neeyik.info
The charts are categorised by the highest DirectX support the model offers - please use the links below to find the appropriate model.  For information about what the specification terms mean, please click here.
	DX6 or earlier	DX7 (TnL)	DX8 (SM1.1)	DX9 (SM2.0)	DX9 (SM3.0)	DX10 (SM4.0)	DX11 (SM5.0)

Model	3dfx
	Voodoo (SST)	Voodoo Rush	Voodoo2	Banshee	Voodoo3   1000	Voodoo3     2000	Voodoo3    3000	Voodoo3   3500
Core
Codename	SST-1	SST-96	SST-2	-	Avenger	Avenger	Avenger	Avenger
Process scale (nm)	500	500	350	350	350	250	250	250
Transistor count (millions)	~1	~1	~4	4	8.2	8.2	8.2	8.2
Core clock speed (MHz)	50	50	90	100	125	143	166	183
Vertex processing pipelines 1	-	-	-	-	-	-	-	-
Vertex shader version (best) 1	-	-	-	-	-	-	-	-
Pixel processing pipelines	1	1	1	1	1	1	1	1
Total effective no. of texture units	1	1	2	1	2	2	2	2
Max pixels output per clock (colour/depth)	1 / 1	1 / 1	1 / 1	1 / 1	1 / 1	1 / 1	1 / 1	1 / 1
Pixel shader version (best)	-	-	-	-	-	-	-	-
Overall DirectX level	6	6	6	6	6	6	6	6
Memory
Interface width (bits)	64	64	64 / 2x 64 2	128	128	128	128	128
Interface type	EDO	EDO	EDO	SDR	SDR	SDR	SDR	SDR
Clock speed (MHz)	50	50	90	100	125	143	166	183
Theoretical Peak Performance
Triangle setup rate (Mtris/sec)	-	-	3	4	5	6	7	8
Pixel processing rate (Mpixels/sec)	50	50	90	100	125	143	166	183
Texel applied rate (Mtexels/sec)	50	50	180	100	125	143	166	183
Pixel output rate (Mpixels/sec)	50	50	90	100	125	143	166	183
Memory bandwidth (GB/sec)	0.40	0.40	0.72 / 1.44 2	1.60	2.00	2.29	2.66	2.96

	Voodoo4     4500	Voodoo4     5500	Voodoo4     6000
Core
Codename	VSA-100	VSA-100	VSA-100
Process scale (nm)	250	250	250
Transistor count (millions)	14	2x 14	4x 14
Core clock speed (MHz)	183	166	166
Vertex processing pipelines 1	-	-	-
Vertex shader version (best) 1	-	-	-
Pixel processing pipelines	40	40	120
Total effective no. of texture units	2	2x 2	4x 2
Max pixels output per clock (colour/depth)	2 / 2	2x 2 / 2	4x 2 / 2
Pixel shader version (best)	-	-	-
Overall DirectX level	6	6	6
Memory
Interface width (bits)	128	2x 128	4x 128
Interface type	SDR	SDR	SDR
Clock speed (MHz)	183	166	166
Theoretical Peak Performance
Triangle setup rate (Mtris/sec)	12	11	11
Pixel processing rate (Mpixels/sec)	366	666 4	1333 4
Texel applied rate (Mtexels/sec)	366	666 4	1333 4
Pixel output rate (Mpixels/sec)	366	666 4	1333 4
Memory bandwidth (GB/sec)	2.93	5.31 3, 4	10.62 3, 4

Model	3DLabs
	Permedia	Permedia 2	Permedia 3
Core
Codename	P1	P2	P3
Process scale (nm)	?	?	?
Transistor count (millions)	?	?	?
Core clock speed (MHz)	25?	50?	125
Vertex processing pipelines 1	-	-	-
Vertex shader version (best) 1	-	-	-
Pixel processing pipelines	1	1	1
Total effective no. of texture units	1	1	1
Max pixels output per clock (colour/depth)	1 / 1	1 / 1	1 / 1
Pixel shader version (best)	-	-	-
Overall DirectX level	6	6	6
Memory
Interface width (bits)	64?	64?	128
Interface type	EDO	EDO	SDR
Clock speed (MHz)	25?	50?	125
Theoretical Peak Performance
Triangle setup rate (Mtris/sec)	0.5	1	11
Pixel processing rate (Mpixels/sec)	25	50	125
Texel applied rate (Mtexels/sec)	25	50	125
Pixel output rate (Mpixels/sec)	25	50	125
Memory bandwidth (GB/sec)	0.20?	0.40?	2.00

Model	ATI
	3D Rage	3D Rage II	3D Rage Pro	Rage 128	Rage 128   Pro	Rage Fury   MAXX	Radeon   '256' VE	Radeon     7000
Core
Codename	-	-	-	-	-	-	RV100	RV100
Process scale (nm)	500	500	350	250	250	250	180	180
Transistor count (millions)	~1?	~5?	~5-8?	8	8	2x 8	~30	~30
Core clock speed (MHz)	?	~25?	75	103	125	135	183	183
Vertex processing pipelines 1	-	-	-	-	-	-	-	-
Vertex shader version (best) 1	-	-	-	-	-	-	-	-
Pixel processing pipelines	1	1	1	2	2	2x 2	1	1
Total effective no. of texture units	1	1	1	2	2	2x 2	3	3
Max pixels output per clock (colour/depth)	1 / 1	1 / 1	1 / 1	2 / 2	2 / 2	2x 2 / 2	1 / 1	1 / 1
Pixel shader version (best)	-	-	-	-	-	-	-	-
Overall DirectX level	6	6	6	6	6	6	6	6
Memory
Interface width (bits)	64	64	64	128	128	2x 128	64	64
Interface type	EDO	SDR	SDR	SDR	SDR	SDR	GDDR	GDDR
Clock speed (MHz)	?	83	100	103	143	155	183	183
Theoretical Peak Performance
Triangle setup rate (Mtris/sec)	?	< 1	1.5	4	8	16 5	30	30
Pixel processing rate (Mpixels/sec)	?	25?	75	206	250	540 5	183	183
Texel applied rate (Mtexels/sec)	?	25?	75	206	250	540 5	549	549
Pixel output rate (Mpixels/sec)	?	25?	75	206	250	540 5	183	183
Memory bandwidth (GB/sec)	?	0.66	0.80	1.65	2.29	4.96 5	2.93	2.93

Model	Intel
	i740	Extreme	Extreme 2
Core
Process scale (nm)	350	?	?
Transistor count (millions)	3.5	?	?
Core clock speed (MHz)	~55	200+	266
Vertex processing pipelines 1	-	-	-
Vertex shader version (best) 1	-	-	-
Pixel processing pipelines	1	1	2
Total effective no. of texture units	1	2	2
Max pixels output per clock (colour/depth)	1 / 1	1 / 1	2 / 2
Pixel shader version (best)	-	-	-
Overall DirectX level	6	6	6
Memory
Interface width (bits)	64	64	64
Interface type	SDR	SDR / GDDR	SDR / GDDR
Clock speed (MHz)	100	100 / 133+	100 / 133+
Theoretical Peak Performance
Triangle setup rate (Mtris/sec)	1.1	~5	~8
Pixel processing rate (Mpixels/sec)	~55	200+	533
Texel applied rate (Mtexels/sec)	~55	400+	533
Pixel output rate (Mpixels/sec)	~55	200+	533
Memory bandwidth (GB/sec)	0.80	0.80+	0.80+

Model	Matrox
	G100	G200	G400	G400 MAX	G450
Core
Codename	?	Eclipse	Toucan	Toucan	Condor
Process scale (nm)	?	350	250	250	180
Transistor count (millions)	?	~5?	~10	~10?	15~25?
Core clock speed (MHz)	100	85	125	150	125
Vertex processing pipelines 1	-	-	-	-	-
Vertex shader version (best) 1	-	-	-	-	-
Pixel processing pipelines	1	1	2	2	2
Total effective no. of texture units	1	1	2	2	4
Max pixels output per clock (colour/depth)	1 / 1	1 / 1	2 / 2	2 / 2	2 / 2
Pixel shader version (best)	-	-	-	-	-
Overall DirectX level	6?	6?	6	6	6
Memory
Interface width (bits)	64	64	128	128	64
Interface type	SDR	SDR	SGR/SDR	SGR	GDDR
Clock speed (MHz)	66	112	166	200	166
Theoretical Peak Performance
Triangle setup rate (Mtris/sec)	?	1.3	6	> 5	> 5
Pixel processing rate (Mpixels/sec)	?	85	300	250	250
Texel applied rate (Mtexels/sec)	?	85	300	250	500
Pixel output rate (Mpixels/sec)	?	85	300	250	250
Memory bandwidth (GB/sec)	0.53	0.90	3.20	2.66	2.66

Model	NVIDIA
	Rva 128 ZX	TNT	TNT2	TNT2       Pro	TNT2       m64	TNT2        Ultra	TNT2     Vanta	TNT2      Vanta LT
Core
Codename	NV3	NV4	NV5	NV5	NV6	NV5	NV6	NV6
Process scale (nm)	350	350	250	220	250	250	250	250
Transistor count (millions)	3.5	7	15	15	~15	15	~15	~15
Core clock speed (MHz)	100	90	125	143	125	150	100	80
Vertex processing pipelines 1	-	-	-	-	-	-	-	-
Vertex shader version (best) 1	-	-	-	-	-	-	-	-
Pixel processing pipelines	1	2	2	2	2	2	2	2
Total effective no. of texture units	1	2	2	2	2	2	2	2
Max pixels output per clock (colour/depth)	1 / 1	2 / 2	2 / 2	2 / 2	2 / 2	2 / 2	2 / 2	2 / 2
Pixel shader version (best)	-	-	-	-	-	-	-	-
Overall DirectX level	5	6	6	6	6	6	6	6
Memory
Interface width (bits)	128	128	128	128	64	128	64	64
Interface type	SDR	SDR	SDR	SDR	SDR	SDR	SDR	SDR
Clock speed (MHz)	100	110	150	166	150	183	125	100
Theoretical Peak Performance
Triangle setup rate (Mtris/sec)	1.5	6	9	10	9	11	7	5.5
Pixel processing rate (Mpixels/sec)	100	180	250	286	250	300	200	160
Texel applied rate (Mtexels/sec)	100	180	250	286	250	300	200	160
Pixel output rate (Mpixels/sec)	100	180	250	286	250	300	200	160
Memory bandwidth (GB/sec)	1.60	1.76	2.40	2.66	1.20	2.93	2.00	0.80

Model	PowerVR
	PCX1	PCX2	Neon 250	Kyro	Kyro II	Kyro II SE
Core
Codename	-	-	250PC	STG4000	STG4500	STG4800
Process scale (nm)	500	350	250	250	180	180
Transistor count (millions)	~1?	~1?	~10?	12	15	15
Core clock speed (MHz)	60	66	125	115	175	200
Vertex processing pipelines 1	-	-	-	-	-	-
Vertex shader version (best) 1	-	-	-	-	-	-
Pixel processing pipelines	1	1	1	2	2	2
Total effective no. of texture units	1	1	1	2	2	2
Max pixels output per clock (colour/depth)	1 / 1	1 / 1	2 / 2	2 / 2	2 / 2	2 / 2
Pixel shader version (best)	-	-	-	-	-	-
Overall DirectX level	6	6	6	6	6~7	6~7
Memory
Interface width (bits)	64	64	128	128	128	128
Interface type	?	SDR	SDR	SDR	SDR	SDR
Clock speed (MHz)	?	?	125	115	175	200
Theoretical Peak Performance
Triangle setup rate (Mtris/sec)	?	~1?	4?	20?	> 25?	> 29?
Pixel processing rate (Mpixels/sec)	?	66	100	230	350	400
Texel applied rate (Mtexels/sec)	?	66	100	230	350	400
Pixel output rate (Mpixels/sec)	?	66	100	230	350	400
Memory bandwidth (GB/sec)	?	?	2.00	1.84	2.80	3.20

Model	S3 Graphics
	ViRGE/325	ViRGE/VX	ViRGE/DX	ViRGE/GX	ViRGE/GX2	Trio3D	Savage 3D	Savage 4	Savage 2000
Core
Process scale (nm)	500	500	350	350	350	350	250	250	180
Transistor count (millions)	?	?	?	?	?	?	~5?	5~10?	12
Core clock speed (MHz)	?						120	125	125
Vertex processing pipelines 1	-	-	-	-	-	-	-	-	1
Vertex shader version (best) 1	-	-	-	-	-	-	-	-	-
Pixel processing pipelines	1	1	1	1	1	1	1	1	2
Total effective no. of texture units	1	1	1	1	1	1	1	2	4
Max pixels output per clock (colour/depth)	1 / 1	1 / 1	1 / 1	1 / 1	1 / 1	1 / 1	1 / 1	1 / 1	2 / 2
Pixel shader version (best)	-	-	-	-	-	-	-	-	-
Overall DirectX level	6	6	6	6	6	6	6	6	6
Memory
Interface width (bits)	64	64	64	64	64	64	64	64	128
Interface type	EDO	EDO	EDO	EDO / SDR	EDO / SDR	EDO	SDR	SDR	SDR
Clock speed (MHz)	?	?	25	?	?		120	125	155
Theoretical Peak Performance
Triangle setup rate (Mtris/sec)	?	?	?	?	?	?	5	> 5	> 12
Pixel processing rate (Mpixels/sec)	?	?	?	?	?	?	120	125	250
Texel applied rate (Mtexels/sec)	?	?	?	?	?	?	120	250	500
Pixel output rate (Mpixels/sec)	?	?	?	?	?	?	120	125	250
Memory bandwidth (GB/sec)	?	?	0.20	?	?	?	0.96	1.00	2.48

Model	SiS						Trident
	6326	300	305				Blade 3D	Blade XP
Core
Process scale (nm)	?	180?	180?				250	250
Transistor count (millions)	~5?	~10?	~10?				?	?
Core clock speed (MHz)	45	125	125				110	166
Vertex processing pipelines 1	-	-	-				-	-
Vertex shader version (best) 1	-	-	-				-	-
Pixel processing pipelines	1	1	1				1	2
Total effective no. of texture units	1	1	1				1	4
Max pixels output per clock (colour/depth)	1 / 1	1 / 1	1 / 1				1 / 1	2 / 2
Pixel shader version (best)	-	-	-				-	-
Overall DirectX level	6	6	6				6	6
Memory
Interface width (bits)	64	128	64				64	128
Interface type	EDO / SDR	SDR	SDR				SDR	SDR
Clock speed (MHz)	100 / 90	83	100				120	166
Theoretical Peak Performance
Triangle setup rate (Mtris/sec)	~1	> 4	> 4				2.5	6
Pixel processing rate (Mpixels/sec)	45	125	125				110	333
Texel applied rate (Mtexels/sec)	45	125	125				110	666
Pixel output rate (Mpixels/sec)	45	125	125				110	333
Memory bandwidth (GB/sec)	0.80 / 0.72	2.66	1.14				0.96	2.66

Notes

1 - DX6 and earlier GPUs do not process vertices - all of the calculations are performed by the CPU.

2 - The frame buffer bus is 64-bits; each texture unit has its own 64-bit bus to the local memory.  Therefore, there is a total of 3 x 64-bit buses in the memory interface.  The bandwidth of 0.72 GB/s is for frame buffer access, and the 1.44 GB/s is the total for texture access.

3 - Bandwidth figures are for frame & vertex buffer access only; texture access is 2.67 GB/s for each VSA chip.

4 -  The 3dfx Voodoo4 5500 and 6000 use two and four GPUs, respectively, on a single board.  The performance figures are calculated using 3dfx's SLI mode.

5 - The ATI Rage Fury MAXX uses two GPUs on a single board.  The performance figures are calculated using AFR (alternate frame rendering) mode.

© 3D Graphics Card Specifications - N D Evanson 2010