<s>
Volta	B-General_Concept
is	O
the	O
codename	O
for	O
a	O
GPU	B-Architecture
microarchitecture	B-General_Concept
developed	O
by	O
Nvidia	O
,	O
succeeding	O
Pascal	B-General_Concept
.	O
</s>
<s>
The	O
architecture	O
is	O
named	O
after	O
18th19th	O
century	O
Italian	O
chemist	O
and	O
physicist	O
Alessandro	O
Volta	B-General_Concept
.	O
</s>
<s>
It	O
was	O
NVIDIA	O
's	O
first	O
chip	O
to	O
feature	O
Tensor	O
Cores	O
,	O
specially	O
designed	O
cores	O
that	O
have	O
superior	O
deep	O
learning	O
performance	O
over	O
regular	O
CUDA	B-Architecture
cores	O
.	O
</s>
<s>
The	O
architecture	O
is	O
produced	O
with	O
TSMC	O
's	O
12	B-Algorithm
nm	I-Algorithm
FinFET	O
process	O
.	O
</s>
<s>
The	O
Ampere	B-General_Concept
microarchitecture	B-General_Concept
is	O
the	O
successor	O
to	O
Volta	B-General_Concept
.	O
</s>
<s>
The	O
first	O
graphics	O
card	O
to	O
use	O
it	O
was	O
the	O
datacenter	O
Tesla	B-Device
V100	O
,	O
e.g.	O
</s>
<s>
as	O
part	O
of	O
the	O
Nvidia	B-Operating_System
DGX-1	I-Operating_System
system	O
.	O
</s>
<s>
It	O
has	O
also	O
been	O
used	O
in	O
the	O
Quadro	O
GV100	O
and	O
Titan	O
V	O
.	O
There	O
were	O
no	O
mainstream	O
GeForce	O
graphics	O
cards	O
based	O
on	O
Volta	B-General_Concept
.	O
</s>
<s>
Architectural	O
improvements	O
of	O
the	O
Volta	B-General_Concept
architecture	O
include	O
the	O
following	O
:	O
</s>
<s>
NVLink	O
2.0	O
:	O
a	O
high-bandwidth	O
bus	B-General_Concept
between	O
the	O
CPU	O
and	O
GPU	B-Architecture
,	O
and	O
between	O
multiple	O
GPUs	B-Architecture
.	O
</s>
<s>
Tensor	O
cores	O
:	O
A	O
tensor	O
core	O
is	O
a	O
unit	O
that	O
multiplies	O
two	O
4×4	O
FP16	O
matrices	O
,	O
and	O
then	O
adds	O
a	O
third	O
FP16	O
or	O
FP32	O
matrix	B-Architecture
to	O
the	O
result	O
by	O
using	O
fused	O
multiply	O
–	O
add	O
operations	O
,	O
and	O
obtains	O
an	O
FP32	O
result	O
that	O
could	O
be	O
optionally	O
demoted	O
to	O
an	O
FP16	O
result	O
.	O
</s>
<s>
Volta	B-General_Concept
's	O
Tensor	O
cores	O
are	O
first	O
generation	O
while	O
Ampere	B-General_Concept
has	O
third	O
generation	O
Tensor	O
cores	O
.	O
</s>
<s>
Volta	B-General_Concept
has	O
been	O
announced	O
as	O
the	O
GPU	B-Architecture
microarchitecture	B-General_Concept
within	O
the	O
Xavier	O
generation	O
of	O
Tegra	B-Operating_System
SoC	B-Architecture
focusing	O
on	O
self-driving	O
cars	O
.	O
</s>
<s>
At	O
Nvidia	O
's	O
annual	O
GPU	B-Architecture
Technology	O
Conference	O
keynote	O
on	O
May	O
10	O
,	O
2017	O
,	O
Nvidia	O
officially	O
announced	O
the	O
Volta	B-General_Concept
microarchitecture	B-General_Concept
along	O
with	O
the	O
Tesla	B-Device
V100	O
.	O
</s>
<s>
The	O
Volta	B-General_Concept
GV100	O
GPU	B-Architecture
is	O
built	O
on	O
a	O
12nm	O
process	O
size	O
using	O
HBM2	O
memory	O
with	O
900GB/s	O
of	O
bandwidth	O
.	O
</s>
<s>
Volta	B-General_Concept
is	O
also	O
reported	O
to	O
be	O
included	O
in	O
the	O
Summit	B-Device
and	O
Sierra	B-Device
supercomputers	I-Device
,	O
used	O
for	O
GPGPU	O
compute	O
.	O
</s>
<s>
The	O
Volta	B-General_Concept
GPUs	B-Architecture
will	O
connect	O
to	O
the	O
POWER9	B-Device
CPUs	O
via	O
NVLink	O
2.0	O
,	O
which	O
is	O
expected	O
to	O
support	O
cache	B-General_Concept
coherency	I-General_Concept
and	O
therefore	O
improve	O
GPGPU	O
performance	O
.	O
</s>
