IBM J9 has
- leadership 1,2,4,8,16,32 & 64-core results (IBM System p and System i)
- leadership 4 and 8-core Intel results
- leadership 8 and 16-core AMD results
- leadership 4-socket result
- leadership x86 2-socket result
Published results on Intel's Harpertown chip provide an interesting comparison of J9, JRockit and HotSpot. (Note the HotSpot number is using Sun's Java 6 _05 update. There hasn't been a publish with their faster _06 Performance update.)
J9's SR1 publish on the 3.16GHz Harpertown chip beats the more recent JRockit publish on the 3.33GHz chip. Also of interest are the uses of 32 and 64-bit JVMs. J9's 64-bit SR3 number beats JRockit by 11% and J9's 32-bit number beats JRockit by 6%.
J9 now has the leadership 8-core Intel number, the leadership 8-core AMD number and the overall leadership 8-core number using an IBM POWER6 chip.
With Sun's Niagara-2 processors having 8 cores per socket, Sun has published a number of leading per-socket scores. With the recent publish on IBM's Bladecenter LS42, IBM and J9 now claim the leadership 4-socket result:
Sun is still leading the 2-socket category ... by 2% ... using twice the number of cores and 16 times the number of threads.
J9 has the overal leadership 16-core result as well as the leading x86 16-core result.
oh .. if you actually read this far and are wondering what we did in SR3. The 10% boost is from the Apache Harmony improvement to TreeMap discussed in David Dagastine's blog.
Derek
The postings on this site are my own and don’t necessarily represent IBM’s positions, strategies or opinions.
SPEC and SPECjbb are registered trademarks of the Standard Performance Evaluation Corporation.
Results referenced are current as of December 10, 2008. The SPECjbb2005 results have been submitted to SPEC for review. Upon successful review, the result will be posted at www.spec.org, which contains a complete list of published SPECjbb2005 results. SPEC, SPECjbb reg tm of Standard Performance Evaluation Corporation.
1-core
IBM System i5 515 (1 chip, 1 core, 2 threads) SPECjbb2005 bops = 20615, SPECjbb2005 bops/JVM=20615
2-core
IBM System p 570 (1 chips, 2 cores, 4 thredas) SPECjbb2005 bops=88089, SPECjbb2005 bops/JVM=88089
4-core
IBM Power 570 (2 chips, 4 cores, 8 threads) SPECjbb2005 bops=205917, SPECjbb2005 bops/JVM=102959
IBM System X3350 (1 chip, 4 cores, 4 threads) SPECjbb2005 bops = 194256, SPECjbb2005 bops/JVM = 97128
8-core
IBM Power 570 (4 chips, 8 cores, 16 threads) SPECjbb2005 bops = 402.923, SPECjbb2005 bops/JVM = 100731.
IBM Bladecenter LS42 (2 chips, 8 cores, 8 threads) SPECjbb2005 bops=382,339, SPECjbb2005 bops/JVM=191,170
IBM System x3650 (2 chips, 8 cores, 8 threads) SPECjbb2005 bops = 344,436, SPECjbb2005 bops/JVM = 86,109.
IBM System x3650 (2 chips, 8 cores, 8 threads) SPECjbb2005 bops = 334,802, SPECjbb2005 bops/JVM = 83,701.
IBM System x3650 (2 chips, 8 cores, 8 thredas) SPECjbb2005 bops = 323172, SPECjbb2005 bops/JVM = 80793.
PRIMERGY RX200 S4 (2 chips, 8 cores, 8 threads,) SPECjbb2005 bops = 316,728, SPECjbb2005 bops/JVM = 79,182.
Sun Fire X4150 (2 chips, 8 cores, 8 threads) SPECjbb2005 bops = 303297, SPECjbb2005 bops/JVM = 75,824.
16-core
IBM Power 570 (8 chips, 16 cores, 32 threads) SPECjbb2005 bops = 867989, SPECjbb2005 bops/JVM = 108499.
IBM Bladecenter LS42 (4 chips, 16 cores, 16 threads) SPECjbb2005 bops = 721843, SPECjbb2005 bops/JVM = 180461.
Transport TX46 (4 chips, 16 cores, 16 threads) SPECjbb2005 bops = 659034, SPECjbb2005 bops/JVM = 164759.
Sun Fire X4450 (4 chips, 16 cores, 16 threads) SPECjbb2005 bops = 464355, SPECjbb2005 bops/JVM = 58044.
Sun Blade T6340 (2 chips, 16 cores, 128 threads) SPECjbb2005 bops = 388456, SPECjbb2005 bops/JVM = 24279.
24-core
Sun Fire X4450 (4 chips, 24 cores, 24 threads) SPECjbb2005 bops = 531669, SPECjbb2005 bops/JVM = 132917.
32-core
IBM Power 595 (16 chips, 32 cores, 64 threads) SPECjbb2005 bops = 1530343, SPECjbb2005 bops/JVM = 95646.
Sun SPARC Enterprise T5440 (4 chips, 32 cores, 256 threads) SPECjbb2005 bops = 692736, SPECjbb2005 bops/JVM = 21648.
64-core
IBM Power 585 (32 chips, 64 cores, 128 threads) SPECjbb2005 bops = 3435485, SPECjbb2005 bops/JVM = 107359.
2 comments:
Core comparisons are getting a bit weak, especially considering CPU density is increasing. How about a socket or CPU count comparison. When it comes to data center footprint and power, that's what matters.
Java is now Open Source. What is the point of IBM maintaining a separate JVM? Why not give it up and contribute the enhacements to the open source version?
as a former OS/2 user I saw IBM's "not invented here" syndrome before.
It's as if supporting Openoffice.org or Sun's Java directly would make IBM somehow weaker.
FC
Post a Comment