Tech titans Oracle and IBM are going at it over Oracle's claims that its new Sparc servers use the world's fastest microprocessor. IBM has taken public exception to the claims and the word battle is heating up.
Oracle's says its new Sparc T5 servers have set 17 world records and are the best platforms for enterprise computing at any scale, delivering the best value for database and enterprise applications. Many of its claims take aim directly at IBM.
Other claims include: Oracle's Sparc T5-8 is the fastest single server for Oracle Database; Oracle's Sparc T5-8 is also the fastest single server for Oracle Middleware with a 12 times cost-performance advantage over the IBM Power 780; Oracle's Sparc T5-8 server has a 2.5 times price-performance advantage over the IBM P780 3-node cluster, when including hardware and software cost; and Oracle's Sparc T5-8 server has a 7 times price advantage over a similar IBM Power 780 configuration for database on a server-to-server basis.
IBM Fights Back
"Against IBM's best published results, using independently confirmed industry-standard public benchmarks, Oracle delivered 5x, 7x, even 12x better cost performance," said John Fowler, Oracle's executive vice president for hardware, in a statement. "Oracle's Sparc T5 performance and cost-performance advantages are so huge that subtle changes in IBM configuration specifics don't make up that gap. IBM customers are being wildly over-charged for the performance they're getting."
Colin Parris, general manager of the Power line of IBM systems, said he was "surprised" and "pleased" and "astonished" by what Ellison said, according to The Wall Street Journal. IBM could not immediately be reached for comment, but Parris was quoted as saying, "This was a frozen-in-time discussion. It was like 2002 -- not at all in tune with the market today."
So who's right and who's hyping it up? We asked Charles King, principal analyst at Pund-IT, for his take on the drama. He told us the new Oracle Sparc T5 looks like a significant upgrade based on the results of tests. That said, he has questions about some of the head-to-head comparisons that Oracle claims around system performance. That's because some of the tests were conducted internally.
Sifting the Rhetoric
"There were some results they submitted to TPC but if you take a look at systems where they are claiming system supremacy, many of the other systems that are on the top 10 list are in some cases 3 and 5 years old," King said.
"In the server world, anything over 6 to 9 months old is extremely old news. So claiming that your system beats systems that are 2 or 3 years old is like a senior in high school beating up on a sixth grader."
King said the rhetoric in Oracle's announcement was very much aimed at IBM, which is understandable considering Big Blue is the leading vendor in the server space today. But Oracle has lost some credibility with such rhetoric.
"Oracle was cited three times last year for making false claims about the performance of its products compared to IBM by the National Advertising Review board," King said. "Given that and the competitive landscape and the unit space right now, I'd be cautious about taking anything from the announcement too literally."
Posted: 2013-03-29 @ 7:53am PT
Can't you tell that IBM has been severely impacted by the launch of these new SPARC servers and that a majority of their responses and claims are out of pure desperation and clearly don't hold any truth? They've even paid their "one person company analysts" to say bad things about Oracle!! Ask a real analyst like Gartner, IDC or Forrester for their analysis & comments!
From where I am standing, IBM is reacting as if its stuck in the headlights and is about to get steamrolled! Get Oracle to comment on IBM's claims for a valid article otherwise its seen as you are being one sided and funded by IBM!
Heres some responses to IBM's claims with questions & facts you can easily look up to question IBM's TRUTHFULLNESS:
"Most Oracle benchmark claims are not audited or published, although they do have some industry benchmarks, including TPC-C (Transaction Processing Performance Council),"
Actually, 12 of the 17 benchmarks that Oracle published are in fact, industry standard benchmarks where there are results from many vendors, although many of them IBM chose not to publish, probably because their Power7 systems don't perform well. These are all audited by independent companies and so are usually considered proper and controlled configurations and testing environments.
Details on all the Oracle benchmarks are fully disclosed with full transparency for Oracles customers.
"I believe some of the systems they cited aren't shipping until September."
Actually, Oracle stated during the launch that *all* systems from the smallest SPARC T5-1B to the largest SPARC M5-32 are shipping by end of April. Why would IBM say September? Probably from some of the benchmarks we're Oracle posted September as availability date which is the 6-month limit you can claim in benchmark results so if anything changes regarding benchmark like configuration, pricing, firmware, etc to run benchmark, can be updated instead of retracted.
Have you actually asked IBM how long it took them to launch the entire Power7 or Power7+ systems? Between 6-months to a year to actually release all models! Clearly, its not easy to launch an entire new range of systems from top to bottom. Oracle has. IBM has not.
"Regarding the TPC-C benchmark as quoted "The IBM Power 780 Model 9179-MHB coming in second. Oracle's T5-8 server, which is scheduled to ship in September, came in third"
Actually, there are two basic types of configurations for TPC-C. Clustered and Non-Clustered. Non-Clustered means single system, clustered means many systems (to get additional performance beyond a single system).
SPARC T3-4 is #1 on TPC-C for "Clustered Systems" and SPARC T5-8 is now #1 for "Non-Clustered" Systems. There are only two benchmarks on IBM's Power7/Power7+ systems and neither are or were world records. IBM's latest/newest/fastest Power7 result is almost 3- years old!
What you should ask IBM is why did they not publish a Power7+ TPC-C result to show performance improvement over Power7?
Why didn't IBM try to beat Oracle in the 3 years since IBM published the #1 TPC-C result? And now of course, its up to IBM to prove they can beat the SPARC T5-8. They just introduced brand new Power7+ systems, and surely if they have so many comments on Oracles TPC-C results, why not just prove their points by publishing?
My guess is IBM can't because their systems are clearly *slower* and more importantly, much worse on *Price/Performance*
"...said Charles King, principal at Pund-IT. "General purpose systems are often purchased to support other vendors' solutions [but] there was little if any mention of other vendors supporting the new T5 servers."
Charles King I believe is a part of IBM PR as he'll only say nice things about IBM. Check his quotes! Check his paychecks!
No mention on other vendors? Solaris was clearly mentioned several times during launch and the latest #'s show over 500 3rd party applications are qualified and supported on the new Solaris/SPARC systems. Sure Oracle focuses on its own software first (just like IBM) but Solaris still has the largest share of 3rd party ISV support for UNIX.
And actually, Oracle announced 2 world record results -for both SPARC T5 *AND* SPARC M5 on the SAP SD 2-Tier benchmark, where clearly, SAP and Oracle are not best of friends, yet Oracle published results because many customers run SAP on SPARC.
Oracle also published SPEC CPU2006, SPECjbb2013, SPECjEnterprise2010 and even STREAM benchmark.
Ask IBM why they did not publish any Power results on the new SPECjbb2013 benchmark or even Power7+ on STREAM proving I/O superiority!
"Oracle sometimes stacked the deck in its tests, using servers with up to twice as many CPU core -- hence more raw processing power -- than the IBM systems it used."
Well, its actually because only Oracle now has a 16-core SPARC T5 CPU and so of course, there will be more cores. If IBM runs a benchmark with a higher GHZ, does that mean that IBM has stacked the deck in their favor because theire CPUs run at a higher GHZ?
The bottom line for servers regardless of # of cores, # of CPUs, # of threads, GHZ, etc is how much does it cost and how well does it run my software and Oracle just announced 17 world record benchmarks on SPARC T5 and SPARC M5 and 12 of them are independent benchmarks, many of which include full price/performance disclosures so nothing can be hidden and systems can be compared.
"However, "Oracle is not the fastest as it claims," said Joe Clabby, president of Clabby Analytics. IBM's System z12 is the fastest microprocessor in the industry, operating at 5.5 GHz; and IBM's Power 7+ processor operates at 4.4 GHz."
Again, Clabby is on IBM's payroll so cannot be trusted as independent. Search Clabby and IBM and you'll see.
Fastest does not equal highest GHZ anymore! Otherwise why would IBM ship 5GHz with Power6 and then come out with 3GHz Power7 and claim Power7 is faster? Even Intel is struggling to go beyond 3GHz with its latest multi-core CPUs.
SPARC T5 is the industry leader with 16-cores @ 3.6GHz and 8 threads/core! A full 128 threads per CPU! Power7+ maxes out at 8 x cores with only 4 threads/core and 3.7GHz. Half the cores and half the threads means 4x fewer threads/CPU means MUCH lower Throughput necessary for todays most demanding enterprise workloads.
Why is Intel working on 15-core CPUs for Ivybridge-EX? because throughput matters. IBM has fallen behind.
"Oracle's benchmark tests were made against three-to-five-year-old IBM servers, not more recent models."
You must be kidding! That's because that is the *LAST time* IBM published a result on Power7 or Power7+! Since Oracle published the SPARC T3 world record, IBM has stopped publishing the benchmark because they can't beat Oracle yet other vendors like Cisco, Dell, etc continue to publish results for transparency on OLTP workload performance.
"IBM contends that Oracle's pricing doesn't include software and storage while its own does"
"Plus, IBM's servers help save enterprises money on software licenses, he says. That's because IBM servers will run more apps on a single server, he says. For enterprise software that's priced per server, companies can pay less for software licenses."
IBM is clearly full of it as many of the benchmarks like TPC-C, TPC-H, SPECjEnterprise2010, SPECjbb2013 require full disclosure of configuration including HW and SW and some even requiring pricing to calculate price/performance. If you do your research, you'll see that SPARC T5 has surpassed IBM and most other vendors on price/performance calculations which includes full OS, Virtualization and SW licensing.
With SPARC having a 0.5 multiplier on Oracle SW for per core based licensing, versus Power7 being a 1.0 multiplier, I can see why IBM is MAD!
Power7/Power7+ requires a minimum of 2x more Oracle SW licenses for the same per core performance as SPARC, clearly making SPARC *SIGNIFICANTLY* lower cost. Oracle wants to sell Oracle SW on OracleHW. Even IBM requires fewer IBM SW PVU's on SPARC than on its own Power7. 100 PVUs vs 120PVUs
Sparc "has been a losing proposition for Oracle for some time," Cross said. "Last year, we had over 350 competitive takeouts worth over $335 million, half of which were from Oracle and half from HP."
These competitive takeout numbers are pure marketing nonsense and don't have any substantiation. Ask IBM to prove which customers have left SPARC for Power7? Call the customers and see if they don't run SPARC anymore. How is that possible when SPARC/Solaris continues to hold #1 (UNIX/RISC) Unit marketshare according to IDC or Gartner.
And clearly IBM isn't doing so well with Power this past year. Maybe its due to SPARC T4 being launched last year (along with Oracles Engineered systems)?
IBM reports 2012 first-quarter results
"Revenue: $24.7 billion, flat"
"Revenues from Power Systems were flat compared with the 2011 period"
IBM reports 2012 second-quarter results
"Revenue: $25.8 billion, down 3 percent"
"Revenues from Power Systems were down 7 percent compared with the 2011 period"
IBM reports 2012 third-quarter results
IBM total revenues for 3rd Qtr of 2012 of $24.7 billion were down 5%
"Revenues from Power Systems were down 2 percent compared with the 2011 period"
IBM REPORTS 2012 FOURTH-QUARTER AND FULL-YEAR RESULTS
"Revenue of $29.3 billion, down 1 percent"
"Revenues from Power Systems decreased 19 percent compared with the 2011 period"
My comments are my own and may not represent those of Oracle.