<s>
Ray-tracing	B-Algorithm
hardware	I-Algorithm
is	O
special-purpose	O
computer	B-Architecture
hardware	I-Architecture
designed	O
for	O
accelerating	B-General_Concept
ray	B-Algorithm
tracing	I-Algorithm
calculations	O
.	O
</s>
<s>
The	O
problem	O
of	O
rendering	O
3D	O
graphics	O
can	O
be	O
conceptually	O
presented	O
as	O
finding	O
all	O
intersections	O
between	O
a	O
set	O
of	O
"	O
primitives	B-Algorithm
"	O
(	O
typically	O
triangles	O
or	O
polygons	B-General_Concept
)	O
and	O
a	O
set	O
of	O
"	O
rays	O
"	O
(	O
typically	O
one	O
or	O
more	O
per	O
pixel	O
)	O
.	O
</s>
<s>
Up	O
to	O
2010	O
,	O
all	O
typical	O
graphic	O
acceleration	O
boards	O
,	O
called	O
graphics	B-Architecture
processing	I-Architecture
units	I-Architecture
(	O
GPUs	B-Architecture
)	O
,	O
used	O
rasterization	O
algorithms	O
.	O
</s>
<s>
The	O
ray	B-Algorithm
tracing	I-Algorithm
algorithm	O
solves	O
the	O
rendering	O
problem	O
in	O
a	O
different	O
way	O
.	O
</s>
<s>
In	O
each	O
step	O
,	O
it	O
finds	O
all	O
intersections	O
of	O
a	O
ray	O
with	O
a	O
set	O
of	O
relevant	O
primitives	B-Algorithm
of	O
the	O
scene	O
.	O
</s>
<s>
Rasterization	O
can	O
be	O
performed	O
using	O
devices	O
based	O
on	O
a	O
stream	B-Architecture
computing	I-Architecture
model	O
,	O
one	O
triangle	O
at	O
the	O
time	O
,	O
and	O
access	O
to	O
the	O
complete	O
scene	O
is	O
needed	O
only	O
once	O
.	O
</s>
<s>
The	O
ray	B-Algorithm
tracing	I-Algorithm
algorithm	O
is	O
inherently	O
suitable	O
for	O
scaling	O
by	O
parallelization	B-Operating_System
of	O
individual	O
ray	O
renders	O
.	O
</s>
<s>
However	O
anything	O
other	O
than	O
ray	B-Algorithm
casting	I-Algorithm
requires	O
recursion	O
of	O
the	O
ray	B-Algorithm
tracing	I-Algorithm
algorithm	O
(	O
and	O
random	O
access	O
to	O
the	O
scene	B-Data_Structure
graph	I-Data_Structure
)	O
to	O
complete	O
their	O
analysis	O
,	O
since	O
reflected	O
,	O
refracted	O
,	O
and	O
scattered	O
rays	O
require	O
that	O
various	O
parts	O
of	O
the	O
scene	O
be	O
re-accessed	O
in	O
a	O
way	O
not	O
easily	O
predicted	O
.	O
</s>
<s>
The	O
complexity	O
of	O
a	O
well	O
implemented	O
ray	B-Algorithm
tracing	I-Algorithm
algorithm	O
scales	O
logarithmically	O
;	O
this	O
is	O
due	O
to	O
objects	O
(	O
triangles	O
and	O
collections	O
of	O
triangles	O
)	O
being	O
placed	O
into	O
BSP	O
trees	O
or	O
similar	O
structures	O
,	O
and	O
only	O
being	O
analyzed	O
if	O
a	O
ray	O
intersects	O
with	O
the	O
bounding	B-Algorithm
volume	I-Algorithm
of	O
the	O
binary	O
space	O
partition	O
.	O
</s>
<s>
Various	O
implementations	O
of	O
ray	B-Algorithm
tracing	I-Algorithm
hardware	I-Algorithm
have	O
been	O
created	O
,	O
both	O
experimental	O
and	O
commercial	O
:	O
</s>
<s>
(	O
1996	O
)	O
Researchers	O
at	O
Princeton	O
university	O
proposed	O
using	O
DSPs	O
to	O
build	O
a	O
hardware	B-Architecture
unit	O
for	O
ray	B-Algorithm
tracing	I-Algorithm
acceleration	O
,	O
named	O
"	O
TigerSHARK	O
"	O
.	O
</s>
<s>
Implementations	O
of	O
volume	O
rendering	O
using	O
ray	B-Algorithm
tracing	I-Algorithm
algorithms	O
on	O
custom	O
hardware	B-Architecture
were	O
carried	O
out	O
in	O
1999	O
by	O
Hanspeter	O
Pfister	O
and	O
researchers	O
at	O
Mitsubishi	O
Electric	O
Research	O
Laboratories	O
.	O
</s>
<s>
(	O
2002	O
–	O
2009	O
)	O
ART	O
VPS	O
company	O
(	O
founded	O
2002	O
)	O
,	O
situated	O
in	O
the	O
UK	O
,	O
sold	O
ray	B-Algorithm
tracing	I-Algorithm
hardware	I-Algorithm
for	O
off-line	O
rendering	O
.	O
</s>
<s>
The	O
hardware	B-Architecture
used	O
multiple	O
specialized	O
processors	O
that	O
accelerated	O
ray-triangle	O
intersection	O
tests	O
.	O
</s>
<s>
Software	O
provided	O
integration	O
with	O
Autodesk	B-Language
Maya	I-Language
and	O
Max	B-Language
data	O
formats	O
,	O
and	O
utilized	O
the	O
Renderman	O
scene	O
description	O
language	O
for	O
sending	O
data	O
to	O
the	O
processors	O
(	O
the	O
.RIB	O
or	O
Renderman	O
Interface	O
Bytestream	O
file	O
format	O
)	O
.	O
</s>
<s>
As	O
of	O
2010	O
,	O
ARTVPS	O
no	O
longer	O
produces	O
ray	B-Algorithm
tracing	I-Algorithm
hardware	I-Algorithm
but	O
continues	O
to	O
produce	O
rendering	O
software	O
.	O
</s>
<s>
(	O
2009	O
–	O
2010	O
)	O
Intel	O
showcased	O
their	O
prototype	O
"	O
Larrabee	O
"	O
GPU	B-Architecture
and	O
Knights	O
Ferry	O
MIC	O
at	O
the	O
Intel	O
Developer	O
Forum	O
in	O
2009	O
with	O
a	O
demonstration	O
of	O
real-time	O
ray-tracing	B-Algorithm
.	O
</s>
<s>
Caustic	O
Graphics	O
produced	O
a	O
plug	O
in	O
card	O
,	O
the	O
"	O
CausticOne	O
"	O
(	O
2009	O
)	O
,	O
that	O
accelerated	O
global	B-Algorithm
illumination	I-Algorithm
and	O
other	O
ray	O
based	O
rendering	O
processes	O
when	O
coupled	O
to	O
a	O
PC	O
CPU	O
and	O
GPU	B-Architecture
.	O
</s>
<s>
The	O
hardware	B-Architecture
is	O
designed	O
to	O
organize	O
scattered	O
rays	O
(	O
typically	O
produced	O
by	O
global	B-Algorithm
illumination	I-Algorithm
problems	O
)	O
into	O
more	O
coherent	O
sets	O
(	O
lower	O
spatial	O
or	O
angular	O
spread	O
)	O
for	O
further	O
processing	O
by	O
an	O
external	O
processor	O
.	O
</s>
<s>
Siliconarts	O
developed	O
a	O
dedicated	O
real-time	O
ray	B-Algorithm
tracing	I-Algorithm
hardware	I-Algorithm
(	O
2010	O
)	O
.	O
</s>
<s>
RayCore	O
(	O
2011	O
)	O
,	O
which	O
is	O
the	O
world	O
's	O
first	O
real-time	O
ray	B-Algorithm
tracing	I-Algorithm
semiconductor	O
IP	O
,	O
was	O
announced	O
.	O
</s>
<s>
Imagination	O
Technologies	O
,	O
after	O
acquiring	O
Caustic	O
Graphics	O
,	O
produced	O
the	O
Caustic	O
Professional	O
's	O
R2500	O
and	O
R2100	O
plug	O
in	O
cards	O
containing	O
RT2	O
ray	B-Algorithm
trace	I-Algorithm
units	O
(	O
RTUs	O
)	O
.	O
</s>
<s>
Nvidia	O
,	O
partnering	O
with	O
Microsoft	B-Application
DirectX	I-Application
,	O
announced	O
the	O
Nvidia	O
RTX	O
developer	O
library	O
in	O
2018	O
,	O
which	O
promised	O
fast	O
real-time	O
ray	B-Algorithm
tracing	I-Algorithm
solutions	O
powered	O
by	O
hardware	B-General_Concept
accelerated	I-General_Concept
ray	B-Algorithm
tracing	I-Algorithm
(	O
ASIC	O
tensor	O
cores	O
)	O
found	O
in	O
the	O
Volta-generation	O
GPUs	B-Architecture
.	O
</s>
<s>
According	O
to	O
the	O
company	O
,	O
the	O
RDNA	O
2	O
micro-architecture	O
supports	O
real-time	O
hardware	B-General_Concept
accelerated	I-General_Concept
ray	B-Algorithm
tracing	I-Algorithm
.	O
</s>
<s>
Intel	O
released	O
Arc	O
Alchemist	O
GPU	B-Architecture
in	O
2022	O
,	O
in	O
which	O
the	O
GPU	B-Architecture
featured	O
ray	B-Algorithm
tracing	I-Algorithm
acceleration	O
core	O
that	O
are	O
performing	O
comparatively	O
with	O
RTX	B-Device
3000	I-Device
series	O
mid-range	O
GPU	B-Architecture
.	O
</s>
