<s>
GeForce	O
is	O
a	O
brand	O
of	O
graphics	B-Architecture
processing	I-Architecture
units	I-Architecture
(	O
GPUs	B-Architecture
)	O
designed	O
by	O
Nvidia	O
.	O
</s>
<s>
As	O
of	O
the	O
GeForce	B-Device
40	I-Device
series	I-Device
,	O
there	O
have	O
been	O
eighteen	O
iterations	O
of	O
the	O
design	O
.	O
</s>
<s>
The	O
first	O
GeForce	O
products	O
were	O
discrete	B-Device
GPUs	I-Device
designed	O
for	O
add-on	O
graphics	B-Device
boards	I-Device
,	O
intended	O
for	O
the	O
high-margin	O
PC	O
gaming	O
market	O
,	O
and	O
later	O
diversification	O
of	O
the	O
product	O
line	O
covered	O
all	O
tiers	O
of	O
the	O
PC	O
graphics	O
market	O
,	O
ranging	O
from	O
cost-sensitive	O
GPUs	B-Architecture
integrated	O
on	O
motherboards	O
,	O
to	O
mainstream	O
add-in	O
retail	O
boards	O
.	O
</s>
<s>
With	O
respect	O
to	O
discrete	B-Device
GPUs	I-Device
,	O
found	O
in	O
add-in	O
graphics-boards	O
,	O
Nvidia	O
's	O
GeForce	O
and	O
AMD	O
's	O
Radeon	B-Device
GPUs	B-Architecture
are	O
the	O
only	O
remaining	O
competitors	O
in	O
the	O
high-end	O
market	O
.	O
</s>
<s>
GeForce	O
GPUs	B-Architecture
are	O
very	O
dominant	O
in	O
the	O
general-purpose	B-Architecture
graphics	I-Architecture
processor	I-Architecture
unit	I-Architecture
(	O
GPGPU	B-Architecture
)	O
market	O
thanks	O
to	O
their	O
proprietary	O
CUDA	B-Architecture
architecture	O
.	O
</s>
<s>
GPGPU	B-Architecture
is	O
expected	O
to	O
expand	O
GPU	B-Architecture
functionality	O
beyond	O
the	O
traditional	O
rasterization	O
of	O
3D	O
graphics	O
,	O
to	O
turn	O
it	O
into	O
a	O
high-performance	O
computing	O
device	O
able	O
to	O
execute	O
arbitrary	O
programming	O
code	O
in	O
the	O
same	O
way	O
a	O
CPU	B-General_Concept
does	O
,	O
but	O
with	O
different	O
strengths	O
(	O
highly	O
parallel	O
execution	O
of	O
straightforward	O
calculations	O
)	O
and	O
weaknesses	O
(	O
worse	O
performance	O
for	O
complex	O
branching	B-General_Concept
code	O
)	O
.	O
</s>
<s>
The	O
company	O
called	O
out	O
to	O
the	O
public	O
to	O
name	O
the	O
successor	O
to	O
the	O
RIVA	B-Device
TNT2	I-Device
line	O
of	O
graphics	B-Device
boards	I-Device
.	O
</s>
<s>
There	O
were	O
over	O
12,000	O
entries	O
received	O
and	O
7	O
winners	O
received	O
a	O
RIVA	B-Device
TNT2	I-Device
Ultra	O
graphics	B-Device
card	I-Device
as	O
a	O
reward	O
.	O
</s>
<s>
Brian	O
Burke	O
,	O
senior	O
PR	O
manager	O
at	O
Nvidia	O
,	O
told	O
Maximum	O
PC	O
in	O
2002	O
that	O
"	O
GeForce	O
"	O
originally	O
stood	O
for	O
"	O
Geometry	O
Force	O
"	O
since	O
GeForce	B-Device
256	I-Device
was	O
the	O
first	O
GPU	B-Architecture
for	O
personal	O
computers	O
to	O
calculate	O
the	O
transform-and-lighting	B-Algorithm
geometry	O
,	O
offloading	O
that	O
function	O
from	O
the	O
CPU	B-General_Concept
.	O
</s>
<s>
Nvidia	O
moved	O
to	O
a	O
twin	O
texture	O
processor	O
per	O
pipeline	O
(	O
4x2	O
)	O
design	O
,	O
doubling	O
texture	O
fillrate	O
per	O
clock	O
compared	O
to	O
GeForce	B-Device
256	I-Device
.	O
</s>
<s>
Later	O
,	O
Nvidia	O
released	O
the	O
GeForce2	O
MX	O
(	O
NV11	O
)	O
,	O
which	O
offered	O
performance	O
similar	O
to	O
the	O
GeForce	B-Device
256	I-Device
but	O
at	O
a	O
fraction	O
of	O
the	O
cost	O
.	O
</s>
<s>
Launched	O
in	O
February	O
2001	O
,	O
the	O
GeForce3	O
(	O
NV20	O
)	O
introduced	O
programmable	O
vertex	O
and	O
pixel	O
shaders	O
to	O
the	O
GeForce	O
family	O
and	O
to	O
consumer-level	O
graphics	B-Architecture
accelerators	I-Architecture
.	O
</s>
<s>
The	O
NV2A	O
developed	O
for	O
the	O
Microsoft	O
Xbox	B-Application
game	O
console	O
is	O
a	O
derivative	O
of	O
the	O
GeForce	O
3	O
.	O
</s>
<s>
Most	O
of	O
these	O
models	O
used	O
the	O
AGP	B-Architecture
4×	O
interface	O
,	O
but	O
a	O
few	O
began	O
the	O
transition	O
to	O
AGP8×	O
.	O
</s>
<s>
The	O
GPU	B-Architecture
was	O
designed	O
not	O
only	O
to	O
support	O
the	O
new	O
Shader	O
Model	O
2	O
specification	O
but	O
also	O
to	O
perform	O
well	O
on	O
older	O
titles	O
.	O
</s>
<s>
However	O
,	O
initial	O
models	O
like	O
the	O
GeForce	O
FX	O
5800	O
Ultra	O
suffered	O
from	O
weak	O
floating	B-Algorithm
point	I-Algorithm
shader	O
performance	O
and	O
excessive	O
heat	O
which	O
required	O
infamously	O
noisy	O
two-slot	O
cooling	O
solutions	O
.	O
</s>
<s>
Launched	O
in	O
April	O
2004	O
,	O
the	O
GeForce	O
6	O
(	O
NV40	O
)	O
added	O
Shader	O
Model	O
3.0	O
support	O
to	O
the	O
GeForce	O
family	O
,	O
while	O
correcting	O
the	O
weak	O
floating	B-Algorithm
point	I-Algorithm
shader	O
performance	O
of	O
its	O
predecessor	O
.	O
</s>
<s>
It	O
also	O
implemented	O
high-dynamic-range	O
imaging	O
and	O
introduced	O
SLI	B-Device
(	O
Scalable	B-Device
Link	I-Device
Interface	I-Device
)	O
and	O
PureVideo	O
capability	O
(	O
integrated	O
partial	O
hardware	O
MPEG-2	O
,	O
VC-1	O
,	O
Windows	O
Media	O
Video	O
,	O
and	O
H.264	O
decoding	O
and	O
fully	O
accelerated	O
video	O
post-processing	O
)	O
.	O
</s>
<s>
The	O
seventh	O
generation	O
GeForce	O
(	O
G70/NV47	O
)	O
was	O
launched	O
in	O
June	O
2005	O
and	O
was	O
the	O
last	O
Nvidia	O
video	B-Device
card	I-Device
series	O
that	O
could	O
support	O
the	O
AGP	B-Architecture
bus	O
.	O
</s>
<s>
The	O
GeForce	O
7	O
also	O
offers	O
new	O
transparency	O
supersampling	B-Algorithm
and	O
transparency	O
multisampling	O
anti-aliasing	O
modes	O
(	O
TSAA	O
and	O
TMAA	O
)	O
.	O
</s>
<s>
The	O
GeForce	O
7950GT	O
featured	O
the	O
highest	O
performance	O
GPU	B-Architecture
with	O
an	O
AGP	B-Architecture
interface	O
in	O
the	O
Nvidia	O
line	O
.	O
</s>
<s>
A	O
128-bit	O
,	O
8	O
ROP	O
variant	O
of	O
the	O
7800	O
GTX	O
,	O
called	O
the	O
RSX	B-Device
'	I-Device
Reality	I-Device
Synthesizer	I-Device
 '	I-Device
,	O
is	O
used	O
as	O
the	O
main	O
GPU	B-Architecture
in	O
the	O
Sony	B-Operating_System
PlayStation	I-Operating_System
3	I-Operating_System
.	O
</s>
<s>
Released	O
on	O
November	O
8	O
,	O
2006	O
,	O
the	O
eighth-generation	O
GeForce	O
(	O
originally	O
called	O
G80	O
)	O
was	O
the	O
first	O
ever	O
GPU	B-Architecture
to	O
fully	O
support	O
Direct3D10	O
.	O
</s>
<s>
Manufactured	O
using	O
a	O
90nm	O
process	O
and	O
built	O
around	O
the	O
new	O
Tesla	B-Operating_System
microarchitecture	I-Operating_System
,	O
it	O
implemented	O
the	O
unified	O
shader	O
model	O
.	O
</s>
<s>
The	O
9800GX2	O
uses	O
two	O
G92	O
GPUs	B-Architecture
,	O
as	O
used	O
in	O
later	O
8800	O
cards	O
,	O
in	O
a	O
dual	O
PCB	O
configuration	O
while	O
still	O
only	O
requiring	O
a	O
single	O
PCI-Express	O
16x	O
slot	O
.	O
</s>
<s>
The	O
9800GX2	O
utilizes	O
two	O
separate	O
256-bit	O
memory	O
busses	O
,	O
one	O
for	O
each	O
GPU	B-Architecture
and	O
its	O
respective	O
512MB	O
of	O
memory	O
,	O
which	O
equates	O
to	O
an	O
overall	O
of	O
1GB	O
of	O
memory	O
on	O
the	O
card	O
(	O
although	O
the	O
SLI	B-Device
configuration	O
of	O
the	O
chips	O
necessitates	O
mirroring	O
the	O
frame	O
buffer	O
between	O
the	O
two	O
chips	O
,	O
thus	O
effectively	O
halving	O
the	O
memory	O
performance	O
of	O
a	O
256-bit/512MB	O
configuration	O
)	O
.	O
</s>
<s>
The	O
later	O
9800GTX	O
features	O
a	O
single	O
G92	O
GPU	B-Architecture
,	O
256-bit	O
data	O
bus	O
,	O
and	O
512MB	O
of	O
GDDR3	O
memory	O
.	O
</s>
<s>
Prior	O
to	O
the	O
release	O
,	O
no	O
concrete	O
information	O
was	O
known	O
except	O
that	O
the	O
officials	O
claimed	O
the	O
next	O
generation	O
products	O
had	O
close	O
to	O
1TFLOPS	O
processing	O
power	O
with	O
the	O
GPU	B-Architecture
cores	O
still	O
being	O
manufactured	O
in	O
the	O
65nm	B-Algorithm
process	I-Algorithm
,	O
and	O
reports	O
about	O
Nvidia	O
downplaying	O
the	O
significance	O
of	O
Direct3D10.1	O
.	O
</s>
<s>
Based	O
on	O
the	O
GT200	O
graphics	B-Architecture
processor	I-Architecture
consisting	O
of	O
1.4	O
billion	O
transistors	O
,	O
codenamed	O
Tesla	B-Operating_System
,	O
the	O
200	O
series	O
was	O
launched	O
on	O
June	O
16	O
,	O
2008	O
.	O
</s>
<s>
The	O
series	O
features	O
the	O
new	O
GT200	O
core	O
on	O
a	O
65nm	B-Algorithm
die	O
.	O
</s>
<s>
The	O
300	O
series	O
cards	O
are	O
rebranded	O
DirectX	O
10.1	O
compatible	O
GPUs	B-Architecture
from	O
the	O
200	O
series	O
,	O
which	O
were	O
not	O
available	O
for	O
individual	O
purchase	O
.	O
</s>
<s>
On	O
April	O
7	O
,	O
2010	O
,	O
Nvidia	O
released	O
the	O
GeForce	O
GTX	O
470	O
and	O
GTX	O
480	O
,	O
the	O
first	O
cards	O
based	O
on	O
the	O
new	B-General_Concept
Fermi	I-General_Concept
architecture	I-General_Concept
,	O
codenamed	O
GF100	O
;	O
they	O
were	O
the	O
first	O
Nvidia	O
GPUs	B-Architecture
to	O
utilize	O
1GB	O
or	O
more	O
of	O
GDDR5	O
memory	O
.	O
</s>
<s>
In	O
November	O
2010	O
,	O
Nvidia	O
released	O
a	O
new	O
flagship	O
GPU	B-Architecture
based	O
on	O
an	O
enhanced	O
GF100	O
architecture	O
(	O
GF110	O
)	O
called	O
the	O
GTX	O
580	O
.	O
</s>
<s>
This	O
GPU	B-Architecture
received	O
much	O
better	O
reviews	O
than	O
the	O
GTX	O
480	O
.	O
</s>
<s>
Nvidia	O
later	O
also	O
released	O
the	O
GTX	O
590	O
,	O
which	O
packs	O
two	O
GF110	O
GPUs	B-Architecture
on	O
a	O
single	O
card	O
.	O
</s>
<s>
In	O
September	O
2010	O
,	O
Nvidia	O
announced	O
that	O
the	O
successor	O
to	O
Fermi	B-General_Concept
microarchitecture	I-General_Concept
would	O
be	O
the	O
Kepler	B-General_Concept
microarchitecture	I-General_Concept
,	O
manufactured	O
with	O
the	O
TSMC	O
28nm	O
fabrication	O
process	O
.	O
</s>
<s>
Earlier	O
,	O
Nvidia	O
had	O
been	O
contracted	O
to	O
supply	O
their	O
top-end	O
GK110	O
cores	O
for	O
use	O
in	O
Oak	O
Ridge	O
National	O
Laboratory	O
's	O
"	B-Device
Titan	I-Device
"	I-Device
supercomputer	I-Device
,	O
leading	O
to	O
a	O
shortage	O
of	O
GK110	O
cores	O
.	O
</s>
<s>
After	O
AMD	O
launched	O
their	O
own	O
annual	O
refresh	O
in	O
early	O
2012	O
,	O
the	O
Radeon	B-Device
HD	I-Device
7000	O
series	O
,	O
Nvidia	O
began	O
the	O
release	O
of	O
the	O
GeForce	O
600	O
series	O
in	O
March	O
2012	O
.	O
</s>
<s>
It	O
introduced	O
significant	O
improvements	O
in	O
performance	O
,	O
heat	O
,	O
and	O
power	O
efficiency	O
compared	O
to	O
the	O
Fermi	B-General_Concept
architecture	I-General_Concept
and	O
closely	O
matched	O
AMD	O
's	O
flagship	O
Radeon	B-Device
HD	I-Device
7970	O
.	O
</s>
<s>
With	O
the	O
GTX	O
Titan	B-Device
,	O
Nvidia	O
also	O
released	O
GPU	B-Architecture
Boost	O
2.0	O
,	O
which	O
would	O
allow	O
the	O
GPU	B-Architecture
clock	O
speed	O
to	O
increase	O
indefinitely	O
until	O
a	O
user-set	O
temperature	O
limit	O
was	O
reached	O
without	O
passing	O
a	O
user-specified	O
maximum	O
fan	O
speed	O
.	O
</s>
<s>
The	O
final	O
GeForce	O
600	O
series	O
release	O
was	O
the	O
GTX	O
650	O
Ti	O
BOOST	O
based	O
on	O
the	O
GK106	O
core	O
,	O
in	O
response	O
to	O
AMD	O
's	O
Radeon	B-Device
HD7790	O
release	O
.	O
</s>
<s>
At	O
the	O
end	O
of	O
May	O
2013	O
,	O
Nvidia	O
announced	O
the	O
700	O
series	O
,	O
which	O
was	O
still	O
based	O
on	O
the	O
Kepler	B-General_Concept
architecture	I-General_Concept
,	O
however	O
it	O
featured	O
a	O
GK110-based	O
card	O
at	O
the	O
top	O
of	O
the	O
lineup	O
.	O
</s>
<s>
The	O
GTX	O
780	O
was	O
a	O
slightly	O
cut-down	O
Titan	B-Device
that	O
achieved	O
nearly	O
the	O
same	O
performance	O
for	O
two-thirds	O
of	O
the	O
price	O
.	O
</s>
<s>
At	O
the	O
same	O
time	O
,	O
Nvidia	O
announced	O
ShadowPlay	O
,	O
a	O
screen	O
capture	O
solution	O
that	O
used	O
an	O
integrated	O
H.264	O
encoder	O
built	O
into	O
the	O
Kepler	B-General_Concept
architecture	I-General_Concept
that	O
Nvidia	O
had	O
not	O
revealed	O
previously	O
.	O
</s>
<s>
No	O
more	O
700	O
series	O
cards	O
were	O
set	O
for	O
release	O
in	O
2013	O
,	O
although	O
Nvidia	O
announced	O
G-Sync	O
,	O
another	O
feature	O
of	O
the	O
Kepler	B-General_Concept
architecture	I-General_Concept
that	O
Nvidia	O
had	O
left	O
unmentioned	O
,	O
which	O
allowed	O
the	O
GPU	B-Architecture
to	O
dynamically	O
control	O
the	O
refresh	O
rate	O
of	O
G-Sync-compatible	O
monitors	O
which	O
would	O
release	O
in	O
2014	O
,	O
to	O
combat	O
tearing	O
and	O
judder	O
.	O
</s>
<s>
In	O
response	O
,	O
Nvidia	O
slashed	O
the	O
price	O
of	O
the	O
GTX780	O
by	O
$150	O
and	O
released	O
the	O
GTX780Ti	O
,	O
which	O
featured	O
a	O
full	O
2880-core	O
GK110	O
core	O
even	O
more	O
powerful	O
than	O
the	O
GTX	O
Titan	B-Device
,	O
along	O
with	O
enhancements	O
to	O
the	O
power	O
delivery	O
system	O
which	O
improved	O
overclocking	O
,	O
and	O
managed	O
to	O
pull	O
ahead	O
of	O
AMD	O
's	O
new	O
release	O
.	O
</s>
<s>
The	O
GeForce	O
800M	O
series	O
consists	O
of	O
rebranded	O
700M	O
series	O
parts	O
based	O
on	O
the	O
Kepler	B-General_Concept
architecture	I-General_Concept
and	O
some	O
lower-end	O
parts	O
based	O
on	O
the	O
newer	O
Maxwell	B-General_Concept
architecture	I-General_Concept
.	O
</s>
<s>
In	O
March	O
2013	O
,	O
Nvidia	O
announced	O
that	O
the	O
successor	O
to	O
Kepler	B-General_Concept
would	O
be	O
the	O
Maxwell	B-General_Concept
microarchitecture	I-General_Concept
.	O
</s>
<s>
It	O
was	O
released	O
in	O
September	O
2014	O
,	O
with	O
the	O
GM10x	O
series	O
chips	O
,	O
emphasizing	O
the	O
new	O
power	O
efficiency	O
architectural	O
improvements	O
in	O
OEM	O
,	O
and	O
low	O
TDP	B-General_Concept
products	O
in	O
desktop	B-Device
GTX	O
750/750	O
ti	O
,	O
and	O
mobile	O
GTX	O
850M/860M	O
.	O
</s>
<s>
Later	O
that	O
year	O
Nvidia	O
pushed	O
the	O
TDP	B-General_Concept
with	O
the	O
GM20x	O
chips	O
for	O
power	O
users	O
,	O
skipping	O
the	O
800	O
series	O
for	O
desktop	B-Device
entirely	O
,	O
with	O
the	O
900	O
series	O
of	O
GPUs	B-Architecture
.	O
</s>
<s>
This	O
was	O
the	O
last	O
GeForce	O
series	O
to	O
support	O
analog	O
video	O
output	O
through	O
DVI-I	B-Protocol
.	O
</s>
<s>
Although	O
,	O
analog	O
display	B-Device
adapters	I-Device
exist	O
and	O
are	O
able	O
to	O
convert	O
a	O
digital	O
Display	B-Protocol
Port	I-Protocol
,	O
HDMI	B-Protocol
,	O
or	O
DVI-D	O
(	O
Digital	O
)	O
.	O
</s>
<s>
In	O
March	O
2014	O
,	O
Nvidia	O
announced	O
that	O
the	O
successor	O
to	O
Maxwell	B-General_Concept
would	O
be	O
the	O
Pascal	B-General_Concept
microarchitecture	I-General_Concept
;	O
announced	O
on	O
6	O
May	O
2016	O
and	O
released	O
on	O
27	O
May	O
2016	O
.	O
</s>
<s>
In	O
Pascal	B-General_Concept
,	O
an	O
SM	O
(	O
streaming	O
multiprocessor	O
)	O
consists	O
of	O
128	O
CUDA	B-Architecture
cores	O
.	O
</s>
<s>
Kepler	B-General_Concept
packed	O
192	O
,	O
Fermi	B-General_Concept
32	O
and	O
Tesla	B-Operating_System
only	O
8	O
CUDA	B-Architecture
cores	O
into	O
an	O
SM	O
;	O
the	O
GP100	O
SM	O
is	O
partitioned	O
into	O
two	O
processing	O
blocks	O
,	O
each	O
having	O
32	O
single-precision	O
CUDA	B-Architecture
Cores	O
,	O
an	O
instruction	O
buffer	O
,	O
a	O
warp	O
scheduler	O
,	O
2	O
texture	O
mapping	O
units	O
and	O
2	O
dispatch	O
units	O
.	O
</s>
<s>
Only	O
the	O
Nvidia	B-Device
Titan	I-Device
X	I-Device
(	O
and	O
Titan	B-Device
Xp	I-Device
)	O
,	O
GTX	B-Device
1080	I-Device
,	O
GTX	B-Device
1080	I-Device
Ti	I-Device
,	O
and	O
GTX	B-Device
1060	I-Device
(	O
6	O
GB	O
Version	O
)	O
support	O
GDDR5X	O
.	O
</s>
<s>
The	O
GTX	B-Device
1070	I-Device
Ti	I-Device
,	O
GTX	B-Device
1070	I-Device
,	O
GTX	B-Device
1060	I-Device
(	O
3GB	O
version	O
)	O
,	O
GTX	B-Device
1050	I-Device
Ti	O
,	O
and	O
GTX	B-Device
1050	I-Device
use	O
GDDR5	O
.	O
</s>
<s>
Unified	O
memory	O
A	O
memory	O
architecture	O
,	O
where	O
the	O
CPU	B-General_Concept
and	O
GPU	B-Architecture
can	O
access	O
both	O
main	O
system	O
memory	O
and	O
memory	O
on	O
the	O
graphics	B-Device
card	I-Device
with	O
the	O
help	O
of	O
a	O
technology	O
called	O
"	O
Page	O
Migration	O
Engine	O
"	O
.	O
</s>
<s>
NVLink	O
A	O
high-bandwidth	O
bus	O
between	O
the	O
CPU	B-General_Concept
and	O
GPU	B-Architecture
,	O
and	O
between	O
multiple	O
GPUs	B-Architecture
.	O
</s>
<s>
16-bit	O
(	O
FP16	O
)	O
floating-point	B-Algorithm
operations	O
can	O
be	O
executed	O
at	O
twice	O
the	O
rate	O
of	O
32-bit	O
floating-point	B-Algorithm
operations	O
(	O
"	O
single	O
precision	O
"	O
)	O
and	O
64-bit	O
floating-point	B-Algorithm
operations	O
(	O
"	O
double	O
precision	O
"	O
)	O
executed	O
at	O
half	O
the	O
rate	O
of	O
32-bit	O
floating	B-Algorithm
point	I-Algorithm
operations	O
(	O
Maxwell	B-General_Concept
1/32	O
rate	O
)	O
.	O
</s>
<s>
In	O
August	O
2018	O
,	O
Nvidia	O
announced	O
the	O
GeForce	O
successor	O
to	O
Pascal	B-General_Concept
.	O
</s>
<s>
This	O
new	O
GPU	B-Architecture
microarchitecture	O
is	O
aimed	O
to	O
accelerate	O
the	O
real-time	O
ray	B-Algorithm
tracing	I-Algorithm
support	O
and	O
AI	B-Application
Inferencing	O
.	O
</s>
<s>
It	O
features	O
a	O
new	O
Ray	B-Algorithm
Tracing	I-Algorithm
unit	O
(	O
RT	B-Algorithm
Core	O
)	O
which	O
can	O
dedicate	O
processors	O
to	O
the	O
ray	B-Algorithm
tracing	I-Algorithm
in	O
hardware	O
.	O
</s>
<s>
It	O
supports	O
the	O
DXR	B-Algorithm
extension	O
in	O
Microsoft	O
DirectX	O
12	O
.	O
</s>
<s>
Nvidia	O
claims	O
the	O
new	O
architecture	O
is	O
up	O
to	O
6	O
times	O
faster	O
than	O
the	O
older	O
Pascal	B-General_Concept
architecture	O
.	O
</s>
<s>
A	O
whole	O
new	O
Tensor	B-Device
core	O
design	O
since	O
Volta	B-General_Concept
introduces	O
AI	B-Application
deep	O
learning	O
acceleration	O
,	O
which	O
allows	O
the	O
utilisation	O
of	O
DLSS	B-Device
(	O
Deep	B-Device
Learning	I-Device
Super	I-Device
Sampling	I-Device
)	O
,	O
a	O
new	O
form	O
of	O
anti-aliasing	O
that	O
uses	O
AI	B-Application
to	O
provide	O
crisper	O
imagery	O
with	O
less	O
impact	O
on	O
performance	O
.	O
</s>
<s>
It	O
also	O
changes	O
its	O
integer	O
execution	O
unit	O
which	O
can	O
execute	O
in	O
parallel	O
with	O
the	O
floating	B-Algorithm
point	I-Algorithm
data	O
path	O
.	O
</s>
<s>
The	O
new	O
GPUs	B-Architecture
were	O
revealed	O
as	O
the	O
Quadro	B-Application
RTX	O
8000	O
,	O
Quadro	B-Application
RTX	O
6000	O
and	O
Quadro	B-Application
RTX	O
5000	O
.	O
</s>
<s>
The	O
high	O
end	O
Quadro	B-Application
RTX	O
8000	O
features	O
4,608	O
CUDA	B-Architecture
cores	O
and	O
576	O
Tensor	B-Device
cores	O
with	O
48GB	O
of	O
VRAM	O
.	O
</s>
<s>
Later	O
during	O
the	O
Gamescom	O
press	O
conference	O
,	O
Nvidia	O
's	O
CEO	O
Jensen	O
Huang	O
,	O
unveiled	O
the	O
new	O
GeForce	B-Device
RTX	I-Device
series	O
with	O
RTX	B-Device
2080	I-Device
Ti	I-Device
,	O
2080	O
,	O
and	O
2070	O
that	O
will	O
use	O
the	O
Turing	O
architecture	O
.	O
</s>
<s>
Nvidia	O
announced	O
the	O
RTX	B-Device
2060	I-Device
on	O
January	O
6	O
,	O
2019	O
,	O
at	O
CES	O
2019	O
.	O
</s>
<s>
On	O
July	O
2	O
,	O
2019	O
,	O
Nvidia	O
announced	O
the	O
GeForce	B-Device
RTX	I-Device
Super	O
line	O
of	O
cards	O
,	O
a	O
20	O
series	O
refresh	O
which	O
comprises	O
higher-spec	O
versions	O
of	O
the	O
RTX	B-Device
2060	I-Device
,	O
2070	O
and	O
2080	O
.	O
</s>
<s>
The	O
RTX	B-Device
2070	I-Device
and	O
2080	O
were	O
discontinued	O
.	O
</s>
<s>
In	O
February	O
2019	O
,	O
Nvidia	O
announced	O
the	O
GeForce	B-Device
16	I-Device
series	I-Device
.	O
</s>
<s>
It	O
is	O
based	O
on	O
the	O
same	O
Turing	O
architecture	O
used	O
in	O
the	O
GeForce	B-Device
20	I-Device
series	I-Device
,	O
but	O
disabling	O
the	O
Tensor	B-Device
(	O
AI	B-Application
)	O
and	O
RT	B-Algorithm
(	O
ray	B-Algorithm
tracing	I-Algorithm
)	O
cores	O
to	O
provide	O
more	O
affordable	O
graphic	B-Device
cards	I-Device
for	O
gamers	O
while	O
still	O
attaining	O
a	O
higher	O
performance	O
compared	O
to	O
respective	O
cards	O
of	O
the	O
previous	O
GeForce	O
generations	O
.	O
</s>
<s>
Like	O
the	O
RTX	B-Device
Super	I-Device
refresh	O
,	O
Nvidia	O
on	O
October	O
29	O
,	O
2019	O
,	O
announced	O
the	O
GTX	B-Device
1650	I-Device
Super	I-Device
and	O
1660	O
Super	O
cards	O
,	O
which	O
replaced	O
their	O
non-Super	O
counterparts	O
.	O
</s>
<s>
Nvidia	O
officially	O
announced	O
at	O
the	O
GeForce	O
Special	O
Event	O
that	O
the	O
successor	O
to	O
GeForce	B-Device
20	I-Device
series	I-Device
will	O
be	O
the	O
30	O
series	O
,	O
it	O
is	O
built	O
on	O
the	O
Ampere	B-General_Concept
microarchitecture	O
.	O
</s>
<s>
The	O
GeForce	O
Special	O
Event	O
introduced	O
took	O
place	O
on	O
September	O
1	O
,	O
2020	O
,	O
and	O
set	O
September	O
17th	O
as	O
the	O
official	O
release	O
date	O
for	O
the	O
RTX	B-Device
3080	I-Device
GPU	B-Architecture
,	O
September	O
24th	O
for	O
the	O
RTX	B-Device
3090	I-Device
GPU	B-Architecture
and	O
October	O
29th	O
for	O
the	O
RTX	B-Device
3070	I-Device
GPU	B-Architecture
.	O
</s>
<s>
With	O
the	O
latest	O
GPU	B-Architecture
launch	O
being	O
the	O
RTX	B-Device
3090	I-Device
Ti	I-Device
.	O
</s>
<s>
The	O
RTX	B-Device
3090	I-Device
Ti	I-Device
is	O
the	O
highest-end	O
Nvidia	O
GPU	B-Architecture
on	O
the	O
Ampere	B-General_Concept
microarchitecture	O
,	O
it	O
features	O
a	O
fully	O
unlocked	O
GA102	O
die	O
built	O
on	O
the	O
Samsung	B-Application
8	B-Algorithm
nm	I-Algorithm
node	O
due	O
to	O
supply	O
shortages	O
with	O
TSMC	O
.	O
</s>
<s>
The	O
RTX	B-Device
3090	I-Device
Ti	I-Device
has	O
10,752	O
CUDA	B-Architecture
cores	O
,	O
336	O
Tensor	B-Device
cores	O
and	O
texture	O
mapping	O
units	O
,	O
112	O
ROPs	O
,	O
84	O
RT	B-Algorithm
cores	O
,	O
and	O
24	O
gigabytes	O
of	O
GDDR6X	O
memory	O
with	O
a	O
384-bit	O
bus	O
.	O
</s>
<s>
When	O
compared	O
to	O
the	O
RTX	B-Device
2080	I-Device
Ti	I-Device
,	O
the	O
3090	O
Ti	O
has	O
6,400	O
more	O
CUDA	B-Architecture
cores	O
.	O
</s>
<s>
Due	O
to	O
the	O
global	O
chip	O
shortage	O
,	O
the	O
30	O
series	O
was	O
controversial	O
as	O
scalpers	O
and	O
high	O
demand	O
meant	O
that	O
GPU	B-Architecture
prices	O
skyrocketed	O
for	O
the	O
30	O
series	O
and	O
the	O
AMD	O
RX	B-Device
6000	I-Device
series	O
.	O
</s>
<s>
On	O
September	O
20	O
,	O
2022	O
,	O
NVIDIA	O
announced	O
its	O
GeForce	B-Device
40	I-Device
Series	I-Device
graphics	B-Device
cards	I-Device
.	O
</s>
<s>
These	O
came	O
out	O
as	O
the	O
RTX	B-Device
4090	I-Device
,	O
on	O
October	O
12	O
,	O
2022	O
,	O
the	O
RTX	B-Device
4080	I-Device
,	O
on	O
November	O
16	O
,	O
2022	O
,	O
the	O
RTX	O
4070	O
Ti	O
,	O
on	O
January	O
3	O
,	O
2023	O
,	O
and	O
the	O
RTX	O
4070	O
,	O
on	O
April	O
13	O
,	O
2023	O
.	O
</s>
<s>
These	O
are	O
built	O
on	O
the	O
Ada	B-General_Concept
Lovelace	I-General_Concept
architecture	O
,	O
with	O
current	O
part	O
numbers	O
being	O
,	O
"	O
AD102	O
,	O
"	O
,	O
"	O
AD103	O
"	O
,	O
and	O
"	O
AD104	O
"	O
.	O
</s>
<s>
The	O
RTX	B-Device
4090	I-Device
is	O
currently	O
the	O
fastest	O
chip	O
for	O
the	O
mainstream	O
market	O
that	O
has	O
been	O
released	O
by	O
a	O
major	O
company	O
,	O
consisting	O
of	O
around	O
16,384	O
CUDA	B-Architecture
cores	O
,	O
boost	O
clocks	O
of	O
2.2	O
/	O
2.5	O
GHz	O
,	O
24	O
GB	O
of	O
GDDR6X	O
,	O
a	O
384-bit	O
memory	O
bus	O
,	O
128	O
3rd	O
gen	O
RT	B-Algorithm
cores	O
,	O
512	O
4th	O
gen	O
Tensor	B-Device
cores	O
,	O
DLSS	B-Device
3.0	I-Device
and	O
a	O
TDP	B-General_Concept
of	O
450W	O
.	O
</s>
<s>
NVIDIA	O
is	O
expected	O
to	O
release	O
a	O
new	O
TITAN	B-Device
RTX	I-Device
,	O
or	O
an	O
RTX	B-Device
4090	I-Device
Ti	O
,	O
sometime	O
around	O
2023-4	O
.	O
</s>
<s>
It	O
will	O
be	O
manufactured	O
by	O
TSMC	O
or	O
Samsung	B-Application
,	O
with	O
parts	O
being	O
codenamed	O
,	O
"	O
GB100	O
"	O
and	O
"	O
GB102	O
"	O
.	O
</s>
<s>
Most	O
of	O
the	O
features	O
present	O
in	O
the	O
desktop	B-Device
counterparts	O
are	O
present	O
in	O
the	O
mobile	O
ones	O
.	O
</s>
<s>
These	O
GPUs	B-Architecture
are	O
generally	O
optimized	O
for	O
lower	O
power	O
consumption	O
and	O
less	O
heat	O
output	O
in	O
order	O
to	O
be	O
used	O
in	O
notebook	O
PCs	O
and	O
small	O
desktops	O
.	O
</s>
<s>
Beginning	O
with	O
the	O
GeForce	O
8	O
series	O
,	O
the	O
GeForce	O
Go	O
brand	O
was	O
discontinued	O
and	O
the	O
mobile	O
GPUs	B-Architecture
were	O
integrated	O
with	O
the	O
main	O
line	O
of	O
GeForce	O
GPUs	B-Architecture
,	O
but	O
their	O
name	O
suffixed	O
with	O
an	O
M	O
.	O
This	O
ended	O
in	O
2016	O
with	O
the	O
launch	O
of	O
the	O
laptop	O
GeForce	B-Device
10	I-Device
series	I-Device
–	O
Nvidia	O
dropped	O
the	O
M	O
suffix	O
,	O
opting	O
to	O
unify	O
the	O
branding	O
between	O
their	O
desktop	B-Device
and	O
laptop	O
GPU	B-Architecture
offerings	O
,	O
as	O
notebook	O
Pascal	B-General_Concept
GPUs	B-Architecture
are	O
almost	O
as	O
powerful	O
as	O
their	O
desktop	B-Device
counterparts	O
(	O
something	O
Nvidia	O
tested	O
with	O
their	O
"	O
desktop-class	O
"	O
notebook	O
GTX	O
980	O
GPU	B-Architecture
back	O
in	O
2015	O
)	O
.	O
</s>
<s>
The	O
GeForce	O
MX	O
brand	O
,	O
previously	O
used	O
by	O
Nvidia	O
for	O
their	O
entry-level	O
desktop	B-Device
GPUs	B-Architecture
,	O
was	O
revived	O
in	O
2017	O
with	O
the	O
release	O
of	O
the	O
GeForce	B-Device
MX150	I-Device
for	O
notebooks	O
.	O
</s>
<s>
The	O
MX150	B-Device
is	O
based	O
on	O
the	O
same	O
Pascal	B-General_Concept
GP108	O
GPU	B-Architecture
as	O
used	O
on	O
the	O
desktop	B-Device
GT	B-Device
1030	I-Device
,	O
and	O
was	O
quietly	O
released	O
in	O
June	O
2017	O
.	O
</s>
<s>
Similar	O
to	O
the	O
mobile	O
GPUs	B-Architecture
,	O
Nvidia	O
also	O
released	O
a	O
few	O
GPUs	B-Architecture
in	O
"	O
small	O
form	O
factor	O
"	O
format	O
,	O
for	O
use	O
in	O
all-in-one	O
desktops	O
.	O
</s>
<s>
These	O
GPUs	B-Architecture
are	O
suffixed	O
with	O
an	O
S	O
,	O
similar	O
to	O
the	O
M	O
used	O
for	O
mobile	O
products	O
.	O
</s>
<s>
Beginning	O
with	O
the	O
nForce	B-Device
4	I-Device
,	O
Nvidia	O
started	O
including	O
onboard	O
graphics	O
solutions	O
in	O
their	O
motherboard	O
chipsets	O
.	O
</s>
<s>
These	O
onboard	O
graphics	O
solutions	O
were	O
called	O
mGPUs	O
(	O
motherboard	O
GPUs	B-Architecture
)	O
.	O
</s>
<s>
After	O
the	O
nForce	O
range	O
was	O
discontinued	O
,	O
Nvidia	O
released	O
their	O
Ion	B-Device
line	O
in	O
2009	O
,	O
which	O
consisted	O
of	O
an	O
Intel	B-Device
Atom	I-Device
CPU	I-Device
partnered	O
with	O
a	O
low-end	O
GeForce	O
9	O
series	O
GPU	B-Architecture
,	O
fixed	O
on	O
the	O
motherboard	O
.	O
</s>
<s>
Nvidia	O
released	O
an	O
upgraded	O
Ion	B-Device
2	O
in	O
2010	O
,	O
this	O
time	O
containing	O
a	O
low-end	O
GeForce	O
300	O
series	O
GPU	B-Architecture
.	O
</s>
<s>
Categoryof	O
graphics	B-Device
card	I-Device
Numberrange	O
SuffixSuffixes	O
indicate	O
its	O
performance	O
layer	O
,	O
and	O
those	O
listed	O
are	O
in	O
order	O
from	O
weakest	O
to	O
most	O
powerful	O
.	O
</s>
<s>
Since	O
the	O
release	O
of	O
the	O
GeForce	O
100	O
series	O
of	O
GPUs	B-Architecture
,	O
Nvidia	O
changed	O
their	O
product	O
naming	O
scheme	O
to	O
the	O
one	O
below	O
.	O
</s>
<s>
Nvidia	O
develops	O
and	O
publishes	O
GeForce	O
drivers	O
for	O
Windows	B-Operating_System
10	I-Operating_System
x86/x86	O
-64	O
and	O
later	O
,	O
Linux	B-Application
x86/x86-64/ARMv7	O
-A	O
,	O
OS	B-Application
X	I-Application
10.5	I-Application
and	O
later	O
,	O
Solaris	B-Application
x86/x86	O
-64	O
and	O
FreeBSD	B-Operating_System
x86/x86	O
-64	O
.	O
</s>
<s>
A	O
current	O
version	O
can	O
be	O
downloaded	O
from	O
Nvidia	O
and	O
most	O
Linux	B-Application
distributions	O
contain	O
it	O
in	O
their	O
own	O
repositories	O
.	O
</s>
<s>
Nvidia	B-Application
GeForce	I-Application
driver	I-Application
340.24	O
from	O
8	O
July	O
2014	O
supports	O
the	O
EGL	B-Application
interface	O
enabling	O
support	O
for	O
Wayland	B-Language
in	O
conjunction	O
with	O
this	O
driver	O
.	O
</s>
<s>
This	O
may	O
be	O
different	O
for	O
the	O
Nvidia	B-Application
Quadro	I-Application
brand	O
,	O
which	O
is	O
based	O
on	O
identical	O
hardware	O
but	O
features	O
OpenGL-certified	O
graphics	O
device	O
drivers	O
.	O
</s>
<s>
On	O
the	O
same	O
day	O
the	O
Vulkan	B-Application
graphics	I-Application
API	I-Application
was	O
publicly	O
released	O
,	O
Nvidia	O
released	O
drivers	O
that	O
fully	O
supported	O
it	O
.	O
</s>
<s>
The	O
support	O
Nvidia	O
's	O
display	O
controller	O
on	O
the	O
supported	O
GPUs	B-Architecture
is	O
centralized	O
in	O
nvidia-modeset.ko	O
.	O
</s>
<s>
Traditional	O
display	O
interactions	O
(	O
X11	O
modesets	O
,	O
OpenGL	O
SwapBuffers	O
,	O
VDPAU	O
presentation	O
,	O
SLI	B-Device
,	O
stereo	O
,	O
framelock	O
,	O
G-Sync	O
,	O
etc	O
.	O
)	O
</s>
<s>
In	O
May	O
2022	O
,	O
Nvidia	O
announced	O
that	O
it	O
would	O
release	O
a	O
partially	O
open	O
source	O
driver	O
for	O
the	O
Turing	O
architecture	O
and	O
newer	O
,	O
in	O
order	O
to	O
enhance	O
the	O
ability	O
for	O
it	O
to	O
be	O
packaged	O
as	O
part	O
of	O
Linux	B-Application
distributions	O
.	O
</s>
<s>
At	O
launch	O
Nvidia	O
considered	O
the	O
driver	O
to	O
be	O
alpha	O
quality	O
for	O
consumer	O
GPUs	B-Architecture
,	O
and	O
production	O
ready	O
for	O
datacenter	O
GPUs	B-Architecture
.	O
</s>
<s>
Currently	O
the	O
userspace	O
components	O
of	O
the	O
driver	O
(	O
including	O
OpenGL	O
,	O
Vulkan	B-Application
,	O
and	O
CUDA	B-Architecture
)	O
remain	O
proprietary	O
.	O
</s>
<s>
In	O
addition	O
,	O
the	O
open	O
source	O
components	O
of	O
the	O
driver	O
are	O
a	O
wrapper	O
for	O
the	O
GPU	B-Architecture
System	O
Processor	O
(	O
GSP	O
)	O
firmware	O
,	O
a	O
binary	O
blob	O
.	O
</s>
<s>
Open-source	O
drivers	O
are	O
developed	O
primarily	O
for	O
Linux	B-Application
,	O
however	O
there	O
may	O
be	O
ports	O
to	O
other	O
operating	O
systems	O
.	O
</s>
<s>
The	O
most	O
prominent	O
alternative	O
driver	O
is	O
the	O
reverse-engineered	O
free	O
and	O
open-source	O
nouveau	B-Application
graphics	O
device	O
driver	O
.	O
</s>
<s>
Nvidia	O
has	O
publicly	O
announced	O
to	O
not	O
provide	O
any	O
support	O
for	O
such	O
additional	O
device	O
drivers	O
for	O
their	O
products	O
,	O
although	O
Nvidia	O
has	O
contributed	O
code	O
to	O
the	O
Nouveau	B-Application
driver	I-Application
.	O
</s>
<s>
For	O
example	O
,	O
nouveau	B-Application
driver	I-Application
lacks	O
support	O
for	O
the	O
GPU	B-Architecture
and	O
memory	O
clock	O
frequency	O
adjustments	O
,	O
and	O
for	O
associated	O
dynamic	O
power	O
management	O
.	O
</s>
<s>
Also	O
,	O
Nvidia	O
's	O
proprietary	O
drivers	O
consistently	O
perform	O
better	O
than	O
nouveau	B-Application
in	O
various	O
benchmarks	O
.	O
</s>
<s>
However	O
,	O
and	O
version	O
3.16	O
of	O
the	O
Linux	B-Application
kernel	O
mainline	O
,	O
contributions	O
by	O
Nvidia	O
allowed	O
partial	O
support	O
for	O
GPU	B-Architecture
and	O
memory	O
clock	O
frequency	O
adjustments	O
to	O
be	O
implemented	O
.	O
</s>
<s>
The	O
software	O
configures	O
the	O
user	O
's	O
system	O
to	O
optimize	O
its	O
use	O
,	O
and	O
the	O
license	O
says	O
,	O
"	O
NVIDIA	O
will	O
have	O
no	O
responsibility	O
for	O
any	O
damage	O
or	O
loss	O
to	O
such	O
system	O
(	O
including	O
loss	O
of	O
data	O
or	O
access	O
)	O
arising	O
from	O
or	O
relating	O
to	O
(	O
a	O
)	O
any	O
changes	O
to	O
the	O
configuration	O
,	O
application	O
settings	O
,	O
environment	O
variables	O
,	O
registry	O
,	O
drivers	O
,	O
BIOS	B-Operating_System
,	O
or	O
other	O
attributes	O
of	O
the	O
system	O
(	O
or	O
any	O
part	O
of	O
such	O
system	O
)	O
initiated	O
through	O
the	O
SOFTWARE	O
"	O
.	O
</s>
