Upload
yatish-bathla
View
217
Download
0
Embed Size (px)
Citation preview
7/25/2019 Future Computing and Information Processing
1/13
Future Directions in
Electronic Computing andInformation Processing
Submitted by:
Yatish Bathla
Hameh !halili
7/25/2019 Future Computing and Information Processing
2/13
Why we need this Step?
Early Failure of Moores Law
Negative Result:
o
IBM has recently given up its PC market.o doubts about the "alidity of the #la$% cannegati"ely influence share prices ofProcessor firms
some eotic new technologies such as nanoelectronics or !uantum computing woul" #ea#le to save us from this slow"own
7/25/2019 Future Computing and Information Processing
3/13
Moores Law
&he la$ is named after Intel co'founder(ordon E) *oore
+umber of transistors that can be placed
ine,pensi"ely on an integrated circuitdoubles appro,imately e"ery t$o years &he capabilities of many digital electronic
de"ices are strongly lin-ed to *oore.s la$: processing speed memory capacity sensors number and sie of pi,els in digital cameras
7/25/2019 Future Computing and Information Processing
4/13
7/25/2019 Future Computing and Information Processing
5/13
Moores Law
e,ponential impro"ement has dramatically
enhanced the impact of digital electronics in
nearly e"ery segment of the $orld economy
*oore.s la$ describes a dri"ing force of
technological and social change in the late
41th and early 45st centuries
Moores second law: Capital cost of asemiconductor fabrication also increases
e,ponentially o"er time
7/25/2019 Future Computing and Information Processing
6/13
Failure Of Moores Law
Failure too- place much earlier in 4116 $henintel7s failed to mo"e from 31 nm to 81 nm
Reason: Decreasing bits'per'9oule energy efficiency
due to the lea-age current synergic effects of the increasing thermal
noise heat dissipation and band$idth duringminiaturiation $ould manifest themsel"es ineither a high bit'error rate or chip o"erheating
7/25/2019 Future Computing and Information Processing
7/13
Potential ways of improvements
y classical information
Parallel Processing it cannot sa"e *oore7s la$ because the la$ is
meant for a single chip) E"en if $e disregard
*oore7s la$ $e are facing serious problems $iththe more than 511 ; of today7s microprocessors)
Parallel processing is useful but the help it can
offer is limited by the total energy dissipation of
the computer)
7/25/2019 Future Computing and Information Processing
8/13
Sin!le"#lectron $echnolo!y
7/25/2019 Future Computing and Information Processing
9/13
%uantum &omputin!
De"ice for computation that ma-es direct use of0uantum mechanical phenomena such as
superposition and entanglement to perform
operations on data
?se 0ubits and represent the state of an n'0ubit
system on a classical computer $ould re0uire
the storage of 4ncomple, coefficients
=arge'scale 0uantum computers could be ableto sol"e certain problems much faster than any
classical computer
7/25/2019 Future Computing and Information Processing
10/13
'dvanta!e: %uantum
&omputin!
@ Auantum Fourier transform algorithm to find the period of afunction that is known in ad"ance to be periodice,ponentially faster than $ith a classical algorithm
@ Shor.s algorithm or the simulation of 0uantum many'body
systems calculate almost 41 times faster integerfactoriation as Classical computers
@ pro"ide a polynomial speedup o"er a classical algorithme,ample: #0uantum search% algorithm disco"ered) &his$ould allo$ one to locate a particular
@ item in a #database% of + entries $ith only of the order ofunder root + 0ueries as opposed to the typical order of0ueries +4 one e,pects $hen dealing $ith classicalcomputers
7/25/2019 Future Computing and Information Processing
11/13
(isadvanta!e: %uantum
&omputin!
De'coherence: $hen a measurement of any type ismade to a 0uantum system decoherence brea-sdo$n and the $a"e function collapses into a single
state 0ubits are not digital bits of data thus they cannotuse as con"entional error correction
#0uantum CP?% $ill ha"e efficiency and heating
problems of their o$n minimum energy re0uirement for the 0uantumlogical operations is fi"e times than classicalcomputer
7/25/2019 Future Computing and Information Processing
12/13
&onclusion
&he e,ponential e"olution of hard$are performance
has ended
space for e"olution is basically in the potential
impro"ement of the efficiency of the soft$are Currently there is no ne$ technology on the horion
to impro"e this efficiency)
In Future near'to'distant future more sophisticated
custom'made parallel'processing clusters as $ell
as con"entional analog optical and 0uantum
computers $ill arise
7/25/2019 Future Computing and Information Processing
13/13
)racious