClockMemoryBus WidthRelease Date
FX 5950 UltraNV38130 nm475 MHz256 MB DDR (950 MHz eff.)256-bitOct 2003
FX 5900 UltraNV35130 nm450 MHz128/256 MB DDR (850 MHz eff.)256-bitMay 2003
FX 5900 XTNV35130 nm400 MHz128 MB DDR (700 MHz eff.)256-bitJan 2004
FX 5700 UltraNV36130 nm475 MHz128 MB GDDR2 (900 MHz eff.)128-bitAug 2003
FX 5700 LENV36130 nm250-300 MHz128/256 MB DDR128-bit2004
FX 5200NV34150 nm250 MHz64 MB DDR (400 MHz eff.)128-bitMar 2003
FX 5500NV34150 nm270 MHz64-128 MB DDR (400 MHz eff.)128-bit2004

Mobile and Embedded Variants

The GeForce FX Go series represented NVIDIA's effort to bring the FX architecture to , focusing on low-power implementations suitable for laptops. These variants prioritized energy efficiency and thermal management over raw performance, using scaled-down versions of the desktop GPUs to fit within the constraints of battery-powered devices. Higher-end models like the Go 5600 used NV31 derivatives on a , while entry-level like the Go 5200 used NV34 on 150 nm. The Go 5000 series encompassed entry- to models such as the Go 5200, Go 5250, Go 5300, Go 5500, and Go 5600, derived from the NV31 and NV34 chipsets. Equipped with 64 to 128 MB of DDR memory and a 128-bit bus, these GPUs operated at modest clock speeds—typically around 300 MHz for the core—to conserve power while supporting 9.0 features in portable environments. For instance, the Go 5200, based on the NV34, featured a 300 MHz core clock and 300 MHz memory clock, making it suitable for light gaming and in early notebooks. Similarly, the Go 5600, an NV31 derivative, offered comparable specifications but with enhanced support for better efficiency in mobile scenarios. The Go 5700, based on NV36 at 130 nm and launched in 2005, provided with ~400 MHz core, 64/128 MB DDR/GDDR2 on 128-bit bus for gaming laptops. High-end mobile offerings included the Go 5800 and Go 5900, targeted at performance-oriented notebooks and based on the NV30 and NV35 architectures, respectively. These models incorporated reduced core clocks, such as 300 MHz, compared to their desktop counterparts, along with thermal throttling mechanisms that dynamically lowered performance to maintain safe operating temperatures under sustained loads. With up to 128 MB of DDR memory on a 256-bit interface for the Go 5900, they delivered cinematic effects via the CineFX engine while addressing heat dissipation challenges in slim laptop chassis. Embedded variants extended the FX series to professional and OEM applications, notably the FX 5700LE, which was deployed in workstations for tasks requiring reliable graphics acceleration. Built on the NV36 at 130 nm, this model featured 128 MB DDR memory and customizations for embedded systems, such as optimized interfaces for industrial OEM integrations. Power efficiency was further enhanced across the mobile and embedded lineup through NVIDIA's PowerMizer technology, which employed dynamic voltage scaling to adjust GPU clocks and voltage based on , extending battery life in notebooks by up to 20% in balanced modes without sacrificing . These adaptations distinguished the mobile FX GPUs from desktop equivalents by emphasizing portability and sustained operation in constrained thermal envelopes.

Performance Analysis

Benchmark Results and Capabilities

The GeForce FX series showcased competitive performance in synthetic benchmarks designed to evaluate 9 capabilities. In 3DMark 2003, the flagship FX 5800 Ultra model scored approximately 5,461 points under standard test conditions with a compatible CPU like the Athlon XP 2500+. This result highlighted the series' ability to handle complex rendering tests, though it was impacted by early driver optimizations specific to the benchmark suite. Theoretical pixel fillrate for the FX 5800 Ultra reached 4 Gpixels/s, enabling strong rasterization performance in scenarios with high counts and basic texturing. In real-world gaming applications, the series delivered playable frame rates in early DirectX 9 titles, particularly those leveraging vertex shading features from its CineFX engine. ran playably in DirectX 8.1 mode to mitigate shader inefficiencies, underscoring the FX's strengths in rasterization-heavy scenes while revealing limitations in full DirectX 9 pixel execution. These results demonstrated the series' suitability for 2003-era games at medium resolutions, with brief utilization of advanced shader pipelines for enhanced geometric detail. Overall, the GeForce FX provided a notable uplift over the preceding GeForce 4 series in 9 environments, often delivering over 100% higher frame rates in shader-intensive tests like Fablemark due to improved vertex throughput. However, it exhibited clear bottlenecks in pixel-heavy workloads, where full-precision floating-point operations halved effective performance compared to theoretical specs. Additionally, the 128-bit interface capped bandwidth at around 16 GB/s, constraining texture streaming at higher resolutions and leading to stuttering with large, high-resolution assets in bandwidth-sensitive scenarios.

Competitive Comparisons

The GeForce FX series introduced full 32-bit floating-point precision in its and vertex shaders, enabling higher accuracy for cinematic effects like dynamic and complex textures in 9 applications, a feature NVIDIA marketed as "CineFX" for superior image fidelity compared to ATI's fixed-point approach in the 9700 and 9800 series. However, this precision came at the cost of slower shader execution on the FX, as ATI's 24-bit fixed-point shaders optimized for speed in gaming workloads, resulting in the 9700 Pro outperforming the FX 5800 Ultra by approximately 10-20% in complex 2.0 tests such as those in 3DMark 03 and early DX9 titles like RightMark . In standard benchmarks without or , the FX 5800 Ultra often matched or slightly exceeded the 9700 Pro, but ATI pulled ahead in quality-enhanced scenarios, where the Radeon's efficient drivers and filtering implementation provided better image quality and frame rates. Against Intel's Extreme Graphics 2, an integrated graphics solution built into 8xx-series chipsets, the GeForce FX series demonstrated overwhelming superiority in gaming performance, delivering playable frame rates in demanding titles like at resolutions up to 1600x1200, while Extreme Graphics 2 struggled below 30 FPS even at lower settings due to its limited 64-bit memory bus and lack of hardware transform and lighting support beyond 7 levels. This gap highlighted the FX's role as a dedicated discrete GPU for enthusiasts, though Intel's integrated option appealed to budget systems for basic multimedia and office tasks at a fraction of the cost, without the need for a separate . The launch delays of the GeForce FX series, attributed to manufacturing yields and design refinements, enabled ATI to capture significant in 2003, surpassing in Q4 with 24.9% to 24.7% in the discrete graphics segment as the Radeon 9700 solidified its position as the first true DX9-compliant GPU. responded by previewing (SLI) multi-GPU technology alongside the FX, promising up to 1.7x performance scaling in supported applications, while ATI's subsequent previews emphasized alternate frame rendering for broader compatibility. These developments intensified the rivalry, with ATI's timely releases and superior single-card performance driving its gains during the FX's rocky debut. In the long term, despite ATI's early dominance in DX9 adoption and market momentum, the GeForce FX series served as a foundational learning ground for 's programming and architecture, directly informing the NV40 () design that addressed FX inefficiencies with hybrid fixed- and floating-point pipelines for better speed-precision balance, ultimately reclaiming the high-end performance lead by mid-2004.

Support and Legacy

Driver Development and Updates

The development of drivers for the GeForce FX series began with the initial release of 's ForceWare 50.xx series in October 2003, coinciding with the launch of the FX hardware lineup. This version, such as 52.16, introduced the Unified Compiler, a GPU-specific optimization tool that enhanced programmable performance across the FX family, addressing early inefficiencies in vertex and processing. It also incorporated PowerMizer technology with dynamic peak power control and Thermal Protection 2.0, which mitigated overheating risks by adjusting clock speeds and fan behavior under load, a common concern for power-hungry FX models like the 5800 Ultra. Launch bugs, including visual artifacts such as scaling corruption in TV tuners and rendering issues like mouse cursor glitches in games such as with enabled, were targeted through registry tweaks and rendering fixes. Subsequent updates in the ForceWare 60.xx series, released in mid-2004 (e.g., version 61.76), focused on bolstering 9 compliance, achieving full support for 9.0c specifications to better align the FX series with emerging shader-heavy applications. These drivers refined the vertex and compilers introduced in 50.xx, improving robustness and texture memory management to reduce compilation errors that had caused instability in complex scenes. Specific bug fixes addressed artifacting-like rendering corruption, such as smearing in on the FX 5950 Ultra with 2xQ and graphical glitches in at high resolutions on the FX 5600 and 5700 Ultra models. The unified driver architecture (UDA), which allowed a single codebase to support multiple GPU generations from TNT to FX, was further solidified here, enabling consistent optimizations without hardware-specific overhauls. By the ForceWare 70.xx series in early 2005 (e.g., version 71.84), driver evolution emphasized multi-GPU configurations alongside newer 6 models. This also included game-specific optimizations, such as improved stability and frame rates in titles like , building on prior fixes for the Battlefield series (e.g., rendering corruption in from earlier updates). Ongoing refinements to the pipeline reduced compiler errors in environments, while enhancements continued to address thermal throttling issues in prolonged sessions. These updates marked the peak of active FX development, prioritizing compatibility with 9.0c and extensions like GL_NV_fragment_program2 for broader application support.

Discontinuation and Modern Compatibility

Official Windows support for the GeForce FX series ended with driver version 175.19 (WHQL) on June 23, 2008, for 32-bit and ; the final driver for was version 96.85, released in October 2007. For Linux users, support ended earlier with the 173.14.39 driver version released on December 6, 2013, after which no further updates or fixes were issued for the FX architecture on Unix-based systems. NVIDIA discontinued active development for the GeForce FX series as part of its broader shift toward the Kepler architecture introduced in 2012, which prioritized unified shader models and higher efficiency that the FX's design could not accommodate. By the , the series' install base had significantly declined due to the rapid evolution of graphics demands, making sustained support uneconomical compared to investments in newer generations. This marked the end of official driver enhancements, optimizations, or compatibility patches, leaving the hardware in a frozen legacy state. In modern computing environments, the GeForce FX series faces substantial limitations, lacking hardware support for APIs like or , which require post-FX architectures for features such as ray tracing and advanced compute shaders. It remains viable for and Vista using the final official drivers, enabling legacy applications, though unofficial installations on via for older drivers may provide limited functionality without official support. No compatibility exists for , 11, or ARM-based systems, which enforce stricter hardware requirements including TPM 2.0 and Secure Boot. Despite its obsolescence, the GeForce FX series holds lasting legacy impact through its pioneering programmable shader pipeline, which introduced high-precision floating-point operations in and vertex shaders, directly influencing early general-purpose GPU (GPGPU) techniques and the evolution toward unified architectures in later products. This foundation contributed to the broader adoption of GPUs for non-graphics computing, paving the way for frameworks like . In retro gaming communities, the series evokes nostalgia for early 2000s titles, with enthusiasts maintaining FX hardware for authentic emulation of era-specific games like those built on 9 engines.

References

  1. https://download.nvidia.com/ndemand/Collateral/Product_Overviews/GeForce_FX/PO_GFFX_5600_5200.pdf
  2. https://www.techpowerup.com/gpu-specs/geforce-fx-5800.c703
  3. https://en.gamegpu.com/video-cards-nvidia-a/geforce-fx-series/
  4. https://www.nvidia.com/content/imagekit/guidelines/GeForce_FX_logo_guidelines.pdf
  5. https://www.techpowerup.com/gpu-specs/?f=generation_NVIDIA%2BGeForce%2BFX
  6. https://www.tomshardware.com/picturestory/715-history-of-nvidia-gpus-2.html
  7. https://www.techpowerup.com/gpu-specs/radeon-9700.c49
  8. https://www.gamespot.com/articles/nvidia-announces-geforce-fx/1100-2897904/
  9. https://m.hexus.net/tech/items/graphics/1993-game-industry-rallies-around-nvidia-geforce-fx/?print=1
  10. https://www.techpowerup.com/gpu-specs/geforce-fx-5900.c77
  11. https://www.gamespot.com/articles/nvidia-unveils-new-geforce-card/1100-6027154/
  12. https://www.techpowerup.com/gpu-specs/geforce-fx-5200.c2351
  13. https://www.gamesindustry.biz/ati-edges-past-nvidia-in-graphics-market-share
  14. https://www.theregister.com/2003/10/30/intel_outsells_nvidia_ati/
  15. https://www.anandtech.com/show/911
  16. https://hardforum.com/threads/the-gpu-nvidia-would-rather-forget-geforce-fx.2029099/
  17. https://www.sec.gov/Archives/edgar/data/1045810/000104581004000010/10k.htm
  18. https://videocardz.net/gpu/nvidia-nv30
  19. https://www.techpowerup.com/gpu-specs/geforce-fx-5950-ultra.c79
  20. https://www.techpowerup.com/gpu-specs/geforce-fx-5800-ultra.c74
  21. https://www.techpowerup.com/gpu-specs/geforce-fx-5200.c60
  22. https://www.techpowerup.com/gpu-specs/nvidia-nv30.g20
  23. https://beyond3d.com/content/reviews/10/2
  24. https://developer.download.nvidia.com/assets/gamedev/docs/CineFX_1-final.pdf
  25. https://www.beyond3d.com/content/reviews/23/2
  26. https://www.beyond3d.com/content/reviews/10/13
  27. https://hothardware.com/reviews/nvidia-geforce-fx-derivatives
  28. https://videocardz.net/nvidia-geforce-fx-5950-ultra
  29. https://www.ign.com/articles/2003/10/27/nvidia-geforce-fx-5950-ultra-review
  30. https://www.techpowerup.com/gpu-specs/geforce-fx-5900-ultra.c704
  31. https://www.tomshardware.com/reviews/nvidia-geforcefx-5900-ultra%2C630.html
  32. https://www.techpowerup.com/gpu-specs/geforce-fx-5900-xt.c76
  33. https://videocardz.net/nvidia-geforce-fx-5900-xt
  34. https://hothardware.com/reviews/msi-geforce-fx-5900-xt
  35. https://videocardz.net/nvidia-geforce-fx-5700-ultra
  36. https://www.techpowerup.com/gpu-specs/nvidia-geforce-fx-5700-ultra-256-mb.b1503
  37. https://forums.tomshardware.com/threads/what-is-geforce-fx5700le.862274/
  38. https://www.gpuzoo.com/GPU-NVIDIA/GeForce_FX_5700_Ultra_GDDR3.html
  39. https://videocardz.net/nvidia-geforce-fx-5500
  40. https://www.evga.com/products/specs/gpu.aspx?pn=F39F8571-E16E-4C1A-8894-70EF8B314766
  41. https://www.techpowerup.com/gpu-specs/geforce-fx-go5700.c1552
  42. http://www.3dmark.com/3dm03/1218148
  43. https://hothardware.com/reviews/nvidia-geforce-fx--its-official?page=2
  44. https://hardforum.com/threads/any-way-to-force-a-geforce-fx-to-run-hl2-in-dx9-mode.833866/
  45. https://www.beyond3d.com/content/reviews/10/21
  46. https://forums.anandtech.com/threads/whats-the-issue-fix-with-geforce-fx-and-dx9.1175743/
  47. https://http.download.nvidia.com/ndemand/Collateral/Retail/Retail_BuyersGuide.pdf
  48. https://www.eetimes.com/the-startup-that-saved-ati/
  49. https://download.nvidia.com/Windows/52.16/ForceWare_Release_50_graphics_drivers_Release_Notes_v0.2.v52.16.pdf
  50. https://download.nvidia.com/Windows/61.76/61.76_ForceWare_Release_60_Graphics_Drivers_Release_Notes_Windows_XP-2000-9x-NT4.pdf
  51. https://www.nvidia.com/en-us/drivers/winxp-2k-4403/
  52. https://download.nvidia.com/Windows/71.84/NVWinRel70Notes_7184.pdf
  53. https://www.techspot.com/drivers/driver/file/information/4628/
  54. https://www.nvidia.com/en-us/drivers/details/5753/
  55. https://www.nvidia.com/en-us/drivers/details/2/
  56. https://www.nvidia.com/en-gb/geforce/drivers/results/71341/
  57. https://nvidia.custhelp.com/app/answers/detail/a_id/3473/~/eol-windows-driver-support-for-legacy-products
  58. https://developer.nvidia.com/vulkan-driver
  59. https://developer.nvidia.com/gpugems/gpugems2/part-iv-general-purpose-computation-gpus-primer
  60. https://hardforum.com/threads/the-gpu-nvidia-would-rather-forget-geforce-fx.2034221/
Add your contribution
Related Hubs
User Avatar
No comments yet.