Slashdot Mirror
IBM Mainframe Running World's Fastest Commercial Processor
dcblogs writes "IBM's new mainframe includes a 5.5-GHz processor, which may be the world's fastest commercial processor, say analysts. This new system, the zEnterprise EC12, can also support more than 6-TB of flash memory to help speed data processing. The latest chip has six cores, up from four in the prior generation two years ago. But Jeff Frey, the CTO of the System Z platform, says they aren't trading off single-thread performance in the mainframe with the additional cores. There are still many customers who have applications that execute processes serially, such as batch applications, he said. This latest chip was produced at 32 nanometers, versus 45 nanometers in the earlier system. This smaller size allows more cache on the chip, in this case 33% more Level-2 cache. The system has doubled the L3 and L4 cache over the prior generation."
Why does the article not mention the name of the CPU? Is only its clock speed faster, or also its execution? Can we also use this CPU in consumer computers or is this for IBM Mainframes?
They seem to be claiming "faster" solely on the basis of clock frequency. In actuality, your normal-ass laptop chips probably have higher performance, except on the specific server-type workloads this was made for.
for IBM Slash-vertizing!!!!!!!!!!
...gives me a bit of a cognitive dissonance sensation. It shouldn't, really, but it does. Is it just me?
How does the L4 cache in these processors work? Generally going to anything off die is going to induce a major latency penalty due to the need to go through a driver stage which can handle outside interference. How can they make the L4 cache fast enough that its small size doesn't make it basically pointless versus just going to main memory?
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
That's a Ming Mecca chip. Those aren't even declassified yet!
So it was my understanding that part of the reason consumer CPUs didn't tend to go above 3-4GHz was that, at those speeds, the electrons can't actually move through the wires fast enough. Specifically for doing memory reads -- at 5.5GHz, I'm calculating about 4cm per clock cycle -- which may be further than the memory is physically located on a normal desktop PC. Meaning it would take not just two, but possibly three or four clock cycles to read a value from main memory.
Granted, on a server, main memory may be closer to the CPU, and the added cache will help as well. But I'm also mostly a software guy -- anyone with some more computer engineering knowledge have any information about this? Is the memory closer? Are they just taking longer to read? And if so is that likely to impact performance significantly (such that this wouldn't be as significant of a gain from 3GHz-5.5GHz as, say, 1GHz-2GHz?)
But their chip isn't overclocked to near death and using a nitrogen cooler. So I've no idea how your comparison works. The headline may as well read "Ford make fastest road car" so you can say but it isn't as fast as a car you've have converted to use a jet engine and drag racing tires.
So, you're comparing a ridiculous configuration of a nitrogen-cooled, over-clocked processor that will maybe run long enough to get a screen shot of it running, to a commercial processor that is designed to run at that speed non-stop for years and years? Yeah, that makes sense.
...those speeds are only achieved with the Turbo button depressed,
otherwise it's 8Mhz for compatibility with legacy software.
[...] But Jeff Frey
That's when I stopped reading and a train of thought departed the station.
Except the 5.5GHz may not be all that fast, as the Z-line of CPUs are the old IBM 360 instruction set, which is is so large, complex, and baroque that it is mostly usually implemented through a thick layer of microcode.
So 5.5GHz may be the speed of the microcode level, the actual "machine instructions" may be a considerable sub-multiple of that.
How about a price list in TFS for budget planning?
There is no right to feel safe thru security vaudeville at the expense of everyone's freedom, privacy and tax money.
No the basic Core i7 Extreme will smoke it too running at normal clocks speeds. However if you're foolish enough to judge the processor by its clock speed, the Core i7 can also win that race too, simply by turning up the cooling and the clock multipler to its maximum supported (which would do 5.7ghz on my older i7). IBM's chip is water cooled BTW.
IBM avoids benchmarking its mainframes for good reason.
5.5Ghz probably makes it about as fast as a 2 year old intel machine. I should know, I have a z114 (previous generation at 3.8Ghz) that i've done extensive benchmarks on. The fact that IBM refuses to publish standard benchmark numbers (specCPU, specVM, etc) should be sufficient proof that they are not pretty.
I can say that the people buying these things are pretty much smoking some fine IBM drugs. Sure, they are actually fairly competitive (but still not class leading) on the high end, but on the low end, which starts at ~200k, after disks and licenses, for 26 MIPS are abysmal. At that price/performance hercules on a midrange desktop PC doing software emulation (and its not even JIT'ed) runs somewhere between 5-15x as fast.
A 26 MIP mainframe is roughly equal to a Pentium 90. A full blown 3.8 Ghz z114 is roughly equal to a 5 year old x86 server.
Worse yet, is FICON, which generally is just a giant layer of inefficiency sitting in front of standard SCSI/SAS disks. So, the IO numbers are pretty abysmal too.
Basically, you have to spend >$400k before the mainframe catches up to what you can do on your desktop with a free emulator.
If your running linux on z, then your really deluded. In fact, your probably better off taking the HMCs, SEs, and CUs that it comes with and running linux on them directly. The only minor saving grace is that IBM doesn't rape people for unlocked processors to run linux (IFL's).
Further, IBM's claims of easier manageability are a joke. I can install ESXi and a half dozen linux machines, in the time it takes an expert system programmer to setup the HCD, install z/vm, and start configuring a linux machine. Oh, and I can migrate the image with a couple mouse clicks. Plus, I don't have to manage my data stores as a bunch of tiny disk images because zOS still prefers to deal with mod3 (~3GB) and mod9 (~9GB) disk partitions. I literally have a few hundred partitions on a machine with just a couple TB of storage. If you think managing a few dozen vmware disks is a problem, multiply it by 3-8x on z to run linux.
Frankly, if you have cobol, JCL, whatever running on these things and your not desperately trying to migrate to another platform, then your must either be extremely rich, or really stupid. The maintenance costs alone over ten years is going to save 7 figure sums, which should more more than enough to hire a couple programmers and a system administrator to port and maintain the apps on a machine that costs $20k every 5 years.
We could have gotten some meaningful benchmarks. According tho this Register arcticle
Back when TurboHercules was still around, in 2009, Tom Lehmann claimed
On the other hand, if your old and creaky code can't be divvied up among a multiplicity of cores, the existence of a far cheaper 64 core, 8 way Nehelem EX machine (or its current equivalent) that's almost as fast as a single zEC12 core doesn't much matter.
OK, here's a benchmark. You're welcome to try running an entire large bank (for example) on one server -- your choice. OK, two servers: I'll allow you one additional for off-site disaster recovery of all development, test, and production workloads, including concurrent batch and online, for all the bank's security zones. Choose wisely, Grasshopper.
I'll believe their claims when I see some test results they can back it up with.
Entry level z114 is indeed 26 mips for $75000.
http://www.tech-news.com/publib/pl2818.html
Which is a joke surely? $400,000 will get you 330 mips, which is erm, surely a mistake???!! It's way way too low:
http://en.wikipedia.org/wiki/Instructions_per_second
Core i7 is 177,730 MIPS at 3.3 Ghz.
What exactly are you basing your claims on? Just pulling things out of thin air?
Here are some things that IBMs customers care about, where are the Core i7 Extreme numbers for these?
How many CICS transactions can I process per second? How many IMS updates? How about DB2 transactions? How many SSL transactions? What differences are there in performance for on-line vs batch processing? Can I tune the system to maximize performance for my particular workload?
So you don't like my benchmark then and want another benchmark? OK. I chose a perfectly reasonable benchmark: number of servers (X) to deliver a particular real-world business outcome, where smaller X is better. A benchmark is simply a measurement to assess particular criteria (such as X) against a particular outcome (such as running a bank). I can agree that that an IBM zEnterprise EC12 server is not the answer to every IT problem. It is, however, the answer to many. And if you can't agree to that, then you simply have more to learn. (How exciting!)
With flash memory you read a block, flip some bits, and write it back to modify that block. Not only that, but Flash memory will wear out after so many reads and writes. That would be devistating to a CPU.
If you are trying to claim that you do telecoms billing on a single x86 server, I am going to claim that you are either at a very small telecom, or are lying.
You claim that mainframes 'can't handle the load', then you claim 'there are no benchmarks'. Well, which is it?
Anyone who decides what machines to use for a major installation such as a telecom based on 'benchmarks' is a complete moron. Unless the benchmark happens to match your workload exactly (and they never do) the benchmark is worthless. The only proper way to benchmark is to try YOUR workload on multiple machines, then make a decision. I am guessing that you are not even aware that you can take your workload to IBM to benchmark it, are you?
It seems the Z114 class at least is benchmarked, their entry level $75k server is 26 mips, which would buy you a big rack of Core i7 servers (each with 177,730 MIP, i.e. 4500 times faster)
http://www.tech-news.com/publib/pl2818.html
So as my own experience of IBM mainframes tells me, they're just too slow. You can claim some magic security gain, but the reality is they don't have enough processing power to do any extra security checks. You can claim extra reliability, but then for the same money I can buy 10 servers and have 10 mirrors running. You can make some vague 'particular real-world business outcome' claim, but I have to prepare real world bills in real world time, and sales talk doesn't crunch numbers.
Given the faults I've seen with telecom billing at every level (POTS, cell, trunk, global) I'm going to go with the "Your example is best used as a counter-example". By the way, how do you know they can't handle the load when you don't have any benchmarks? They're all bullshit, but there's a benchmark relevant to most business problems.
No, no typo. There's indeed Flash Express -- and yes, IBM's engineers have figured out a way to add yet another memory tier using (very high quality) flash memory. The processor can directly address it -- it's all mapped within the 64-bit virtual address space from what I've read. Yes, it's slower than DRAM but it's faster than storage-attached SSD (which at least has a longer distance to travel). Flash Express is great for things like paging, memory dumps, gigantic in-memory databases, and certain things that Java wants, so that's how operating systems and databases will use it. IBM even encrypts everything that lands on this memory-addressable flash, just in case someone tries to physically rip it out of the server. (Yes, they thought of that.)
Hmmm, six cores with each running at 1 ghz equals 6 ghz with a 5% overhead makes it 5.7 ghz maximum... IBM Marketing!!!!
And the published information supporting your assumption that the cores are only running at 1GHz, and the 5.7 GHz comes from multiplying the clock rate by the number of cores and subtracting 5% as overhead, rather than each core truly running at 5.7 GHz, is?
This ius what I've seen woring in a mainframe shop as well. The performance is not great, and the OS and tools were horrid (this is z/OS, not Linux). The costs were astonomical for the performance as well. The only thing they can really clain is very good reliability, but in the end, it's human error that gets you every time. We had well administered Windows servers running database, etc, that were kicking the mainframe's ass in both performance and in uptime (systems, not hardware). If you never change *anything* that might cause problems, you'll have a great uptime, but so would a cluster of linux boxes, with better price and performance. There are very few workloads where a mainframe is a benefit, and the only thing keeping most peple there is the difficulty in leaving, requiring re-writing software and tools.
Long live CICS!
"Seven years of college down the drain. Might as well join the f-ing Peace Corps." - John 'Bluto' Blutarsky
Points 1,2,3 apply to this chip too. At the end of the day,its a chip running Linux timeslices and Java. It can be benchmarked and it can be compared. Even if IBM runs away from comparisons.
Point 4, The table lists an entry server of 26 MIPS for $75000 which will buy you a big rack of Corei7s. You mention 780 MIPS, the register article (mentioned in a comment lower down) estimates 1600 for the top of the range chip. i.e. 1% of the processing power of the i7.
Presumably that's a top of the range price in millions, but lets ignore that for a second.
If I switched computers to this IBM mainframe, from its current rack of 4 Corei7s I would have 0.25% of the processing power. I would firstly probably ditch the integrity checks, and security checks, they're expensive to calculate and I don't have the processing power. If the floating point is as bad as the MIPs then I would probably have to switch some of the calculations from float to integers or fixed point math, and round, again with lots of problems and contractual headaches. I would calculate the bill less often and not be able to update the live online bill, again a result of the lack of processing power. We would raise prices, this is due to the high cost of the mainframe.
The consequence of vague claims, not backed up by hard reality would be devastating.
After a lot of vague talk from you, when forced to you finally make a benchmark claim. 780MIPS and its pitiful, even if it was a $1000 computer it would be pitiful.
I'm now wondering if my tablet PC (Asus tf700 quad core 1.6Ghz) is faster than your mainframe, because these numbers from you and others who've benchmarked IBM kit, are sooo low.
I would argue that the jump from machine code to assembly to C is much smaller than the jump from C to lisp/bash/perl/prolog
Seems someone is astro-turfing quite a lot. Some of the cowards sound suspiciously like him as well. And the lingo is exactly hte one you hear from IBM sales people.
While I do think that the Mainframe is underappreciated and am happy about the attention on Slashdot, I'm not very fond of such shady tactics... Either state that you're an IBM employee trying to market your product or don't pretend to just be a Slashdoter without an agenda and shut up. Thanks
The x86_64 cpus support a mode called x32 where they use the 64-bit CPU mode but with 32-bit pointers. The hardware supports it, but there is relatively little software support. A Linux port is currently in progress.
Wrong. See http://hardware.slashdot.org/comments.pl?sid=3078075&cid=41151983
MODS: mod it down until it is at score 1 (no need to go to -1 and burn the guy's Karma).
And to BBC Watcher: nothing personal.
Wrong. See http://hardware.slashdot.org/comments.pl?sid=3078075&cid=41151983
MODS: mod it down until it is at score 1 (no need to go to -1 and burn the guy's Karma).
And to jthill: nothing personal.
Here are some things that IBMs customers care about, where are the Core i7 Extreme numbers for these?
How many CICS transactions can I process per second? How many IMS updates?
Well, you're unlikely to get numbers for the first of those, given that IBM apparently killed off CICS for Windows and I'm not sure which x86 UN*Xes, if any, got versions of CICS. I'm not sure to what extent TXSeries for Multiplatforms would let you, for example, run CICS on Windows Server or Linux.
As for the second, as far as I know, IBM's never ported IMS to any non-mainframe OS.
How about DB2 transactions?
About 13,000 XML transactions per second in at least one benchmark - but those were Xeons, not Cores (server rather than desktop/laptop processors).
Looks like a little someone had all their karma go by by cause he is a fucking asshole and now is trying to burn just whomever.
jOrgePeixoto, nothing personal, but you really should just kill yourself.
Oh and MODS, just burn this asshole down.
Well, yes, that was my point. The workloads run on mainframes are different than the workloads run on other processors. Therefore, there are no benchmarks which can be accurately used to compare them.
These people that keep harping on the 'IBM won't do benchmarks' theme are completely missing the point. It is like saying that because GE Locomotives does not publish 'MPG' or '0 to 60' figures that means that they are hiding something and that obviously a Toyota is a better vehicle. Well, where are the 'fuel consumed per 1000 ton/miles' measurements for Toyota?
Well, yes, that was my point. The workloads run on mainframes are different than the workloads run on other processors.
Some workloads run on mainframes are different from the workloads run on other systems. A workload using software that only runs on, say, z/OS would be such a workload, as per my comment on IMS.
Other workloads are run on many different types of large systems, whether the processors happen to execute a descendant of the System/360 instruction set and the I/O subsystem happens to run S/3x0-style channel programs or not. DB2 workloads could be such a workload, as per my comment on DB2.
You don't even try to make sense.
Please go troll somewhere else.
And by the way, my Karma is good.
POWER7 runs at 4.25 GHz under 45nm design rules so 5.2 GHz or 5.5 GHz in a more expensive machine using a better manufacturing process with better system cooling does not seem impossible. PCs use air cooling and are supposed to have lower power consumption so of course the clock rates are not so high. Still there are Intel Ivy Bridge desktop CPUs sold that hit 3.9 GHz on Turbo mode and server CPUs that hit 4.1 GHz on Turbo mode.
The don't use nitrogen but they use water cooling.
From another person on another forum: He said that he has actually seen one of these. And that the "Flash Express" is actually an SSD of some sort.
Aside from the Apples to Oranges comparison that others have pointed out, which current CPU do you think is the fastest CPU out there? Core i7? POWER7? Itanium? Some MIPS or Sparc? Which one? And just equalizing their frequencies is not valid - one of the CPUs has far more registers, pipelines and cache than the other, so how would you equalize those?