“IBM Corp is lining up the launch of its first xSeries server based on Intel’s 64-bit Itanium-2 processor for the first quarter of 2003, and has confirmed that it has dropped plans to bring its AIX Unix operating system to the Itanium processor platform.” Read the article at TheRegister.
IBM Readies Itanium 2 xSeries, Dubs Linux the Unix for Intel
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Okay, Itanium arguably isn’t going to be anywhere near the “biggest” of the big iron, even in a quad-procy config as the article mentions, but still — not bad! From what I heard, Intel finally got Itanic 2’s performance Right(tm), even if the price is still most definitely Wrong(c).
Here’s to hoping IBM will be able to make a *very* large fat margin on the servers, and continue with their Linux strategy!
Aside: how are GCC’s Itanic 2 optimizations? Or would IBM likely end up using the Intel compiler?
From what i’ve read Itanium2 is giving an improvement of between 1.5 and 2 times the performance of the standard Itanium running apps compiled on Itanium (I), while if the programs are recompiled on Itanium2 the performance gain is suppose to be 2.5 times the Itanium. This is on a same clock speed. The one problem with the Itanium2 for the moment is that it’s a power hog. This will hopefully be resolved next year when they switch from .18 process to a .13 process.
Also Itanium2 has alot of Cache on board (L1, L2 and L3) along with the die shrink next year there will be a cheaper version with only about 1mb of L3 cache (as oppose to either 3mb or 6mb of L3)
<RANT>
why is IBM supporting linux in that way, IBM is one of the biggest TCPA players.
hence the question will there be a palladium for lee nooks…
scarey..
</RANT>
Apparently HP already offers some Itanium servers (not sure about Itanium 2) but people aren’t buying them. The good’old PA-RISC ware is much more in demand, no contest.
reminds me of the fate the Windows NT for Alpha had: people just didn’t buy the blasted thing. VMS and Tru64, that was OK with Alpha, but not NT.
If you wonder where is the analogy: the problem was that people didn’t have any software to run on the WinNT for Alpha. The exact same problem is with all the unices ported to Itanium: no software available.
Plus, in the big-iron market you want a very reliable, time-tested CPU architecture. Itanium is not.
HP has been offering Itanium 2 for at least couple of months now. I was looking over Summer especially once the chip was released. I couldn’t find anyone else do did at the time. Got to find some way (easy as well) to access huge chunks of memory (>4GB for a single process) since you can not do this with Intels 32 bit design.
Wonder why they didn’t mention that AMD’s Hammer series is also due about then…
Thanks Fred. Needless to say, we’re not too hot on purchasing Itanium-based HP-UX machines, either. that would account for my ingorance.
Eugenia, shouldn´t it be “view down moderated comments” (or down-moderated)? Not that we care a lot about them but anyway 🙂
Why does IBM need to invest in Intel when it has the power4 and other processors that it manufactures?
“the problem was that people didn’t have any software to run on the WinNT for Alpha. The exact same problem is with all the unices ported to Itanium: no software available.”
I remember something called fat binaries (normally mentioned in talks about macos moving to intel arch.) where the same software works on more than one architecture, x86,ppc,Itanium2,etc.
Does linux has this feature, and is it a possible fix to for the software problem?
Fat binaries are simply binaries that contain code for more than one processor architecture. That code still has to be compiled to run on each platform. The only real benifet to fat binaries is that that they make distribution of multiplatform software easier. While Linux doesn’t have a similar feature (though it would be trivial to hack it, since Linux already supports multiple executable formats) the relience in the Linux world of package managers (which automates the distribution process) kind of makes fat binaries redundent. Besides, the original poster is pretty off. Linux software by and large is open source. You can build a pretty good mid-range server (what Itanium seems to be aimed at for the moment) with just Open Source tools. Since most open source products already run on multiple architectures, and UNIX software (unlike NT software) is generally designed to be portable, a minimal amount of porting work will result in Itanium systems having a huge software base right off the bat. For the really heavy duty stuff, only few closed-source applications (which by and large already have Linux ports) would have to be ported to Itanium, which seems quite likely thanks to existing industry support.
..very sueful Linux applications that just have to be proted to Itanium to make Itanium servers useful… why is that people don’t want them?
“the problem was that people didn’t have any software to run on the WinNT for Alpha. The exact same problem is with all the unices ported to Itanium: no software available.”
Actually, it’s not as much of a problem for the Unices as it is with NT. Much of the “interesting” software for Unix, such as Apache or Samba, is open-source, so it can, at least theoretically, be recompiled for Itanium. Also, some of the interesting closed-source software, such as DB2, is owned by IBM, which is perfectly capable of porting its own software onto its own machines. There is enough software for early adopters to use on these new Itanium machines.