Slashdot Mirror
Intel 64-bit Announcements at IDF
cribb writes "Some fascinating stuff is going on over at the IDF. Ever since the first sneak previews of the Opteron, there has been lots of uncertainty around its future, and that of AMD. AMD have bet everything on the success of their new 64-bit CPU, and with Microsoft severely delaying the release of a 64-bit Windows, and Intel complaining that 64-bit processing has no place in the desktop market, things were starting to look dim for AMD. However, after rumours around the 64-bit extensions of the Pentium 4 EE, it became clear that Intel is not willing to lag behind AMD in the 'innovation' department. Now comes the shocker: Intel boss Craig Barrett today anounced that Xeon-class 64-bit server CPUs codenamed Nocona will be coming out the second half of 2004. It isn't clear whether they will support AMD's Opteron AMD64 extensions. Barrett is quoted saying, 'There will be one operating system that will support all (64-bit) extended systems.' Maybe 64-bit computing is right around the corner after all, and we may even see compatible instruction sets from Intel and AMD! And does this mean that Intel will be dumping Itanium, which never caught on as expected in the server market, and forget the billions spent on developing it?" See some other articles at EE Times, and EWeek.
There will be one operating system that will support all (64-bit) extended systems.
He's right. It's called Linux.
This is a good thing, whenever someone plays catch up, they alwasy seem to develop a better product than if they were at the top. Take for example how IE6 has slowed improvements while other browsers continue to create. A little competition is a good thing.
"Intel complaining that 64-bit processing has no place in the desktop market," Can't 64-bit processors use more memory at a time than 32-bit processors? I don't understand...
Intel had to play 'catch-up" by incorporating MMX into the Pentium when NexGen was plotting on incorporating their own SIMD system (which became 3DNow!) but this time, they really got screwed over. They had planned on Itanium taking the 64-bit market over, and did not figure on AMD's x86-64 at all. What really did Intel in this time around was that AMD was doing what Intel had traditionally done, continue the backwards compatibility long past any logical point and not only making it work, but making it attractive. This is the mis-step that brought Motorola down from it's "king of the desktop CPU" position, when they released the 88k as the "next-generation" CPU rather than focus on delivering better 68k's. The division of resources back then is a step Motorola never really recovered from. I wonder how Intel will do on it.
Karma Whoring for Fun and Profit.
I think it is a pity, that the alpha processor (that was once the best processor) had to die, just because HP and Intel wanted to succeed with their Itanium processor (and are now failing).
Felix
/dev/earth not found. Reboot?
It's nice to see that Intel's not just sitting back on past glories and thinking that'll solve everything for them. With AMD and Intel getting so competitive, and comparable products from both coming out so close together, it can only benefit the consumers.
"Although this means that Intel could bring a 32/64-bit chip to PCs soon, Barrett said the company has no plans do so in the near future."
Right, so introducing a 32bit/64bit "server chip" is absolutely NOTHING like introducing a "desktop chip". They still clearly are pretending that they are not competing with AMD's strategy. Who are they kidding?
It'll be one hell of a backtrack if they do drop Itanium. Yet it will be hard to keep Itanium viable with another 64bit chip that is (presumably) much better at handling x86 code.
What this really signals is that Opteron, and AMD64 are really quite impressive indeed. It's billions that Intel will be dropping so they can compete with it, and you don't make that sort of move unless you're really very very worried.
As to whether they will be compatible with AMDs extensions: I suspect Intel won't be ale to bring themselves to that. The "One operating system will support all 64bit extensions" sounds more like a deal has been cut with Microsoft to make the 64bit version of windows work with Intel's 64bit extensions as well of those AMD. In practice I suspect that means Intel will be very close to AMDs extensions, with a few quirks, and the intention of trying to grab the market and drag things away with their own extra extensions with newer chips.
Could this be behind the slowness of 64bit windows for Opterons?
Jedidiah.
Craft Beer Programming T-shirts
And it's not only a intellectual exercice. Intel also had released Pentium Pro with no success but Pentium II III IV are based on it. I'm sure there's lot of stuff that can be reused...
Try again, 64 bit is very useful for lots of things. Keep in mind that when you 'offload to a 3D card' as you so easily put it, you're using a largely specific-purpose processor. This means that you've got to be in the canon of algorithms that the hardware-maker thought you'd use. A general purpose 64bit is very useful.
Well, many calculations in academia are done on "desktop" computers. Some of the calculations done in the lab I work in can easily gobble up more 4GB of RAM. A couple of weeks ago we were looking into our options to address this problem at a reasonable price. Speed would also be nice when you have to cruch that many numbers, but if you don't have the RAM you can't even wait longer to get the results.
Not counting legacy PC architecture goofiness, 32 bits currently provide a 4G addressable space. So, apart for power-users, servers, hardcore gamers and trendy techno-posers, what's the advantage of running 64-bit systems? Sure you can make biggest calculations in one instruction, but overall you have to move twice as much data around to achieve the same thing if you have less than 4G or RAM.
Yes I know 4G of RAM is getting increasingly common, but is it really needed? just because Windows is as thick as a whale omelette doesn't mean you need that much to achieve the same result.
Honestly, I could understand the need to have more than 8 and 16 bit processors, to make multiprecision calculations less necessary for common things and to avoid segmentation kludges, but for the majority of people (i.e. people running Word and Excel, and playing Minesweeper a little), I don't see the interest at all. Better have good fast cheap 32-bit systems than expensive, underused 64-bit ones. Unless of course future versions of the Windows require that much power, which doesn't even seem likely for the short term.
"A door is what a dog is perpetually on the wrong side of" - Ogden Nash
There was a time when people got by with 8-bit and 16-bit processors. True, 32-to-64 is a much bigger leap than those jumps were, but technology keeps advancing.
My old modem had two processors and a slew of analog filters - now modems are just codec that allow the main cpu to do all the processing. Same is true with many sound cards (they used to have DSP coprocessors or specialized hardware). My old laptop had a MPEG decoder card, but now that's done by hardware. (get the trend I'm trying to show yet?). Pretty soon a video cards that is just a D/A will satisfy most users, like the equivalent WinModems.
I've always thought it unlikely that Intel would be caught off guard by AMD's Opteron. I think Intel could have announced this earlier, but wanted AMD to become overconfident with its Opteron and spend oodles of cash etc. on developing public awareness of 64-bit computing, explaining what it is, convincing people that it's worth the upgrade, etc. Then, after AMD (who is already cash-strapped) puts all its eggs into the 64-bit basket, Intel finally comes out and says "Thank you for raising public awareness about 64-bit computing for the desktop for the past year, AMD. Now that you have no more money, we will now announce our 64-bit chip and compete with yours." Here's a list-form of Intel's strategy:
1. AMD comes out with Opteron.
2. Intel waits.
3. AMD spends all its money and resources on promoting 64-bit computing, thinking this will make Intel look obsolete and make themselves the chip-maker of the future.
4. Intel waits.
5. Intel releases own 64-bit computing and takes over the market that AMD spent all its money developing.
6. (AMD pulls out empty pockets and holds them like wings and wonders what happened:) ?????
7. Profit for Intel!
8. I cry.
Wrong. 64-bit computing is ten years old with the Alpha, including PCs running GNU/Linux. Not to mention the later UltraSPARC, PA-RISC 2 and MIPS workstations.
And today we already have the PowerPC G5.
This all proves Wintel is the biggest drag in Informatics.
Leandro Guimarães Faria Corcete DUTRA
DA, DBA, SysAdmin, Data Modeller
GNU Project, Debian GNU/Lin
Sure, 32 bit is fine NOW but in the future there will be more intensive software that will require more RAM.
This is just a guess but I wouldn't be suprised in "Longhorn" will be native 64 bit. Maybe Microsoft is waiting for the hardware to catchup so their inefficient code can take advantage of more memory. (I know, it's a cheap shot at Microsoft)
So, with this, more or less, Itanium may or may not be out of the big picture. From what we're seeing now, it appears as though Itanium will remain a high-end enterprise class chip.
But you have to wonder... what on earth was Intel thinking? Ever since its announcement, intel has hinted that Itanium would eventually migrate down to the low-end desktop market.
But, it wasn't x86 compatible by a longshot, and had no intentions of ever being hardware-compatible with plain old x86 CPUs. Without backward compatiblity, there was close to zero chance of intel ever capturing the desktop market with it (it was a completely new architecture. there had been no software written for its new instruction set to date).
But then you realize that intel broke their most sacred tradition by breaking backwards compability. Suddenly, "intel-compatible" wasn't "intel-compatible" anymore. Moving from x86 to Itanium would be like moving from x86 to SPARC/Alpha/PowerPC.
And SPARC, Alpha, and PowerPC are all more powerful than the current Opteron chips, and cheaper by several orders of magnitude (specifically PowerPC).
In other words, it would be more likely for everybody to migrate over to Apple that it would be to move to Itanium. And it would be cheaper too.
With these latest announcements, I'm hoping that intel has finally adopted the x86-64 bandwagon and cooperated with Microsoft and AMD. (Imagine if WinXP-64 worked on two architectures.... and the compatibility nightmares it would cause...)
Either way, the scores are as follows:
Sun/DEC - 6/10 Have been using 64-bit for years. Yet, nobody seems to want it.
IBM/Apple: 8/10 - Successfully brought 64-bit to market, but launched without a supported full-fledged 64bit OS
AMD - 10/10 - Openly allowed developers to develop with the x86-64 sim years ago. Launched x86-64 before anyone else backed with full Linux support, and windows support in open beta. Successfully penetrated consumer, mobile, and enterprise markets simultaneously.
Intel - -5/10 (yes, negative) - Created an expensive proprietary system with no backward compatiblity, and is cumbersome to work with. It flops. They still don't have a 64-bit desktop processor. Their only successes are made by copying AMD.
-- If you try to fail and succeed, which have you done? - Uli's moose
This is what I think happened: Intel thought their server reputation would trump AMD's 64-bit offering. They thought their Itanium would be the only reputable and reliable 64-bit platform for x86. They were wrong: they started having problems and delays, while AMD was well ahead with backward compatible Opterons. Intel was not going to have major market share with Itaniums. However, while AMD most likely bet their entire existence on x86-64, Intel took into account the worst-case scenario with Itaniums and developed 64-bit Pentiums on the side.
After the Itanium failure, they came to a realization that MS was not willing to develop and support 2 different instruction sets. Praise as you may how advanced MS' NT kernel is, and how portable it is - it just doesn't make business sense, even for Microsoft, to support it on more than one platform (remember Alpha?). Also, as I remember, Linus also expressed his preference was AMD's solution and hoped Itanium would lose out. Intel is not so powerful after all. So, now they are forced to execute their plan B, and introduce their 64-bit Xeons.
Also, the statement "one operating system" was made by MS spokesperson, not Intel, as suggested by the
Intel's approach is compatible with AMD's, the Microsoft representative said. "There will be one operating system that will support all (64 bit) extended systems," the representative said.
Wait, the fact that someone wants superior tech makes them a "fanboi?" We can tell where your loyalties lie. That being said, you obviously don't understand the culture of consumption. 64 is clearly better than 32, which is certainly better that 16. Why? Who cares why it just is. And as for NEED, since when has that had anything to do with it. Most people don't NEED a computer at all, but look at how many own one.
So now we have to port to IA32, IA64, AMD, and some other instruction set too? And we have to do separate ports for Windows, Linux, Solaris, HPUX, etc etc?? Guess which ones are going to get dropped because the industry says enough is enough! Itanium, because nobody has bought any machines, and the new Intel instruction set, if it's not 100% AMD compatible, because it's last in an already-crowded niche. IA32 will be the low-end architecture and AMD64 will be the high-end.
That sounds suspiciously like PCI Express...
In the mean time my 1998 vintage Mesh/Alpha desktop system (no, it's not a server, it was sold via consumer magazines in the UK) is still running happily with 64 bit Linux... and that was hardly the first either, an honour that probably belongs to someone like Sun.
Actually, there were Alpha desktops long before that, and the Alpha chip was certainly around before Sun had any 64-bit machines. As were the 64-bit MIPS chips, which ran in desktop machines.
At any rate, it depends what you mean by "desktop", although your system sounds like it'd qualify. SGI machines make fantastic desktops (IRIX is very well-designed) and they're mostly 64-bit, but they also cost upwards of $10,000 (much upwards, quite often) when they came out. So they aren't consumer grade by any standard.
I think Apple's campaign has some truth to it in the sense that theirs is the first 64-bit desktop that normal people will actually buy and use. And it's definitely the first that's explicitly designed to do normal desktop computing stuff, as opposed to high-end graphics or engineering apps.
As for this whole thing, I'm not suprised and this is a MAJOR boon for AMD, because now people have no reason not to port their software to x86-64. The companies get customers now (Athlon64 and Opteron) and more later when Intel releases their chip. The other big win for AMD is that their chips is out NOW. So when the software starts to come, people who want/need that 64bits will get Opterons and they can gain some real market share before Intel's processor comes out (especially the desktop one since Intel is releasing the server chip first). As long as AMD is willing to cut back on their prices a little now to trade for future gains, this could be a MAJOR opportunity for them.
As for us consumers, this is a win. Intel trying to push Itanic (or even worse a THIRD arch) down our throats would be terrible. Now we have one clear "winner" in the 64 bit wars (don't reply with stuff like the G5, I'm talking the Intel/AMD/Transmetta/etc. side).
And where is Transmetta's announcment? They should make one two! I bet they could get a good chip out the door before Chipzilla gets a good mobile x86-64 chip out there. This would be a great chance for them too, they could grab a good chunk of the laptop market becuase it would only be them and AMD, and AMD isn't marketing towards low power ultra-lite laptops.
Comment forecast: Bits of genius surrounded by a sea of mediocrity.
Classic business mistake: telling your customers what they want.
Religion is a gateway psychosis. -- Dave Foley
Programs can take advantage of 64-bit memory and such, but really, theres no point at making every loop that counts from 1 to 10 into a 64-bit integer.
You nailed that. Heat is a bitch. Current shipping chassis for desktop vendors are barely able to even dissapate the heat from a P-IV 3.4 140 watt and Itaniums put out at least twice as much heat.
Compare this with the current Opeterons that are running at around 80 watts or less. Heat (and bus architecture) now seem to be the limitig facor for a given system. So for the same heat costs you can run 2 or more Opterons for one P-IV and more vs. and Itanium.
As to the 100,000 number, you might want to check out the register's version. Somewhere around 10,000 CPU's is a more realistic system (with 4000 of them coming from a single system).
It's hard to picture why there will ever be a need for 128-bit computing.
I'm sure plenty of people said the same thing about 32-bit (and even 8 and 16 bit) at some point in the past.
Aside from just the addressable memory is the ability to do larger math calculations in the larger registers. I've done some side by side comparisons of 32 and 64 bit compiled openssl on opterons and the 64bit version has a huge speed increase, very likely due to the additional size of the registers, and the additional registers that were added for 64bit mode.
Besides.. you can't say that the addressable memory that 64bit gives is more then enough... sure it's huge now, but in another 20 years or so it might not be. I remember getting my first PC 17 years ago... an AT&T PC6300, 8088 CPU (12MHz I think) with 640K RAM and 2x360K floppies. When I finally got a 20Meg (yes, meg) harddrive a year or so later I thought I'd never be able to use all that space. Now anything less then 250 or so gigs is a waste of time. RAM usage grows slower then drive, but it still grows. I don't buy a system with less than 2Gig of RAM anymore, and that's something like 3000x the amount of ram I had in that first pc. Keep extrapolating that out for another 20, 40, or 60 years.
- My favorite error message: xscreensaver, running on an old Sparc 5 w/ 8bit color: bsod: Couldn't allocate color Blue
As for Itanium not selling, That's funny. Itanium sold over 100,000 cpus last year which is a big number for the enterprise server market (That's more than some other major RISC processors sold in 2003 (like Power 4)). If you don't believe me Google "Itanium" "100,000" and "Otellini" and you'll see lots of links to Intel pres Paul Otellini's announcement back in Nov that Intel would ship over 100,000 Itanium processors in 2003.
Yes, except that Itanium's biggest competitor in the enterprise server market isn't the Power4, its G5 cousin or any other RISC chip. The Itanium's lunch is being eaten by the Xeon. If you'd Googled on the less specific "itanium sales" your first hit would be IDC Waterfalls its Itanium Sales. As that article observes, "The [100,000] number may seem relatively huge, unless we do not take into account sales of Intel Xeon processors that amount in millions."
The problem, when push comes to shove, is that for "enterprise" customers, 64-bit CPUs are still a solution in search of a problem. As of right now there aren't any applications I can think of that most businesses use where the Itanium has a pure performance advantage that outweighs the Xeon's much higher price-performance advantage. The High Performance Computing market, which is what you really referred to above, is not the enterprise market, and as flashy as HPC is, it's not where the money is, either -- go into any business using Intel architecture machines and you will see server rooms filled with HP ProLiants and Dell PowerEdges, and all of those will be P4/Xeon boxes.
It doesn't matter whether Mr. Otellini tells people he's happy with "over" 100,000 Itanium processors being shipped or not. Compared to the amount of money Intel sank into the processor, this is peanuts. If they deliver a 64-bit x86 processor and it outsells the Itanium by an order of magnitude in its first year or two, which is not unlikely, it's going to be very hard to justify not end-of-lifing the Itanium line and migrating customers to the new processor.
Wrong list.
432 and Itannic had full management support. The 432 failed becuase the market rejected it. Itannic appears destined for the same fate. The 960 wasn't rejected by the market, it was rejected by Intel management.
well the i960 isnt a complet failure. i often se this cpu in computers well atleast on the expansion cards in computers. many raid controllers ues this cpu and i even found some switches/routers that do
so the chip is used
Also keep in mind that even though a 64-bit processor with a 64-bit memory manager in an OS will allow for increased memory addressing, the chipset must also support the addressable memory. 8 GB is the limit in most boxes. I think what might be interesting in the future is the synergy between MRAM and full 64-bit processing . . . with MRAM cheap enough and abundant enough, virtual memory will be eliminated with applicatioins getting free reign across the entire addressable space that doubles as the entire storage space.
Yeah, spot #1 is held by IA-64 with 64 processors, and #2 by Power4 with 32 processors, as are all except #10 on that list (where every power4 is a 32 processor box)
Not to mention TPC-C is something for which vendors tweak heavily, and it is a fairly exclusive and expensive club to get into.
The only TPC comparison between Itanium & Opteron can be found in the 300GB TPC-H with a 2GHz 16-way opteron cluster (13,194) vs a 1GHz Itanium 2 (4,774) SMP box. Unless a 1.5GHz Itanium2 has a significant core change it isn't going to deal with the almost 4x lead, assuming the benchmarks are good, which I have some doubt of.
64 bit CPUs are really good at addressing -lots- of memory, allowing some databases to become memory resident. For some applications this would be a very big win.
The timeline for the Itanium processors was unlucky. When the first roadmap for the Itanium was put up (around 1997), AMD's K7 was not even on the horizon. At this time a 800-1000 MHz 64bit processor looked good for the next 10 years. Suns UltraSparc II was at 150 MHz at this time, PowerPC about the same range, and the topselling PC processors were at 200 MHz. It was slowly growing up to 500MHz with the P3, and the other architectures maxed out somewhere at 450 MHz. Then suddenly AMD came and put up the first 1GHz processor, and a race started between Athlons and the P3s, leaving everything behind in core frequency. Eventually the 800MHz of the Itanium didn't look that impressive anymore.
Intels P4 was poised to gain the frequency crown, and it made the Itanium look even worse (though it was not true performancewise), and you could buy Xeon III with 933MHz cores, good enough for most business servers with their large caches, with sophisticated chipsets for SMP, and with OS support for clustering. So Itanium suddenly was only a processor for numeric applications, because Intel had to fight AMD in the bread-and-butter PC business.
The Itanium II is still an impressive processor with its number crunching abilities and its integer performance. But it's cheaper for most people just to throw more P4 Xeons at the same problem, because the underlying technology has been implemented in millions of systems. Tightly packed blade servers are mostly based on P4 architectures, and increased redundancy by having more processors and systems clustered can't be easily beaten by less processor cores for more processing power.
The processor race between Intel and AMD has cannibalized the possible markets for the Itanium, and the number of fields, where Itanium make sense from a price/performance ratio are getting smaller.
With the upcoming of AMD's x86-64 even the More Address Space argument is looking weak, because you can get the same with AMD's architecture, which seems to have the smoother migration path due to its outstanding x86-32 performance. Again less business cases for Itanium.
So were are the application Itanium fits best in, if the alternative is to take a blade server with twice the numbers of P4 processors running at top speed? Or get an SMP Opteron system, which is cheaper, more widely available and seems to have the better OS support?
You cannot compare a cluster system with a SMP system in TPC-H. That is stupid.
I won't debate that point. It may very well be so and is not worth researching just now.
Hmm ... I suppose one could say things like Lear Jets, Mercedes Benz autos and fancy yachts are market failures in the same sense. They do things out of the ordinary and require a deep pocket. (Of Ada, this is true of pre-1995 compilers and still true if one wants support, special features or just a validated (certified) compiler.
Not since 1995 or 1996. The US DoD did not extend the Ada mandate and in fact closed the AJPO in 1996. To repeat part of my original post, Ada has been in extensive use worldwide for aeronautics and astronautics (in fact my current knowledge is that every current airframe (civilian or military) runs on code written in Ada). The Paris Metro and part of the NYC Subway system runs on Ada code.
This may still be true, but I expect it to change. Ada-95 is eminently suited for general commercial development as well as systems programs and since GNU Ada has been available there are less stats available on market share than there are on Linux deployment.
To the best of my knowledge, Ada is being taught as a first programming language in some 200+ universities around the world, including in the US Military Academies.
It's only the most readable modern language and the one that is most likely to catch programmer errors at compile-time. These two factors alone make it the most cost-effective language to use for any project of substantial size, and for any type of project from writing commercial off-the-shelf apps (short time-to-market) to software that will live for decades (high maintainability), including system apps (like an operating system).
Speaking of which, I am in the first stages of designing a new OS (*nix-like) in Ada targeted to Intel/AMD architecture processors. I expect to make an announcement here when there is more in writing, but anyone that wants to discuss it or work on it from the first may reach me at toolmakr at buzco dot nyct.net.