POWER6 err, try >10GB/s peak

Wednesday Jun 27, 2007

I finally saw an IBM power6 presentation that says:

"Balanced design with highest system bandwidth
– 2X Memory Bandwidth (> 10 GB/sec)

What happened to the implied IBM press release of 300GB/s bandwidth that could download iTune library in 30 sec... oh, that was meaningless marketing fluff(is that the right word?).

So to move the iTunes library from power6 memory to POWER6 CPU at the 10GB/s listed above, it would really take 15 hours at p570 peak memory speed(not-measured!).

postscript: Another doc listed 45GB/s SMP+IO bandwidth (whatever that means), time for IBM to publish the measured STREAM bcopy performance so we can put an end to this nonsense.

On a Sun Fire E25K going to disk(Yes IO, which is slower than IBM's memory) it would take half that 15 hour time. Sun demonstrated a delivered 21 GB/sec of delivered disk to CPU bandwidth. (yes I need to repeat 'delivered' twice as IBM has a tendency to only mention peak numbers and then omits the word 'peak'.

OK we don't have IBM delivered IO performance measured on an IBM p595, as IBM doesn't trust to share those numbers with the public...

For news article about bigger systems see this:
http://www.betanews.com/article/Sun_We_Can_Build_a_Faster_Supercomputer_Than_IBM/1182889189

JaveOne: More performance results

Tuesday May 08, 2007

Sun US-IV+ vs. POWER5+:

• Sun Fire E2900 with dual-core US-IV+ (24 threads) beats the the fastest IBM POWER5+ p5 570 result (2.2 GHz 32 threads) of 326,651 bops.
• How do they compare on pricing? Well IBM doesn't seem to post pricing online for expensive products over about $500,000 or 32 threads, so I can't get the public pricing for the IBM p570.
The only online pricing (a bit old) is from IBM p570 TPC-C disclosure report. IBM TPC-C result of 1,025,169 tpmC at $4.42/tpmC on a 16-core (8 processors, 32 threads) 2.2 GHz IBM System p5 570 (configuration planned to be available 05/31/06). look carefully at the line items, just turning on the processors is expensive: "MODEL 570 PERMANENT PROCESSOR ACTIVATION FEATURE *16 = $343,040" (note: that does NOT count any memory costs or actually getting the processors any of the 48 other p570 required line items). My comments on the TPC-C benchmark.
• Also see this for IBM hardware per core pricing on the high end - wow!

The Sun Fire E2900 with 1.95GHz US-IV+ achieved 332,917 SPECjbb2005 bops and 27.743 SPECjbb2005 bops/JVM on the SPECjbb2005 benchmark. The Sun Fire E2900 used Solaris 10.

SPECjbb2005 Performance (ordered by performance bops : SPECjbb2005 Business Operations per Second, bigger is better)

System	Date	Processors		Performance
		(Chips, Cores, Threads)	GHz Type	bops	JVMs	bops/JVM
Sun Fire E2900	5/07	12, 24, 24	1.95 US-IV+	332,917	12	27,743
IBM p5 570	1/06	8, 16, 32	2.2 POWER5+	326,651	8	40,831

Sun results have been submitted to SPEC for review and are on track for publication.

Benchmark Description

SPECjbb2005 (Java Business Benchmark) measures the performance of a Java implemented application tier (server-side Java). The benchmark is based on the order processing in a wholesale supplier application. The performance of the user tier and the database tier are not measured in this test. The metrics given are number of SPECjbb2005 bops (Business Operations per Second) and SPECjbb2005 bops/JVM (bops per JVM instance).

Disclosure Statement:

SPECjbb2005 Sun Fire E2900 (12 chips, 24 cores, 24 threads1.95 GHz) 332,917 SPECjbb2005 bops, 27,743 SPECjbb2005 bops/JVM submitted for review; IBM eServer p5 570 (8 chips, 16 cores, 32 Threads 2.2 GHz) 326,651 SPECjbb2005 bops, 40,831 SPECjbb2005 bops/JVM. SPEC, SPECjbb reg tm of Standard Performance Evaluation Corporation. Results as of 5/8/07 on www.spec.org

Certified Results
	Performance:		332,917 SPECjbb2005 bops
			27,743 SPECjbb2005 bops/JVM
	Reference Date:		May 8, 2007
Systems:			Sun Fire E2900
Processor/GHz:			12 US-IV+ 1.95 GHz
Operating System:			Solaris 10
JVM:			Java HotSpot(TM) 32-Bit Server, Version 6.0_02