Slashdot Mirror
Linux Leads MS in Itanium Support
lizrd writes "The New York Times is reporting (yeah, yeah, you gotta sign to read it) that several Linux distros will be shipping stable versions of Linux for Intel's new 64-bit Itanium chip on the day that it is released to the public. Microsoft however will not be supplying a version of Windows for Itanium until sometime in the fall of next year, several months after the expected May release of the new processor."
What I'm saying is, what if Itanium+Linux=crap just gets shortened to Linux=crap in the minds of some folks, even though the shaky new Itanium platform was really at fault?
STOP . AMERICA . NOW
Unfortunately, it's mostly irrelevant that Linux will be available first. How many hardware vendors are going to actually develop systems to use this chip and market them without Windows? Will Dell accept the overhead of carrying Linux-only hardware? I don't think so.
On the other hand, this may be a chance for VA Linux to have a product that you can't buy elsewhere.
The 1/2/2001 issue of Red Herring magazine has a cover-page article titled Limping Giant -
The much-delayed Itanium chip, Intel's most ambitious project in 32 years, is frustrating the chip maker's plans for life after the PC by Dean Takahashi.
I summarized the article as follows:
* Itanium is 2 years late
* Itanium was not ready for its official October launch event
* Itanium delay has:
o benefited Sun and frustrated HP and SGI
o created investor doubts
o reduced Intel's 4Q profit
o resulted in management shakeup
o Albert Yu was "the first head to roll"
* Intel has been diversifying but uP is still 75% of revenue
* Itanium is key to maintaining high uP margin
* Itanium delay has been attributed to:
o loss of manic focus on chip design
o loss of engineers to communication chip makers
o poor execution
o frequent spec changes to keep pace with competitors
o departure of Lew Platt, Richard Sevcik, David House
o layout delays caused by off-shore design group
o an over-inflated design team (300 to 500 engineers)
o overspecialization of circuit designers
o difficulty in coordination
o high number of inexperienced engineers
o high engineer turnover rate (30%)
o frequent abandonment of large blocks of finished work
o a committee that demanded approval for everything
o declining engineer morale
o bad blood between Intel and HP engineers
o changes in target manufacturing process
o failure to manage chip complexity
Download Mazes and Puzzles from www.puz.com
Actually, WinNT on Alpha ran in 32-bit mode IIRC.
Also, I've had a chance to use a Linux-powered Itanium system for a while. Not as impressive as it could have been, since GCC is not up to snuff so far in dealing with instruction-level parallelism (a very high percentage of the output instructions were noops). It can only get better though.
The people who do run into problems are using hardware where there are problems.
For example: I can pretty much state that you weren't trying to do a multi processor install on a machine with an embedded Adaptec SCSI chip. If you did that you would find out very quickly what people were complaining about.
It remains to be seen if the Itanium is really where personal computing is headed.
Perhaps the PC industry has always remained stuck in the x86 rut not out of choice, but because it couldn't escape.
The important software was from Microsoft. i.e. DOS. Win 3.1. Win 9X. It only runs on x86 processors. In recent years the very latest MS OS's could run on other processors. Only this year does MS even have an OS (Win2K) that they could port to other processors, like Alpha, MIPS and PPC. AFAIK they haven't yet. So even with MS's capability to become processor neutral (nevermind the software packaging issues that arise) it will just take time, and demand for things to slowly move out of the x86 rut.
I've often thought (when not distracted by cute guys needing tech support) that if all major software were processor neutral, and the over the counter software packaging issues were solved that the hardware designers would be truly free to innovate once again instead of trying to remain compatible with not only 20 year old hardware, but misdesigned 20 year old hardware.
If initially non x86 processors are expensive, it's just supply and demand. So we may have a serious inertia problem to get unstuck from the x86-only rut even if the other impediments are removed.
I'll see your senator, and I'll raise you two judges.
Microsoft's compiler is a total joke.
If this is true, and I'm not disputing it; then I find it simply amazing.
If MS is keeping their eye on the ball then they should realize the importance of having a good compiler, development tools, etc. I don't use VC, but I know people who swear by it (not at it) and seem to think MS has invested huge resources into it. So why not good fundamental compiler technology?
Well, geez, with MS's resources, couldn't they just hire some gorgeous compiler dudes?
I'll see your senator, and I'll raise you two judges.
So MS would wisely wait a few weeks or months to make sure their OS installs 99+ percent of the time.
But immediately upon the release of a new processor, just how great of a diversity of different types of hardware do you anticipate there to be?
Me: not that many. Distro vendors could likely support them all well. It's not like trying to support the diversity of legacy x86 boxers. MS has a huge collection of knowledge about such boxes -- often boxes which no longer even have documentation available, giving them a tremendous advantage autodetecting older hardware. And they've had a long time to perfect their autodetection techniques. (Well, polish, not perfect.) But for the future Linux gets to start off on a more level playing field. As time marches on and legacy boxes go to heaven (or hell, maybe only non-x86 boxes go to heaven?) this only works more and more to Linux's advantage.
I'll see your senator, and I'll raise you two judges.
Win2k isn't all that much better. It's an order of magnitude more stable and scalable than NT but installation is still a major pain, and despite what the M$ flacks will tell you there are still blue screens galore.
What a complete pile of bullshit. It is remotely possible that you're running 2000 on a complete POS PC in which case perhaps it's possible that either faulty hardware or crap drivers (for example ATI still hasn't figured out how to make stable video drivers), however BSOD's are incredibly rare on 2000, and on any variant of NT 4 as of SP4. While my job is software architecture, incidentally I run several NT4 machines that are heavily loaded with web, file, and database services and they are only ever rebooted when serious security fixes come out (in which case it isn't a big deal to install and reboot in off hours. If it was a 24/7 machine I'd have it clustered and it still wouldn't be an issue). My development machine has never BSOD'd in 2000. EVER. And I guarantee I beat my machine as hard or harder than anyone out there. Again if you run POS hardware well...
Is Windows 2000 perfect? Hardly. explorer.exe crashes on me every now and then. Actually on a GeForce MX equipped machine it crashes at a frequency that I would term "often" : Maybe once a day. Ctrl-Alt-Del...find explorer.exe...kill process....Run "explorer.exe". Only harm is that ICQ no longer has the cute little system tray icon.
I guess the problem is that ATI's shoddy work shouldn't be in a position to be crashing the machines. That would be a design decision if I'm not mistaken..
The same design decisions are taken with Linux where any number of device drivers are fully capable of taking the ship down with them. While it would be nice to have a fully microkernel architecture ala QNX where nothing but a tiny segment of core code can take down the machine and things like video drivers can be killed an restarted, the performance hit was unwanted so Microsoft made what I'd consider a pretty reasonable compromise and decided that as of NT 4 they'd integrate several outer ring drivers in ring 0. The idea is that the vendors would be expected to pursue a microkernel type system in their drivers (i.e. not sticking everything in the DD), and what they do put in there is super-ultra-mondo checked. I have found that Matrox, for example, does an awesome job and they have excellent video drivers. ATI on the other hand has single handedly, IMHO, given Windows NT/2000 a bad reputation. I like ATI however they need a serious shit kicking for the crappy, unstable drivers that they put out. This is all anecdotal and is IMHO.
It also helps refute the "chasing taillights" argument.
Perhaps the PC industry has always remained stuck in the x86 rut not out of choice, but because it couldn't escape
Er, no. Transitioning to NT for Alpha with FX!32 was no more technically difficult than the 68k-PPC transition. The reason that the PC industry has "remained stuck in the x86 rut" is because it didn't have a dictator like Apple to force a transition, and because RISC isn't magically superior.
Software designed and optimized for a consumer-grade desktop RISC chip doesn't perform any better than software designed and optimized for a consumer-grade desktop x86 chip. Relative "elegance" is irrelevant -- there is no material benefit to a switch.
Nor is it an effect of market position giving higher-volume consumer chipmakers an advantage -- otherwise a top-of-the-line Motorola chip would certainly outperform a top-of-the-line AMD chip, given the 1995 positions of each company. But the Athlon manages to kick the consumer-grade PPC's ass anyway.
All the bullshit about RISC superiority is just that -- bullshit. Inertia is an excuse, not an explanation.
There's no "we" in team, only "me"
(Posting as AC so Intel doesn't take my IA64 boxes away for talking too much) I've been developing on/for IA64 hardware for several years now. The hardware is not ready for release. Linux is not either, no matter what Turbolinux claims. I wouldn't say whistler/64 is behind Linux. It seems quite a bit more stable than Linux on multiproc machines. I'll probably get modded down for this but the stability and maturity of the MS compiler is way ahead of GCC for IA64. C++ support is one area where GCC for IA64 is severely lacking. Check out the mailing lists on linuxia64.org to keep up to date and find out where you can help.
There are a lot of hightly specialised Linux distros out there (check out the distros page on lwn.net) with all sorts of uses (routers, terminals, servers, workstations, ...). There is no real shortage of resources for Linux developement. When something new is needed, it's generally done by the group that needs it which is often a new group, not an old one abandoning their previous project.
Bill - aka taniwha
--
Bill - aka taniwha
--
Leave others their otherness. -- Aratak
How backwards compatible is the Itanium?
The article mentions it contains a 32-bit section which allows it to run older x86 instructions.
What I'm wondering is if the reason for the new versions is just to take advantage of the new 64-bit world, or if you could actually just install say Windows 95 onto it and live crippled?
Just kind of curious how important this OS battle is to the adoption of the processor.
I'm not sure what kind of market demand there will be anyway. I don't see most computers today as being CPU-bound.
It's not really even significant. I doubt there are going to be a tremendous number of Itanium sales next year, anyway. It's nice that early adopters use Linux, but not Windows, but not very significant.
The more interesting question is about gcc. How is support for Itanium coming with gcc? The EPIC architecture probably requires a lot from the compiler to take good use of it. I assume that gcc *does* support Itanium, since Linux is running on it, and porting Linux to another compiler would probably be more effort than porting it to another platform that gcc targets.
If Microsoft has a significantly better compiler, Windows will probably be a much better system for Itanium. I've heard of Intel's involvement with gcc, so I doubt that MS will do much better, but still, support is just a baby step in the battle for the best system.
If you are modding me down because you disagree with me, use the "Flamebait" category, not the "Troll" one.
Nothing!
The only thing of interest is if Intel actually releases and pushes the chip in any way before MS is ready, if so then perhaps we can safely say WIntel is dead. If they hold off anything other than a "you can get it now" release then it's another point in the split MS argument.
Never underestimate the dark side of the Source
Has anyone moved up to a 128 bit processor? DEC had a good long head start, having introduced a 16 bit procesor back in 1970 (My DEC assembly language book hypothesizes that a 32 bit chip should enable faster processing but would be prohibitively expensive to make :-) but they haven't kept that margin, though I seem to recall that they have a great wide memory bus.
I'm trying to teach myself to set people on fire with my mind... Is it hot in here?
AMD has an incoming 64-bit architecture which will be binary compatible with the x86 (much like the 386 coud run 8086 software unchanged). I'd put my money on them over the Itanium. Haven't read much about it lately, other than a Slashdot announcement awhile back. Anyone has further news on that?
I couldn't give you the details, but if it was all that simple, Linux/MS would not NEED to release a new version for it, would they? They'd get around to an optimized version later.
-bugg
The whole POINT of the "PC REVOLUTION" was in "commodity hardware". That is, get tons of manufacturers churning out parts that ran to the PC spec set forth by IBM (forth - um, open firmware joke, get it? nevermind).
The point was to develop this cheap-ass piece of junk platform to the point where people didn't need to pay extortionate fees to Sun, DEC, SGI, HP, Intergraph, and the mother-ship, IBM, etc. Now, DEC is gone, HP is just another Packard-Bell, so is Intergraph, and SGI, is acquisition fodder. Only Sun and IBM really remain as strong players. I'm guessing that has nothing at all to do with the PC revolution, and more to do with the Internet revolution and the need for bulletproof servers.
Until Intel got a monopoly in chips (AMD was a nice try, but are they REALLY positioned to harm intel? Last I checked, intel was still dictating platform standards) - it was an open platform and the dream was alive. Someday, there was going to be a beefy and robust PC that could replace expensive minis at commodity hardware prices, and run an OS grandma could admin. Then Intel figured out that with a monopoly, they wouldn't have to compete with any other players, they could set the standards, and block this insanity from happening. Sure, they'll still be productin commodity hardware, but they'll be using the enterprise pricing model. And using their IA-32 market dominance to crowbar Itanium into the enterprise server market, no matter how inferior it is, technically. If it runs Win32, it's golden. No matter how overpriced it is. No matter how much laughter it generates when placed next to REAL enterprise hardware.
It's called market segmentation. The Celeron/Xeon thing was a small-scale application, and proof of concept. Look at the technical difference between Celeron and Xeon. Then look at the price difference. You could put a Xeon in a desktop machine, and benefit, but the price made it not worth it. Granted, Itanium will have a big technical difference - PCs DO need to go 64 bit to be serious in the enterprise server market. But they need MUCH more than that - in a practical sense, less performance for more $? Crazy. That's market segmentation. A tool designed to artificially constrain supply in the marketplace, to drive up prices, while not suffering from constrained supply (and high costs) on the manufacturing end. The results? Pure profit. Bring lots of vaseline.
These are my friends, See how they glisten. See this one shine, how he smiles in the light.
After I did the install an MCSE wiped the disk and spent 3 days trying to do the reinstall. (NT Server). He refused to believe me when I told him what was wrong. After he gave up a third party then spent 2 weeks trying to do the same install. After I wrote down the procedure (which involved doing more than just booting from the diskettes - you had to install even newer drivers from the ASUS web site - and yes I tried to install them with a CD boot - Nothing but Blue Screens) he was able to get it to work also.
Nothing 'piece of cake' about it - it was a major hassle.
As I said - you were working with equipment which didn't have any problems with NT - the other people weren't.
And no - the instructions on the ASUS web site didn't mention using the diskettes. Or even that there were other drivers to install.
When you are a young smart ass with little experience everything looks easy to you. You assume that other people are just stupid - when the fact is that you just haven't been presented with the same set of problems that they have seen.
When you run into a rip roaring bitch of an install you won't be quite so much of a smart ass anymore. One of the great lessons of life is that it knocks the arrogance out of you.
No login required!
m l
http://news.cnet.com/news/0-1003-200-4236527.ht
Okay, so Intel is about to ship it's Itanium, and Microsoft doesn't have an OS to power the new architecture. But Linux is ready to support the new chip...
...hmmm. I just don't see a problem here.
(Not, however, first post.)
If you can read this, then I forgot to check "Post Anonymously".
I wonder if this is at all a negative for the Wintel corporate structure
Careful now! The judge could just decide to split the Wintel monopoly into two seperate corporations.
I'll see your senator, and I'll raise you two judges.
Note the CNET in the URL. It means the the NY Times just reprinted CNET's Original Article...at least you don't have to sign up for regristration!
Doh!
It's nice that Linux and NT have been ported to those processors. Many applications will benefit from that. Having a 64bit address space available greatly changes how you can approach many problems in OS design and implementation. To take real advantage of 64bit (and beyond) processors, we should have new approaches, not just old code widened.
However, having said that, the on the proven technology front it can only be good for Linux to be ahead - as then it gets the label of being proven sooner than MS Windows, which will be uppermost in IT managers minds.
If you ask me, the battle on this front will not be decide next year, but the year after, when the Itanium is expected to start pushing into the mainstream server market.
KTB:Lover, Poet, Artiste, Aesthete, Programmer.
KTB:Lover, Poet, Artiste, Aesthete, Programmer.
There is no
Wait, doesn't RedHat distribute a "Personal" and "Server" version of the same fucking product with a couple config tweaks and a couple extra deamons?
I'm a loner Dottie, a Rebel.
I wonder if this is at all a negative for the Wintel corporate structure...
Just think, Linux boxes will be running Itanium right when it comes out. But, big surprise, Itanium turns out to be a little buggy. By the time Windows supports it, the bugs have been ironed out.
Bravo for Windows! Suddenly running more stable than Linux!
If RH et al play their cards right, this could be a massive PR coup.
Alternatively, MS could use the information to locate potential problems, and use that for their OWN PR coup.
But the bottom line is thus - whoever gets to market first has a strategic advantage. And the longer they hold that, the more defendable it becomes.
The itanium is a dead processor already, but it's strategic value, in terms of mindshare, is very significant.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
Uh.... how incompatible.... in so mach as it runs any x86 code, back so far as the 8086 back in the 70's?
Really? So the Itanium isn't some new-fangled CPU that breaks with the past after all? It can run x86 code as is, natively, without emulation? I had no idea.
IA64 is definitely not intended for the personal computing market - at least not for a long time yet. This is a server processor
Ah, that brings back memories. Every new processor is always "initially targeted at the server market" going all the way back to the 386 at least. Makes me smile every time I see it!
LOL!
I tried to set up a Windows NT gateway for someone and tried for HOURS to get two network cards working (a 3com, and a generic ne2k). After countless! reboots, I gave up, installed a copy of 98, and came back the next day and installed linux. Both 98 and Linux worked fine w/ the two network cards, but NT, a ``professional, workstation, server, etc'' OS would not use both cards (I tried switching the slots, everything, to no avail)
He who knows not, and knows he knows not is a wise man
That's a bit optimistic. Try Spring 2002. Trust me.
These are my friends, See how they glisten. See this one shine, how he smiles in the light.
Itanium would surely be rolled out targeting the Advanced Server market and thats where MS wanted to be. And not just MS, add Sun and several others to the list too.
If Linux could just target the server market and ignore everything else, with the new distros running on Itanium Servers would mean a lot to firms who would look for solid servers.
One up for Tux
Microsoft - sorry guys, do better next time!.
Rapid Nirvana
Linux supports every damn appliances on the market. You name it, if it has a processor, Linux runs on it. This is news only insofar as the first Itanium boxes MUST have Linux to run, but you know being the first to the market does NOT mean you will keep that market. Just ask Woz how powerful the company he co-founded became.
Burn Hollywood Burn
...because there's not exactly going to be an immediate, huge demand for a greatly overpriced, unproven processor that's incompatible with just about everything that's been built up in the PC-clone era of the last 19 years. Similarly, you can't run Windows 98 on a PowerPC or Alpha. Does it matter?
It remains to be seen if the Itanium is really where personal computing is headed. After all, Intel has introduced other non-x86 processors in the past and had very high hopes for them. The RISC i860 processor, introduced in 1989, is a good example. The successor, the i960 is still available. But these chips are outside the x86 realm, and there's reason to think that the Itanium could be as well. Moving to an entirely new processor *family*, not just the next generation of what's currently available, is not to be taken lightly. This is doubly true when the benefits of such a change are not at all obvious.
I have had Windows2000 for several months now and have yet to be able to install it on either my laptop or my desktop. It always tries to load loads of drivers for hardware I don't have and then stops, requiring a reboot. On my dad's PC it installed first time without any hassles and my PCs are newer than his. Contrast that with Linux where I have never had a problem installing RH5.2, 6.0, 6.1, Mandrake 7.0, 7.1 and SuSE 6.4. I've been told it's Compaq's fault that I can't install 2000, but then that's the MS way; blame everyone but themselves for the problem.
I've done it on a Dell with dual-CPU's, embedded Adaptec SCSI, and Dell's PowerEdge RAID controllers.
Piece-of-cake.
Why does MS insist on this lunacy? How different are the codebases? Or is it just a bunch of registry settings that change the versions as in NT 4.0 (according to O'Rielly at least)?
Why do they make such a big stink over something that should be a Radio Box at install time?
The only reason is $$$ that I can think of. Kinda like the IDE cards that sold for $60, but you put a resistor on them, and they were the companies $200 IDE raid card.
We have seen these kinds of stories over and over again on just about every Linux friendly site out there. I love Linux, and I have the utmost respect and admiration for all of the people involved with this, especially intel for taking such a pro-linux stance the whole way.
..." do nothing but make us look just as arrogant as Microsoft marketing. Stop talking shit about Windows, and just spread the good word. People don't need to be told that Windows sucks, they all learn it eventually.
But I need to say something to all of the web admins/authors/.etc out there.
STOP GLOATING YOU ASSHOLES!!!
Story titles like "Liunux leads MS in
Linux distros typically do not install perfectly everywhere. Even Mandrake, which I think is the best, cannot use my external CDRW.
Linux distros can ship a product faster because if it installs properly 95 percent of the time that is good enough. 95 percent may not be good enough for Windows, though.
So MS would wisely wait a few weeks or months to make sure their OS installs 99+ percent of the time. Linux users are a forgiving bunch with respect to features that don't install or work. I installed a half dozen Linux's until finally Mandrake 7.2 worked with my ATI graphics chipset.
My 78 year old father is not so forgiving. If a piece of electronics doesn't work right out of the box he returns it and buys something else. That is the target market for Windows. (Maybe not quite, but I think that Windows aspires to more hardware compatibility than Linux does.)
Marjo Wycam, Master of the Programming Arts