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	ATI
	Radeon     5450	Radeon     5570	Radeon     5670	Radeon     5750	Radeon     5770	Radeon     5830	Radeon     5850	Radeon     5870
Core
Codename	Cedar	Redwood	Redwood	Juniper	Juniper	Cypress	Cypress	Cypress
Process scale (nm)	40	40	40	40	40	40	40	40
Transistor count (millions)	292	627	627	1040	1040	2150	2150	2150
Core clock speed (MHz)	650	650	775	700	850	800	725	850
Component execution units 1	80	400	400	720	800	1120	1440	1600
Vertex shader version (best)	5.0	5.0	5.0	5.0	5.0	5.0	5.0	5.0
Total effective no. of texture units	8	20	20	32	40	56	72	80
Max pixels output per clock (colour/depth)	4 / 16	8 / 32	8 / 32	16 / 64	16 / 64	16 / 64	32 / 128	32 / 128
Pixel shader version (bset)	5.0	5.0	5.0	5.0	5.0	5.0	5.0	5.0
Overall DirectX level	11	11	11	11	11	11	11	11
Memory
Interface width (bits)	64	128	128	128	128	256	256	256
Interface type	DDR2 / DDR3	DDR3	GDDR5	GDDR5	GDDR5	GDDR5	GDDR5	GDDR5
Clock speed (MHz)	800	900	1000	1150	1200	1000	1000	1200
Theoretical Peak Performance
Vertex processing rate (Gverts/sec)	3.25	16.3	19.4	31.5	42.5	56.0	65.3	85.0
Pixel processing rate (Gpixels/sec)	13.0	65.0	77.5	126.0	170.0	224.0	261.0	340.0
Texel applied rate (Gtexels/sec)	5.2	13.0	15.5	28.0	34.0	44.8	52.2	68.0
Pixel output rate (Gpixels/sec)	2.6	5.2	6.2	11.2	13.6	12.8	23.2	27.2
Memory bandwidth (GB/sec)	6.4	28.8	64.0	73.6	76.8	128.0	128.0	153.6

Model	ATI
	Radeon     5890
Core
Codename	Hemlock
Process scale (nm)	40
Transistor count (millions)	2 x 2150
Core clock speed (MHz)	725
Component execution units 1	2 x 1600
Vertex shader version (best)	5.0
Total effective no. of texture units	2 x 80
Max pixels output per clock (colour/depth)	2 x 32 / 128
Pixel shader version (bset)	5.0
Overall DirectX level	11
Memory
Interface width (bits)	2 x 256
Interface type	GDDR5
Clock speed (MHz)	1000
Theoretical Peak Performance
Vertex processing rate (Gverts/sec)	145.0 2
Pixel processing rate (Gpixels/sec)	580.0 2
Texel applied rate (Gtexels/sec)	116.0 2
Pixel output rate (Gpixels/sec)	46.4 2
Memory bandwidth (GB/sec)	256.0 2

Model	NVIDIA
	GeForce   GTX 470	GeForce   GTX 480
Core
Codename	GF100	GF100
Process scale (nm)	40	40
Transistor count (millions)	~ 3000	~ 3000
Core clock speed (MHz)	? / ?	? / ?
Component execution units 1	512 ?	512
Vertex shader version (best)	5.0	5.0
Total effective no. of texture units	60 ?	60
Max pixels output per clock (colour/depth)	48 / 384 ?	48 / 384
Pixel shader version (bset)	5.0	5.0
Overall DirectX level	11	11
Memory
Interface width (bits)	384 ?	384
Interface type	?	?
Clock speed (MHz)	GDDR5	GDDR5
Theoretical Peak Performance
Vertex processing rate (Gverts/sec)	?	?
Pixel processing rate (Gpixels/sec)	?	?
Texel applied rate (Gtexels/sec)	?	?
Pixel output rate (Gpixels/sec)	?	?
Memory bandwidth (GB/sec)	?	?

Notes

1 - All GPU vendors use a unified architecture for their DX11 chips - this means execution units are used to process all kinds of components (vertex, geometry and pixel).  Previous architectures had    pipelines of units dedicated for each type of component.

2 - The ATI Radeon HD 5890 is composed of two GPUs.  The performance figures are calculated for when operating in AFR (alternate frame rendering) mode.

© 3D Graphics Card Specifications - N D Evanson 2010