<s>
Nvidia	B-General_Concept
NVENC	I-General_Concept
(	O
short	O
for	O
Nvidia	O
Encoder	O
)	O
is	O
a	O
feature	O
in	O
Nvidia	B-Architecture
graphics	I-Architecture
cards	I-Architecture
that	O
performs	O
video	B-General_Concept
encoding	I-General_Concept
,	O
offloading	O
this	O
compute-intensive	O
task	O
from	O
the	O
CPU	B-General_Concept
to	O
a	O
dedicated	O
part	O
of	O
the	O
GPU	B-Architecture
.	O
</s>
<s>
It	O
was	O
introduced	O
with	O
the	O
Kepler-based	O
GeForce	B-Application
600	O
series	O
in	O
March	O
2012	O
.	O
</s>
<s>
The	O
encoder	O
is	O
supported	O
in	O
many	O
livestreaming	O
and	O
recording	O
programs	O
,	O
such	O
as	O
vMix	B-Language
,	O
Wirecast	B-Operating_System
,	O
Open	B-Language
Broadcaster	I-Language
Software	I-Language
(	O
OBS	O
)	O
and	O
Bandicam	B-Protocol
,	O
as	O
well	O
as	O
video	O
editing	O
apps	O
,	O
such	O
as	O
Adobe	B-Operating_System
Premiere	I-Operating_System
Pro	I-Operating_System
or	O
DaVinci	B-Language
Resolve	I-Language
.	O
</s>
<s>
It	O
also	O
works	O
with	O
Share	O
game	O
capture	O
,	O
which	O
is	O
included	O
in	O
Nvidia	O
's	O
GeForce	B-Application
Experience	O
software	O
.	O
</s>
<s>
Until	O
March	O
2023	O
consumer-targeted	O
GeForce	B-Application
graphics	O
cards	O
officially	O
support	O
no	O
more	O
than	O
3	O
simultaneously	O
encoding	O
video	O
streams	O
,	O
regardless	O
of	O
the	O
count	O
of	O
the	O
cards	O
installed	O
,	O
but	O
this	O
restriction	O
can	O
be	O
circumvented	O
on	O
Linux	B-Application
and	O
Windows	O
systems	O
by	O
applying	O
an	O
unofficial	O
patch	O
to	O
the	O
drivers	B-Application
.	O
</s>
<s>
Doing	O
so	O
also	O
unlocks	O
NVIDIA	O
Frame	O
Buffer	O
Capture	O
(	O
NVFBC	O
)	O
,	O
a	O
fast	O
desktop	O
capture	O
API	B-Application
that	O
uses	O
the	O
capabilities	O
of	O
the	O
GPU	B-Architecture
and	O
its	O
driver	B-Application
to	O
accelerate	O
capture	O
.	O
</s>
<s>
Nvidia	O
chips	O
also	O
feature	O
an	O
onboard	O
decoder	O
,	O
NVDEC	B-General_Concept
(	O
short	O
for	O
Nvidia	O
Decoder	O
)	O
,	O
to	O
offload	O
video	O
decoding	O
from	O
the	O
CPU	B-General_Concept
to	O
a	O
dedicated	O
part	O
of	O
the	O
GPU	B-Architecture
.	O
</s>
<s>
NVENC	B-General_Concept
has	O
undergone	O
several	O
hardware	O
revisions	O
since	O
its	O
introduction	O
with	O
the	O
first	O
Kepler	B-General_Concept
GPU	B-Architecture
(	O
GK104	O
)	O
.	O
</s>
<s>
The	O
first	O
generation	O
of	O
NVENC	B-General_Concept
,	O
which	O
is	O
shared	O
by	O
all	O
Kepler-based	O
GPUs	B-Architecture
,	O
supports	O
H.264	B-Application
high-profile	O
(	O
YUV420	O
,	O
I/P/B	O
frames	O
,	O
CAVLC/CABAC	O
)	O
,	O
H.264	B-Application
SVC	O
Temporal	O
Encode	O
VCE	O
,	O
and	O
Display	O
Encode	O
Mode	O
(	O
DEM	O
)	O
.	O
</s>
<s>
Actual	O
throughput	O
varies	O
on	O
the	O
selected	O
preset	O
,	O
user-controlled	O
parameters	O
and	O
settings	O
,	O
and	O
the	O
GPU/memory	O
clock	O
frequencies	O
.	O
</s>
<s>
The	O
published	O
8×	O
rating	O
is	O
achievable	O
with	O
the	O
NVENC	B-General_Concept
high-performance	O
preset	O
,	O
which	O
sacrifices	O
compression	O
efficiency	O
and	O
quality	O
for	O
encoder	O
throughput	O
.	O
</s>
<s>
Introduced	O
with	O
the	O
first-generation	O
Maxwell	B-General_Concept
architecture	I-General_Concept
,	O
second	O
generation	O
NVENC	B-General_Concept
adds	O
support	O
for	O
the	O
high-performance	O
HP444	O
profile	O
(	O
YUV4:4:4	O
,	O
predictive	O
lossless	B-Algorithm
encoding	I-Algorithm
)	O
,	O
and	O
increases	O
encoder	O
throughput	O
up	O
to	O
16×	O
realtime	O
,	O
which	O
corresponds	O
to	O
about	O
1080p	O
@	O
480Hz	O
with	O
the	O
high-performance	O
preset	O
.	O
</s>
<s>
Maxwell	B-General_Concept
GM108	O
does	O
not	O
have	O
NVENC	B-General_Concept
hardware	O
encoder	O
support	O
.	O
</s>
<s>
Introduced	O
with	O
the	O
second-generation	O
Maxwell	B-General_Concept
architecture	I-General_Concept
,	O
third	O
generation	O
NVENC	B-General_Concept
implements	O
the	O
video	O
compression	B-General_Concept
algorithm	I-General_Concept
High	B-Algorithm
Efficiency	I-Algorithm
Video	I-Algorithm
Coding	I-Algorithm
(	O
a.k.a.	O
</s>
<s>
HEVC	B-Algorithm
,	O
H.265	B-Algorithm
)	O
and	O
also	O
increases	O
the	O
H.264	B-Application
encoder	O
's	O
throughput	O
to	O
cover	O
4K-resolution	O
at	O
60Hz	O
(	O
2160p60	O
)	O
.	O
</s>
<s>
However	O
,	O
it	O
does	O
not	O
support	O
B-frames	B-General_Concept
for	O
HEVC	B-Algorithm
encoding	O
(	O
just	O
I	B-General_Concept
and	I-General_Concept
P	I-General_Concept
frames	I-General_Concept
)	O
.	O
</s>
<s>
The	O
maximum	O
NVENC	B-General_Concept
HEVC	B-Algorithm
coding	B-Algorithm
tree	I-Algorithm
unit	I-Algorithm
(	O
CU	O
)	O
size	O
is	O
32	O
(	O
the	O
HEVC	B-Algorithm
standard	O
allows	O
a	O
maximum	O
of	O
64	O
)	O
,	O
and	O
its	O
minimum	O
CU	O
size	O
is	O
8	O
.	O
</s>
<s>
HEVC	B-Algorithm
encoding	O
also	O
lacks	O
Sample	O
Adaptive	O
Offset	O
(	O
SAO	O
)	O
.	O
</s>
<s>
Adaptive	O
quantization	O
,	O
look-ahead	O
rate	O
control	O
,	O
adaptive	O
B-frames	B-General_Concept
(	O
H.264	B-Application
only	O
)	O
and	O
adaptive	O
GOP	O
features	O
were	O
added	O
with	O
the	O
release	O
of	O
Nvidia	O
Video	O
Codec	O
SDK	O
7	O
.	O
</s>
<s>
These	O
features	O
rely	O
on	O
CUDA	B-Architecture
cores	O
for	O
hardware	O
acceleration	O
.	O
</s>
<s>
SDK	O
7	O
supports	O
two	O
forms	O
of	O
adaptive	O
quantization	O
;	O
Spatial	O
AQ	O
(	O
H.264	B-Application
and	O
HEVC	B-Algorithm
)	O
and	O
Temporal	O
AQ	O
(	O
H.264	B-Application
only	O
)	O
.	O
</s>
<s>
Nvidia	O
's	O
consumer-grade	O
(	O
GeForce	B-Application
)	O
cards	O
and	O
some	O
of	O
its	O
lower-end	O
professional	O
Quadro	B-Application
cards	O
are	O
restricted	O
to	O
three	O
simultaneous	O
encoding	O
jobs	O
.	O
</s>
<s>
Its	O
higher-end	O
Quadro	B-Application
cards	O
do	O
not	O
have	O
this	O
restriction	O
.	O
</s>
<s>
Fourth	O
generation	O
NVENC	B-General_Concept
implements	O
HEVC	B-Algorithm
Main10	O
10-bit	O
hardware	O
encoding	O
.	O
</s>
<s>
It	O
also	O
doubles	O
the	O
encoding	O
performance	O
of	O
4K	O
H.264	B-Application
&	O
HEVC	B-Algorithm
when	O
compared	O
to	O
previous	O
generation	O
NVENC	B-General_Concept
.	O
</s>
<s>
It	O
supports	O
HEVC	B-Algorithm
8K	O
,	O
4:4:4	O
chroma	O
subsampling	O
,	O
lossless	B-Algorithm
encoding	I-Algorithm
,	O
and	O
sample	O
adaptive	O
offset	O
(	O
SAO	O
)	O
.	O
</s>
<s>
Nvidia	O
Video	O
Codec	O
SDK	O
8	O
added	O
Pascal	O
exclusive	O
Weighted	O
Prediction	O
feature	O
(	O
CUDA	B-Architecture
based	O
)	O
.	O
</s>
<s>
Weighted	O
prediction	O
is	O
not	O
supported	O
if	O
the	O
encode	O
session	O
is	O
configured	O
with	O
B	B-General_Concept
frames	I-General_Concept
(	O
H.264	B-Application
)	O
.	O
</s>
<s>
There	O
is	O
no	O
B-Frame	O
support	O
for	O
HEVC	B-Algorithm
encoding	O
,	O
and	O
the	O
maximum	O
CU	O
size	O
is	O
32×32	O
.	O
</s>
<s>
The	O
NVIDIA	O
GT	O
1030	O
and	O
the	O
Mobile	O
Quadro	B-Application
P500	O
are	O
GP108	O
chips	O
that	O
do	O
n't	O
support	O
the	O
NVENC	B-General_Concept
encoder	O
.	O
</s>
<s>
In	O
laptop	O
graphics	O
,	O
NVIDIA	O
MX	O
Graphics	O
do	O
not	O
include	O
NVENC	B-General_Concept
as	O
they	O
are	O
based	O
on	O
a	O
Maxwell-generation	O
GM108	O
or	O
a	O
Pascal-generation	O
GP108	O
chip	O
.	O
</s>
<s>
The	O
GeForce	B-Application
MX350	O
is	O
a	O
GP107	O
chip	O
whose	O
NVENC	B-General_Concept
encoder	O
is	O
disabled	O
during	O
manufacture	O
.	O
</s>
<s>
Volta	B-General_Concept
NVENC	B-General_Concept
has	O
similar	O
performance	O
as	O
Pascal	O
's	O
NVENC	B-General_Concept
.	O
</s>
<s>
It	O
does	O
not	O
offer	O
support	O
for	O
HEVC	B-Algorithm
B-Frames	B-General_Concept
.	O
</s>
<s>
In	O
mobile	O
graphics	O
,	O
as	O
with	O
most	O
other	O
GeForce	B-Application
MX-series	O
graphics	O
,	O
the	O
GeForce	B-Application
MX450	O
does	O
not	O
support	O
NVENC	B-General_Concept
as	O
it	O
is	O
a	O
TU117	O
chip	O
whose	O
hardware	O
encoder	O
is	O
permanently	O
disabled	O
in	O
its	O
manufacture	O
.	O
</s>
<s>
The	O
GeForce	B-Application
MX550	O
,	O
however	O
,	O
does	O
support	O
NVENC	B-General_Concept
as	O
its	O
hardware	O
encoder	O
remains	O
enabled	O
at	O
manufacturing	O
level	O
.	O
</s>
<s>
Sixth	O
generation	O
NVENC	B-General_Concept
implements	O
HEVC	B-Algorithm
8K	O
encoding	O
at	O
30	O
FPS	O
,	O
HEVC	B-Algorithm
B-frames	B-General_Concept
and	O
HEVC	B-Algorithm
B-frames	B-General_Concept
as	O
reference	O
(	O
with	O
support	O
for	O
each	O
and	O
middle	O
modes	O
)	O
and	O
Alpha	O
HEVC	B-Algorithm
support	O
and	O
provides	O
up	O
to	O
25%	O
bitrate	O
savings	O
for	O
HEVC	B-Algorithm
and	O
up	O
to	O
15%	O
bitrate	O
savings	O
for	O
H.264	B-Application
.	O
</s>
<s>
The	O
initial	O
launch	O
of	O
the	O
Nvidia	B-Application
GeForce	I-Application
GTX	O
1650	O
was	O
exempt	O
from	O
this	O
generation	O
however	O
,	O
as	O
it	O
used	O
Volta	B-General_Concept
NVENC	B-General_Concept
instead	O
of	O
Turing	O
.	O
</s>
<s>
Nvidia	O
updated	O
the	O
NVENC	B-General_Concept
encoder	O
of	O
the	O
GTX	O
1650	O
cards	O
in	O
2020	O
to	O
also	O
use	O
the	O
Turing	O
engine	O
.	O
</s>
<s>
The	O
GTX	O
1650	O
Super	O
uses	O
the	O
Turing	O
NVENC	B-General_Concept
engine	O
as	O
it	O
is	O
based	O
on	O
the	O
TU116	O
rather	O
than	O
the	O
TU117	O
used	O
in	O
the	O
original	O
GTX	O
1650	O
.	O
</s>
<s>
Ampere	B-General_Concept
has	O
essentially	O
the	O
same	O
NVENC	B-General_Concept
generation	O
engine	O
as	O
Turing	O
.	O
</s>
<s>
The	O
only	O
substantive	O
difference	O
is	O
that	O
NVDEC	B-General_Concept
received	O
support	O
for	O
AV1	B-Application
decoding	O
(	O
with	O
film	O
grain	O
)	O
.	O
</s>
<s>
In	O
entry-level	O
mobile	O
graphics	O
,	O
the	O
GA107-chip-based	O
GeForce	B-Application
MX570	O
comes	O
in	O
two	O
versions	O
,	O
one	O
of	O
which	O
(	O
the	O
GeForce	B-Application
MX570	O
A	O
)	O
has	O
the	O
hardware	O
decoder	O
and	O
encoder	O
permanently	O
disabled	O
during	O
manufacturing	O
.	O
</s>
<s>
Nvidia	O
announced	O
the	O
next-gen	O
NVENC	B-General_Concept
with	O
8K	O
10-bit	O
60FPS	O
AV1	B-Application
fixed	O
function	O
hardware	O
encoder	O
in	O
Ada	B-General_Concept
Lovelace	I-General_Concept
GPUs	B-Architecture
.	O
</s>
<s>
The	O
Nvidia	B-General_Concept
NVENC	I-General_Concept
SIP	B-Architecture
core	O
needs	O
to	O
be	O
supported	O
by	O
the	O
device	B-Application
driver	I-Application
.	O
</s>
<s>
The	O
driver	B-Application
provides	O
one	O
or	O
more	O
interfaces	B-Application
,	O
(	O
e.g.	O
</s>
<s>
OpenMAX	B-Application
IL	I-Application
)	O
to	O
NVENC	B-General_Concept
.	O
</s>
<s>
The	O
NVENC	B-General_Concept
SIP	B-Architecture
core	O
can	O
only	O
be	O
accessed	O
through	O
the	O
proprietary	B-Application
NVENC	B-General_Concept
API	B-Application
(	O
as	O
opposed	O
to	O
the	O
open-source	O
VDPAU	B-Application
API	B-Application
)	O
.	O
</s>
<s>
It	O
is	O
bundled	O
with	O
Nvidia	O
's	O
GeForce	B-Application
driver	I-Application
.	O
</s>
<s>
NVENC	B-General_Concept
is	O
available	O
for	O
Windows	O
and	O
Linux	B-Application
operating	I-Application
systems	I-Application
.	O
</s>
<s>
The	O
free	O
and	O
open-source	O
nouveau	O
device	B-Application
driver	I-Application
does	O
not	O
support	O
Nvidia	B-General_Concept
NVENC	I-General_Concept
.	O
</s>
