These two layers are separated by a magnetic junction
tunnel (MTJ) as a carrier. If they are not parallel, MTJ
resistivity is high and used to code “1”; if they are parallel,
MTJ incurs low resistivity which is coded as “0”. The cell is
accessed by an NMOS transistor that is connected to the wordline
and
bit-line
(Figure
1).
As
the
resistance
difference
between
these
two
states is large,
applying
a small
voltage
between
source
and bit
lines
followed
by
sensing
conducted
current
leads
to a fast
and low-energy
read
operation.
To
write,
the
magnetic
orientation
of the
free layer
needs being
altered
by
conducting
a large
and directional
current.
Consequently,
write
operation
takes
long
time
and consumes
large
energy
that
is
source
of STT-RAM
inefficiency.
Besides,
an STT-RAM cell
lifetime
is
typically
limited
by
10
13
cycles. Fortunately, no
current or refresh cycle is required to hold data in the free layer
which incurs no intrinsic leakage. Therefore regarding access
times, the access latency of random writes on STT-RAM is
slower than SRAM, although read access latency is
comparable. Besides, STT-RAM is a promising candidate for
energy saving cache design, if some techniques are used to
mitigate write inefficiencies.
These two layers are separated by a magnetic junctiontunnel (MTJ) as a carrier. If they are not parallel, MTJ resistivity is high and used to code “1”; if they are parallel, MTJ incurs low resistivity which is coded as “0”. The cell isaccessed by an NMOS transistor that is connected to the wordlineandbit-line(Figure1).Astheresistancedifferencebetweenthesetwostates is large,applyinga smallvoltagebetweensourceand bitlinesfollowedbysensingconductedcurrentleadsto a fastand low-energyreadoperation.Towrite,themagneticorientationof thefree layerneeds beingalteredbyconductinga largeand directionalcurrent.Consequently,writeoperationtakeslongtimeand consumeslargeenergythatissourceof STT-RAMinefficiency.Besides,an STT-RAM celllifetimeistypicallylimitedby1013 cycles. Fortunately, nocurrent or refresh cycle is required to hold data in the free layerwhich incurs no intrinsic leakage. Therefore regarding accesstimes, the access latency of random writes on STT-RAM isslower than SRAM, although read access latency iscomparable. Besides, STT-RAM is a promising candidate forenergy saving cache design, if some techniques are used tomitigate write inefficiencies.
การแปล กรุณารอสักครู่..
