IBM introduces new powerful mainframe computers

Aug 28, 2012
IBM logo

IBM on Tuesday introduced a new line of mainframe computers the company calls its most powerful and technologically advanced ever.

IBM said its zEnterprise EC12 mainframe server is designed to help clients securely and quickly sift through massive amounts of data, meeting the demands of retail and other clients in the age of "Big Data." Running at 5.5 GHz, IBM said the that powers the mainframe is the fastest chip in the world. Processing speed is 25 percent faster than the previous model.

Mainframes are used by ranging from banks to chain stores. IBM says the new model could be used by retailers to manage online transactions and analyze clients' buying habits and then use the information to create a "more customized ," such as a custom coupon issued during a transaction.

"Whether its retail or whether its transportation, making reservations, whatever it is, the system has been built really to help clients do those new types of new-age transactions," said Doug Brown, an IBM vice president of marketing.

IBM says more than $1 billion was spent on for the system at 18 sites worldwide, with most of it in Poughkeepsie, about 40 miles (65 kilometers) north of its headquarters in Armonk in Westchester County.

The new mainframe is being promoted as one of the most secure systems ever with a tamper-resistant cryptographic co-processor to provide privacy for sensitive transactions.

IBM has been focusing on its software and services divisions, which are more profitable than selling the mainframe computers that made the company famous decades ago. But the sales of those mainframes help feed demand for IBM services.

Explore further: Ineda developing low power companion processors to increase battery life for wearables

4.2 /5 (5 votes)
add to favorites email to friend print save as pdf

Related Stories

IBM Mainframes Go 3-D

Apr 27, 2007

By integrating its Cell processors with the mainframe, the company hopes to create virtual worlds on the Web that benefit gamers as well as the enterprise.

IBM offers concession in EU antitrust case

Sep 20, 2011

(AP) -- U.S. computer maker IBM Corp. on Tuesday offered to make it easier for competitors to provide maintenance services for its mainframe computers, a concession to get European regulators to close an antitrust probe.

EU accepts IBM concessions in antitrust probe

Dec 14, 2011

(AP) -- The European Union's competition watchdog says it has accepted concessions made by IBM to make it easier for rivals to perform maintenance on its mainframe computers and is closing its investigation of the company.

IBM Rolls Out New Mainframe

Feb 26, 2008

IBM today announced the System z10 mainframe to help clients create a new enterprise data center. The system z10 is designed from the ground up to help dramatically increase data center efficiency by significantly improving ...

Recommended for you

FIXD tells car drivers via smartphone what is wrong

2 hours ago

A key source of anxiety while driving solo, when even a bothersome back-seat driver's comments would have made you listen: the "check engine" light is on but you do not feel, smell or see anything wrong. ...

Watching others play video games is the new spectator sport

8 hours ago

As the UK's largest gaming festival, Insomnia, wrapped up its latest event on August 25, I watched a short piece of BBC Breakfast news reporting from the festival. The reporter and some of the interviewees appeared baff ...

User comments : 13

Adjust slider to filter visible comments by rank

Display comments: newest first

Phil DePayne
1 / 5 (1) Aug 28, 2012
god damn
Deathclock
2.3 / 5 (12) Aug 28, 2012
5.5ghz is 90 picoseconds (90 trillionths of a second) between clock transitions... that's pretty damn fast.
Physmet
5 / 5 (9) Aug 28, 2012
I love the technology. I think it a crazy commentary on our society though that with all this power, one of the touted uses is to "create a 'more customized shopping experience,' such as a custom coupon issued during a transaction." :D
Deathclock
2.6 / 5 (10) Aug 28, 2012
5.5ghz is 90 picoseconds (90 trillionths of a second) between clock transitions... that's pretty damn fast.


I don't know what the 1 ratings for this are all about, but if it's because you think I am wrong:

1ghz is 1ns for a full cycle, thus 5.5ghz is 181.8ps for a full cycle, hence one half cycle (or one clock transition) would take 90.9ps.

So either you think I am wrong (and are wrong yourselves) or the 1 ratings you gave that post are dishonest and based on being pussy hurt over some time in the past when I embarrassed you by proving you woefully wrong (which I have done to many people on this site...). It won't make you feel better to know that I no doubt don't even remember the incident that you're all upset about.
Aloken
3.4 / 5 (5) Aug 28, 2012
So either you think I am wrong (and are wrong yourselves) or the 1 ratings you gave that post are dishonest and based on being pussy hurt over some time in the past when I embarrassed you by proving you woefully wrong (which I have done to many people on this site...). It won't make you feel better to know that I no doubt don't even remember the incident that you're all upset about.


Either that or people just don't care about an inane comment derived from basic math. There's no need to get defensive about it or care about ratings at all.
Deathclock
2.7 / 5 (7) Aug 28, 2012
So either you think I am wrong (and are wrong yourselves) or the 1 ratings you gave that post are dishonest and based on being pussy hurt over some time in the past when I embarrassed you by proving you woefully wrong (which I have done to many people on this site...). It won't make you feel better to know that I no doubt don't even remember the incident that you're all upset about.


Either that or people just don't care about an inane comment derived from basic math.


I'd be surprised if most posters on this website could count to 20 without removing their socks...
El_Nose
1 / 5 (5) Aug 28, 2012
@deathclock

I gave you a 1 case you are technically wrong -- i figured that why everyone else did it too ... was I wrong?

anyway working backwards through your calculation 90 trillionths is one half cycle.

0.000 000 000 000 9
...............................^-90 trillionths
............................^--trillionths
......................^-- billionths
..............^-- millionths
.......^-- thousandths

so 0.000 000 000 000 9 * 2 = 0.0000000000018
1/0.0000000000018 = 555,555,555,555 Hz
^- kilo
^- mega
^giga

you are stating a speed of 555 GHz not 5.5

-- take (1/5,500,000,000)/2 = 9.0*10^-11

you were off by a couple of orders of magnitude ^^

happy trails ( i am always nice -- it feels good to be snarky)
Tiranasta
5 / 5 (4) Aug 28, 2012
@El_Nose Uh, no. You're the one off by two orders of magnitude. The number you specified is 0.9 trillionths, not 90 trillionths.
Deathclock
2.6 / 5 (5) Aug 28, 2012
@deathclock

I gave you a 1 case you are technically wrong -- i figured that why everyone else did it too ... was I wrong?


Yeah you are wrong.

I know for a fact that 1ghz has a cycle time of 1ns. I know this because I know that 100mhz has a cycle time of 10ns, because I use this daily in my work with a 100mhz processor and time-critical algorithms that measure propagation time of a pulse of light in an optical fiber... (also, t = 1/f)

if 1ghz is 1ns, then we divide 1ns by 5.5 to get 0.181ns @ 5.5ghz... which is 181ps. Half cycle is 90.9ps, since I was talking about a single clock transition.
Eikka
5 / 5 (2) Aug 28, 2012
Regardless of which magnitude of power it really is, here's an interesting fact about processors: they're not completely synchronous.

Parts of the processor have to operate faster than the "global" clock rate, because they're supposed to do their jobs between the clock pulses. When you set the inputs of, let's say an ALU to add two numbers together, the result has to appear in the outputs before the next clock cycle issues the next command to read them out. That means the transistors in that part of the chip have to operate many times faster to settle down into a steady state before the data is read out. Otherwise you get data corruption and the machine crashes.

And that's why you read news that scientists have managed to produce a transistor that switches at 150 GHz, yet the processors themselves are 50 times slower.
Deathclock
1 / 5 (1) Aug 28, 2012
Parts of the processor have to operate faster than the "global" clock rate, because they're supposed to do their jobs between the clock pulses. When you set the inputs of, let's say an ALU to add two numbers together, the result has to appear in the outputs before the next clock cycle issues the next command to read them out.


What? No...

Instructions can take more than one cycle... most do. An addition operation might take 2 or 3 cycles depending on the architecture... and you have a multi-level pipeline feeding instructions at all times... there are multiple instructions in process (fetch/decode/execute) at all times... new ones can start before previous ones end, if there is no dependency between the result of the first and the second.

Modern processor architecture addressed the problem you're talking about a long time ago with pipelining.
Deathclock
1 / 5 (2) Aug 28, 2012
new ones can start before previous ones end, if there is no dependency between the result of the first and the second.


...and when there is a dependency between two contiguous instructions it's called a pipeline stall and compilers can usually optimize code to greatly minimize these occurrences. It seems your knowledge of this (if not merely intuition) is outdated by two decades, at least.
Mike_Massen
1 / 5 (3) Aug 29, 2012
So now a microprocessor is considered a mainframe - LOL!