Digit-Life :: Computer Hardware In Detail
  Articles Subscribe to the reviews RSS feed  Subscribe to reviews via Feedburner  Subscribe to reviews by e-mail
July 3, 2008
AMD 780G/780V/740G Integrated Socket AM2+ Chipsets

Hybrid CrossFire and High-Definition video.

July 2, 2008
AMD Phenom X4 In Real-Life Applications

How memory speed affects CPU performance.

June 28, 2008
Corsair Dominator DDR2-1142 (PC2-9136) 4GB Kit

High capacity, high frequency and Green design.

June 27, 2008
Foxconn GeForce 9800 GTX / GX2 2x512MB

Reference cards in nice boxes.

June 26, 2008
NVIDIA GeForce GTX 260 896MB

What does it offer for 399 USD?

June 25, 2008
AMD Phenom X3 8750

Stakes on the odd.

June 23, 2008
NVIDIA nForce 790i and Intel X48 Chipsets

Transition to DDR3, 1600 MHz FSB support, fully-fledged PCI-E 2.0.

NVIDIA GeForce GTX 280 1024MB

Will it outperform 9800 GX2? (Updated: now also with synthetic test results.)

June 17, 2008
x64 CPU Performance Testing Methodology

Version 3.0.

June 16, 2008
i3DSpeed, May 2008

Added test results for 2 x GeForce 9800 GX2 Quad SLI.

More articles »

3D Graphics Chipsets Comparision

Discuss


Comments


All chipsets support the following features:

  • Alpha Blending
  • Bilinear filtering
  • Dithering
  • Flat Shading
  • Fogging
  • Gouraud Shading
  • Perspective Correction
  • Specular Lightning (excl. TAZ-3D)
  • Rendition Verite 2100/2200 features per-polygon MIP-mapping only

"DVD Decoder" stands for the software MPEG2 decoding. The standard DVD decoding resolution is at least 800x600, in other words scaling is supported along with YUV->RGB transformation. The "+" means the support was announced, but wasn't verified in applications.


Chipset 3D Features RAM,
MB
Bus AGP2x DME
(DiME)
Multi-texturing 2 pixels/
clock
Fillrate
(million
pixels/s)
(million
texels/s)
Throughput
(million
polygons/s)
DVD
dec.
API Max. 3D
resolution
with Z-buffering
Color
depth
Windowed
rendering
Type Additional information
AF AA Edge AA Bump
maps
Trilinear
filtering
Z buffer D3D OGL
3Dfx Voodoo2 n/a n/a + n/a + 16bit 6-12 AGP/
PCI
n/a n/a + n/a 90/180 3 n/a + MCD 800x
600
16bit n/a 3D 175Mhz RAMDAC
3Dfx Voodoo2 SLI n/a n/a + n/a + 16bit 12-24 AGP/
PCI
n/a n/a + n/a 180/360 3 n/a + MCD 1024x
768
16bit n/a 3D 175Mhz RAMDAC
3Dfx Voodoo Banshee n/a n/a + + + 16bit 4-16 AGP/
PCI
n/a n/a n/a n/a 100/100 4 ? + + 1600x
1200
16bit/
24bit
+ 2D/3D 128-bit VGA core, Texture compression,Environment Mapping
3Dfx Voodoo Graphics n/a n/a + n/a n/a 16bit 4-8 PCI n/a n/a n/a n/a 45/45 1 n/a + MCD 640x
480
16bit n/a 3D 135Mhz RAMDAC
3Dfx Voodoo Rush n/a n/a + n/a n/a 16bit 4-8 PCI n/a n/a n/a n/a 45/45 1 n/a + MCD 800x
600
16bit + 2D/3D  
3Dlabs PERMEDIA2 n/a + n/a n/a + 16bit 4-8 AGP/
PCI
n/a n/a n/a n/a 42/42 1 ? + ICD 1024x
768
8bit/
16bit/
24bit
+ 2D/3D 230/250Mhz RAMDAC, stencil buffer
3Dlabs PERMEDIA3 + + + + + 32bit 8-16 AGP/
PCI
+ + + n/a 125/250 8 + + ICD 1920x
1200
8bit-
32bit
+ 2D/3D 270Mhz RAMDAC, Stencil buffer, virtual textures, procedural textures, light maps, environment maps, voxel rendering, 32-bit, 16-bit and n/an-linear Z-buffer
Intel i740 n/a + + n/a + 16bit 4-8/
12-24
AGP/
PCI
+ + n/a n/a 66/66 1.5 Yes3 + MCD,
ICD
1280x
1024
8bit/
16bit
+ 2D/3D 203/220Mhz RAMDAC
Matrox G200 n/a + + n/a + 32bit 4-32 AGP/
PCI
+ + n/a n/a 100/100 1.5 Yes + + 1600x
1200
8bit-
32bit
+ 2D/3D 250/320Mhz RAMDAC, VCQ Rendering
NEC PowerVR PCX2 n/a n/a n/a n/a n/a 32bit 4 PCI n/a n/a n/a n/a 66/66 ? ? + MCD 1024x
768
8bit/
16bit/
24bit
+ 3D  
NEC PowerVRS2 + + + + + 32bit 4-32 AGP/
PCI
+ + ? ? 120/? 4 ? + + 1600x
1200
8bit/
16bit/
24bit
+ 3D è 2D/3D Texture compression, Stencil buffer, Atmospheric effects
nVidia RIVA 128 + + + n/a + 16bit 4 AGP/
PCI
n/a + n/a n/a 100/100 5 Yes + ICD 1024x
768
8bit/
16bit/
24bit
+ 2D/3D 230Mhz RAMDAC
nVidia RIVA 128ZX + + + n/a + 16bit 8 AGP/
PCI
+ + n/a n/a 100/100 5 Yes + ICD 1024x
768
8bit/
16bit/
24bit
+ 2D/3D 250Mhz RAMDAC
nVidia RIVA TnT + + + + + 16bit/
24bit/
4-16 AGP/
PCI
+ + + + 180/180 6 Yes + ICD 1600x
1200
8bit/
16bit/
24bit/
32bit
+ 2D/3D 250Mhz RAMDAC, procedural textures, texture modulation, light maps environment maps, destination alpha channel
Rendition V2100 n/a n/a + n/a n/a 16bit 4-16 AGP/
PCI
n/a n/a n/a n/a ?/? ? ? + ICD 1280x
1024
8bit/
16bit/
24bit
+ 2D/3D 170Mhz RAMDAC
Rendition V2200 n/a n/a + n/a n/a 16bit 4-16 AGP/
PCI
n/a n/a n/a n/a 50/50 2 + + ICD 1280x
1024
8bit/
16bit/
24bit
+ 2D/3D 200/220Mhz RAMDAC
Rendition RRedline ? + + ? + 24bit 4-32 AGP/
PCI
+ + + + 250/? ? + + + ? 8bit/
16bit/
32bit
+ 2D/3D 250Mhz RAMDAC, Multiple monitors
S3 Savage3D + n/a + + + 16bit/
24bit
2-8 AGP/
PCI
+ + n/a n/a 125/125 5 Yes4 + + 1280x
1024
16bit/
32bit
+ 2D/3D 250Mhz RAMDAC, AGP x4, Stencil buffer, texture compresion, procedural textures, reflection maps, environment maps
ATI RagePro n/a n/a + n/a + 16bit 2-16 AGP/
PCI
+ + + n/a 45/90 1.2 Yes + ICD 1280x
1024
16bit/
32bit
+ 2D/3D 230Mhz RAMDAC, texture compresion
ATI RAGE 128 n/a + + + + 16bit/
24bit/
32bit
4-32 AGP + + + + 200/? 3.6 + + ICD 1920x
1600
16bit/
32bit
+ 2D/3D 250Mhz RAMDAC, AGP x4, Stencil buffer, environment maps, Vertex and Z-based reflections, Vertex and Z-based fog, LOD biasing, Texture morphing, Twin-Cache architecture
Silicon Reality TAZ-3D n/a + n/a n/a + 32bit 4-12 AGP/
PCI
n/a n/a n/a n/a 100/100 1.7 n/a + ICD 1600x
1200
32bit ? 3D Stencil buffer
#9 T2R4 n/a + n/a n/a + 32bit 4-32 AGP/
PCI
+ + n/a n/a ?/? ? ? + ICD 1920x
1200
16bit/
32bit
+ 2D/3D Atmospheric effects, interpolated fogging, 10 Levels-of-Detail Per-Pixel Mip Mapping
Stellar Semi. Aquilla PX n/a n/a n/a + ? 32bit 4 AGP/
PCI
+ + n/a n/a 100/100 ? n/a + + 1024x
768
16bit/
32bit
+ 2D/3D Source Color Keying,
Tile, Mirror, Clamp Texture Chunking; Linear, exponential, table, range-based fog; Texture Compression, support for square and n/an-square textures
Stellar Semi.TEX n/a + n/a + ? 32bit 4-16 AGP/
PCI
+ + n/a n/a 250/250 ? n/a + + ? 16bit/
32bit
? 3D Source Color Keying,
Tile, Mirror, Clamp Texture Chunking; Linear, exponential, table, range-based fog; Texture Compression, support for square and n/an-square textures
Stellar Semi. VelaTX n/a + n/a + + 24bit 2.5 Emb
+ 4-16
AGP/
PCI
+ + + n/a 200/? ? n/a + + ? 16bit/
32bit
? 3D Source Color Keying,
Tile, Mirror, Clamp Texture Chunking; Linear, exponential, table, range-based fog; Texture Compression, support for square and n/an-square textures
ARK
Tiger 3D
n/a n/a + + + 24bit 4-16 AGP/
PCI
n/a n/a n/a n/a 110/110 1.7 n/a + + 1600x
1200
16bit/
32bit
+ 2D/3D  
BitBoys Glaze3D n/a + + + + 16bit/
32bit
4-16
or more
AGP/
PCI
? ? + + 400/800 3 n/a + + 1600x
1200
8bit/
16bit/
32bit
+ 2D/3D TV-Out, 400Mhz Rambus DirectRDRAM, Projective textures, Stencil buffer, Texture edge clamp, Texture borders, 4 independent simultaneous textures with trilinear filtering in a single pass
BitBoys Glaze3D SLI n/a + + + + 16bit/
32bit
4-16
or more
AGP/
PCI
? ? + + 800/1600 3 n/a + + 1600x
1200
8bit/
16bit/
32bit
+ 2D/3D TV-Out, 400Mhz Rambus DirectRDRAM, Projective textures, Stencil buffer, Texture edge clamp, Texture borders, 4 independent simultaneous textures with trilinear filtering in a single pass

Notes

Texture read performance is 190MP/s1 for RivaTNT and 250MP/s for PERMEDIA3 (2 texels per clock). But unlike PERMEDIA3, RivaTNT (as well as Rage128) can process 2 pixels per clock, e.g. with bilinear filtering its fillrate is 190MP/s1, while PERMEDIA3 has it at 125MP/s.

Theoretical fillrate of TNT is 250Mpps. The matter is that nVidia has planned to set internal GPU clock rate to 125MHz, but is currently producing 95MHz 0.35-micron chips 95Mhz1. The 250Mpps products are due next year, so nVidia will be producing 125MHz TNT2 using the 0.25-micron technology. At the moment fillrate and triangle setup engine of RivaTNT is 190Mpps1 6 million1 triangle/s, respectively.

With advanced 3D features enabled the fillrate is:

  Riva TNT PERMEDIA 3
Anisotropic filtering 95 MP/s1 125 MP/s
Bump map embossing 95 MP/s1 125 MP/s
Dot product bump map n/a 125 MP/s
Bump environment map n/a 62.5 MP/s

The NEC PowerVRS2 supports both bump map embossing and dot product bump map. There are many rumours that PowerVRS2 supports multitexturing (similar to Permedia3).

Without multitexturing (e.g. with bilinear filtering) RivaTNT performs twice as fast as - at 190MP/s1, while the performance of PERMEDIA3 remains the same - 125MP/s.

At the moment we believe there will be two types of graphics chipsets:

  • Able to process 2 pixels per clock (e.g. RivaTNT, Rage128 and RRedline) and performing twice as fast in this mode.
  • Not able to process 2 pixels per clock (e.g. PERMEDIA 3).

This is not critical at the moment, since most games will in the near future use either multitexturing, or trilinear filtering (not just bilinear), but performance will be equal in such situations. But when more powerful chips arrive (e.g. Glaze 3D) this will become important, since quality of games will grow to 2 textures per clock and 8 textures per clock.

The fillrate of Rage128 is 200MP/s2. The texture read speed is 1 GT/s (gigatexel per second). Simple calculations indicate that the read speed is 800 MT/s and another 200MT/s are gained by caching. As you remember, Rage128 features pixel and texture caches.

Little is known about Rendition RRedline. We just know that its fillrate will be 250MP/s.

1 - This relates to RivaTNT 95MHz. The ready products operate at 90MHz, higher clock rate being a matter of overclocking at your own risk.

2 - The first revisions of ATI chip will work at 100MHz. After October 1998 ATI is going to boost clock rate to 125MHz.

3 - i740 supports software DVD decoding only at 720x480.

4 - Savage3D supports software DVD decoding in all resolutions starting with 800x600 as well as at 720x480 (MPEG2 original resolution).

In cooperation with Alexey Alexeenko (viktor@ces.kamchatka.su)
Additions by Yuri Klimov (klimov@mccme.ru)
May 8, 1998
Last updated: October 13, 1998



Discuss    Comments


Comments

Report bugs   Register       


  Total: 0

Platform & Cooling · Graphics Cards · Multimedia & ProAudio · Notebooks & Handhelds · Other Devices · Shopping


Advertise With Us · About Us · Affiliates · Forum


Copyright © 1997-2008: Byrds Research & Publishing Ltd. All rights reserved.
Design by Explosion & Artem Pavlenko. Programming by Sergey Anokhin.