CSC 456 Spring 2012/ch1 BC: Difference between revisions

From Expertiza_Wiki
Jump to navigation Jump to search
No edit summary
No edit summary
 
(15 intermediate revisions by 2 users not shown)
Line 1: Line 1:
From 2006-2012 the increase in the number of transistors on a chip has grown from 167 million to 2.6 billion, a 15x increase.
==Processor Improvement==


From 2006-2012 the clock frequency has increased from 2.4ghz to 5.2, a 2.2x increase.
From page 2 of Solihin, the change in transistor counts from 1971-2006 is mentioned to have increased from 2,300 to 167 million, a 72,608x increase. From 2006-2012 the increase in the number of transistors on a chip has grown from 167 million to 2.6 billion, a 15x increase.<ref name="trans count"/>


IBM now has the 16-core processor Power PC A2, Intel has the 10 core Xeon E7, AMD has the 16 Opteron Interlagos, and Sun has the 8-core Niagara.
Also from page 2, it is mentioned that the clock frequency of processors grew from 750 KHz to 2.4 GHz between 1971-2006, a 3,200x increase. From 2006-2012 the clock frequency has increased from 2.4ghz to 5.2, a 2.2x increase.<ref name="proc chrono"/>


The following facts have been updated from the original list on page 4 of Solihin:
*IBM now has the 16-core processor Power PC A2
*Intel has the 10 core Xeon E7
*AMD has the 16 Opteron Interlagos
*Sun has the 8-core Niagara.<ref name="xeon"/><ref name="powerpc"/><ref name="amd"/><ref name="sun"/>
==Tables==
The following table shows the improvement in Intel processors since 1971. The bold items are Intel processors that have come out since the book was written. Original table can be found on page 4 of the book.
{| class="wikitable"
{| class="wikitable"
|-
|-
!Evolution of Intel Processors
!Evolution of Intel Processors<ref name="intel procs"/>
|-
|-
!From
!From
Line 16: Line 27:
|1971
|1971
|4004
|4004
|740 KHz, 2300 transistors, 10 micrometers, 640B addressable memory, 4 KB program memory
|740KHz, 2300 transistors, 10 micrometers, 640B addressable memory, 4KB program memory
|
|
|-
|-
|1978
|1978
|8086
|8086
|16 bit, 5-10 MHz, 29000 transistors at 3 micrometers, 1MB addressable memory
|16-bit, 5-10MHz, 29000 transistors at 3 micrometers, 1MB addressable memory
|
|
|-
|-
|1982
|1982
|80286
|80286
|8-12.5 MHz
|8-12.5MHz
|Virtual memory and protection mode
|Virtual memory and protection mode
|-
|-
|1985
|1985
|386
|386
|32 bit, 16-33MHz, 275K transistors, 4 GB addressable memory
|32-bit, 16-33MHz, 275K transistors, 4GB addressable memory
|Pipelining
|Pipelining
|-
|-
Line 41: Line 52:
|1993
|1993
|Pentium
|Pentium
|60-200 MHz
|60-200MHz
|On-chip L1 caches and SMP suport
|On-chip L1 caches and SMP suport
|-
|-
Line 51: Line 62:
|1997
|1997
|Pentium MMX
|Pentium MMX
|233-450 MHz, 32KB L1 cache, 4.5M transistors
|233-450MHz, 32KB L1 cache, 4.5M transistors
|Dynamic branch prediction, MMX instruction sets
|Dynamic branch prediction, MMX instruction sets
|-
|-
|199
|1999
|Pentium III
|Pentium III
|450-1400MHz, 256KB L2 cache on chip, 28M transistors
|450-1400MHz, 256KB L2 cache on chip, 28M transistors
Line 61: Line 72:
|2000
|2000
|Pentium IV
|Pentium IV
|1.4-3 GHz, 55M transistors
|1.4-3GHz, 55M transistors
|Hyperpipelining and SMT
|Hyperpipelining and SMT
|
|
Line 67: Line 78:
|2006
|2006
|Xeon
|Xeon
|64 bit, 2 GHz, 167M transistors, 4MB L2 cache on chip
|64-bit, 2GHz, 167M transistors, 4MB L2 cache on chip
|Dual-core and virtualization support
|Dual-core and virtualization support
|-
|-
|2008
|'''2008'''
|Intel Core i7
|'''Intel Core i7'''
|64bit, 3.2GHz, 730m transistors, 4 core
|'''64-bit, 3.2GHz, 730M transistors, 4 core'''
|
|
|-
|-
|2010
|'''2010'''
|Intel Xeon "Nehalem-EX"
|'''Intel Xeon "Nehalem-EX"'''
|64bit, 2.66ghz, 2300m transistors, 8 core
|'''64-bit, 2.66GHz, 2300M transistors, 8 core'''
|
|
|-
|-
|2011
|'''2011'''
|Intel Xeon E7
|'''Intel Xeon E7'''
|64bit, 2.67ghz, 2600m transistors, 10 core
|'''64-bit, 2.67GHz, 2600M transistors, 10 core'''
|first Intel chip with 10 processors
|'''First Intel chip with 10 processors'''
|}
 
 
 
The following table is a revised version of the one on page 8 of Solihin listing some examples of current high end multicore processors.
{| class="wikitable"
|-
!Examples of Current Multicore Processors<ref name="z196"/><ref name="xeon"/>
|-
!Name
!# Cores
!Clock Freq
!Clock Type
!Caches
!Chip Power
|-
|IBM z196
|4 cores
|5.3GHz
|OOO Superscalar
|128KB L1, 1.5MB L2, 24MB L3, 192MB L4
|1800W
|-
|Intel Xeon E
|10 cores
|2.67GHz
|SIMD
|64KB L1, 256KB L2, 30MB L3
|130W
|}
|}


==Top Trends==
===Super Computers===
The top rated super computer for November 2011 is Japan's "K Computer", rated by top500.org.<ref name="top2011"/> The K Computer operates at 10.51 petaflop/s with 705,024 cores and is the first to break the 10 petaflop/s barrier. Just one year earlier the highest rated supercomputer was China's Tianhe-1A at 2.57 petaflop/s, 1/4 as fast as the K Computer.
This is the second top500 list in a row that the K Computer stayed in first place. In fact, none of the top 10 changed, which is mentioned here by TOP500 editor Erich Strohmaier, "This is the first time since we began publishing the list back in 1993 that the top 10 systems showed no turnover".<ref name="top2011"/> This shows that the improvement in potential processing power has started to slow down. This observation combined with the reduction in transistor count and clock frequency growth described earlier would suggest that we are approaching the "wall" of processing capabilities.
===Cluster Computing===
Since June 2006, the trending supercomputer architecture has gone towards cluster computing. Cluster architecture usage has gone from 72% in 2006 to 82% in 2011, while Constellation has gone from 7.6% to 0%.<ref name="topstats"/>


z196
==References==
4 cores
5.3ghz
OOO Superscalar
128kb L1
1.5mb L2
24mb L3
192mb L4
1800 watts


Xeon E
<references>
10 cores
<ref name="trans count">http://en.wikipedia.org/wiki/Transistor_count </ref>
2.67ghz
<ref name="proc chrono">http://en.wikipedia.org/wiki/Microprocessor_chronologyref </ref>
SIMD
<ref name="intel procs">http://en.wikipedia.org/wiki/List_of_Intel_microprocessors </ref>
64kb l1
<ref name="z196">http://en.wikipedia.org/wiki/IBM_z196_(microprocessor) </ref>
256kb l2
<ref name="xeon">http://en.wikipedia.org/wiki/Nehalem_(microarchitecture)#Westmere </ref>
30mbl3
<ref name="powerpc">http://en.wikipedia.org/wiki/PowerPC_A2 </ref>
130W
<ref name="amd">http://www.tomshardware.com/news/interlagos-bulldozer-opteron-16-core-valencia,13984.html </ref>
<ref name="sun">http://en.wikipedia.org/wiki/UltraSPARC_T1 </ref>
<ref name="top2011">http://top500.org/lists/2011/11</ref>
<ref name="topstats">http://i.top500.org/stats</ref>
</references>

Latest revision as of 18:58, 20 February 2012

Processor Improvement

From page 2 of Solihin, the change in transistor counts from 1971-2006 is mentioned to have increased from 2,300 to 167 million, a 72,608x increase. From 2006-2012 the increase in the number of transistors on a chip has grown from 167 million to 2.6 billion, a 15x increase.<ref name="trans count"/>

Also from page 2, it is mentioned that the clock frequency of processors grew from 750 KHz to 2.4 GHz between 1971-2006, a 3,200x increase. From 2006-2012 the clock frequency has increased from 2.4ghz to 5.2, a 2.2x increase.<ref name="proc chrono"/>


The following facts have been updated from the original list on page 4 of Solihin:

  • IBM now has the 16-core processor Power PC A2
  • Intel has the 10 core Xeon E7
  • AMD has the 16 Opteron Interlagos
  • Sun has the 8-core Niagara.<ref name="xeon"/><ref name="powerpc"/><ref name="amd"/><ref name="sun"/>

Tables

The following table shows the improvement in Intel processors since 1971. The bold items are Intel processors that have come out since the book was written. Original table can be found on page 4 of the book.

Evolution of Intel Processors<ref name="intel procs"/>
From Procs Specifications New Features
1971 4004 740KHz, 2300 transistors, 10 micrometers, 640B addressable memory, 4KB program memory
1978 8086 16-bit, 5-10MHz, 29000 transistors at 3 micrometers, 1MB addressable memory
1982 80286 8-12.5MHz Virtual memory and protection mode
1985 386 32-bit, 16-33MHz, 275K transistors, 4GB addressable memory Pipelining
1989 486 25-100MHz, 1.5M transistors FPU integration
1993 Pentium 60-200MHz On-chip L1 caches and SMP suport
1995 Pentium Pro 16KB L1 caches, 5.5M transistors OOO execution
1997 Pentium MMX 233-450MHz, 32KB L1 cache, 4.5M transistors Dynamic branch prediction, MMX instruction sets
1999 Pentium III 450-1400MHz, 256KB L2 cache on chip, 28M transistors SSE instruction sets
2000 Pentium IV 1.4-3GHz, 55M transistors Hyperpipelining and SMT
2006 Xeon 64-bit, 2GHz, 167M transistors, 4MB L2 cache on chip Dual-core and virtualization support
2008 Intel Core i7 64-bit, 3.2GHz, 730M transistors, 4 core
2010 Intel Xeon "Nehalem-EX" 64-bit, 2.66GHz, 2300M transistors, 8 core
2011 Intel Xeon E7 64-bit, 2.67GHz, 2600M transistors, 10 core First Intel chip with 10 processors


The following table is a revised version of the one on page 8 of Solihin listing some examples of current high end multicore processors.

Examples of Current Multicore Processors<ref name="z196"/><ref name="xeon"/>
Name # Cores Clock Freq Clock Type Caches Chip Power
IBM z196 4 cores 5.3GHz OOO Superscalar 128KB L1, 1.5MB L2, 24MB L3, 192MB L4 1800W
Intel Xeon E 10 cores 2.67GHz SIMD 64KB L1, 256KB L2, 30MB L3 130W

Top Trends

Super Computers

The top rated super computer for November 2011 is Japan's "K Computer", rated by top500.org.<ref name="top2011"/> The K Computer operates at 10.51 petaflop/s with 705,024 cores and is the first to break the 10 petaflop/s barrier. Just one year earlier the highest rated supercomputer was China's Tianhe-1A at 2.57 petaflop/s, 1/4 as fast as the K Computer.

This is the second top500 list in a row that the K Computer stayed in first place. In fact, none of the top 10 changed, which is mentioned here by TOP500 editor Erich Strohmaier, "This is the first time since we began publishing the list back in 1993 that the top 10 systems showed no turnover".<ref name="top2011"/> This shows that the improvement in potential processing power has started to slow down. This observation combined with the reduction in transistor count and clock frequency growth described earlier would suggest that we are approaching the "wall" of processing capabilities.

Cluster Computing

Since June 2006, the trending supercomputer architecture has gone towards cluster computing. Cluster architecture usage has gone from 72% in 2006 to 82% in 2011, while Constellation has gone from 7.6% to 0%.<ref name="topstats"/>

References

<references> <ref name="trans count">http://en.wikipedia.org/wiki/Transistor_count </ref> <ref name="proc chrono">http://en.wikipedia.org/wiki/Microprocessor_chronologyref </ref> <ref name="intel procs">http://en.wikipedia.org/wiki/List_of_Intel_microprocessors </ref> <ref name="z196">http://en.wikipedia.org/wiki/IBM_z196_(microprocessor) </ref> <ref name="xeon">http://en.wikipedia.org/wiki/Nehalem_(microarchitecture)#Westmere </ref> <ref name="powerpc">http://en.wikipedia.org/wiki/PowerPC_A2 </ref> <ref name="amd">http://www.tomshardware.com/news/interlagos-bulldozer-opteron-16-core-valencia,13984.html </ref> <ref name="sun">http://en.wikipedia.org/wiki/UltraSPARC_T1 </ref> <ref name="top2011">http://top500.org/lists/2011/11</ref> <ref name="topstats">http://i.top500.org/stats</ref> </references>