Kelvin (microarchitecture)

Nvidia Kelvin
Release date	2001-2003 2001
Architecture	Kelvin
Fabrication process	150 nm
API support
Direct3D	8.0
OpenGL	1.2 (1.5)
History
Predecessor	Celsius
Successor	Rankine
Support status
	Unsupported

Kelvin is the codename for a GPU microarchitecture developed by Nvidia, and released in 2001, as the successor to the Celsius^[1] microarchitecture. It was named with reference to William Thomson, 1st Baron Kelvin and used with the GeForce 3 and 4 series.

Graphics features

DirectX 8.0 (8.1 in the original Xbox)
OpenGL 1.2 (1.5;^[2]^[3] 1.4 in the original Xbox)
Shader Model 1.3
Vertex Shader 1.1
Max VRAM size bumped to 128MB
New memory controller with Z compression^[2]

Chips

GeForce 3 (3xxx) series

NV20, 57 million transistor

GeForce 4 (4xxx) series

Crush11, 20 million transistor^[4]
NV17, 29 million transistor^[5]
NV18, 29 million transistor
NV2A (Xbox GPU), 57 million transistor^[6]
NV25, 63 million transistor
NV28, 36 million transistor^[7]
NV2A (Xbox GPU), 57 million transistor^[6]

GPU list

GeForce 3 (3xxx) series

Model	Launch	Code name	Fab (nm)	Bus interface	Core clock (MHz)	Memory clock (MHz)	Core config¹	Fillrate				Memory
Model	Launch	Code name	Fab (nm)	Bus interface	Core clock (MHz)	Memory clock (MHz)	Core config¹	MOperations/s	MPixels/s	MTexels/s	MVertices/s	Size (MB)	Bandwidth (GB/s)	Bus type	Bus width (bit)
GeForce3 Ti200	1 October 2001	NV20	TSMC 150 nm	AGP 4× PCI	175	200	4:1:8:4	700	700	1400	42.75	64 128	6.4	DDR	128
GeForce3	27 February 2001				200	230		800	800	1600	50	64	7.36
GeForce3 Ti500	1 October 2001				240	250		960	960	1920	60	64 128	8

GeForce 4 (4xxx) series

Model	Launch	Code name	Fab (nm)	Bus interface	Core clock (MHz)	Memory clock (MHz)	Core config¹	Fillrate				Memory				Supported API version
Model	Launch	Code name	Fab (nm)	Bus interface	Core clock (MHz)	Memory clock (MHz)	Core config¹	MOperations/s	MPixels/s	MTexels/s	MVertices/s	Size (MB)	Bandwidth (GB/s)	Bus type	Bus width (bit)	Direct3D	OpenGL
GeForce4 MX IGP + nForce2	1 October 2002	NV1F	TSMC 150 nm	FSB	250	133 200^{[citation needed]}	2:0:4:2	500	500	1000	0	Up to 128 system RAM	2.128 6.4^{[citation needed]}	DDR	64 128	7.0	1.2
GeForce4 MX420	6 February 2002	NV17		AGP 4× PCI		166						64	2.656	SDR DDR	128 (SDR) 64 (DDR)
GeForce4 MX440 SE	2002	NV17		AGP 4× PCI								64 128	5.312	DDR	128
GeForce MX4000	14 December 2003	NV18B		AGP 8× PCI								64 128	2.656		64
GeForce PCX4300	19 February 2004	NV18B		PCIe ×16								128	2.656		64
GeForce4 MX440	6 February 2002	NV17		AGP 4× PCI	275	200		550	550	1100		64 128	6.4		128
GeForce4 MX440 8x	25 September 2002	NV18		AGP 8× PCI	275	250		550	550	1100			8
GeForce4 MX460	6 February 2002	NV17		AGP 4× PCI	300	275		600	600	1200			8
GeForce4 Ti4200	16 April 2002	NV25		AGP 4×	250	222 (128 MB) 250 (64 MB)	4:2:8:4	1000	1000	2000	125		7.104 (128 MB) 8 (64 MB)			8.0a	1.3
GeForce4 Ti4200 8x	25 September 2002	NV28		AGP 8×	250	250		1000	1000	2000	125	128	8
GeForce4 Ti4400	6 February 2002	NV25		AGP 4×	275	275		1100	1100	2200	137.5		8.8
GeForce4 Ti4400 8x (Ti4800SE²)	20 January 2003	NV28		AGP 8×	275	275		1100	1100	2200	137.5		8.8
GeForce4 Ti4600	6 February 2002	NV25		AGP 4×	300	325		1200	1200	2400	150		10.4
GeForce4 Ti4600 8x (Ti4800³)	20 January 2003	NV28		AGP 8×	300	325		1200	1200	2400	150		10.4
Model	Launch	Code name	Fab (nm)	Bus interface	Core clock (MHz)	Memory clock (MHz)	Core config¹	Fillrate				Memory				Supported API version
Model	Launch	Code name	Fab (nm)	Bus interface	Core clock (MHz)	Memory clock (MHz)	Core config¹	MOperations/s	MPixels/s	MTexels/s	MVertices/s	Size (MB)	Bandwidth (GB/s)	Bus type	Bus width (bit)	Direct3D	OpenGL

Model	Features
Model	nFiniteFX II Engine	Video Processing Engine (VPE)
GeForce4 MX420	No	Yes
GeForce4 MX440 SE	No	Yes
GeForce4 MX4000	No	Yes
GeForce4 PCX4300	No	Yes
GeForce4 MX440	No	Yes
GeForce4 MX440 8X	No	Yes
GeForce4 MX460	No	Yes
GeForce4 Ti4200	Yes	No
GeForce4 Ti4200 8x	Yes	No
GeForce4 Ti4400	Yes	No
GeForce4 Ti4400 8x	Yes	No
GeForce4 Ti4600	Yes	No
GeForce4 Ti4600 8x	Yes	No

Console GPUs

Model	Launch	Code name	Fab (nm)	Bus interface	Core clock (MHz)	Memory clock (MHz)	Core config^1,2,3	Fillrate				Memory				Latest supported API version
Model	Launch	Code name	Fab (nm)	Bus interface	Core clock (MHz)	Memory clock (MHz)	Core config^1,2,3	MOperations/s	MTexels/s	MPixels/s	MTri/s	Size (MB)	Bandwidth (GB/s)	Bus type	Bus width (bit)	Direct3D	OpenGL
XGPU (Xbox) ^[8]^[9]	November 15, 2001	NV2A	TSMC 150 nm	Integrated	233	200	4:2:8:4¹	5,800	1,864	932	116.5	64	6.4	DDR	128	8.1	1.4

References

^ "CodeNames". nouveau.freedesktop.org. Retrieved 15 February 2021.
^ ^a ^b "GPU chips — envytools git documentation". envytools.readthedocs.io. Retrieved 15 February 2021.
^ "TechPowerUp". TechPowerUp. Retrieved 15 February 2021.
^ "NVIDIA GeForce2 MX + nForce 220 Specs". TechPowerUp. Retrieved 15 February 2021.
^ "NVIDIA NV17 GPU Specs". TechPowerUp. Retrieved 15 February 2021.
^ ^a ^b "NVIDIA Xbox GPU Specs". TechPowerUp. Retrieved 15 February 2021.
^ "NVIDIA GeForce4 Ti 4800 Specs". TechPowerUp. Retrieved 15 February 2021.
^ "Anandtech Microsoft's Xbox". Anandtech.com. Archived from the original on 4 November 2010. Retrieved 11 November 2010.
^ Smith, Tony (14 February 2001). "TSMC starts fabbing Nvidia Xbox chips". The Register. Retrieved 20 November 2019.