<s>
Nvidia	B-Application
Optimus	I-Application
is	O
a	O
computer	O
GPU	B-Device
switching	I-Device
technology	O
created	O
by	O
Nvidia	O
which	O
,	O
depending	O
on	O
the	O
resource	O
load	O
generated	O
by	O
client	O
software	O
applications	O
,	O
will	O
seamlessly	O
switch	O
between	O
two	O
graphics	O
adapters	O
within	O
a	O
computer	O
system	O
in	O
order	O
to	O
provide	O
either	O
maximum	O
performance	O
or	O
minimum	O
power	O
draw	O
from	O
the	O
system	O
's	O
graphics	O
rendering	O
hardware	O
.	O
</s>
<s>
A	O
typical	O
platform	O
includes	O
both	O
a	O
lower-performance	O
integrated	O
graphics	B-Architecture
processor	I-Architecture
by	O
Intel/AMD	O
and	O
a	O
high-performance	O
one	O
by	O
Nvidia	O
.	O
</s>
<s>
Optimus	O
saves	O
battery	O
life	O
by	O
automatically	O
switching	O
the	O
power	O
of	O
the	O
discrete	O
graphics	B-Architecture
processing	I-Architecture
unit	I-Architecture
(	O
GPU	B-Architecture
)	O
off	O
when	O
it	O
is	O
not	O
needed	O
and	O
switching	O
it	O
on	O
when	O
needed	O
again	O
.	O
</s>
<s>
The	O
technology	O
mainly	O
targets	O
mobile	O
PCs	O
such	O
as	O
notebooks	B-Device
.	O
</s>
<s>
When	O
an	O
application	O
is	O
being	O
launched	O
that	O
is	O
determined	O
to	O
benefit	O
from	O
the	O
performance	O
of	O
the	O
discrete	O
GPU	B-Architecture
,	O
the	O
discrete	O
GPU	B-Architecture
is	O
powered	O
up	O
and	O
the	O
application	O
is	O
served	O
by	O
a	O
rendering	O
context	O
via	O
that	O
GPU	B-Architecture
.	O
</s>
<s>
Otherwise	O
the	O
application	O
is	O
served	O
by	O
a	O
rendering	O
context	O
that	O
uses	O
the	O
integrated	O
GPU	B-Architecture
.	O
</s>
<s>
Switching	O
between	O
the	O
graphics	B-Architecture
processors	I-Architecture
is	O
designed	O
to	O
be	O
completely	O
seamless	O
and	O
to	O
happen	O
"	O
behind	O
the	O
scenes	O
"	O
.	O
</s>
<s>
Official	O
supported	O
operating	O
systems	O
by	O
Nvidia	O
are	O
Microsoft	B-Application
Windows	I-Application
and	O
Linux	B-Operating_System
.	O
</s>
<s>
A	O
project	O
called	O
Bumblebee	O
is	O
an	O
alternative	O
open	B-License
source	I-License
implementation	O
of	O
Optimus	O
support	O
for	O
Linux	B-Operating_System
.	O
</s>
<s>
When	O
a	O
user	O
launches	O
an	O
application	O
,	O
the	O
graphics	O
driver	B-Application
tries	O
to	O
determine	O
whether	O
the	O
application	O
would	O
benefit	O
from	O
the	O
discrete	O
GPU	B-Architecture
.	O
</s>
<s>
If	O
so	O
,	O
the	O
GPU	B-Architecture
is	O
powered	O
up	O
from	O
an	O
idle	B-Algorithm
state	O
and	O
is	O
passed	O
all	O
rendering	O
calls	O
.	O
</s>
<s>
Even	O
in	O
this	O
case	O
,	O
though	O
,	O
the	O
integrated	O
graphics	B-Architecture
processor	I-Architecture
(	O
IGP	O
)	O
is	O
used	O
to	O
output	O
the	O
final	O
image	O
.	O
</s>
<s>
Under	O
Windows	O
the	O
Nvidia	O
driver	B-Application
also	O
provides	O
the	O
option	O
to	O
manually	O
select	O
the	O
GPU	B-Architecture
in	O
the	O
right-click	O
menu	O
upon	O
launching	O
an	O
executable	O
.	O
</s>
<s>
Within	O
the	O
hardware	O
interface	O
layer	O
of	O
the	O
Nvidia	O
GPU	B-Architecture
driver	B-Application
,	O
the	O
Optimus	O
Routing	O
Layer	O
provides	O
intelligent	O
graphics	O
management	O
.	O
</s>
<s>
The	O
Optimus	O
Routing	O
Layer	O
also	O
includes	O
a	O
kernel-level	O
library	O
for	O
recognizing	O
and	O
managing	O
specific	O
classes	O
and	O
objects	O
associated	O
with	O
different	O
graphics	O
devices	O
.	O
</s>
<s>
This	O
Nvidia	O
innovation	O
performs	O
state	O
and	O
context	O
management	O
,	O
allocating	O
architectural	O
resources	O
as	O
needed	O
for	O
each	O
driver	B-Application
client	O
(	O
i.e.	O
,	O
application	O
)	O
.	O
</s>
<s>
In	O
this	O
context-management	O
scheme	O
,	O
each	O
application	O
is	O
not	O
aware	O
of	O
other	O
applications	O
concurrently	O
using	O
the	O
GPU	B-Architecture
.	O
</s>
<s>
By	O
recognizing	O
designated	O
classes	O
,	O
the	O
Optimus	O
Routing	O
Layer	O
can	O
help	O
determine	O
when	O
the	O
GPU	B-Architecture
can	O
be	O
utilized	O
to	O
improve	O
rendering	O
performance	O
.	O
</s>
<s>
Specifically	O
,	O
it	O
sends	O
a	O
signal	O
to	O
power-on	O
the	O
GPU	B-Architecture
when	O
it	O
finds	O
any	O
of	O
the	O
following	O
three	O
call	O
types	O
:	O
</s>
<s>
Optimus	O
avoids	O
usage	O
of	O
a	O
hardware	O
multiplexer	B-Protocol
and	O
prevents	O
glitches	O
associated	O
with	O
changing	O
the	O
display	O
driver	B-Application
from	O
IGP	O
to	O
GPU	B-Architecture
by	O
transferring	O
the	O
display	O
surface	O
from	O
the	O
GPU	B-Architecture
frame	B-Algorithm
buffer	I-Algorithm
over	O
the	O
PCI	O
Express	O
bus	O
to	O
the	O
main	O
memory-based	O
framebuffer	B-Algorithm
used	O
by	O
the	O
IGP	O
.	O
</s>
<s>
The	O
Optimus	O
Copy	O
Engine	O
is	O
a	O
new	O
alternative	O
to	O
traditional	O
DMA	B-General_Concept
transfers	O
between	O
the	O
GPU	B-Architecture
framebuffer	B-Algorithm
memory	O
and	O
main	O
memory	O
used	O
by	O
the	O
IGP	O
.	O
</s>
<s>
Using	O
the	O
IGP	O
to	O
output	O
data	O
from	O
the	O
discrete	O
GPU	B-Architecture
may	O
become	O
a	O
bottleneck	O
at	O
high	O
framerates	O
.	O
</s>
<s>
On	O
desktop-replacement	O
or	O
gaming	O
laptops	B-Device
(	O
where	O
performance	O
is	O
prioritized	O
over	O
energy	O
efficiency	O
)	O
,	O
Optimus	O
technology	O
has	O
been	O
heavily	O
criticized	O
.	O
</s>
<s>
Some	O
vendors	O
have	O
provided	O
hardware	O
multiplexer	B-Protocol
solutions	O
to	O
completely	O
shut	O
off	O
the	O
Optimus	O
architecture	O
.	O
</s>
<s>
The	O
operation	O
may	O
be	O
hosted	O
in	O
the	O
BIOS	B-Operating_System
,	O
and	O
often	O
require	O
a	O
reboot	O
.	O
</s>
<s>
The	O
binary	O
Nvidia	O
driver	B-Application
added	O
partial	O
Optimus	O
support	O
May	O
3	O
,	O
2013	O
in	O
the	O
319.17	O
.	O
</s>
<s>
The	O
open-source	B-License
project	O
tries	O
to	O
provide	O
support	O
for	O
graphics-chip	O
switching	O
.	O
</s>
<s>
As	O
in	O
the	O
Windows	O
implementation	O
,	O
by	O
default	O
all	O
applications	O
run	O
through	O
the	O
integrated	O
graphics	B-Architecture
processor	I-Architecture
.	O
</s>
<s>
one	O
can	O
only	O
run	O
a	O
program	O
with	O
improved	O
graphical	O
performance	O
on	O
the	O
discrete	O
GPU	B-Architecture
by	O
explicitly	O
invoking	O
it	O
as	O
such	O
:	O
for	O
example	O
,	O
by	O
using	O
the	O
command	B-Application
line	I-Application
or	O
through	O
specially	O
configured	O
shortcut	O
icon	O
.	O
</s>
<s>
Automatic	O
detection	O
and	O
switching	O
between	O
graphics	B-Architecture
processors	I-Architecture
is	O
not	O
yet	O
available	O
.	O
</s>
<s>
Work	O
in	O
progress	O
on	O
a	O
graphical	B-Application
interface	I-Application
-	O
bumblebee-ui	O
-	O
aims	O
to	O
allow	O
more	O
convenient	O
starting	O
of	O
programs	O
for	O
improved	O
graphical	O
performance	O
when	O
necessary	O
.	O
</s>
<s>
Steam	B-Application
for	O
Linux	B-Operating_System
can	O
be	O
set	O
up	O
to	O
run	O
games	O
using	O
the	O
discrete	O
GPU	B-Architecture
(	O
)	O
.	O
</s>
<s>
The	O
Bumblebee	O
Project	O
continues	O
to	O
evolve	O
as	O
more	O
necessary	O
software	O
changes	O
are	O
made	O
to	O
the	O
graphics	O
architecture	O
of	O
Linux	B-Operating_System
.	O
</s>
<s>
To	O
make	O
most	O
use	O
of	O
it	O
,	O
it	O
is	O
best	O
to	O
use	O
a	O
recent	O
Linux	B-Operating_System
distribution	O
.	O
</s>
<s>
,	O
Bumblebee	O
software	O
repositories	O
are	O
available	O
for	O
Arch	O
Linux	B-Operating_System
,	O
Debian	O
,	O
Fedora	O
,	O
Gentoo	B-Application
,	O
Mandriva	O
,	O
OpenSuSE	B-Operating_System
and	O
Ubuntu	B-Operating_System
.	O
</s>
<s>
An	O
attempt	O
by	O
Nvidia	O
to	O
support	O
Optimus	O
through	O
DMA	B-General_Concept
BUF	O
,	O
a	O
Linux	O
kernel-mechanism	O
for	O
sharing	O
buffers	B-General_Concept
across	O
hardware	O
(	O
potentially	O
GPUs	B-Architecture
)	O
,	O
was	O
rebuffed	O
by	O
kernel	B-Operating_System
developers	O
in	O
January	O
2012	O
due	O
to	O
license	O
incompatibility	O
between	O
the	O
GPL-licensed	O
kernel-code	O
and	O
the	O
proprietary-licensed	O
Nvidia	O
blob	O
.	O
</s>
<s>
When	O
no	O
software	O
mechanism	O
exists	O
for	O
switching	O
between	O
graphics	O
adapters	O
,	O
the	O
system	O
cannot	O
use	O
the	O
Nvidia	O
GPU	B-Architecture
at	O
all	O
,	O
even	O
if	O
an	O
installed	O
graphics	O
driver	B-Application
would	O
support	O
it	O
.	O
</s>
<s>
Many	O
linux	B-Operating_System
distributions	O
now	O
support	O
Nvidia	O
offloading	O
,	O
where	O
the	O
nvidia	O
card	O
does	O
all	O
rendering	O
.	O
</s>
<s>
Since	O
the	O
internal	O
laptop	B-Device
display	O
is	O
physically	O
connected	O
to	O
the	O
intel	O
driver	B-Application
,	O
the	O
nvidia	O
card	O
renders	O
to	O
the	O
intel	O
display	B-Algorithm
memory	I-Algorithm
.	O
</s>
<s>
To	O
avoid	O
tearing	O
,	O
the	O
xorg	O
server	O
has	O
a	O
mechanism	O
called	O
Prime	O
Synchronization	O
to	O
time	O
these	O
buffer	B-General_Concept
updates	O
to	O
avoid	O
tearing	O
,	O
similar	O
to	O
vsync	O
;	O
the	O
nvidia	O
driver	B-Application
must	O
be	O
loaded	O
as	O
a	O
kernel	B-Operating_System
module	O
for	O
this	O
to	O
work	O
.	O
</s>
<s>
The	O
leading	O
implementation	O
of	O
this	O
approach	O
is	O
Ubuntu	B-Operating_System
's	O
'	O
prime-select	O
'	O
package	O
,	O
which	O
has	O
a	O
command	B-Application
line	I-Application
and	O
graphical	B-Application
tool	I-Application
to	O
turn	O
the	O
nvidia	O
card	O
off	O
.	O
</s>
<s>
Ubuntu	B-Operating_System
's	O
prime-select	O
script	O
is	O
available	O
on	O
Ubuntu	B-Operating_System
derivatives	O
,	O
which	O
in	O
some	O
cases	O
add	O
their	O
own	O
graphical	B-Application
tools	I-Application
.	O
</s>
<s>
Some	O
older	O
and	O
high-end	O
laptops	B-Device
contain	O
a	O
BIOS	B-Operating_System
setting	O
to	O
manually	O
select	O
the	O
state	O
of	O
the	O
hardware	O
multiplexer	B-Protocol
to	O
switch	O
output	O
between	O
the	O
two	O
video	O
devices	O
.	O
</s>
<s>
In	O
this	O
case	O
,	O
a	O
Linux	B-Operating_System
user	O
can	O
place	O
the	O
laptop	B-Device
in	O
hardware	O
configurations	O
where	O
there	O
is	O
only	O
once	O
graphics	O
device	O
.	O
</s>
<s>
Since	O
driver	B-Application
version	O
435	O
the	O
proprietary	O
driver	B-Application
supports	O
render	O
offloading	O
of	O
a	O
single	O
window	O
.	O
</s>
