NVIDIA GeForce FX 5900XT Performance Review

ok.....so here goes....
I've have a GeForce FX 5900XT card in my com....by MSI and was itching to make a review....so over the past few days I've slogged over this I'm proud to present my first review:

The GeForce FX 5900XT is based on the full-fledged NV35 core, which is the same eight pipeline GPU found on the GeForce FX 5900 and 5900 Ultra models (both of which have since been discontinued). The reference clock speed specifications for it is 390/700 MHz. The previous GeForce FX 5900 was clocked at 400/800 MHz, so the GPU clock speeds are quite similar, but not the memory. However, testing reveals that the GeForce FX 5900XT hardly differs much in performance from the GeForce FX 5900. Despite the lower memory clock speed on the XT version, it has been optimized for lower latency operation; hence the overall performance differential is small. Finally, the pricing is at a comfortable S$320 - S$380 price point (depending on the card model and its other features/bundles) that most gamers and graphics card enthusiasts are willing to fork out for a good performance card. In the following pages, you'll soon find out how well the GeForce FX 5900XT fares.

Here's the full technical specification list of a typical NVIDIA GeForce FX 5900XT product:-

NVIDIA GeForce FX 5900XT 128MB Technical Specifications

Graphics Engine	NVIDIA GeForce FX 5900XT GPU High Precision 3D Rendering Engine 8 pixels per clock rendering engine Full 128-bit Precision Graphics Pipeline Native support for 128-bit / 64-bit floating point and 32-bit integer rendering modes Up to 16 textures per pass CineFX 2.0 Engine Advanced Pixel Shaders allow floating-point pixel shader operations to run 2x faster than CineFX 1.0 Engine. Support for DX 9.0 Pixel Shader 2.0+ Support for DX 9.0 Vertex Shader 2.0+ Very long pixel programs up to 1024 instructions Very long vertex programs with up to 256 static instructions and up to 65536 instructions executed before termination Subroutines in shader programs Dynamic flow control Procedural shading Z-correct bump-mapping Hardware-accelerated shadow effects with shadow buffers UltraShadow Technology to accelerate shadow computations Two-sided stencil Keyframe animation Programmable matrix palette skinning Custom lens effects: fish eye, wide angle,fresnel effects, water refraction Intellisample HCT Technology Loss-less Color Compression and Z Compression (up to 4:1 ratio in real time) Fast Z-clear Adaptive texture filtering Dynamic Gamma Correction Fast antialiasing and compression performance High Compression Technology (HCT) increases visual quality at high resolutions with fast performance DirectX and S3TC texture compression Video Mixing Renderer (VMR) DVD and HDTV-ready MPEG-2 decoding up to 1920x1080i resolutions Dual DVO ports for interfacing to external TMDS transmitters and external HDTV encoders Support for dual-link DVI with resolutions support of 1600x1200. High-performance 2D rendering engine Digital Vibrance Control 3.0 (DVC) nView Multi-display Technology Default clock speed in 2D environment: 300/700 MHz (core/DDR memory) Default clock speed in 3D environment: 390/700 MHZ (core/DDR memory) 256-bit memory interface 22.4GB/sec Memory Bandwidth AGP 8X including fast writes and sideband addressing Advanced thermal monitoring and thermal management
Graphics Memory	128MB DDR SDRAM of 350MHz clock rate (effective speed of 700MHz DDR)
RAMDAC	Dual 400MHz RAMDACs that support resolutions of 2048x1536@85Hz
Video Decoder / Encoder	GPU integrated NTSC/PAL TV Encoder support resolutions up to 1024x768 with built-in Macrovision copy protection Video decoder by optional 3rd party ASIC part.
Connectors	1x analog RGB connector 1x mini-DIN connector (for Video-out) 1x DVI-I connector (for Digital Flat Panel displays)
Drivers	Driver support for Microsoft Windows 95/98/98SE/Me/NT 4.0/2000/XP, Linux IA32 / IA64 / AMD64 and FreeBSD NVIDIA Unified Driver Architecture (UDA) Fully compliant OpenGL 1.5 API and lower suport Complete DirectX support including 9.0 and lower
Other Information	AGP 2.0/3.0 slot required 1 PCI-assigned IRQ Required 1 four-pin power connector

GPU/VPU Comparison

Here's a comparison of similar caliber and/or similarly priced GPUs/VPUs. I've done this because some people think that the ATI RADEON 9800 can be conpared to FX 5900XT but the ATI RADEON 9800 is not a competitor to the GeForce FX 5900XT, it is added here just for comparison sake. From the GPU/VPU comparison, the RADEON 9600XT doesn't seem like a competitor to the GeForce FX 5900XT, but that's exactly the segment where NVIDIA was pitching their product. Oddly, the GeForce FX 5700 Ultra is in the same space with similar pricing, but NVIDIA had no comments with regards to its awkward pricing with respect to the GeForce FX 5900XT . Take special note of the technical differences between the GeForce FX 5900XT and the more powerful but phased out GeForce FX 5900 as I present the results on the upcoming pages.

GPU / VPU Core Comparison

GPU/VPU	NVIDIA GeForce FX 5900XT	NVIDIA GeForce FX 5900	NVIDIA GeForce FX 5700 Ultra	NVIDIA GeForce FX 5800	ATI RADEON 9600XT	ATI RADEON 9800
Core Code	NV35	NV35	NV36	NV30	RV360	R350
Manufacturing Process (microns)	0.13	0.13	0.13	0.13	0.13	0.15
Core Clock	390MHz	400MHz	475MHz	400MHz	500MHz	325MHz
Rendering (Pixel) Pipelines	8	8	4	8	4	8
Pixel Fill Rate (MPixels/s)	3,120	3,200	1,900	3,200	2,000	2,600
Texture Units per Pipeline	1	1	1	1	1	1
Peak Texture Fill Rate (Mtexels/s)	3,120	3,200	1,900	3,200	2,000	2,600
Vertex Shaders	Array	Array	Array	Array	2	4
Memory Clock	350MHz (700MHz DDR)	400MHz (800MHz DDR)	450MHz (900MHz DDR II)	400MHz (800MHz DDR II)	300MHz (600MHz DDR)	290MHz (580MHz DDR)
DDR Memory Bus	256-bit	256-bit	128-bit	128-bit	128-bit	256-bit
Memory Bandwidth	22.4GB/s	25.6GB/s	14.4GB/s	12.8GB/s	9.6GB/s	18.6GB/s
ATI SMARTSHADER	-	-	-	-	2.0	2.1
ATI SMOOTHVISION	-	-	-	-	2.1	2.1
ATI HYPER Z	-	-	-	-	III+	III+
ATI TRUEFORM	-	-	-	-	2.0	2.0
NVIDIA LMA II	Yes	Yes	Yes	Yes	-	-
NVIDIA CineFX Engine	2.0	2.0	2.0	1.0	-	-
NVIDIA Intellisample technology	Yes with HCT	Yes with HCT	Yes with HCT	Yes	-	-
Pixel Shader Support	2.0+	2.0+	2.0+	2.0+	2.0	2.0
Vertex Shader Support	2.0+	2.0+	2.0+	2.0+	2.0	2.0
DirectX Support	Up to DirectX 9.0	Up to DirectX 9.0	Up to DirectX 9.0	Up to DirectX 9.0	Up to DirectX 9.0	Up to DirectX 9.0
FSAA Mode	Multi Sampling(up to 8x)	Multi Sampling(up to 8x)	Multi Sampling(up to 8x)	Multi Sampling(up to 8x)	Multi Sampling(Up to 6x)	Multi Sampling(Up to 6x)
Anisotropic filtering Modes	Up to 8x	Up to 8x	Up to 8x	Up to 8x	Up to 16x	Up to 16x
Other Features	NView DVC 3.0 Support for NVIDIA Cg	NView DVC 3.0 Support for NVIDIA Cg	NView DVC 3.0 Support for NVIDIA Cg	NView DVC 3.0 Support for NVIDIA Cg	VideoShader HYDRAVISION	VideoShader HYDRAVISION
RAMDAC	Dual 400MHz	Dual 400MHz	Dual 400MHz	Dual 400MHz	Dual 400MHz	Dual 400MHz
TV Output	Int., 1024x768	Int., 1024x768	Int., 1024x768	Int., 1024x768	Int., 1024x768	Int., 1024x768
TMDS transmitter	optional	optional	Int. 165MHz	optional	Int. 165MHz	Int. 165MHz
AGP speed	8x	8x	8x	8x	8x	8x
AGP bus Compatibility	AGP 3.0, 2.0 and 1.0	AGP 3.0, 2.0 and 1.0	AGP 3.0, 2.0 and 1.0	AGP 3.0, 2.0 and 1.0	AGP 3.0 and 2.0	AGP 3.0 and 2.0

Taken form hardwarezone are some test they conducted on the graphics card:

Results - 3D Mark 2003 Pro (version 340)

Results observed here are pretty much to what we found when running 3D Mark 2001SE Professional.






Overall, the GeForce FX 5900XT is still a wonderful performer, operates at reasonable temperatures, occupies only a single slot (depends on ODM’s cooler designs), has a shorter depth than the GeForce FX 5700 Ultra graphics card and has a tantalizing price. The RADEON 9600XT graphics cards are in the S$285 – S$390 price range while most GeForce FX 5900XT graphics cards can be bought at the price point of S$320 – S$380.

So....how was it?