BM Seer Unofficial thoughts from an anonymous Sun employee

Sun T5220/T5120 Faster than 4.7GHz 4-core power6 IBM p 570 (p550/p520 only 4.2GHz)

Wednesday Apr 02, 2008

IBM and those who comment on this blog (employees? or just fans?) are trying to ignore that the UltraSPARC T5120/T5220 simply outperforms the 4-core 4.7GHz (2-chip) IBM p 570. They try to say that we should compare not to the price of 4.7GHz IBM p570 (which is slower than UltraSPARC T5120/T5220), but compare to even lower GHz 4.2GHz IBM p 550 (4-core)/4.2GHz IBM p 520 (4-core) so IBM doesn't look so bad? What, this makes no sense! When they lose, do they just make stuff up and ignore that people can think on their own?

Simply put Sun SPARC Enterprise T5220/T5120 are faster than the 4.7GHz IBM p 570 (4-core). See the benchmarks below:

SPECjbb2005: The Sun T5220 server (single UltraSPARC T2) demonstrated 9% better performance than the 4-core 4.7GHz IBM p570 (POWER6) result of 175,474 SPECjbb2005 bops 87737 SPECjbb2005 bops/JVM. The Sun T5220 server has 2.5x better power-performance and has 4.9x better SWaP than the IBM 4-core p570.

SAP-SD: The 2RU Sun SPARC Enterprise T5120 server with a 1.4 GHz UltraSPARC T2 processor outperformed the 4RU IBM System p570 with two 4.7 GHz POWER6 dual-core (quad-thread) processors by 7%.

SPECjAppServer2004: One Sun T5220 server (single UltraSPARC T2 chip) demonstrated 67% better performance over the IBM result of 1197.51 SPECjAppServer2004 JOPS@Standard which used 4-core IBM p570 with 4.7GHz POWER6 processors. The Sun SPARC Enterprise T5220 used as Application server has 3.8x better power-performance and has 7.3x better SWaP than the IBM p570 power6. Enterprise T5120 used as database servrer has 3.4 better power-performance and has 13.5x better SWaP as the IBM p550.

Siebel: The benchmark demonstrates that the Sun SPARC Enterprise T5120 and T5220 Servers provide the highest performing and the most cost-effective business solution for Siebel CRM applications. This is achieved through a powerful combination of Oracle's Siebel CRM Release 8.0 architecture with Oracle 10g R2 database running on a 8-core / 64 threads UltraSPARC T2 processor, Sun Java System Web Server, and the industry leading, free and open Solaris 10 OS. http://www.oracle.com/apps_benchmark/doc/Sun_Siebel8_10000_PSPP_On_Solaris.pdf

Pointer to IBM p 570 paper: http://www.oracle.com/apps_benchmark/doc/IBM_Siebel8_7000_PSPP_On_AIX_POWER6%20Final.pdf

SPECweb2005: NO results from 4.7GHz IBM p570!

Disclosure statement:

SPECjbb2005 Sun SPARC Enterprise T5120 (1 chip, 8 cores) 192055 SPECjbb2005 bops, 24007 SPECjbb2005 bops/JVM, Sun SPARC Enterprise T5220 (1 chip, 8 cores) 192055 SPECjbb2005 bops, 24007 SPECjbb2005 bops/JVM, IBM p570 (2 chips, 4 cores) 175474 SPECjbb2005 bops, 87737 SPECjbb2005 bops/JVM, SPEC, SPECjbb reg tm of Standard Performance Evaluation Corporation. Results as of 10/08/2007 on www.spec.org.

Two-tier SAP Standard Sales and Distribution (SD) standard SAP ERP 2005 application benchmark: SPARC Enterprise Model T5120 (1-way, 1 proc, 8 cores, 64 threads) 1 x 1.4 GHz UltraSPARC T2, 64GB memory, 2175 SD Benchmark users, 1.91 sec avg response time, Cert#2007059, Oracle 10g, Solaris 10; IBM System p 570 (2-way, 2 processors, 4 cores, 8 threads) 2 x 4.7 GHz POWER6+, 32GB memory, 2035 SD Benchmark users, 1.99s avg resp time, Cert#2007037, Oracle 10g, AIX 5L Version 5.3; SAP, R/3, mySAP reg TM of SAP AG in Germany and other countries. More info www.sap.com/solutions/benchmark.

SPECjAppServer2004 1 Sun SPARC Enterprise T5220 (8 cores, 1 chip) and 1 Sun SPARC Enterprise T5120 (8 cores, 1 chip) 2000.92 SPECjAppServer2004 JOPS@Standard. 1 IBM p570 (4 cores, 2 chips) and 1 IBM p550 (4 cores, 2 chips) 1197.51 SPECjAppServer2004 JOPS@Standard. 1 HP rx2660 (4 cores, 2 chips) and 1 HP rx2660 (4 cores, 2 chips) 874.17 SPECjAppServer2004 JOPS@Standard. SPEC, SPECjAppServer reg tm of Standard Performance Evaluation Corporation. Results from www.spec.org as of 10/10/2007.

SPEC, SPECint registered trademarks of Standard Performance Evaluation Corporation. Sun SPARC Enterprise T5120 server (1xUltraSPARC T2 1.4GHz) (1 chip, 8 cores, 64 threads) 83.9 SPECint_rate2006. Competitive results from www.spec.org as of February 12, 2008. IBM System p570 (4.7GHz POWER6 processor, 1 chip, 2 cores, 4 threads) 60.9 SPECint_rate2006. HP Proliant DL360 G5 (3.16 GHz, Intel Xeon processor X5460,1 chip, 4 cores, 4 threads) 73.0 SPECint_rate2006.

[6] Comments
Like this post? del.icio.us | furl | slashdot | technorati | digg
Comments:

I'm still curious to see a T5120 Oracle database benchmark without the surrounding fluff, compared to a Power6 machine. Given how parallel most modern database applications are, it really should be an impressive system.

The T5120 is a bit of a rocket, but the benchmarks you quote above are heavily skewed in terms of price, mostly because IBM didn't ship the Power6 p520 machine in time for it to be used in benchmarking.

When you're talking about building systems which have only got 2 power6 dual-core processors (as all of these benchmarks used), the p520 is the best price/performance power6 machine, as you pay a fairly big premium for the flexibility of the p570 to scale up to 8 processors/16 cores.

Posted by Ewan on April 03, 2008 at 01:54 AM PDT #

p570 - hi-end, on smb market IBM and SUN use X86. T5220 - smb market, any 2CPU (8 cores) x86 Xeon with 64Gb RAM cost $20K it is 2 times cheaper T5220 ($40K), but already 8 cores faster on SPECjbb2005, SAP-SD, SPECjAppServer2004 ... 16 Xeon cores Sun Fire X4450 cost only $31K

Posted by Triffids on April 03, 2008 at 07:51 AM PDT #

Ewan & Triffids: Good of you guys to ignore performance and to avoid comparisons by just making up what competes against what. How convenient but totally baseless.

Ewan: I would be quite surprised if the fastest IBM p 520 4-core (which is only 4.2GHz) could beat the performance of a 4.7GHz IBM p 570 4-core.
Right now in terms of performance: Sun T5120 >> IBM p570 4.7GHz 4-core 2-chip >> IBM p520 4.2GHz 4-core. But we're still waiting for IBM to publish the kinds of benchmarks on the p520 which would show what you want (if it was fast that would be easy). NOTICE IBM says with rPerf that the p520 is 28% slower than the p570 above, which would make it MUCH slower that the Sun T5120.

Triffids: The above benchmarks show that the Sun T5120/T5220 is faster that the 4-core p570. Also lots of data showing that T5120 beats 4-socket QC. Remember that the 8-core p570 is cabled together 4-core p570s, IBM has yet to publish the NUMA latency differences.

Posted by BM Seer on April 03, 2008 at 10:21 AM PDT #

Hi BM, I don't think I did ignore performance, I actually said how quick the T5120 is? It's a fantastic box for it's market-segment, but if it was the Greatest Box on Earth, Sun wouldn't bother making the M8000 or the V125 would they?

Would the T5120 outperform a 2 processor 4 core M8000 on all the benchmarks you mention above? I imagine it would, but that doesn't make the M8000 a bad box that noone should buy does it?

I specifically only talked about price/performance, which is perhaps the key metric when customers buy systems. When a customer comes out and asks me what the SWaP rating is of a server, whether it's Sun, HP, or IBM, I'll come back and let you know.

Posted by Ewan on April 03, 2008 at 10:44 AM PDT #

I agree that T5120 is interesting box, but I can't understand on which market it is great ? on SMB market it can't compete with 8-16 cores Xeons. 16 cores Xeons 2-3x cheaper and faster on most of benchmarks, on mid-market T5120 can't compete because T5120 can't scale as it can p570...

Posted by Triffids on April 05, 2008 at 05:52 AM PDT #

16 core Xeons (4-socket QC) at full speed CPUs and full speed memory 64GB aren't that cheap.

Let's please specify config & price together. There is a lot talk about cheaper but people play lots of entry level games, then you get $5k for this and that (modern equivalent of nickeled & dimed) and then you see a big bill.

Here is some real $/perf comparisons.
http://blogs.sun.com/ritu/entry/mysql_benchmark_us_t2_beats

Posted by BM Seer on April 10, 2008 at 10:05 AM PDT #

Post a Comment:
Comments are closed for this entry.