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.
As for one operating system, who? They in cahoots with Microsoft, after Microsoft dragged it's feet on AMD? Sounds like collusion, anti-competitiveness, and all that.
A feeling of having made the same mistake before: Deja Foobar
Linux? FreeBSD?
So after the Apple 'first 64 bit desktop' campaign we get to see an AMD 'first 64 bit desktop' _and_ an Intel 'first 64 bit desktop' campaign?
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.
Beep beep.
Maybe they're saying that home users have no need for that "extra memory at a time". To some extent, they are right. THe avergae user will probably look at a 64-Bit computer and go "ooooh, 64-Bit" and see it as a selling point. However, the average user will probably never need the extra memory or power of the 64-Bit.
EE Times is also reporting that Intel may be pushing a new kind of RAM interface to compete with existing DDR and RDRAM. At 2 Gbit/sec per wire, this is about twice the speed of current RDRAM and four times the speed of DDR SDRAM. But, more interestingly, this is a point-to-point architecture - unlike the traditional bus architecture, when you add more memory modules you can get more bandwidth. Also notable is that simultaneous bi-directional communications happens over a single wire. Infineon and Samsung have made test chips, and results are to be released at the International Solid State Circuits Conference today.
I wonder how this figures into their processor/chipset roadmap...
HIV Crosses Species Barrier... into Muppets
Posting anon due to affiliations.
Anyone that didn't think Intel had 64-bit Workstation and Desktop chips "in the pipeline", as it were, must be sitting in a cave humming with their fingers in their ears.
The production pipeline on these sorts of products. take years, so this was not a knee jerk reaction. If you look very carefully at what Intel has actually officially said the whole time, you'll see that they simply said they would provide a solution when the appropriate OS support and perceived need becomes available, and that is EXACTLY what has happened here. What do you know, Steve Balmer announces Windows XP 64 now has support for these "Xeon" extensions. These things don't happen over night.
It is still a fact that most people DO NOT need 64-bit computing in any way shape or form, but one mistake that Intel did make is the fickleness of the vocal minority and AMD fanbois.
Also, if you think that the existing Prescotts don't already have these extensions (just disabled at the moment), you are also kidding yourself.
I know 64 bit CPU's can access lots more memory.
How else does 64 bits make things better/faster?
(simple language please, i'm a techno idiot)
It may also be that MS is waiting for the Intel product so that they don't have to make massive code changes if Intel's implementation is somewhat different than AMDs.
EA David Gardner -"... but the consumers have proven that actually what they want is fun."
The Register compares Itanic to the i432: "Bob Colwell, chief architecture honcho for the chip that saved Intel in the mid-1990s, the P6 (Pentium Pro), described the i432 as 'a wonderful research project masquerading as a bad product'."
...any more than IBM would ditch Power4/5 architecture, just because they have a commodity market x86 chip with 64-bit address extensions (Opteron).
In the 'big iron' enterprise market against RISC where Itanium is beating everything handily (check out the latest TPC-C list Top 10 where Itanium holds spots 1,3,4,7,10 (5 out of the Top 10 are Itanium systems running a mix of Linux, HP-UX and Windows on HP and NEC systems), Itanium is gradually out-selling all of the big RISC opponents like Power4. Note that IBM is certainly not spending the money to put up an Opteron cluster for the TPC-C test(no 32-way or 64-way scaled solutions for it on the horizon) even if they got good enough results (which they wouldn't) if they can't beat Itanium 2 right now with the high-margin Power 4. No doubt they'll have a run at Itanium again this year with Power 5.
But there's no way that Opteron OR a 64-bit Xeon plays in the big high thoughput space, so people that assume Intel would get rid of Itanium simply don't know what they're talking about.
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.
That's exactly right.
Intel (and Barrett specifically) can't stand Microsoft. They were 'forced' into supporting AMD64 by Microsoft, because MS only wants one 64-bit OS.
The last time Microsoft strong-armed Intel, Intel created the OSDN (the current employer of Linus Torvalds.) Intel helped write the IA64 (a.k.a. Itanium) port of Linux, and had it up and running months before Windows was running on Itanium systems.
Intel likes Linux. Specifically because it isn't Microsoft.
Another non-functioning site was "uncertainty.microsoft.com."
The purpose of that site was not known.
An example from Reuters:
"Analysts have said they expect a major announcement from Intel this week on 64-bit computing, a technology that lets computers churn through doubly large chunks of data than the current 32-bit computers."
AMD turning the heat up
64-bit does not double the memory bandwidth, it simple means a bigger address space + 64bit integers, (although the Opteron does have high bandwidth this is nothing to do with 64-bitness)
Maybe Microsoft is waiting for the hardware to catchup so their inefficient code can take advantage of more memory.
Does it really matter anymore what Microsoft is doing?
Where I live, opterons are selling like hotcakes and something must be going onto them.
"Nine times out of ten, starting a fire is not the best way to solve the problem." - my wife
64-bit instructions not being used? Maybe not under Win64 or "early version of a linux distro supporting it"....
NetBSD had full support for the Hammer architecture before IT WAS EVEN BURNED ON SILICON. It is also a true 64-bit operating system unlike Debian/SPARC64 which utilizes a 64-bit kernel w/ 32-bit userland.
NetBSD is definately 64-bit clean for the most part.
And no, BSD is *NOT* dying.
While AMD have alway talking about new developpment and invits others to share, Intel keep all secret and try to act like no others exists (including there customers sometimes).
l ogy/64bitextensi ons/30083401.pdfn ology/64bitextensi ons/30083501.pdf
o ntent_type/white _papers_and_tech_docs/24592.pdfo m/us-en/assets/content_type/white _papers_and_tech_docs/24593.pdfo m/us-en/assets/content_type/white _papers_and_tech_docs/24594.pdfo m/us-en/assets/content_type/white _papers_and_tech_docs/26568.pdfo m/us-en/assets/content_type/white _papers_and_tech_docs/26569.pdf
Sorry Intel. There is no AMD words in your doc, but now all the worlds known that your IA32-e is no more than the AMD X86-64. For me you just act like a child!
Intel IA32-e documentation:
http://developer.intel.com/techno
http://developer.intel.com/tech
AMD x86-64 documentation:
http://www.amd.com/us-en/assets/c
http://www.amd.c
http://www.amd.c
http://www.amd.c
http://www.amd.c
How long Intel while wait before it make the same kind "new extention" compatible with HyperTransport ?
It was supposed to replace X86. Itanic will go same route. Repositioned and slowly fade into the sunset.
Help fight continental drift.
Nothing was stopping Linux from supporintg AMD and releasing it early. As pointed out, BSD did. In fact, AMD probably could have thrown some engineers at Linux and helped it progress to release a bit faster. Imagine that.. MS's perceived competition releasing an OS that takes advantage of faster processors before MS does. Can't imagine MS would let themselves get left in the dust.. Intel conspiracy or not. So, if you want to believe in a MS/Intel conspiracy, then you also have to include AMD in it for going along with it.
I know.. an AMD/MS/Intel/SCO conspiracy agasint Linux.. yeah, that's the ticket. And the only ones fighting the good fight are OSS developers and.. IBM?
Welcome to the net of 1000 lies. Upgrades are scheduled soon that should bring us to the 10,000 lies mark.
I would not be surprised if OS-2 suddenly got a new lease of life in 64-bit form, as there is a window of opportunity while Bill's pathetic product is, as is always the case, late and buggy. Another outsider, which would really be fast on 64 bit, would be BeOS. I am not up to date on developments there, but an OSS equivalent was in development last time I looked.
But, whatever happens, I expect there will be a choice of *nixes before Bill has anything working reliably. It could be the beginning of the end for M$, after all if the 64 bit market is mainly servers, then the actual OS is of no importance, only the stability, security and cost. A server serves files, it does not usually need to run anything OS specific, it is (simplistically) only the protocol on the network that matters, hence the failure of IIS in the web server market. It failed on all 3, stability, security and cost.
I personally don't need high performance servers, but when they are affordable I will want a 64-bit machine for circuit simulation because I am (usually) an analogue designer, and transient (time domain) simulation of certain types of circuits is an enormous computational load. Hopefully SPICE will be ported to 64-bit, and I will at a guess be using SuSE. My network is unusual, it only supports me, so the oldest, slowest machine (K6-II/500) is the main server. I predict that there will be a strong minority interest in 64-bit desktops for similar reasons, a fair number of people also do difficult digital simulation work and other tasks which really need the speed, so I am guessing that 20% will be workstations, 80% servers. The chances are that Bill will lose money heavily, he would likely get the greatest share of the workstations, but why would he expect to sell many copies of a server OS, if current trends continue? Servers are becomming commodity items (SAN, NAS....) and the OS cost really needs to be negligible.
"There is no surprise either that heavy weights such as IBM, Dell, SUN and even HP -- who pretty much designed Itanium -- put some of their eggs in their AMD busket."
Good post, except for the above quote. IBM has one (eServer 325) machine that is based on the Opteron. This machine is being marketed to high performance computing environments and not really as a general purpose machine. Not sure if it was The Register or The inquirer, but one of the sites had quotes from a top IBM executive on their lack of plans for more Opteron based models (i.e. a 4 way and some other form factors).
Dell has outright said they are not supporting the Opteron and were holding out to see what happens in the market (which could mean they were holding out for Intel).
HP's Opteron systems are a rumor and have yet to be substantiated.
That leaves Sun. Sun is the only major vendor with both an available Opteron system and another (4 way, 6 hdd) on the way. They are also the only major vendor that has made long term commitments to Opteron on a variety of setups/form factors, including workstations and 8 way servers. Sun has also promised to deliver 32/64 bit Linux (third party) and 32/64 bit Solaris on Opteron.
In fact when Sun announced all of there support, that should have tipped us all off that the Intel annoucement was coming. When would Sun do what everybody else is doing, unless everybody else was actually gonna be doing something else? Don't be surprised two years from now when Sun is still the only major vendor of Opteron systems.
It is directly compatible with AMD's 64-bit implementation. Intel and AMD have a sharing arrangement with x86 dating back to when AMD first licensed x86 from Intel, which basically allows Intel to use whatever AMD adds to the instruction set. And vice versa of course. Generally whoever implements it first gets in a generation ahead (Intel with the SSEs, AMD with x86-64).
On another note, these new Xeons are based on the Prescott core, so it is now extremely likely that the existing Prescott cores all have the capability, just not turned on, like what Intel did with hyperthreading on the Northwoods. It's been clear from the start that Prescott is hiding some functions up its sleeve, as there are at least 10 million transistors that can't be accounted for with the increased cache and other added functionality, even when being very generous with the estimations.
I use Linux AND BSD extensively so I'm not going to try to be offensive and start a flame war but....
It depends on your definition of stable linux port. BSD developers tend to have a severe case of obsessive-compulsive disorder. They thrive on efficiency and simplicity. That's why I can run a modern BSD-based OS on my old DEC MicroVAX 3300 w/ 20 megs of RAM and still have it be usable. (I know there is/was a Linux/VAX port but it's flopped.)
Nothing in the BSD world is considered stable by the various *BSD projects until it has been beaten to death for years. Though the FreeBSD project makes "unstable" releases like the entire 5.x branch, they at least inform the world that it's unstable. (for what it's worth I've heard it's not THAT bad now)
Look at how often NetBSD actually does a release, early NetBSD releases (not to mention 386BSD) predate Linux yet NetBSD's most recent "stable" release is 1.6.1.
My experience with the latest and greatest Linux distro "stable releases" would definately qualify them as "experimental".
Linux has a lot of commercial support and drivers are available a lot faster for the latest video cards for playing Quake which has advantages, lately the Linux community doesn't have to bust ass as much to reverse engineer drivers.
To sum it up, the BSD projects tend to do things in the name of good old fashioned text book computer science and Linux tries to push the envelope. Both approaches have their advantages and performance is roughly the same that I've seen.
I prefer the BSD environment because it's simple, comfortable and just works (providing I don't buy crap hardware like WinModems, Broadcom 802.11 cards or brainless USB printers).
If I really want to run Linux binaries, I can with virtually no performance hit. Same with SCO UNIX, Ultrix/VAX, Ultrix/MIPS, Solaris, Xenix, DEC UNIX/OSF1, MacOS X/Darwin (to a point), etc.
Not to mention, the NetBSD machine-independant driver architecture is too cool for words.
Enough about BSD though. I do use Linux quite a bit for some things, our corporate web/mail server runs Debian Linux because NetBSD/sparc64 doesn't support SMP on UltraSPARC machines yet. Multimedia support is a bit better under Linux too, Video4Linux is very cool. The only real video capture cards supported under NetBSD and FreeBSD are the BT848/878/etc, Matrox Meteor (ancient) and QuickCams.
It's all a trade-off really. Though *BSD suffers because of it, I actually like the fact that the projects do not play to media hype, politics and goofy marketing (*cough* LIN*GASP*DOWS *cough*).
Alpha failed not because of Itanium but because it didn't run x86. The key to processor success is fast x86 performance. Intel knew this before Itanium and everyone bitched about x86 being archaic and impossible to design compilers around. So they start from scratch, develop Itanium and are now being laughed at for trying to develop something new instead of patching the old.
Yes REX prefix byte is a small change in the instruction set, but a very nice one: it allow to extends the architecture to 64 bits (more longer registers, more addresses, etc...) while permit to execute IA32 userland code without problems.
The most nice part it that the change do more cleaning that hacking: Intel have a long history of adding new opcodes to IA32, but almost no one use it because binary programs have to run on all IA32 chips including the older one. So now the difference between older and new IA32 chips is so big that for time critical functions, programmers have to test the capability of the chip and to dynamicaly branche to a dedicated functions optimized for that chip. x86-64 make a hug reset to this entropy. This don't stop the process as new opcodes will exists in the futur, but at least it voids the last 10 years or so ugly hacks to the IA32.
In a wonderfull world, where all uses Debian source package and recompile it for each of his machines, the binary compatibility problem will be less important. But this even don't solve the problem, as the detection and specialization to a dedicated chip will be done while compiling.
The lesson is that it is not supportable to indefinitely add hack to an architecture. Sometimes it better to make a new backward compatible architecture. I doubt that 128bits computing will be the next thing that will trigg an architecture change. But without doubt architectur change will happens again in the futur to clean the older one and enable new capability.
And at this time the market will certainely follow the proposition that will have the bigger performance and smallest change ratio. This is exactly why IA64 is dead and x86-64 shine, forcing Intel to copy it and rename it IA32-e...
the G5 is not the power4, maybe but they do share almost the same architecture minus some cache and another processor on the die. The exacution units in the 970 are FROM the Power4, it is not really a different beast and it is a better processor than itanium because it can scale to different markets.
I agree about CMPXCHG16B. AMD should have including it as it seem to help creating fast, explicit lock free algo. But to be atomic the CMPXCHG16B opcodes have to do bus locking. Because of that this don't make a hug difference at the end: The AMD implementation will have more opcodes but in a path where the bus locking is the slowest thing anyway...