IBM's iron grip on its big iron business just got a little tighter. Three small companies that have been trying for years to carve out little niches in the System z mainframe market have all withdrawn their complaints with the European Union's antitrust authorities. The most interesting of the dropped complaints against Big Blue …
...of the buggy-whip market.
Mainframes have lasted longer than anyone might have expected. But now that the Nehalem-EX has RAS features, the handwriting is on the wall, and soon there will be no reason to process data on anything but an x86 chip any longer.
So the real manaical laughter is coming from Sunnyvale.
IBM sold more Z systems last quarter than before.
When x86 can handle VM worlkoads like 'Z' then I'll happily by Mr Prickett Morgan a pint of wallop.
RAS is all very well in theory. However if Intel copy too much and make x86 (and there is an amobination of an Architecture if there ever was one) work too much like a MF I'd expect the Nazgul to hit intel right where it hurts. Remember IBM has been doing H/W VM a lot longer than pretty well anyone. They have the H/W patents to back it up.
IMHO, Intel need to drop this x86 emulation. The code translation takes up so much of the CPU die that you could easily trade that off for another two cores of RISC. Drop the X86 instuctions then you will have an architecture that can challenge 'Z'. Until then, by keeping the compatibility with past generations of processor (that might have been adequate for their time) you are frankly p*****g in the wind.
And coming up on the blind side is ARM. Te-he
"...When x86 can handle VM worlkoads like 'Z' then I'll happily by Mr Prickett Morgan a pint of wallop..."
No one believes that x86 can replace any Mainframe workload: Mainframes have superior I/O. But if we only talk about cpu performance, then any decent x86 solution can replace Mainframe workloads.
You need two IBM Mainframe Z196 cpus to match one old 45nm Intel Nehalem-EX in terms of cpu performance. And the Z196 runs at 5.26GHz and has close to half a GB of cache (L1+L2+L3). The new 32nm Westmere-EX is much faster than the old Nehalem-EX. And next year the 22nm Ivy Bridge version will be 40% faster than Westmere-EX.
"...Drop the X86 instuctions then you will have an architecture that can challenge 'Z'..."
Do you mean RAS will improve so x86 can challenge Z? Well, I believe that x86 RAS is not there yet. Far from it. As I have understood it, Mainframes have much better RAS. Mainframes are dog slow in terms of cpu performance. What you buy them for, is the RAS. And I/O performance.
Mainframes and high-end RISC servers have supposedly been on the verge of death by x86 since the Pentium Pro. At this point, repeating those claims is just comical.
What do you mean "on the verge of death by x86 since..."?
They've been 'on the verge of death' since they were invented and the very first computer geek profoundly announced "there will be no need for more than 3 or perhaps 4" computers.
do i have to start paying for cable now?
I ran the wire to the box myself so I don't think I should pay to watch TV.
I nominate you for worst analogy of the year
You don't pay for cable, you pay for content. it's not the same as a bit of hardware.
Marketing is fun to watch...
This aspect of Intel/IBM/whomever marketing is a barrage of opinions that are comical to watch. They do affect our careers, so some turns are painful.
Other commentator's musings about making x64 more RISC-like seem to ignore Intel's rocky Itanium experience. Intel seems locked into x86/x64 for the foreseeable future.
Let the prognostications begin!
Wrong lessons... bigger ultimate fail...
Itanium failed for a few reasons, one was using a ridiculous amount of die for poor x86 emulation. Another was building very cool looking stuff into the ISA that no compiler could actually target.
x64 won't become more RISC-like, it will become irrelevant and simply replaced by an ARM-like architecture running native ARM with the odd emulated back-up.
@Tinker Tailor Soldier?
A question, weren't the x86 emulation removed from later Itanium cpus? In the beginning Itanium needed hardware, but later the Itanium got fast enough to emulate x86 at decent speeds?
I like Itanium because it is much cleaner architecture than the bloated and buggy x86. Intel also concurs, otherwise Intel would not have released an rivaling cpu architecture aimed at 64bit and high end servers.
Too bad Intel did not bet on Itanium. Maybe because Intel wanted to learn RAS from HP, and now when Intel knows about RAS, Intel has incorporated Itanium RAS like features into Xeon. And now Intel has dropped interest in Itanium. I suspect this was the Intel plan.
@Kebabbert ...yes the x86 emulation was removed
But that didn't stop the fact that the first gen itanics really disappointed on the performance front which didn't help the marketing of later chips against x64. And although some parts of IA64 could be targeted by a decent compiler... e.g. the large register banks, some predication, register calling; other parts just really couldn't be, at least not generically, e.g. ld.chk, register bank rotation, multiple floating point states. Arguably also, code that isn't fairly carefully written isn't that amenable even to VLIW. One does wonder whether the power/performance ratios couldn't have been better with a less is more ISA that was very carefully targeted on what compilers could do with existing large code-bases....
It really seems as if a bunch of gee-wiz computer science-y types built this as a science project rather than a set of bloody minded "How many transistors are we spending for that, what does that really get us?" computer engineers.
Ultimately, the thing that will unseat x86 is power, you can look at it as losing cores (lost performance) or you can look at it as a bunch of logic that can't be independently power-managed since its difficult to shut down a part of your instruction decode pipeline. ARM hasn't really been targeted yet at the performance end of the spectrum, it's just not their niche, but, you can't help wondering what happens as the performance of ARM inches upward and more software gets ported to it.
"We are using the summer period to explore new business opportunities."
at the bottom of a lake, I presume...
Paging Florian Mueller
Florian Mueller, your humble pie is ready.
"I could swear I could hear an echoing bwah-haa-haa-haa off in the distance..."
Naw... You're just picking up McCoy Pauley through your Bluetooth headset.
Mainframe != IBM only.
Perhaps not, but
nobody ever got fired for buying IBM.
An emulator wont hurt
I do not see why an emulator will hurt them, it would be a very valuable tool for developing and testing software. They should ease the access to mainframes to help porting software.
I've been asked by our IBM outsourced customers to test/port our software to AIX and eventually also Z/OS. IBM provides access to hardware for free. And running AIX/ZOS seem to be cheaper than linux when outsorced to IBM, when you already have those already.
You can get far better IO on most other high end unix boxes, they just do not provide so many levels of fault tollerant microcode as the mainframe. Something IBM touts you head full of.
I've tested hot swappable memory in X3590 x86 servers and it is slow as hell, almost unusable. If you need this feature you have to sacrifice latency or pay the extra for the mainframe.
"...You can get far better IO on most other high end unix boxes..."
I didnt understand this. Do you mean that high end Unix boxes can have much better I/O than Mainframes?
Regarding the emulator. No, IBM will not allow emulators, because the emulators rival a true Mainframe. IBM does not want emulators to be a real alternative to light Mainframe load. Read here for more information. TurboHercules emulator can give up to 3.200 MIPS which is quite respectable, when you consider that software emulation is 5-10x slower than running native code.