<s>
The	O
Green500	B-Application
is	O
a	O
biannual	O
ranking	O
of	O
supercomputers	O
,	O
from	O
the	O
TOP500	B-Operating_System
list	B-Operating_System
of	I-Operating_System
supercomputers	I-Operating_System
,	O
in	O
terms	O
of	O
energy	O
efficiency	O
.	O
</s>
<s>
The	O
list	O
measures	O
performance	O
per	O
watt	O
using	O
the	O
TOP500	B-Operating_System
measure	O
of	O
high	O
performance	O
LINPACK	O
benchmarks	O
at	O
double-precision	O
floating-point	O
format	O
.	O
</s>
<s>
And	O
number	O
two	O
on	O
the	O
list	O
(	O
the	O
current	O
fastest	O
on	O
TOP500	B-Operating_System
)	O
is	O
also	O
over	O
50%	O
more	O
efficient	O
than	O
the	O
currently	O
most	O
efficient	O
(	O
and	O
much	O
smaller	O
)	O
Nvidia-based	O
system	O
.	O
</s>
<s>
No	O
large	O
Nvidia-based	O
system	O
make	O
the	O
top	O
10	O
positions	O
of	O
Graph500	O
(	O
smaller	O
ones	O
in	O
7th	O
to	O
10th	O
,	O
nor	O
any	O
longer	O
any	O
(	O
small	O
or	O
large	O
)	O
ARM-based	O
(	O
Fugaku	B-Device
was	O
at	O
the	O
top	O
of	O
the	O
list	O
in	O
June	O
2021	O
)	O
.	O
</s>
<s>
,	O
an	O
Appro	O
International	O
,	O
Inc	O
.	O
Xtreme-X	O
supercomputer	O
(	O
Beacon	O
)	O
topped	O
the	O
Green500	B-Application
list	O
with	O
2.499	O
LINPACK	O
GFLOPS/W	O
.	O
</s>
<s>
,	O
the	O
Eurotech	O
supercomputer	O
Eurora	O
at	O
Cineca	B-Device
topped	O
the	O
Green500	B-Application
list	O
with	O
3.208	O
LINPACK	O
GFLOPS/W	O
.	O
</s>
<s>
The	O
Cineca	B-Device
Eurora	O
supercomputer	O
is	O
equipped	O
with	O
two	O
Intel	O
Xeon	O
E5-2687W	O
CPUs	O
and	O
two	O
PCI-e	O
connected	O
NVIDIA	B-Device
Tesla	I-Device
K20	O
accelerators	O
per	O
node	O
.	O
</s>
<s>
,	O
the	O
L-CSC	O
supercomputer	O
of	O
the	O
Helmholtz	O
Association	O
at	O
the	O
GSI	O
in	O
Darmstadt	O
Germany	O
topped	O
the	O
Green500	B-Application
list	O
with	O
5.271GFLOPS/W	O
and	O
was	O
the	O
first	O
cluster	O
to	O
surpass	O
an	O
efficiency	O
of	O
5GFLOPS/W	O
.	O
</s>
<s>
It	O
runs	O
on	O
Intel	O
Xeon	O
E5-2690	O
Processors	O
with	O
the	O
Intel	B-Device
Ivy	I-Device
Bridge	I-Device
Architecture	O
and	O
AMD	O
FirePro	O
S9150	O
GPU	B-Architecture
Accelerators	O
.	O
</s>
<s>
,	O
the	O
Shoubu	O
supercomputer	O
of	O
RIKEN	O
outside	O
Tokyo	O
Japan	O
topped	O
the	O
Green500	B-Application
list	O
with	O
7.032GFLOPS/W	O
.	O
</s>
<s>
The	O
then-top	O
three	O
supercomputers	O
of	O
the	O
list	O
used	O
PEZY-SC	O
accelerators	O
(	O
GPU-like	O
that	O
use	O
OpenCL	B-Application
)	O
by	O
PEZY	B-General_Concept
Computing	I-General_Concept
with	O
1024	O
cores	O
each	O
and	O
6	O
–	O
7GFLOPS/W	O
efficiency	O
.	O
</s>
<s>
,	O
DGX	O
SaturnV	O
Volta	B-General_Concept
,	O
using	O
"	O
NVIDIA	O
DGX-1	O
Volta36	O
,	O
Xeon	O
E5-2698v4	O
20C	O
2.2GHz	O
,	O
Infiniband	O
EDR	O
,	O
NVIDIA	B-Device
Tesla	I-Device
V100	O
"	O
,	O
tops	O
Green500	B-Application
list	O
with	O
15.113GFLOPS/W	O
,	O
while	O
ranked	O
only	O
469th	O
on	O
Top500	B-Operating_System
.	O
</s>
<s>
It	O
's	O
only	O
a	O
little	O
bit	O
more	O
efficient	O
than	O
the	O
much	O
bigger	O
Summit	B-Device
ranked	O
2nd	O
while	O
1st	O
on	O
Top500	B-Operating_System
with	O
14.719GFLOPS/W	O
,	O
using	O
IBM	O
POWER9	B-Device
CPUs	O
while	O
also	O
with	O
Nvidia	B-Device
Tesla	I-Device
V100	O
GPUs	O
.	O
</s>
