IBM punts first z196 mainframes Big Blue said today that it has shipped its first System zEnterprise 196 mainframe, and it is a whopper. In fact, IBM shipped two zEnterprise 196 boxes to Swiss Re, the Zurich-based re-insurer that is one of the few mainframe users in existence that predates IBM itself - Swiss Re is 147 years old, and IBM turns 100 next year …
I wondered that myself until I RTFA!
"These machines were launched in July and can have 80 cores spinning at 5.2 GHz and 3 TB allocated to a single system image running z/OS. With 80 cores dedicated to z/OS, the z196 is rated at more than 50,000 MIPS of aggregate processing capacity. At 1,200 MIPS of raw performance each, each single engine in the box has about 30 percent more MIPS than its z10 predecessors. (Customers can allocate non-z/OS cores to do other work, such as manage I/O or clustering software, run Linux, or accelerate Java or DB2 workloads.)"
80 cores dedicated to z/OS
16 cores that can be allocated to I/O, clustering, linux or to accelerate java or DB2 tasks
= 96 cores of computational power
=P
My question would be.. if the 80 cores run at 5.2GHz what do the other 16 run at? and if its the POWER architecture, how much cache does each cpu have?
The z196 cpus have something like ~400MB cache in total, for each cpu. When you count the L1+L2+L3+L4 cache. All cache adds up to several GB. Yes that is true. And the CPU runs at 5.2GHz! That is a real monster cpu!
And yet any modern x86 is 5-10x faster than these z196 monstrosities. I have proved this in my earlier posts. These z196 cpus are really slow dinousaur cpus that should never have left the lab. An abomination indeed.
"These z196 cpus are really slow dinousaur cpus that should never have left the lab. An abomination indeed."
Care to back that claim up ? with more than just Oracle marketing bullsh*t.
Saying that a x86 CPU is 5-10 times faster than a CPU in the z196 is so wrong, that it's ridiculous.
I am no great fan of Mainframes, I do respect them, cause I've done migrate off mainframe projects where they dorks that did the sizing didn't understand what the F*** they were dealing with.
So one thing is Standard SAP software or Oracle databases or... another thing is when all the different accelerators inside such an old dinosaur gets something to tear at. And being such a great fan of the SUN T3 processor, I would have expected you to understand that. But.. I fear you don't really get it..
// Jesper
Here are three links that show how slow the Mainframe cpus are.
.
Here is a source from Microsoft
http://www.microsoft.com/presspass/features/2003/sep03/09-15LinuxStudies.mspx?
"we found that each [z9] mainframe CPU performed 14 percent less work than one [single core] 900 MHz Intel Xeon processor running Windows Server 2003."
The z10 is 50% faster than z9, and the z196 is 50% faster than z10, which means a z196 is 1.5 x 1.5 = 2.25 times faster than a z9. This means a z196 corresponds to 2.25 x 900MHz = 2 GHz Intel Xeon. But todays modern server x86 cpus have 8 cores, which means they have in total 8 cores x 2 GHz = 16 GHz. We see that x86 at 16GHz is more than z196 at 2GHz. This shows that a z196 is really slow
.
Here is another source from a famous Linux expert that ported Linux to IBM Mainframe, who says 1MIPS == 4MHz x86.
http://www.mail-archive.com/linux-390@vm.marist.edu/msg18587.html
This shows that a z196 with 1400 MIPS corresponds to 5,6GHz x86. But a modern x86 has 8 cores, that means it has in total 16GHz, which is 3x faster than 5.6GHz. Again, we see that the Mainframe is quite slow
.
Here is another link where the cofounder of TurboHercules says that a 8-way Nehalem-EX gives 3.200 MIPS using software emulation:
http://en.wikipedia.org/wiki/TurboHercules#Performance
But software emulation is 5-10x slower. This means a 8-way Nehalem-EX running native code should be 5-10x faster, that is, 16.000 - 32.000MIPS. This big MIPS number matches a fully equipped z196 mainframe with 64 cpus. Again, we see that the Mainframe cpu is slow.
In fact, you need two 8-way Intel Nehalem-EX servers to outperform the biggest z196 Mainframe with 64 cpus costing 30 million USD or more - if we talk about cpu performance.
All this above are not Oracle marketing b*llsh*t. I show credible sources, one developer who ported Linux to IBM Mainframes. Another developer who wrote a Mainframe emulator. And lastly, MS doing benchmarks.
In fact, it is IBM marketing that is b*llsh*tting fooling everyone the Mainframes are fast. Well, they Mainframes are slooooow. That is because the IBM Mainframes cpus are a derivative from the POWER6 which we all know is really slow.
You need four POWER6 to match two ordinary Intel Nehalem on TPC-C. Of course the Mainframe cpus are slow too, then.
Lets see.... a Management report from Microsoft paid for by Microsoft from 2003 ?
You can't be serious.
And another 2003 link from some dude using MHz of the processor at the time. That you then try to relate to current Nehalem processors.. what in 2003 I guess you were in what kindergarden ?
It makes no sense what so ever.
And then there is the Nehalem-ex thing with TurboHercules. Bzzzzz.. Only problem is that the dude assumes that a 8 Core 2.26GHz Nehalem-EX chip is x2 faster than a 4 Core 2.93 GHz Nehalem-EP chip. Now that isn't really the case now is it ?
268 SPEC-INT for 2 Nehalem-EP chips with a total of 8 cores. The number for 376 for the same blade using Nehalem-EX. Now that is 42% increase in chip speed not x2..Ok ? You know math and checking facts, we have talked about this before. So his calculations more likely say around 2300 Mips not 3200.
And no you are not using credible sources.. they are old and you are doing strange pseudo math to. Now the worst part is that for some workloads I would say that 2x64 core Big Iron Intel boxes could most likely match a zSeries, on native x86 workload. So I actually agree with you, there.
BUT...
I would say with my experience in porting things off the mainframe, that this number would be something totally different if you were trying to port a Mainframe native stack to a native Windows stack on x86. Then you might as well go buy 10 machines. And then the mainframe all of a sudden seems like a cheap alternative.
It's all about TCO.. TCO.. not TCA.
// Jesper
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
