Integrovany radic je pre NUMA klucovy a teda pre skalovanie
Barcelona je serve chip pre nominlane 8 CPU systemy
Most VMMs can pin processes to a CPU an then to local memory, so with this box you get about 100 per cent scaling from two to four CPUs, and about 350 per cent from two to eight.
(url)http://www.theinquirer.net/?article=4...
ppvedane inak 8 Barcelona ma vykon 4,5 nasobku vykonu dvoch Barcelon
4 barcelony maju vykon 2x taky ako dve Barcelony
a preto sa pouizva v superpocitacoch
Supercomputer users were among Opteron's earliest adopters, and that is one piece of Opteron history that may be repeated with release of the quad-core Barcelona. What may well become the world's largest supercomputer is now being built at the Texas Advanced Computing Center with Barcelona chips.
More than 15,000 Barcelona processors will be used in this system, capable of computer power in excess of 500 teraflops. A teraflop is 1 trillion floating-point operations per second.
Tommy Minyard, assistant director of the Austin-based computing center, said that when the US$30 million system is finished early next year, it will have 63,000 cores in total. This supercomputer, called "Ranger," is using Sun Microsystems Inc.'s Sun Blade 6000 servers
http://www.pcworld.com/article/id,137025-c,amd/article.html
Cray XT4(TM) Systems With New Quad-Core AMD Opteron(TM) Processor to Provide Petaflops Capability
Global supercomputer leader Cray Inc. (NASDAQ: CRAY) today reported that it expects to begin shipping its industry-leading Cray XT4(TM) systems with new, ultra-powerful Quad-Core AMD Opteron(TM) processors by year-end.
The Cray XT4 system is a massively parallel processor (MPP) system that provides unrivaled sustained performance by balancing performance between powerful processors, memory, interconnect, and IO. Both hardware and software are designed to scale to more than 120,000 processor cores and over one petaflops of peak performance. Both the Cray XT3 and Cray XT4 systems can be upgraded to Cray XT4 systems with quad-core support.
"We upgraded from single-core to dual-core in the field very smoothly, and we anticipate being able to do the same with the Cray XT4 system with quad-core support," said Buddy Bland, Project Director of the Leadership Computing Facility at Oak Ridge National Laboratory. "We are already running machines with 23,000 processor cores, and we have high expectations about our impending quad-core delivery."
http://investors.cray.com/phoenix.zhtml?c=98390&p=irol-newsArticle&ID=1049549&highlight=
zoznam najvykonnejsich superpocitacov sveta v mesiaci jun 2007
Rank Site Computer Processors Year Rmax Rpeak
1 DOE/NNSA/LLNL
United States BlueGene/L - eServer Blue Gene Solution
IBM 131072 2005 280600 367000
2 Oak Ridge National Laboratory
United States Jaguar - Cray XT4/XT3
Cray Inc. 23016 2006 101700 119350
3 NNSA/Sandia National Laboratories
United States Red Storm - Sandia/ Cray Red Storm, Opteron 2.4 GHz dual core
Cray Inc. 26544 2006 101400 127411
4 IBM Thomas J. Watson Research Center
United States BGW - eServer Blue Gene Solution
IBM 40960 2005 91290 114688
http://www.top500.org/list/2007/06/100
Rmax and Rpeak values are in GFlops
No intel 1. a 4. miesto IBM Power/PowerPC 2. a 3. meisto AMD Opteron
Lenze v top vedecky pocitacovh sa Windows neinstaluje
Operating System Count Share %
Linux 347 69.40 %
SuSE Linux Enterprise Server 8 1 0.20 %
HP Unix (HP-UX) 15 3.00 %
MacOS X 3 0.60 % 32989
Solaris 4 0.80 %
http://www.top500.org/stats/list/29/os
A simple mathematical model for the approximate theoretical performance of a multi-processor system relative to the single processor base is as follows:
Pn / P1 = S ** (log2(n))
Pn is the performance with n processors, P1 is the performance with one processor, S is the scale factor and n in the number of processors. In the ideal situation, S = 2. In practice S is less than 2 and hopefully (but not always) greater than 1. A more transparent interpretation of the above formula is: for every doubling of the number of processors, performance increases by a factor of S.
aby sme mali vzorec na S z 2 CPU
(Pn1/P1)/(Pn2/P1)=Pn1/Pn2=(S**log2(n1))/(S**log2(n2))
log(n1)/log(n2)
za predpokladu, ze sa kvalita skalovania nemeni
lenze pre skalovanie z 2 na 8 pre Barcelonu
je to scestne
4,5 !=3
t.j pri zvyseni poctu CPU z 2 na 8 by mal narast vykon 3x, ale u Barcelony narastol 4,5 teda S pre 8 CPU je vacsie ako pre 2 CPU a to 1,5
lenze skalovanie z 2 na 4 CPU je podla nomry
2!=2
ak berzie sklaovanie z 4 na 8
2,25 namiesto teoretickeho 1,5!!!!!
toto je zahada
kedze AMD sa skaluje lepsier nez je idealny stav!!!
http://osr507doc.sco.com/en/PERFORM/multiprocessor_systems.html
re 1 na 8
9=S*log2(8)/log2(1)=S*3 cize S u Barcelony je 3
In the ideal situation, S = 2
to znamena, ze s pohladu skolavanie s poctom CPU je Bracelona lepsia nez idealna
The scale factor for Itanium 2 is in the range of 1.60, which is considered very good. IBM has not released sufficient information to derive the scaling factor for the Power 4 processor used in the pSeries 690 systems, but it is probably is the range of 1.7, which is exceptionally high
The scale factor from 4 to 8 CPUs is 1.54. Beyond 8 CPUs, the scale factor falls to 1.41
takze o com sa bavime 2,25 namiesto 1,54 u Intel Xeon scaling faktor zo 4 na 8