Slashdot Mirror
Intel Quietly Adopts AMD's x86-64
HishamMuhammad writes "The rumors reported earlier at /. are confirmed. The latest offerings in the Pentium 4 family now support AMD's x86-64 architecture, even though Intel is not willing to admit it very openly, by using cryptic names like EM64T and (gasp) IA-32e.
(The naming issue was discussed on lkml, and the consensus there was to use 'x86-64,' even though sometimes AMD refers to it as 'AMD64'). Intel's FAQ admits their implementation is basically compatible with x86-64, except for the minor differences that have always set Athlons and P4s apart. It's about time Intel jumped on AMD's bandwagon, since its homegrown 64-bit architecture seems not to be doing
very well."
Is this the slight amount of force needed to budge the processor tug-of-war?
Excuse me, I don't mean to impose, but I am the ocean
Although LKML has appeared to agree on x86-64 the folks over at Debian appear to have gone the other way and name the arch amd64.
Your hair look like poop, Bob! - Wanker.
How will it perform compared to AMD's chips? AFAIK AMD usually performs better clock to clock?
Buy Intel, Buy Quality.
Don't you mean Bye, Intel. Buy Quality.
I suppose in most technical circles that always pull for the underdog and cheer when the big dog stumbles that items like this come as great news. But its appearing more and more like Intel is the one playing catch up. They may still have market share and a far wider range of products to support them, but AMD has taken the Intel bull by the horns and is beginning to bring it to its knees. Problem is, its the competition that has driven the market and without Intel, AMD has no identity. I just hope Intel can turn things around.
This comment was generated by a squadron of trained super elite albino ninja chickens for you.
... an Intel guy I chatted with last fall said that they did not expect to put 64-bit processors in desktop machines for at least a decade. I smiled politely. ;) -ghostis
Computer Science is all about trying to find the right wrench to bang in the right screw. -T.Cumbo?
I for one am glad to see Intel recognizing that they are no longer the only game in town.
If the world works like I believe it should, the competition between Intel and AMD will provide consumers with better products at lower prices.
Sure, AMD is ahead right now. There is incompatibility between the two 64 bit architectures, and developers may choose to design for one or the other. But the Intel FAQ is right in that Intel processors support SSE3 and HyperThreading, for which AMD has no counterpart. This is in addition to Intel performance-enhancing compilers. If developers choose to develop around Intel's 64 Bit processor, then AMD may soon find itself behind again.
A NYC lawyer blogs. http://www.chuangblog.com/
As an interesting side note, check out this story. It says that Intel reverse engineered the AMD64 architecture (which isn't terribly surprising) but then flat-out copied the documentation, even though some of their implementation didn't match up!
Nice one, guys.
EM64T also known as 64 bit memory extensions, have been in the P4 Spec back to the Willamette Days, they have simply "Turned them on" (Heard from an Intel Channel Conference)
And is there anyone -- anyone who reads Slashdot at least -- that didn't already know this?
"It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
I love it. The slashad in this story for me was SGI's pushing their Altix on Itanium2!
HA-HA!
with a x86-64 can i run every software writted for a i386 cpu type?
\n.\n
You make me laugh...
I like the cut of your giblets...
- Your stupidity got you into this mess, why can't it get you out? -Will Rogers
Granted it does not perform well comared to AMD.
Trust me. IA64 is not going to die any faster than *BSD or Apple. As soon as Intel lowers the price and the low voltage higher clock rate chips come out. IA64 is good. Trust us.
Sure, you can't build a $1500 Itanium box, but at the same time, the second fastest computer in the world is powered by Itanium processors. So is the fifth. AMD Opterons power #17.
their glory days are now more or less behind them. No computer in my house uses Intel processors. My family has running AthlonXPs, 1 running a Sempron and 1 Powerbook with a G4. The 32bit AMD hardware is very, very affordable and perfect for tossing together something that just works and needs to be run by someone who doesn't have a lot of disposable income.
No one I know of talks about Intel and 64bit processors except to make fun of the Itanic. The Athlon64 and Opteron processors on the other hand are the objects of lust for many of the geeks I know. When they think 64bit that they can own, they think either AMD or Apple, not Intel.
Click here or a puppy gets stomped!
in early 2005 Anandtech
Back on December 26, 2002, Robert X. Cringely stated this would happen.
If someone says he and his monkey have nothing to hide, they almost certainly do.
More like Buy Intel, Bye Quality
Intel's been talking about this for some time, and it's been posted on /. numerous occasions.
I guess we're trying to paint them with a bad brush, just because. I don't see anything quiet about it.
Do you mean quiet as in they aren't saturating the market with bullshit about how much more amazing the internet will be with 64 bit extensions and other nonsense claims designed to sucker the technically illiterate into upgrading for no reason?
I don't need no instructions to know how to rock!!!!
Uh, dual cores on a single die?
"It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
Yeah Let me see which one the gray unwashed masses is going to pick.
The one supported by MS's Windows or the one that is not?
The one that came out first and by 1Q2005 will have dual processors on chip or the one that will have same arriving one years later?
The one that has a memorable name that is fast catching on AMD64 or the one called
SomethingOrOther-64-notquitesurewhatwewillstickint heend?
Hard question indeed.
Help fight continental drift.
One would think that Intel, better than anyone else, should know one simple fact about the computer universe:
Try as you might, you just can't get rid of x86.
RISC vendors failed. Intel's own RISC efforts failed. Itanium is an overengineered design that nobody wants. What did they think was going to happen?
In the world of computers, especially PC type computers, backwards compatibility is king. That's what keeps incumbents like Intel and (especially) Microsoft on top. You'd think they'd know this better than anyone else. Has AMD beaten Intel at its own game? Time will tell.
Look on the bright side: the complete failure of Itanium in the marketplace (let's call it what it is, even though Intel hasn't officially thrown in the towel yet) means that we won't be stuck with an entire generation of computing where Intel calls the shots. Can you imagine what would have happened if Itanium prevailed and nobody else was allowed to produce a compatible processor?
Tired of FB/Google censorship? Visit UNCENSORED!
single core on a single die?
Wha ha ha ha ha that's HT for you.
Did I mention that your Compy sees 2 cores for the price of one? What a bargain!
(its meant to be funny)
Is it true that more people vote for the winner of American Idol, than vote for the president? -Ali G.
So does anyone know what Intel left out of their AMD-64 (Intel will hate that reference) instruction set implementation?
"It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
Perhaps I am over reacting, but this is huge. AMD use to fight to keep with Intel, and now *Intel* is playing catch up.
...Michael...
AMD may soon be in a spot to dictate processor advances and even bus specifications. This could have siginicant impact for future PC's.
What's next? Microsoft losing the desktop?!?!
I now wish I would have held on to my AMD stock!
Where would the car industry, and the American economy, be if we had headlines like "GM Quietly Adopts Ford's Gascap Diameter"? These interoperability issues might make short-term profits for Intel, and offer marketdroid simplified lockin strategies, but they're inefficient limits to scaling the market to encompass everyone. So longterm profits are sacrificed, as well as usability. This fruit of the Intel/AMD crosslicense agreements should be congratulated and promoted as a "best practice" that's best for everyone touched by the industry - which is practically everyone.
--
make install -not war
I got my AMD64 @ 3200+ 2 months ago. It was time for a new system and in the past I would of gone with Intel but preformance and price drove me to AMD. And I wanted to be ready for win64 and when VMware has better support for 64bit on linux *waits...*
:)
as a side note if anyone is woundering no problems really just minor little things in games (WC3 has a colour problem in the mini-map) everything else i've used is damn fast
also when making a p4 3.4ghz for a friend at the same time as my system, he went with a 7200rpm hdd I went with a 10 000rpm sata hdd and it was worth every penny. I'm sure we all know how the hdd seems to be the bottle neck now.
It was only a matter of time for them to use AMD's extention. Common now, why would they want to have 2 different 64-bit extentions on the market competing. This would just be extremely annoying to developers and such. On top of that, AMD's has been out now for quite a while, so trying to jump into the market NOW with their own wouldn't be very smart ...
Thats my 2 dollars...
http://developers.slashdot.org/article.pl?sid=04/
or was it when they started shipping 64 bit Prescotts?5 &tid=118&tid=137&tid=126
http://slashdot.org/article.pl?sid=04/08/06/00025
Just because it shows up on the Register it is now news again.
I think that's a bit overstated. They didn't just double the width of the registers and data paths. They upped the address size beyond what I'll be able to afford in the rest of my life, added more registers overall in 64-bit mode, and generally seem to have dealt with the worst constraints imposed by backward compatability with the original 8086/88 processors.
It's hard to call an Opteron an x86 chip. More accurately it's a superset of the x86 archtecture.
What I really wish they'd do next is what IBM pioneered with their 400 series mid-frames. In those systems with 44-bit addressing, every byte of data -- including every byte on every disc drive -- had a unique address. I thought that was a groundbreaking idea at the time.
"It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
I like how theres an ad for Itanium2 processors on the slashdot article links for HP and Microsoft not supporting Itanium. :P
"Intel® EM64T is one of a number of platform innovations Intel is delivering"
So... copying somebody else is "innovation". So that's the definition Microsoft has been using all these years!
The global economy is a great thing until you feel it locally.
Talk about old stuff. Man slashdot is getting behind. This stuff was anounced at Intel's development conference in like April.
Should be enough for anybody.
Agent K: A *person* is smart. People are dumb, stupid, panicky animals, and you know it.
So much for "quiet"...
The Intel chips don't have an IOMMU. This means that unless the chipset provides one (none currently do) 32 bit PCI cards cannot do DMA into memory beyond 4GB, forcing the use of bounce-buffers.
In short, 32 bit PCI cards on systems with > 4GB memory will be G L A C I A L L Y S L O W.
On AMD64 the IOMMU remaps memory for 32 bit DMA below 0x10000000, thereby allowing 32 bit cards to access the full 64 bit address space.
The lesson: Buy the original. Buy AMD.
If the big advantage of these new 64-bit processors is nominally found in servers, then AMD will clean house because their systems scale and perform VERY well in the server role compared to Intel. Sure, you may not be able to tell the difference between AMD and Intel on the desktop, but for most types of server loads, there is no contest. The Opterons are very, very good server systems, and for many types of loads e.g. database servers, they run rings around Xeon processors for a very low cost.
Unless Intel matches a very competent ccNUMA and I/O fabric to their EMT64 cores, they will not be competitive where it matters.
It's funny, b/c they know they don't have "Anything Inside". Now their little stickers are going to say "Powered by AMD Goodies".
Help me, help you. - Jerry McGuire
Intel forgot what made it successful--backward compatibility. With the huge amount of programs now, switching to an architecture that is not compatible with its x86 line came to bite it in the ass. I guess a better architecture is less important than being to run your old programs.
You mean the "rumors" aren't officially "news" until they appear on /.? Forget what we've been reading since Febuary on http://www.anandtech.com, http://www.tomshardware.com, http://www.theinquirer.net, http://www.arstechnica.com, http://www.hardocp.com, http://www.aceshardware.com, and of course http://www.intel.com, it's not true until it appears on /. ...
/.'s FP.
PSSST!!! I've heard the rumor that Apple is planning on ditching Motorola's chips for IBM processors in their upcoming Macintoshes. Has anyone elseo heard about something called a "G5"? Some say it might also be 64 bit? Heavens-to-Betsy, let's post it to
Get off my virtual lawn, you damned virtual kids!
We've known about this for months. Why is it news now?
"How will it perform compared to AMD's chips? AFAIK AMD usually performs better clock to clock?"
Comparing processors "clock for clock" has never meant a lot, and is meaning less and less all the time. Different designs do things so differently that clock rate has about as much to do with actual performance as the color of the chip package.
The best measure of CPU performance remains the price/performance ratio. That is, for a given amount of money, how fast will a CPU perform a given task? In other words, how much bang for the buck. AMD has consistantly been beating Intel in that department for years. Sure, you might find a chip from Intel that is 10% faster, but it will cost you 80% more.
Even comparing price/performance on just CPUs has become difficult to impossible. Core logic (especially the memory subsystem and periperal bus) have become so important, and so differentiated, that establishing an apples-to-apples CPU comparison is hard. So instead of comparing just CPUs, you have to compare CPU/chipset/memory combinations.
dragonhawk@iname.microsoft.com
I do not like Microsoft. Remove them from my email address.
For those who missed it last time around, Linus was also tempted to call it amd64 in reaction to intel's handling of the subject but decided to stick with the vendor neutral x86-64.
:)
And yeah, this moved from the realm of rumor to fact nearly a year ago
This looks good on them, suckers! lol
They chose AMD64 because that is the name of the platform. AMD came up with the platform, and thus named it how they chose. Plenty of people supported AMD64 before Intel made compatable chips, and it would be stupid to renamed the arch just to please Intel after the fact. Kinda like how i386 is called i386, since Intel made it.
And given that AMD at least supports open source, and donates hardware to linux distros and BSD projects, and intel are complete assholes about even trying to get docs for hardware, much less donations, I think supporting AMD in naming their arch is the least we can do.
The Intel guy has a point. Most people have no use for a 64-bit address space, at least for right now. Indeed, increasing the address space can actually slow things down.
It is only the fact that the AMD64/x86-64/whatever design also happens to add more general purpose registers that makes it useful for every day computing. Of course, you can have more registers without having a 64-bit address space, but that's too complex for most people. So "64-bit is better" wins out.
dragonhawk@iname.microsoft.com
I do not like Microsoft. Remove them from my email address.
This is very old news, this was decided about two to three months ago (or even longer) on the gcc list.
HyperThreading is quite different from a dual-core chip.
HT is more like dual-pipeline but still only one arithmetic units (and some other executions unit).
So when your first pipeline is stalled you maybe able to use the free cycles for the second pipeline.
Slightly offtopic, but a few days ago Valve's Steam stuff (the bit installed on the victim's PC) ran a poll about gamer's hardware, in which I participated.
/proc/cpu) aside, it's been AMD all the time. I hope they keep doing well.
I was very surprised by the intermediate results: 47% was running an AMD CPU (lots of them 64 bit), Intel at 51% and the rest other wacky stuff. Considering that gaming is a major drive (maybe only windows upgrades are more important --- and those are few and far between lately) in processor upgrades, I'd be worried if I were intel.
Personally, I've been a happy AMD user since their 386-40MHz. A brief flirt with a Pentium Pro and even a fling with a CentaurHauls (or something, I remember that name from
We had eight of these (smp so 16 actually) installed in one of our racks for about 2 months before we realised that we could just install an x86_64 linux distro onto them. The performance increase we saw for IO was remarkable.
Now, I'm not up on IA-64, so I can't comment on whether it's a good design or not. However, I think it's unfortunate for intel that they are in a situation where they are continually blasted by IT community.
The primary complaint with the iterations IA-32 is that the chip was overly complex (from a design and programming perspective) because Intel chose to maintain backwards compatibility. However, when they decided to start completely from scratch and create a completely new architecture with IA-64, they were roundly criticized for (wait for it) not maintaining backwards compatibility.
Now they are being written off as an also-ran that is now trying to "catch up" with AMD, which is a bit laughable when you consider that they only did what AMD has been doing for years (namely, copying a competitors design).
I think it's great that there's competition in the market, and that the consumer has largely benefitted from having two x86 vendors. But make no mistake: AMD is not the "good guy". They are not innovators by any stretch of the imagination. Many of their employees suffer deplorable work conditions. They offer a similar product at cheaper prices. End of story.
I would prefer systems that are quiet enough to be in the same room with a TV, for example.
So what you want is an AMD64 system implementing Cool'n'Quiet?
I have to agree with Intel that 64-bit desktops don't make a lot of sense right now.
You were saying? :-P
The once-dominant and still-superior 64 bit Alpha processor architecture (which is unfortunately now owned by Intel) has also been suffering a decreasing market interest under Intel's evil and terminal stewardship.
After Intel bought the rival Alpha technology, perhaps they should have phased out the Pentium instead.
Comment removed based on user account deletion
All recent (around 2 months) servers we've
been getting from HPaq, with 3.4 GHz Xeons, are all 64 bit capable. SuSE SLES 9 even says on install "you are installing 32 bit software on a 64 bit computer".
By that reasoning, a Civic has better "performance" than a Porsche, because it can go half as fast as the Ferrari, but only costs 1/5 as much.
All chip makers (yes, even AMD) charge a premium for their fastest chips. If you want a chip that's 10% faster, you'll usually pay more than 10% extra. And there are people who are willing to pay that, either to impress their friends or because the extra speed improvement is actually useful to them.
...remember when he claimed to have a Ph.D?
...like price/performance.
sure you can get faster intel chips, but you will pay far more for the privilege.
price/performance, amd completely wastes intel on all levels.
Um but the amd64 can run 32-bit code [and 16-bit code] in full speed without "emulation layers" like the ia64 does. Sure the amd64 does more than an 8086 does but it still does what an 8086 does [and more].
Hell, everytime you boot an amd64 into winxp you're basically starting up a glorified 32-bit cpu with zero 64-bit extensions.
Tom
Someday, I'll have a real sig.
These super computers are nice for bragging rights, but they are only a very small fraction of what makes a chip successful or not.
It's like claiming that a car make is better than the other because it's top of the line model is good. If they sell 200 of those a year, it's only a PR asset. You have to look at what people are actually using.
Treehugger? Treehugger... Treehugger!
> Can you imagine what would have happened if Itanium prevailed and nobody else was allowed to produce a compatible processor?
No, I can't.
Because it already happened in the software world.
And this is so evident I almost have to acknowledge trademarks.
Note that currently the architecture supports 48 bits of addressing (see the wikipedia, though you'll have to do your own calculations to figure it out (256 terabytes = 2^48).
I figure we'll exhaust that probably sometime around 2028 or so (doubling memory requirements each 18 months). I may be approaching retirement age by then, but statistically am likely to still be alive.
It may be possible to extend the architecture to support full 64-bit addressing, however, which would be likely to do us for almost another 50 years from now.
Oceania has always been at war with Eastasia.
They've got, what, 85% of all desktop and server CPUs? And they got there with some pretty unethical (though never provably illegal) techniques and tricks. No Intel executives are in danger of starving to death. They don't need your pity.
Inside a few Dell PowerEdge servers. Of course Dell gets first dibs since they're the only major manufacturer who is Intel-only...
Intel® Celeron® processor, 325J, 2.53GHz, 256KB Cache, 533MHz FSB
Intel® Pentium®4 processor, 520, 2.8GHz, 1MB Cache, 800MHz FSB [add $99]
Intel® Pentium® 4 processor, 3.4GHz,1MB Cache,800MHz FSB, EM64T [add $249]
Intel® Pentium® 4 processor, 3.6GHz,1MB Cache,800MHz FSB, EM64T [add $349]
The Doormat
If you're not outraged, then you're not paying attention.
Way back when I read that Intel's 64Bit processors would not run 32bit code, and that one would need an software emulator -- but AMD was going to have both 32Bit and 64Bit in one chip - I was surprised and couldn't believe it.
So Intel basically let AMD do all the work and spend the R&D dollars, and Intel comes in and assimulates it.
Sounds like what AMD did with the 386 (aka 486) Intel processor. What goes around comes around.
One day, soon...
we will all look back on thes AMD vs Intel arguments and find them quaintly silly, since the two companies will have merged to try to stave off tremendous competition from cheap Chinese processors.
You know I'm right.
I think why it is a story is that this is the first time Intel is shipping in their "normal" processors as opposed to their expensive processors.
All the main processors (2.4 GHz upwards with HT) now have this available as opposed to before where it was Xeon or Extreme Edition only which were >$1000.
All I can say is that I wouldn't buy a PC without 64 bit extensions now...
It's much worse than that, unfortunately. The bounce buffers must be allocated in the low memory (below 4G for sure), and the only way to ensure that is to allocate them at boot time. Linux kernel does it with the SWIOTLB buffer. You can specify the size at boot, but after that it's fixed. If DMA ever requests more memory than the buffer has, the kernel will panic (apparently latest 2.6 kernels have some more graceful way to handle it, but in any case DMA requests cannot be fulfilled once there is no memory for bounce buffers). On the other hand, SWIOTLB memory effectively disappears from the system.
So, if you have a nice gaming system with 256MB video card, you may need at least that much memory just for bounce buffers, or more: I'm not sure what the exact requirements are, but I've seen EM64T boxes which would be stable only if SWIOTLB is twice the size of video RAM. Half a gig of RAM not available to the system. So at least for gaming boxes, buy AMD64, don't buy EM64T.
It's hard to call an Opteron an x86 chip. More accurately it's a superset of the x86 archtecture.
Kinda like how the 80286 is a superset of the 8086.... and the 80386 is a superset of the 80286.....and how the Pentium MMX is a superset of the 80386.... and how SSE makes the..... get the drift?
Opteron is an x86 chip. It's just been too long since we've had a 286-to-386 scale change in the architecture.
The bounce buffer is only if you want to DMA data from above 4GB down to the 32 bit range the card sits at, no?
Since you have 4 GB of address space that the 32 bit card can get to, the worst case to me would be if you have around 4GB or more in a system. In this case you'd have to waste some of your RAM in the 32 bit region for the bounce buffer.
Otherwise, say you have a 256MB of RAM and a 256MB video card. You can simply have your video card reside at the 256-512MB memory region.
I'm probably missing something...check my work.
Does it hurt to hear them lying? Was this the only world you had?
WE MUST NOT HAVE A BIT GAP!!!
Lameness filter encountered. Post aborted!
Reason: Don't use so many caps. It's like YELLING.
I think Intel is quietly adding support for the x86-64 architecture due to the fact that Microsoft will soon release a version of Windows XP that will fully support the x86-64 architecture. I believe that the target ship date of this new release is some time in the first quarter of calendar year 2005.
That 268 435 456 megabytes, if read from address 0 to the end, would take 3.034 DAYS to finish at 1 GIGABYTE per second.
Damn, that is a lot of memory
I've seen it mentioned here that the Intel stuff has SSE3 and Hyperthreading, and AMD has Hypertransport and pretty good I/O in general. What nobody seems to have mentioned is that Intel was planning on leaving out support for the "page table NXE bit" which enables some nifty security features, on OpenBSD anyhow:
http://www.openbsd.org/amd64.html
Does anyone know if or when Intel will remedy this? I seem to remember reading that it wasn't a permanent problem, and eventually they would add the feature or something.
Now before I get modded down, I be to remind whoever might read this that what I am saying is FACT. - bogaboga
SSL has it's performance double when using 64 bit. That's one common application that anyone will find useful, besides the greater registers and increased memory space available for applications and operating system both.
Us gamers are already looking to 64 bit. I think Microsoft might be right about 2005 being it's year of implementation. Of course, that is assuming they ever ship 64 bit windows!!
Their 64 bit architecture is Itanium, and that's not in desktops yet. Prescott & other EM64T-enabled chips are "32 bit chips with 64 bit extensions" - hence their terminology of IA-32e. It's just a normal 32 bit architecture with some 64 bit address & data registers bolted on and a few new instructions to access them, not a real 64 bit architecture.
So by that logic, I figure that Prescott is really a 4 bit CPU, with extensions...
Why would anyone engrave "Elbereth"?
AMD64 will never be adoped by Intel. em64t is ridiculous. IA-32e is equally stupid and misleading to boot. x86-64 is too long. Why not just say x64? The x has become associated with x86 so it's easy to tell what we're referring to. It also makes it painfully obvious that this is the successor to x86 instead of the Itanic's IA64 architecture.
Not just "may be possible to extend", but it is possible. The current silicon does not support address space over 48 bits (actually I think I've read it doesn't support over 40 bits), but the architecture is there and x86-64 software must use full 64-bit pointers already so they will work. The CPU design itself can be upgraded quite easily when more than 40 or 48 bits actually becomes a necessity.
Perhaps HURD will be worth something again with x86-64 systems. Go AMD.
okay, agreed. the adequacy.org link was a bit obscure. It had a blurb about how AMD chips would turn your son into an evil hacker and so I went with it. 'Cus it's funny, see? Okay, maybe not.
Still, though... flamebait? Of all days I choose to post, I do it on the day that Jerry Sanders gets mod points. Damnit!
Yes, they could make it so that SWIOTLB is not needed when you have less than 4G RAM, but, unfortunately, it's not done this way. I really wish it was, nay be future Linux kernels will change this. But SWIOTLB is always used on EM64T systems, so the problem is not for folks with over 4GB of RAM, it's for those with 1GB of RAM and 256MG video card.
As far as having the card reside at 256-512MB memory region, that's exactly what bounce buffers do. The problem is, once you allow kernel to give out memory from that region to other applications, there is no way to get it back, so the only safe way is to hoard this region in case DMA ever needs it.
If you run an x86 operating system, then the cpu behaves exactly like a normal x86 chip. All the 16-bit and 32-bit software that you normally expect to work on x86 will work. All x86 operating systems should run on this cpu just as well as they run on a traditional x86 cpu.
If you run an x86-64 operating system, then within that operating system you can run 32-bit x86 applications, but not x86 device drivers. In this mode of operation, the OS and the device drivers must be 64-bit. The applications can be either 32-bit x86 or 64-bit x86-64. If your 32-bit (resp. 64-bit) application needs shared libraries (DLLs or .so's), then you need the appropriate 32-bit (resp. 64-bit) libraries as well. For most people this means you need both 32-bit and 64-bit versions of system libraries like the C library in order to run mixed code. It is expected that 64-bit OS vendors will provide both versions.
One note about compatibility that needs to be stressed: the 32-bit compatibility is very very good. You may remember back in the windows 95 days that it was possible to run 16-bit MS Word on 32-bit windows, and it just worked. The same is true for the 32 to 64 bit transition: 32-bit office suites run correctly in a 64-bit OS and in fact this is literally what Red Hat does (they rely on a 32-bit version of openoffice.org in their AMD64 distribution). Even JIT bytecode-to-x86 Java runtime environments work correctly in x86-64.
Lastly, you probably won't care, but 16-bit x86 software does NOT work in 64-bit mode. You have to run a 32-bit OS if you want to run 16-bit applications.
I don't see anyone lauding AMD for extending the registers to 64 bits. What AMD did goes far beyond that. They have effectively doubled the number of general purpose registers, upgraded the memory management to a full modern feature set, moved the memory controller to the CPU, and did all of that at a lower price with less heat.
The x86 has long been handicapped by too few registers. Intel could have fixed that long ago, but didn't. The hated segmented memory architecture still shows up in certain aspects of the memory management design. Intel could have fixed that, but didn't. And Intel could have made these improvements at any of the 32 bit interations, but they didn't.
I've never been a fan of the x86. It has always had numerous flaws, but Intel's process technology has enabled them to deliver good performance anyway. AMD saw the flaws and fixed them. Intel saw the flaws, wrote off decades of experience, and switched to a new design with a new set of equally crippling flaws. That's why people applaud AMD and dump on Intel. AMD did what Intel should have done years ago.
I can't help but feel smug about this. Intel stood on the sidelines saying
"Oh you don't NEED 64 bits and please note how we have more mhz"
while AMD was actually doing something and said
"Well, you know, 64 is TWO TIMES 32, and you can get some RIGHT HERE."
Intel roadmap:
1. Whine
2. Bitch
3. Moan
4. Conform
I regularly report MSN spam to the Hotmail admins.
Moving the memory controller (Northbridge) on-chip certainly helps memory latency for things that blow the cache, but a second issue is the wonderful TLB that AMD implemented.
Normally, a TLB has something like 64 entries, meaning that it can map 256K of memory (at 4Kbyte pages) at a time. When you go outside that, not only do you have to take the cache hit, but you ALSO have to load the address translations into the TLB from the in-memory page tables.
Major ouch. AMD added a 512-entry L2 TLB that greatly reduces the penalty. It's not single-cycle, but it's a lot better than going to memory. They also added a bunch of cohernecy logic so they don't have to flush it every context switch.
Anyway, it is a Big Win in large-memory workloads, and I don't think that Intel has anything like it. They copied the instruction set, but AMD made some nice architectural optimizations as well.
Back in the day, DEC engineers put a "gotcha" on their chip masks after seeing their designs pop up in Russian made fabs. Magnified sufficiently, you could actually read the words, "VAX: when you care enough to steal the very best."
Sounds like AMD has earned the right to use that line...
*** *** You're just jealous 'cause the voices talk to me... ***
"The rumors reported earlier at /. are confirmed."
They were not rumours!. They were facts.
I don't see this as any news...
This once again proves that Pentiums are for faggots.
Microsoft are pushing the AMD architecture now. They want people to be running 64-bit in two years' time, and they're not talking about Itanium 2. Intel themselves have admitted that Itanium 2 is now only aimed at big iron, as it was a flop in the desktop and standard server markets.
When I am king, you will be first against the wall.
As I understand it, initial dual-core AMD64 chips will have dual caches feeding off the same single or dual RAM channels previously accommodating a single-core chip. With half as much memory bandwidth available per core, does that imply more frequent pipeline stalls and therefore more benefit from SMT?
Comment removed based on user account deletion
Intel did overlook one important aspect. They do not honor the NX bit. AMD Opterons and AMD64 CPUs do. That is one very big difference for alot of people.
We all remember how cumbersome memory segmentation was when programming under DOS/Win16. 4GB(?) segments may be less a kludge than 64K ones, but what's the point to go down that route when you can code in flat space under x86-64?
People who dislike China tend to mention Tiananmen Square a lot, but they always forget the Tank Man is also a Chinese.
Do you have it in standard units like Libraries of Congress, 'Trips to the moon and back if each bit was a penny and each penny piled' or light years of volkswagen-length?
;)
Bonus marks will be awarded for a statement in terms of lengths of string.
Also, were you counting big endian or little endian?
alpha - belongs to HPQ (hp+compaq)
amd64 - AMD
arm - Intel
hppa - HPQ
i386, ia64 - Intel
m68k - Motorola
mips, mipsel - MIPS
powerpc - IBM/Motorola/Apple
sparc - Sun
s390 - IBM
It's better to be the foot on the boot than the face on the pavement. ~~ tkx Kadin2048
http://www.theinquirer.net/?article=4668
.. it says so right in that weks article.
Whacko Cringely fans dont know anything outside of what he says when he's merely stating what others said before.
Like "his" Baxter idea. There's exact implementations of his idea out there today, but of course he knew that but realized that few knew about it. In fact the week after that he was going to admit it but decided not to
Let's not give Richard Stallman any reason to complain.
You can be a genius in the computing field by merely reading old papers and re-implementing their ideas. Most people don't read the literature. Save your brain cycles for inventing stuff that's really new. Stand on the feet of giants!
I attended a one day intel channel conference last week and they talked about this when presenting the CPU roadmap for the next few quarters. They were calling it EM64.
What was more interesting is how they seriously played it down as unimportant. It was like, "we now have 64 bit!" "But there are only 2 versions of linux and a beta version of winXP that use it so it's not really that important." "and all your apps are still 32 bit so it doesn't matter anyway"
Basically, it's not important that we had to copy the other guys stuff and not offer it til almost a year later because nothing really important *cough*NON-microsoft*cough* runs 64 bit anyway. But we have it!! And the itanium had it a year ago! (was amusing how he threw that in too)
My coworker and i tried not to laugh out loud.
No unauthorized use. Trespassers will be shot. Survivors will be shot again.
It's also approximately 7.5 years of DVD-quality video (based on 4 gigs an hour, 24/7). In memory.
Oceania has always been at war with Eastasia.
It's Intel's fault for not naming their chips ending in '86' any more. You can say 8086 assember or 80386 assembler and people will know what you mean... the 64-bit version ought to be called an 'i786' or something, but Intel marketing decided to drop the numbers after the 486.
-- Ed Avis ed@membled.com
There is still difference between EM64T (Intel's x86-64 implementation) and AMD64. EM64T was derived from the long-time undisclosed project, codenamed as Yamhill. From sources I have read (x-bit labs? could not remember), Intel actually reverse-engineered AMD's CPU and AMD's early-openly development documents to get Yamhill to work. But in AMD's document, AMD made a mistake of leaving one instruction out of documents. However, Intel followed this mistake. So EM64T lacks one instruction that AMD64 has. But since this instruction is almost for internal use, such difference is not critical.
Why there is no yet another case?
AMD did reverse-engineering before to 386, 486 and Pentium before (AMD and Intel had filed case for this). And they two are all messed with numerous cross-license after so many years' R&D, manufacture and acquisition. So any case is just wasting and results in lose-lose case.
I just felt shame for Intel the whhole process of rolling out EM64T. Hope both of them can be honest. As consumer, let us enjoy more price-cutting.
>>LEAFyoung