Follow Slashdot stories on Twitter


Forgot your password?

Intel Shipping Merced Engineering Samples 91

JDC writes "Apparently, Intel has first silicon on their IA-64 processor. I wonder how long till I can play Descent on one of those babies... ;) " Linux has been working on the simulators, and is reported to work on the actual silicon. However, contrary to prior reports, Windows does boot and run on the engineering prototype.
This discussion has been archived. No new comments can be posted.

Intel Shipping Merced Engineering Samples

Comments Filter:
  • 1) Windows doesn't have 16bit code for hardware support. At least it shouldn't. Who uses that kind of hardware these days? 32bit processers have been around since the *386*. It's about time Microsoft stopped supporting the 286, don't you think?
    2) How many drivers has Microsoft written. I'd suggest that they've probably written fewer than the Linux folks have. Remember, we've only very recently got support from hardware vendors. Everything had to be reverse engineered in the past. It's a wonder we support as much as we do!
    3) Linux is Y2K clean. Remember, Linux is the kernel. (Duh). GNU/Linux may not be, but that's up to the application writers. If ipopd doesn't work, then get one that does. And again, we don't know how many of Microsofts *applications* aren't going to run. (Regardless of the fact that Windows doesn't even come with a POP daemon. But I think I've made my point by now).

    Yes, if it comes from Microsoft it's almost always bad. And if it's good, it's because Microsoft bought it from somebody else. (VC++, IE, etc)
  • Well, yes and no...

    1. 64 Bit Memory and IO
    IA32 has 64 bit memory bus and instruction that can access with 1 atomic operation. However,
    there's no 64-bit single cycle arithmetic...
    (unless you count CMPXCHG8B, CDQ, FILD, FISTP, MOVQ, SHLD, SHRD, and MMX packed bit instructions)

    2. Virtual/Real memory access. With IA32, you can access more memory than 2^32 *bytes* of memory
    (up to 2^36 *bytes* just not linearly), but not as much 2^64 *bytes*...

    However, for the most part, those are the main differences between 32bit and 64bit architectures.

  • It's not just 64 bits in this case; there are several new processor technologies. The most significant of which are: very long instruction words (the compiler packs multiple operations into a single instruction word); asynchronous memory reads, writes and asynchronous stack management (tell the processor to do a load several instructions before you need it and then check if it was successful later, stack management is done by allocating registers and the memory stack writes are automatically done in the background); true speculative evaluations (conditions are built into the instruction set so there's no loss when a speculative instruction never executes); register rotation (a group of number registers can be rotated so that several passes of a loop can be evaluated at the same time(using VLIWs)).

    Despite the processor changes, the harder challenge has been to write a compiler that can take advantage of these features which are actually the cause of the greater performance.
  • G4 info here: uctor/cpu/7400.html

    and lots of it. Anyone know estimated spec, MIPS, or GigaFLOPS for Merced at a certain MHz?
  • There is no need for 128bit chips, now or ever.
    64 bits is enough to address 18 million terabytes.
    More than the number of atoms in the earth...

    Don't fall into the trap of thinking 64 bits
    is twice 32 bits... 33 bits is twice 32 bits, each
    bit doubles what you address.
  • the only intresting comments posted where about Gcc and parallel nature of the code

    IA64 is all about paralleling the intructions
    (althougth im intrested in the switching registers bit V funky)

    has anyone looked at the backend part of gcc that intel are donateing ?

    they have spent just over 2 years on the compiler and I want to see what its like

    linux relies on gcc

    will an intel compiled linux be better than gcc ?
    (I think so is this right)



    a poor student @ bournemouth uni in the UK (a deltic so please dont moan about spelling but the content)
  • > Typical Linux drone. If it comes from Microsoft it's bad, if it's Open Source it's good.

    Sorry; most "Linux drones" that I know are more honest than that: our actual position is If it comes from Microsoft it's probably bad; if it's Open Source it's probably good.

    And when open source writers do screw up they admit it and fix it, rather than (say) blaming it on the actions of an elite hacker group with esoteric knowledge of obscure system internals like (say) http strings.

    It's all a matter of track record. Microsoft doesn't have a bad reputation because they happen to be called 'Microsoft'; they have a bad reputation because they make crapware and lie their asses off about it. And pay astroturfers to help them mislead the public.

    p.s. -- Notice, if you will, that we scrupulously avoid comma splices as well. [Ouch! Sorry! I couldn't help myself!]

  • OK, but there are even more atom in th universe, and sooner or later we are gonna simulate a chunk of it
  • an R&D facility for the bigname company that I work(ed) for. Can't say who, but there's been a patent flap lately...

    The guys at the plant hate the stupid thing. 32 bit software runs pitifully on it. Got to see an 8-by Tanner box running though, that was cool...
  • Minor quibble with your post:

    "the reason the first Pentium Pro's didn't do so well when they came out was due to the fact that they didn't support 16-bit code very well."

    Not exactly, but close. The Pentium Pro's run 16-bit code like a screaming demon. The problem is that Win95/8 frequently switch between 16 and 32 bit mode (thunking?) and that operation was very slow on the PP's.
  • Wow, get with the program.

    Alpha, Sparc, MIPS, PPC are all shipping 64bit and are cheap to boot, are you really that far behind the times?

    Hell, Cobalt have dirt cheap 64bit systems shipping in volume right now.

    What... You're waiting for Intel to say 64bit is good?

  • Just see it as an expansion of your home network. Those slow ass Intel P3-450s can always be made to do something like printspooling
  • Since when does a comment about PR spin constitute FUD?

    Talk about knee-jerk reactions.

    The really funny point to all of this is that we're talking about 'samples'... which means things can and will change. Personally, I'll wait till the boxens are out, the OS's are shipped and we can really bang on things then.

    Use the right tool for the job...

  • Anybody know if there are 128bit chips on the way?
  • Uhm right, it is a bit offtopic, but if you liked the mmx extensions, you are just gonna love this baby.

    (And if you thought mmx was "too little, too late", then G4 might have just that little extra to make it a natural entrypoint to vector processing on a chip)
  • My favorite book on VLIW is 'Bulldog: A Compiler for VLIW Architecture' by John R. Ellis. It's out of print, unfortunately, but it should be in your University library.

    It describes the first VLIW system and the compiler that made it work. This was a machine from a company called Multiflow (RIP). The book starts out good and just gets better, as all the pitfalls and problems of creating a good parallelizing compiler are examined. The book is very well written, too -- easy to understand and fun.

    The most astonishing parts of the project (to me) was just how hard it was to get relatively optimized programs for VLIW machines. The Bulldog compiler took *hours* to run on reasonably sized programs. Now, this was a few years ago, and typical workstations are now much faster -- but programs have gotten bigger, too.


  • After reading the URLs posted in this article, Microsoft has stopped Win64 for alpha chips... not Merced. So I fall to understand the reasoning for the "However, contrary to prior reports, Windows does boot and run on the engineering prototype. ". Contrary to what reports? That article had nothing to do with Win64 running under Intel's Merced.
  • Can anyone see this story on the main page? I can't. I only got to it by going from my user info to my previous comment to the story. What the fuck?
  • About the hotmail hack... my understanding is that right now, the street is saying that there was no real hack, and that indeed, it was all about a backdoor left open. Now some claims it was passport, some other talk about a simple cgi script.
    Whatever the truth is, hole or not, the bashing started on this forums BEFORE people were even aware about NT server, passport, or whatever.

    At the end of the day, the only thing I notice is that I'm myself confused by all the information that comes out of this forums, and that most of it is generated by fanatics who should know better...

    Anyway... happy threads!



  • I share your sentiments. I wonder if it's just an animation that makes the computer look like it's actually doing something.

    They did that originally for Windows. I seriously doubt they've improved their morality since then.
  • Wrong. The thunking was pretty much as fast on PPro and a P2. The reason the PPro was slow was the 16 bit code does segment/selector switching a lot. The Pentium had a cache for the entries for each selector in the GDT/LDT whereas the PPro didn't. This cache was readded in the P2 to speed up the performance in 16 bit code.

    The reason (in case it isn't obvious) that 16 bit code does a lot of segment/selector switching is that each one can only access 64k of data, which isn't very much in modern OSes.

    John Wiltshire
  • If thats the case, my computer is going to be a workstation. I really want to move beyond 32bit, with the security of having a fair idea of the architecture.

  • I think Intel have really painted themselves into a corner with Merced. Everyone knows that it hasn't been achieving the kind of performance that it was expected to, and it's being overtaken by other chips more rapidly than originally forecast. It's pretty much a dead end. I can wait for the next generation, thank you.

  • May be this explains why there is no longer an alpha-NT alliance. It's not needed in MS mind, may be?
  • What does 64 bit offer that 32 bit hasn't got? What I'm asking is, is there a big difference in raw performance because of the increase in bits, or does it have more to do with addressing etc. I was never very good at hardware at University, so can someone enlighten me?
  • Anyone else notice the article said they demonstrated Merced running on W2K and Linux? Surely it should be the other way round -- W2K and Linux running on Merced?
  • as far as saying the evaluation period was over.

    at least linux ran an application as well (Apache)

    see the register []'s note on it.
  • well.. maybe showed some stuff is putting it a little better.. but it probably ran because merced is still backward compatible. it has a 32bit part IIRC.
  • Do not underestimate the powers of engineers in brightly coloured suits dancing RAP and braking doors...
  • Try cards with 64 bit design or 64 bit oriented design - packet engines GigE, Tulip 100MB Ether, ITT SCSI, etc.
    Most of these are nightmare on 32 bit and kick ass on alpha and other 64 bit machines. I mean kick ass to the point when 32 bit machines are out of the question.

    Of course: try them under linux or BSD ;-)
  • No point to be sorry, let's run the Mindcraft becnhmark again. And with 1GB Ethernets. They are 64 Bit ;-)

    I have done some prteliminaries on non-interl arch and the results were very mindcrafting ...
  • I mostly agree with your post but take exception to your complaint about the Hotmail bashing. Microsoft deserved to get bashed for the Hotmail fiasco. Whether it was running on Solaris or NT is beside the point. They bought the company that provided the service and opened up a humongous security hole when they tried to "integrate" it into MSN with their passport "technology." If my understanding is incorrect please correct me. But if it is correct, then Microsoft deserves all the bashing delivered to it. Can you honestly think of a larger security hole? A simple URL that just about anyone could remember quickly and BOOM! you have instant access to anyone's email account? Come on, you can't complain about complaints about that!
  • There are some direct advantages , especially if you're working with (for instance) bitmaps, because you can grab twice as much data at a time, put it all inone register and (eg) xor it with a mask in one operation. Also you get bigger integers without needing special software.

    The main advantage is being able to handle more memory. 32 bit architectures are (basically) limited to 4GB, which is not very much by todays standards.

    Finally, since this chip is a complete redesign, Intel get the chance to do a lot of things differently/better than they were done in IA32.
  • Come on, Intel is full of crap. This is just a marketing ploy to get our attention away from the superior chip Athlon.

    When have we seen news about this sort of thing before? Oo, we just sampled the first 486, oo, we have beta silicon on Pentium II it runs some apps, but others it won't!

    Who cares!! When you can run a Quake3 timedemo on that sucker then call me up!

    Everyone go make some noise at the motherboard manufacturers to get those Athlon boards out so we can see some real performance NOW!! The chips are ready, if AMD had the marketing clout that Intel does then the boards would be too and I would have a brand spanking new Athlon running on my machine at home now.

  • A Summary of 64bit specific Benefits:

    1. 64 Bit Memory and IO
    Merced can access memory and IO Buses with 64 bit width. That is it can transfer 64bits with 1 atomic operation. On i32 it would take 2 seperate atomic operations to transfer 64bits.
    (I believe the PII has a 64bit memory bus/but only 32bit adressing)

    2. Virtual/Real memory access. With 64bits Merced can acces as much as 2^64 bits of ram or virtual memory. (i32 can only access 2^32 bits of memory)


  • What? You mean you want to re-use the same hardware over and over. That would be way to logical. Don't damn us yet. I still have to make it to retirement.
  • Here's a good question: Are we going to finally get some TRUE general purpose registers?! Say, like the aging (but nice) 68K processor?

    I'm not a big fan of "Optimize code by choosing the right register" Intel x86 docs keep implying, as well as specific combinations of adressing modes.
  • Unfortunately MS has so richly deserved severe bashing on sooo many occasions that they have lost their right to appeal...
  • 1) Windows NT never had any 16 bit OS code in it. It has a 16 bit user mode emulation thingy (NTVDM) which is similar to Wine in operation, but no 16 bit code in the kernel. NT has never run on a 286.

    2) Microsoft have written quite a lot of drivers. Look in %winnt%/system32/drivers and in the version info of *most* of the drivers you will find 'Microsoft' as the author. I'd suggest that Microsoft have definitely written more drivers than Linux people. They only got hardware vendor support when their OS became popular (remember before Windows?)

    3) MS have a Y2K statement. You obviously haven't read it.

    I still haven't figured out what your point was.

    Microsoft bought VC++? Who from? Just curious.

    John Wiltshire
  • Didn't the 68K have address registers and data registers as separate things (A0, A1, D0, D1,...)?
    I think you mean general purpose registers like "RISC" registers...

    BTW, the IA64 has *lots* general purpose registers.... and a funky register offset address
    mode (ala 29K processor), but better... Hopefully the compilers can take advantage...

    Unfortunatly, this looks more and more like another i432 chip from intel...


  • Microsoft has supposedly been working on Win64 (for the alpha) since 1994 or so. One would hope that they nearly have something by now!
  • Most of the reviews I read seemed to use the Matrox Millenium (one of the fastest at the time) which did write combining in the driver. It was only DOS games that really let you use FASTVID.

    Actually, my PPro BIOS lets me enable write combining in the BIOS which really messes up your 16 colour modes.

    The P6 core has had a few improvements over the PPro version.

    John Wiltshire
  • Do you think we can tell if someone is fired or quit a job?
  • There is no need for 128bit chips, now or ever.
    64 bits is enough to address 18 million terabytes.
    More than the number of atoms in the earth...

    64 bits is barely enough to mmap every atom in one cubic millimetre of material, let alone the earth.

    Remember, Moore's law is exponential. 32->64 bits will only take twice as long as 16->32 bits.
    Expect 128 bit processors by 2015.
  • The Register has some pictures of the Merced box: -000036.html []
  • When will gcc be able to produce efficient code for Merced? I heard it was tricky to exploit the new features in Merced.
  • by Anonymous Coward on Wednesday September 01, 1999 @03:02AM (#1712173)
    I only want to hear bad things about Microsoft. I don't want to hear about MS software running! I want to hear about security problems and how they are clueless about writing and kind of decent software. I want to see stories of how they gave Mother Theresa a hard time about bundling software in the Vatican. I want to see Bill Gate getting a pie in the face. Ziff Davis has a story a day about Microsoft evil rampage against good people in the computer industry. Why cant you! DAMM YOU LIBERAL SLASHDOT MONKEYS
  • Um, is it just me, or is the typical reaction to a first-silicon of a supposedly very fast CPU not that "gee, this will be the new engine for e-business"? Sounds extremely silly to me. Instead of attempting to refute Apple's praise of the G4 (hrm, with its Velocity Engine -- where do these people get their terms?) they start talking about engines for e-business. Weird.
  • I believe Intel [] is working with (and invested in) Cygnus [] to help GCC with Merced compatibility and performance. Anyone else know of any official statements?
  • It seems that point of view is migrating from achitecture to OS... of course tanks to linux.
  • Yes, but with no base+offset address mode, there'll be LOTS of parallel code! :)


  • >>... a big error message popped up.

    So, that's the big advantage of 64 bit architecture - easy-to-read dialog boxes!
  • It seems like everyone *except* Intel is 64 bit now. Even my company's router accelerator uses a couple of embedded Mips 64-bit processors in it. 64 bit is everywhere now except Intel (and I guess Mac).
  • /.ed already? That's gotta be some kinda RECORD!!?
  • > Since Windows doesn't fully run in 64-bit mode on the alpha even after all these years, are we supposed to believe that it suddenly runs in 64-bit mode on a brand new chip?

    um, yes. 32-bit windows not "fully running in 64-bit mode" on the alpha doen't make me blow any fuses over the notion that 64-bit windows runs in 64-bit mode on a new processor. it would have done the same thing on the alpha platform as well, if alpha support hadn't been dropped.

    for a while, it was a major possibility that the alpha version of win64 would be out before merced was available. microsoft isn't married to intel (politically) in the same way they once were.
  • And another astroturfer boldly steps up to defend Bill's honor and fortune. (Or is he merely defending his own stock options?)

    Since Windows doesn't fully run in 64-bit mode on the alpha even after all these years, are we supposed to believe that it suddenly runs in 64-bit mode on a brand new chip?

    Oh, yeah. Intel provided backward support hardware for 32-bit software (by demand of the Neanderthal lobby). You can probably run Windows 95 on it. (Oops - I take that back -- Wind9x still isn't 32-bit clean, is it?)

    I'll be surprised if W2K is Y2K clean by this time next year, let alone 64-bit clean.

  • It doesn't matter that Win64 was running. The original poster's point is still valid: A large reason why W2K ran at all is probably due to the large amount of legacy 32-bit code that's supported by the IA32 emulation on the chip.

    After all, the reason the first Pentium Pro's didn't do so well when they came out was due to the fact that they didn't support 16-bit code very well. So, even though everyone was running a Win32 platform, most people didn't get much bang for their Pentium Pro buck due to the 16-bit code laying around under the hood.

  • Exactly. I think the Linux community is just hurting itself with its now constant Microsoft bashing. What was a couple of Bill Gates jokes is now becoming a political agenda, from which no one benefits.
    This permanent immature attitude is making Linux fans looks like real morons, I take for example the Redhat fiasco, and now the disinformation on the compatibility of the Merced.

    When Hotmail crash, 90% of the post are MS bashing. When someone mention any MS product, he get bashed. Anyting actually posted here is always compared to Microsoft, then bashed, no matter how related it is. A post about Pentiums? 3Dfx? Hard drives? It always comes back to MS bashing.

    I think Rob once pointed out to those who were sending 'hate mail' to Mindcraft were making a mistake. Maybe it's time some of the people here listen a bit.

    After all, you want your community to grow, no?


  • well forget that new Motherboard that I just bought, here comes a new socket probably. And probably new ram, new slots, new case, etc... oh well (dam intel)

    ClarkBar :)
  • This isn't meant to bash MS...but they are the kings of mis-direction and marketing hype. they can get it it to boot...I'll believe that. But does it run ??? Is it even close to being functional ?? Heck..they can get NT to boot on Alpha..but they pulled the proverbial plug on that because they couldn't get it anywhere near right.

    So for me..this is just another story of the little boy who cried wolf.. after beign deceived so many times I refuse to believe anything that MS says until I SEE it.

Today is a good day for information-gathering. Read someone else's mail file.