What will happen with the introduction of the 64bit desktop processors?

[Structural Integrity]

I'm wondering what will happen to the desktop world with the introduction of the 64bit desktop processors, which will happen somewhen this year. I'm particulary worried about the OS-es and support of programs under these OS-es.

I believe that Windows Blackcomb will be the first 64bit windows desktop OS and that Linux already has a few 64bit distros, but does anyone know what will happen to all the 32 bit applications? Will they remain compatible with these OS-es?

And how about current OS-es? I heard the Athlon64 will be backward compatible with 32bit OS-es, but the first Intel 64bit desktop CPU won't. I really don't like being forced to switch to something like blackcomb because 32bit OS-es aren't supported.

Ofcourse, a full change to 64bit desktop systems will take a few more years, but it's ONLY a few more years.

[Leshy]

To be honest, I think very little will happen. The 64-bit processors will most likely be introduced initially on the server market, and make their way to the desktop market from there. For 64-bit processors to become mainstream will probably take another few years. Remember that there is no use for Intel to release a 64-bit only processor if the software to support it/take advantage of it isn't there.

AMD doesn't have that issue, as it's processor should be able to run both 64- and 32-bit software, simultaneously if needed. This could very well give AMD an edge over Intel, as it's fully backward compatible with the software that home users and companies already have.

At any rate, once systems with 64-bit processors start appearing, operating systems used will be WinXP64 and Blackcomb, which will be distributed in much the same way as any new OS is.

[Atamur]

iirc, Intel's Itanic is able to run 32-bit code through emulation albeit at a significant performance penalty.

[BesigedB]

Looks like intel have shot themself in the foot then.

[Structural Integrity]

Leshy, there are numerous 64bit processors out there that are specially made for servers. The AMD hammer however (the athlon64) will be especially aimed towards the desktop market.

Does anyone know if the 32bit desktop platform will remain to exist and will be developed further after the introduction of the 64bit platform? Or can we all wave byebye to it?

[MT]

64 bit on the desktop? what rubbish have you been reading?

For a start, its not as clear cut as "64 bit" software. The two potential competitors are Itanium 2 (already here, clocked at 1 GHz iirc) using IA-64 architecture, and AMD Opteron (not here, probably not going to make it tbh) which uses x86-64 architecture. The two are not compatible. Microsoft havent decided whether to compile an OS for x86-64, whereas it is (iminent|here) for Itanium. Then there is software. Will companies really start developing 3 versions (x86, x86-64, Itanium2) of their software? I doubt it.

Then consider that Intel sees Itanium as a workstation/server OS only, and it is considerabley pricey for what it delivers. The important thing that will start a move to 64 bit si when a LOT of programs need to address more than 4 GB of memory - theres very few applications around that require that atm.

AMD have shot themselves in the foot with Opteron - unless they come up with some clever technology like Intel's HT to split the processor into two logical processors.

MT's top reasons why x86-64 will not have great uptake on the desktop

1.) Cost. It will cost a LOT.
2.) Performance. Compared to what people are already runnign atm, there will be negligible increase in performance. Its quite likely that performance will be degraded severely from their x86 chips
3.) It just isnt required atm.

Top reasons why IA-64 wont have a great uptake on the desktop

1.) Intel dont want it to.

AMD would have been better off in developing low power cpu's for laptops etc.

[Structural Integrity]

MT, have you ever taken a look at the AMD ClawHammer, also called the Athlon64? From what I have read it will be a 64bit desktop processor.

http://www.amdboard.com/amdroadmap.html

[Nodrog]

Why would the average user currently need a 64-bit processor for a desktop PC?

[queball]

So they can run GNU/HURD with partitions bigger than 2GB.

[Nodrog]

Originally posted by queball
So they can run GNU/HURD with partitions bigger than 2GB.

Why would the average user currently need a 64-bit processor for a desktop PC?

[NEWSBOT3]

whats the advantage of 64 over 32 anyway ?

[Not_RIT]

Originally posted by Nodrog
Why would the average user currently need a 64-bit processor for a desktop PC?

Because it sounds better than 32bit and its 'new' and will be marketed by Dell et al.

[meglamaniac]

Originally posted by NEWSBOT3
whats the advantage of 64 over 32 anyway ?

Purely and simply, it can do bigger numbers at once in hardware.
I'm sure you remember the days when a C64/Spectrum/Amstrad was a cool home PC. They were 8 bit, and couldn't do numbers bigger than 255 in hardware. In fact if you wanted SIGNED numbers they couldn't go over 127 (-128 through to 127 = 256 combinations).
So yes you can work bigger numbers in software (combining bytes and stuff) but it's many times slower, and when even a calendar has to rely on software maths routines (365 days in the year) its a severe limitation.

However todays machines are 32bit.
They can handle numbers from 0 to 4,294,967,296, or -2,147,483,648 to 2,147,483,647. For everyday 'desktop' needs this is fine and (as has been mentioned) allows up to 4Gb of directly addressable memory.
There simply is no reason to switch to 64Bit for the desktop because there are no benifits to the vast majority of users. If you need huge ammounts of ram for your server then there are various 64Bit processors allready about and various OS's compiled for them (there's a linux distro for most of them for a start).

[Nodrog]

Originally posted by NEWSBOT3
whats the advantage of 64 over 32 anyway ?

What megla said. The 'bit size' refers to the 'width' of the data bus in the CPU. In a 32 bit processor, you can only transmit binary numbers smaller than 2^32 (4294967296) at once. Therefore it is only possible to directly address 2^32, or 4 gigabytes, of memory. As memory sizes are rapidily increasing, it is desirable in many cases to have more than 4 gig of RAM in your computer without having to piss about with segmentation and suchlike, hence 64 bit processors are needed. The average home user does not currently need over 4 gig of memory, but obviousy certain servers do. As memory requirements are always increasing, 64 bit processors will become widespread in several years time when 4 gigabytes of RAM becomes a 'realistic' amount to find in a desktop computer.

[Nodrog]

Originally posted by Not_RIT
Because it sounds better than 32bit and its 'new' and will be marketed by Dell et al.

lol

[meglamaniac]

Originally posted by Nodrog
The 'bit size' refers to the 'width' of the data bus in the CPU.

In general it refers to how large a number can be stored in the processors registers for direct manipulation. Clever tricks can be pulled with the architechture to allow different bus widths.
If it's any indication of how long 32bit has been satisfactory, the first 32bit processor was the 386 and we havn't moved on since in terms of numerical capacity.
Intel also pulled some clever tricks to make sure that the 32bit mode of the 386 wouldn't interfere with older apps - hence the dreaded "386 protected mode" (protected mode was, and still is, the only way to access the 32bit capabilities) and all the problems that tended to happen with old dos games and windows 3.11.
It really isn't viable to do the same thing again, unless you want protected protected mode for 64bit features. The whole story of the Intel chip and the PC is one long series of bodge jobs trying desperatly to make sure everything would still work and no company was going to loose money or sue because thier apps wouldn't work any more.
For 64bit, it's probably best to start again (at least on the processor level).
Unfortunately many other things not directly related to the processor are, in computing terms, held together with masking tape and prit-stick. This includes memory addressing (which is 24bit at the moment - yay for segments), the various different 'levels' of memory (base, upper, extended, etc), the fact that to write to your IDE CD burner the system has to in effect 'pretend' it's an SCSI drive, and so on and so forth.

PCs are a mess.

[Leshy]

Originally posted by meglamaniac
There simply is no reason to switch to 64Bit for the desktop because there are no benifits to the vast majority of users.

Without the addition of 'right now' or 'in the near future' in that sentence, I'll class this in the same category as Bill Gates' statement that we'll never need more than 64k memory

[meglamaniac]

I thought it was obvious that I meant 'currently'.

Obviously in a few years windows is going to need 2Gb of ram just to get past the bootloader, never mind into the GUI, so we'll get there in the end.

[JetLinus]

I know at least one use for 64bit Processors:

Chess AIs!

As the chess board has 8*8 = 64 fields, we can have a "bit"-map for lets say possible moves and for different "figures//characters" (thew things you play with).
Then you can process those "arrays" in one step -- quite fast.

I think Deep Blue has 64bit? Not?

[Atamur]

Originally posted by Not_RIT
Because it sounds better than 32bit and its 'new' and will be marketed by Dell et al.

Right on. How many ppl proudly announce that they have 128meg video memory but have no idea what you are talking about when you ask them what kind of 3d games they play? Numbers sell. P4 1.4 wiht PC600 memory Ghz was slower than P3 1Ghz with PC133 but it sold. computer store salespeople will tell you, "this computer is 32-bit but that one is 64-bit. twice as many bits means it's twice as fast. the 32-bit stuff will be obsolete in 2 months, you know..."

Realistically, very few desktop users will need 64-bit platform anytime soon. maybe CAD or heavy-duty graphics applications will benefit...

[meglamaniac]

Originally posted by JetLinus
I know at least one use for 64bit Processors:

Chess AIs!

As the chess board has 8*8 = 64 fields, we can have a "bit"-map for lets say possible moves and for different "figures//characters" (thew things you play with).
Then you can process those "arrays" in one step -- quite fast.

I think Deep Blue has 64bit? Not?

Deep Blue was a multi-processor machine as far as I know - I don't know if it was 64bit or not (I imagine it probably was). However, when you have a few dozen processors or however many it had, it kind of nullifies that particular problem.

[Atamur]

Originally posted by JetLinus
I know at least one use for 64bit Processors:

Chess AIs!

As the chess board has 8*8 = 64 fields, we can have a "bit"-map for lets say possible moves and for different "figures//characters" (thew things you play with).
Then you can process those "arrays" in one step -- quite fast.

I think Deep Blue has 64bit? Not?

Just because you can theoretically address 2^64 or 16mil TB of memory, doesn't mean you have that much of it . Besides, the only time this sort of logic is used is 4 or 5 piece endings. But those are much more effecient to store by specifying 6-bit position of each piece or something.

[Nodrog]

Originally posted by JetLinus
I know at least one use for 64bit Processors:

Chess AIs!

As the chess board has 8*8 = 64 fields, we can have a "bit"-map for lets say possible moves and for different "figures//characters" (thew things you play with).
Then you can process those "arrays" in one step -- quite fast.

I almost came the first time I read that that was how the board was represented in chess because it was so unbelievably clever.

Originally posted by Atamur
Just because you can theoretically address 2^64 or 16mil TB of memory, doesn't mean you have that much of it . Besides, the only time this sort of logic is used is 4 or 5 piece endings. But those are much more effecient to store by specifying 6-bit position of each piece or something.

Thats how its represented at almost every stage in the game as far as I know, although Im not particularly familiar with the subject.

[queball]

Originally posted by Structural Integrity
I'm wondering what will happen to the desktop world with the introduction of the 64bit desktop processors, which will happen somewhen this year. I'm particulary worried about the OS-es and support of programs under these OS-es.

I believe that Windows Blackcomb will be the first 64bit windows desktop OS and that Linux already has a few 64bit distros, but does anyone know what will happen to all the 32 bit applications? Will they remain compatible with these OS-es?

And how about current OS-es? I heard the Athlon64 will be backward compatible with 32bit OS-es, but the first Intel 64bit desktop CPU won't. I really don't like being forced to switch to something like blackcomb because 32bit OS-es aren't supported.

Ofcourse, a full change to 64bit desktop systems will take a few more years, but it's ONLY a few more years.

Why is it important? The Itanium will have IA32 compatibility, see http://www.intel.com/eBusiness/products/itanium/faqs/ Q10.

[MT]

Originally posted by Structural Integrity
MT, have you ever taken a look at the AMD ClawHammer, also called the Athlon64? From what I have read it will be a 64bit desktop processor.

http://www.amdboard.com/amdroadmap.html

What you call (Claw|Sledge)Hammer was rebadged as Opteron. Also, show me where MS say they will support x86-64.

[Coffee]

Originally posted by MT
What you call (Claw|Sledge)Hammer was rebadged as Opteron. Also, show me where MS say they will support x86-64.

I think they will.

AMD are offering (with the x84-64) a low cost 64bit processor - they are going for the server market and already people on average now ask 'what?' rather than 'who?' with regards to AMD.

If MS refuse to port to x84-64 then they dont give the server's any choice but IA-64. OTOH if the server market wants a cheap 64bit cpu then they will be forced to use Linux - they couldnt buy MS if they wanted to and in todays economic climate once a few benchtests get out 'hey guys, 1/4 of the cost for the same performance' and AMD et al gets their stuff together and actually provides a decent chipset MS will be forced to release of x86-64 or lose the low cost server market completely.

MS will not let Linux be the only OS to take advantage of x86-64, its just too big a risk as if it does take off it'll lose them huge amounts of market share in one of their key areas, the server market.

Well thats my take on it anyway.

[Structural Integrity]

Originally posted by MT
What you call (Claw|Sledge)Hammer was rebadged as Opteron. Also, show me where MS say they will support x86-64.

I read on tweakers.net about MS supporting 64bit processing in their OS. I can't find the link so sudden.

I had a discussion about this with a fellow student, and he told me that the big advantage of a 64bit processor was having a larger instruction set. Operations done in multiple cycles now, could be done in one cycle.
The clawhammer has a 64bit addon instruction set, but supports the original 32bit instructions by disabeling half of the bus. So, no emulation, like the Intel Itanium.

So, besides being able to work with larger numbers and greater precision, you also have more instructions available.

[Luckeh!!!!]

Originally posted by Structural Integrity

The clawhammer has a 64bit addon instruction set, but supports the original 32bit instructions by disabeling half of the bus. So, no emulation, like the Intel Itanium.
.

Apparently it can do both on the fly, now if i could find the blasted documentation...

[Structural Integrity]

Originally posted by Luckeh!!!!
Apparently it can do both on the fly, now if i could find the blasted documentation...

http://www.tweakers.net/ext/i.dsp/1044838650.jpg

http://www.x86-64.org/documentation_...esentation.pdf

http://www.extremetech.com/article2/0,3973,39131,00.asp

[MT]

Q: Will Windows .NET Server's 64-bit versions be compatible with the AMD Athlon 64 and AMD Opteron processors?

A: No. This collaboration is ongoing, and future Windows operating systems have not been contemplated.

Q: When will the AMD-compatible version of 64-bit Windows be available?

A: This is an ongoing collaboration with AMD, and the results of this collaboration will occur in a 64-bit version of the Windows operating system after Windows .NET Server's release to manufacturing.

Q: Does AMD know the official name of the Windows version that will work with the Athlon 64 and Opteron processors?

A: The official name has not been decided yet, but it will be a version after Windows .NET Server.

Q: Do current versions of Windows 32-bit server and client operating systems work with the Athlon 64 and Opteron processors?

A: The AMD Athlon 64 and Opteron processors are designed to support Windows XP, Windows 2000, and other current and future 32-bit advanced server products.

Q: Do current versions of Windows 64-bit server and client operating systems work with the AMD Athlon 64 and AMD Opteron processors?

A: No, this collaboration deals with future 64-bit Windows operating systems and the development of technology and products that optimize the benefits of the x86-64 architecture.

Doesnt really tie in with that roadmap you've got (especially as .NET server just got a 3 month delay, making it appear 2H03)

[hAl]

Funny thing is that HP/compaq already produce extremly fast 64 bits alpha processers that can be used in multiprocessing. Those are extremly succefull in breaking speadrecords but not commercially it seems so HP have signed contracts to port its 64 bits operating systems Thru64 unix version and OpenVMS back to intells main processorline even though intell who are investing tons have not even come close to current 64 bit performances of that Alpha range processors.

I think Windows NT already runs nicely on those alpha processors for quite a while. Not sure about w2000 but that probably does run on 64 bit alpha's as well.

hAl

[Nodrog]

woah, just found out today I have to do a project comparing 64 bit architectures for uni

[Cyp]

With the introduction of 64 bit desktop processors, it will become nessecary to use 7 bits to store the number of bits of desktop processors, compared to previously only requiring 6. This represents a 16⅔% increase in the required number of bits to store the number of bits used in desktop processors.

[queball]

Originally posted by Cyp
With the introduction of 64 bit desktop processors, it will become nessecary to use 7 bits to store the number of bits of desktop processors, compared to previously only requiring 6. This represents a 16⅔% increase in the required number of bits to store the number of bits used in desktop processors.

But requiring 7 bits implies that there are at least 65 different number-of-bits, whereas the only common values 32 or smaller are 2-bit (for processors that manipulate the number of bits in other processors), 8-bit, 16-bit, and 32-bit. This can be represented in 2 bits. So a fifth integer width increases the number of bits required from 2 to 3, a 50% increase.

[Cyp]

Originally posted by queball
But requiring 7 bits implies that there are at least 65 different number-of-bits, whereas the only common values 32 or smaller are 2-bit (for processors that manipulate the number of bits in other processors), 8-bit, 16-bit, and 32-bit. This can be represented in 2 bits. So a fifth integer width increases the number of bits required from 2 to 3, a 50% increase.

I was assuming that, for some reason, that a human might decide to write the number of bits on a piece of paper, in binary, and that the human would want to be able to represent a 0 bit computer (which wouldn't be a very useful computer). Didn't think there was really much point in a computer writing the number of bits down in binary, so it would have to be a mad human doing it. An even madder human might write 32 and 64 in fibonacci coding, resulting in 00101011 and 1000100011, resulting in a 25% increase in the number of bits required to represent 64 instead of 32.

[Laze]

Originally posted by MT
[b]64 bit on the desktop? what rubbish have you been reading?

Most of the electronics and computing press - and the technical docs

For a start, its not as clear cut as "64 bit" software. The two potential competitors are Itanium 2 (already here, clocked at 1 GHz iirc) using IA-64 architecture, and AMD Opteron (not here, probably not going to make it tbh) which

The not going to make bit will clearly come as a suprise to AMD. The beta silicon has been looking rather good.

uses x86-64 architecture. The two are not compatible. Microsoft havent decided whether to compile an OS for x86-64, whereas it is (iminent|here) for Itanium. Then there is software. Will companies really start developing 3 versions (x86, x86-64, Itanium2) of their software? I doubt it.

Microsoft announced support for the x86-64 about the same time that the AMD CEO testified for Microsoft in the anti trust trial.
The same version of windows will probably be used for both x86 and x86-64 (at least initialy) as the main operating system support is easy to add (64 bit memory addressing and allocation and switching processor modes when task switching between 32 bit and 64 bit programs). It only took a few days work to make a Linux kernel that worked like that.

Hopefully in the future MS will produce a version of windows that is more optimised for the X86-64.

Then consider that Intel sees Itanium as a workstation/server OS only, and it is considerabley pricey for what it delivers.

The Itanium is indeed a disaster. This is why people are seriouly wondering if AMD will overtake Intel

The important thing that will start a move to 64 bit si when a LOT of programs need to address more than 4 GB of memory - theres very few applications around that require that atm.

Databases,
Ultra high performance web and fileservers using memory mapped data.
Movie editing tools (it is simpler if you can memory map the uncompressed movie to work on it)

MT's top reasons why x86-64 will not have great uptake on the desktop

1.) Cost. It will cost a LOT.
2.) Performance. Compared to what people are already runnign atm, there will be negligible increase in performance. Its quite likely that performance will be degraded severely from their x86 chips
3.) It just isnt required atm.

1) Not really - the 64 bit chip has a suprisingly small increase in silicon area. When they move to 100 nm fab technologies it will probably be smaller than the old Athlon dies.
2) The performance of the floating point unit in 64 bit mode totaly screams. Finaly we have a non stack based floting point unit. The SIMD instruction sets should also benefit greatly.
The memory mapping advantages do not apply to the desktop yet - but they will soon.
The DDR memory controler built into the chip makes a very considerable difference to performance.
3) It soon will be. As soon as > 4GB of ram becomes usefull. The current 36 bit addressing hack using multiple segments on the X86-32 is horrible. Programming with far pointers just leads to pain and bugs.

[Laze]

Originally posted by meglamaniac

Intel also pulled some clever tricks to make sure that the 32bit mode of the 386 wouldn't interfere with older apps - hence the dreaded "386 protected mode" (protected mode was, and still is, the only way to access the 32bit capabilities) and all the problems that tended to happen with old dos games and windows 3.11.
It really isn't viable to do the same thing again, unless you want protected protected mode for 64bit features.

Windows 3.11 was a mess that was mostly 16 bit.
The old dos programs were also used to low level hardware access. It is not suprising things did not work. Now most programs run under real operating systems (win 2K/XP or Linux) and hence do not make use of low level hardware access (windows will not let them). This means doing another extension from 32 bits to 64 bits will be vastly less painfull than the 16 to 32 bit transition.

The whole story of the Intel chip and the PC is one long series of bodge jobs trying desperatly to make sure everything would still work and no company was going to loose money or sue because thier apps wouldn't work any more.
For 64bit, it's probably best to start again (at least on the processor level).

It is a series of bodge jobs. However there are hundreds of billions of $ invested in applications that run on the X86 architecture. If inventing a new architecture worked then we would all be on the Alpha chip by now.

[meglamaniac]

Originally posted by Laze
Windows 3.11 was a mess that was mostly 16 bit.
The old dos programs were also used to low level hardware access. It is not suprising things did not work.

The architechture of the processor doesn't reflect on the stability of the programs. The fact is that MS-DOS was a very good operating system for what it was originally designed to do; that is, to run singlethreaded programs on a 16bit processor, on the standard ram of the day.
The first problems came with the 32bit 386 and (seperately) in the way extended memory was introduced. The whole "protected mode" thing was a big issue for DOS, because it effectively meant shutting down dos every time you wanted to run a 32bit application since the processor shifted modes. The whole memory thing (hello the A20 gate etc (think that's the right number, you know what i mean anyway)) was also a mess.
All that was ok until Microsoft decided to try and bolt a multitasking interface that operated half in 16bit and half in 32bit on top of it all, and the memory wasn't protected. The whole mess that was Windows 1.0 through till the (none too soon) demise of the line with Windows ME is the cause of most of the instability associated with computers.

Low level access has nothing to do with it - you can still have all the low level access you like with 32Bit ASM, and so long as you're a competant programmer it will be perfectly stable.
The problem was the attempt by Microsoft to bodge together a cheap solution. Trying to run a 32bit GUI on a 16bit platform was allways a dumb idea.

Windows 'NT' branch has allways operated purely in 32bit protected mode and made proper use of the memory - hence the distinct lack of crashes. DOS under that branch is an emulation (hence you'll see NTVDM (NT Virtual DOS Machine) in the process list if you run an old DOS application under XP for example).
I take it I don't have to mention that linux uses protected mode too, and has pretty much since it's introduction.

Originally posted by Laze
It is a series of bodge jobs. However there are hundreds of billions of $ invested in applications that run on the X86 architecture. If inventing a new architecture worked then we would all be on the Alpha chip by now.

The stuff I said illustrates what happened last time the core architecture was changed. Do we really want that to happen again? If you carry on taping one quick fix on top of another, eventually its all going to come off in one big sticky lump and take the paint off the wall with it.

[MT]

[note to self: dont argue with Laze, he >> you]

Rebuttals;

Databases,
Ultra high performance web and fileservers using memory mapped data.
Movie editing tools (it is simpler if you can memory map the uncompressed movie to work on it)

Not atypical desktop usage (you may argue the movies, but even that is EXTREMELY high end desktop usage.)

Microsoft announced support for the x86-64 about the same time that the AMD CEO testified for Microsoft in the anti trust trial.

Thats as may be. Itanium is already supported in XP 64-bit edition. AMD's own press release states they expect support for x86-64 in the OS release after .NET server - so not this year then - a year is a long time in IT.

2) The performance of the floating point unit in 64 bit mode totaly screams. Finaly we have a non stack based floting point unit. The SIMD instruction sets should also benefit greatly.

Excellent - didn't know that about the FPU. However, with the first x86-64 set to debut at 2 GHz (in a years time!), do you really think that, given that 90% of the time it will be running legacy 32 bit code, the extra performance of a 64 bit chip will counter act whatever Intel has in its bag for the desktop. At the end of the day, AMD is a hobbyist chip MAINLY. They look for bang for buck, and how well it performs in Quake 57. The other benefits of 64 bit chips wont be tested by common day usage.

1) Not really - the 64 bit chip has a suprisingly small increase in silicon area. When they move to 100 nm fab technologies it will probably be smaller than the old Athlon dies.

Memory prices seem to be holding reasonabley steady - not the ridiculous prices of a while ago. 1 GB DIMMS are already pretty thin on the ground, and pricey at that - how many systems integrators will be willing to fork out ~£500 just to get the amount of memory that benefits a 64 bit chip

The Itanium is indeed a disaster. This is why people are seriouly wondering if AMD will overtake Intel

Save McNealy! Buy a Sun!

Most of the electronics and computing press - and the technical docs

I read the 64 bit special in PC Pro last month! (probably explains a lot tbh..)

It is a series of bodge jobs. However there are hundreds of billions of $ invested in applications that run on the X86 architecture. If inventing a new architecture worked then we would all be on the Alpha chip by now.

Quite. Why make this move now when it patently isn't required, confusing the marketplace, and, imo, potentially ruining AMD. Intel will keep plugging away with their 32 bit options on the desktop, and I hope the x86-64 chips will keep up.

Doesnt this go against company line anyway Laze? ;p

[Laze]

Originally posted by meglamaniac
The architechture of the processor doesn't reflect on the stability of the programs. The fact is that MS-DOS was a very good operating system for what it was originally designed to do; that is, to run singlethreaded programs on a 16bit processor, on the standard ram of the day.
The first problems came with the 32bit 386 and (seperately) in the way extended memory was introduced. The whole "protected mode" thing was a big issue for DOS, because it effectively meant shutting down dos every time you wanted to run a 32bit application since the processor shifted modes. The whole memory thing (hello the A20 gate etc (think that's the right number, you know what i mean anyway)) was also a mess.

Correct - you are basicaly saying dos was designed for single tasking 16 bit programs. The A20 gate hack was introduced to enable a 286 to support more tham 1Mb of ram. The hack was emulated on later machines to enable old software using it to still run

All that was ok until Microsoft decided to try and bolt a multitasking interface that operated half in 16bit and half in 32bit on top of it all, and the memory wasn't protected. The whole mess that was Windows 1.0 through till the (none too soon)

And how is that the fault of the 32 bit 386 ? The problem is Microsoft using totaly inapropriate software to provide a multi tasking environment. The chip runs the old stuff (16 bit dos) and the new(est) stuff (NT/2000/XP) just fine. It is the nasty hack trying to merge dos with a multi tasking environment that went horribly wrong.

Low level access has nothing to do with it - you can still have all the low level access you like with 32Bit ASM, and so long as you're a competant programmer it will be perfectly stable.

Windows (NT/2K/XP) will not let you randomly write to the hardware registers and random memory locations no matter what you program with. That is what I mean by low level access.

The problem was the attempt by Microsoft to bodge together a cheap solution. Trying to run a 32bit GUI on a 16bit platform was allways a dumb idea.

Correct - not the fault of the chip that MS bodged it

The stuff I said illustrates what happened last time the core architecture was changed. Do we really want that to happen again?

Let us look at the differences.
When the 386 came in our primary operating system (DOS) was not designed to multi task. It was not designed to enforce any sort of memory allocation or hardware virtulisation. Trying to make that run a multi tasking operating system enforcing hardware virtulisation was bound to be a disaster.

Now we have Win2K/XP etc which are full multi tasking operating systems enforcing hardware virtulisation. The problems of moving that to 64 bit will be vastly smaller than the problems of moving DOS to 32 bit hardware.

If you carry on taping one quick fix on top of another, eventually its all going to come off in one big sticky lump and take the paint off the wall with it.

Almost all of the old nasty hacks are not in the main code path now. All of the old nasty X86 instructions are emulated in microcode. A lot of them disapear when you enable protected mode rather than virtual x86 modes.
Even more nastyness vanishes when you put a X86-64 into 64 bit mode (stack based floating point becomes register based \o/ - you have twice as many general perpose registers so register renaming is much less of an issue etc)

[Laze]

Originally posted by MT

Not atypical desktop usage (you may argue the movies, but even that is EXTREMELY high end desktop usage.)

I will grant you databases and webservers are not really desktop use. Movie editing is however. My next door neigbour hooks his digital video camera up to his PC to edit his home videos. I would argue this is starting to become mainstream.

Thats as may be. Itanium is already supported in XP 64-bit edition. AMD's own press release states they expect support for x86-64 in the OS release after .NET server - so not this year then - a year is a long time in IT.

The Itanium has been on sale since the 29th May 2001 (almost 2 years). It is hardly suprising that operating systems support it !
Remember current versions of windows will run on the X86-64 in 32 bit mode (and will benefit from many of the general perfomance enhancements of the chip)
I expect a 64 bit version of windows to follow the launch reasonably quickly. i.e. months not years.
btw a 64 bit version of windows for the X86-64 is supposed to be out in September.

Excellent - didn't know that about the FPU. However, with the first x86-64 set to debut at 2 GHz (in a years time!)

Never believe the claimed debut speed untill a few weeks before launch. They usualy quote the speed it runs in the labs + a little bit.
The launch speed is almost certain to be competitive at the time

90% of the time it will be running legacy 32 bit code, the extra performance of a 64 bit chip will counter act whatever Intel has in its bag for the desktop.

The X86-64 has many performance enhancements that will benefit 32 bit code. The memory subsystem is one that springs to mind. Btw what does Intel have in the bag for the desktop ?
I must have blinked and missed it.

At the end of the day, AMD is a hobbyist chip MAINLY. They look for bang for buck, and how well it performs in Quake 57. The other benefits of 64 bit chips wont be tested by common day usage.

AMD have finaly designed a chip that is suitible for more than just the hobbyst just at the time Intel seems to have droped the ball. That is why it will be so interesting to see what happens

Memory prices seem to be holding reasonabley steady - not the ridiculous prices of a while ago. 1 GB DIMMS are already pretty thin on the ground, and pricey at that

Memory prices drop in long surges with plataues in between.

The price drops usualy seem to be triggered when the next generation of DRAM chips becomes commonly available.
The next generation of dram chips used on the new 2GB DIMMS is currently very rare and unbeliveably expensive. It will soon start to drop. It is better to have systems that are capable of taking the extra memory before it becomes available. After all when you have had a system for a year or so a ram upgrade can provide a welcome boost.

Why make this move now when it patently isn't required, confusing the marketplace, and, imo, potentially ruining AMD. Intel will keep plugging away with their 32 bit options on the desktop, and I hope the x86-64 chips will keep up.

It is required now - high end desktops and all servers are starting to bump against the X86-32 limits. What is high end now is standard in 1-2 years time.

Doesnt this go against company line anyway Laze? ;p

Naturaly all opinions are my own (I dont work directly for IBM anyway)

However IBM is working closely with AMD since AMD has licensed the IBM Silicon on Insulator process to build the X86-64 chips. That would also make IBM the logical fab partner if AMD finds it does not have enough fab capacity.

What is interesting is how far IBM has backed away from the itanium chip. It is deffinitely pushing the Power IV chip for performance systems.

[MT]

Ah, it must have been Hewpaq that decided to drop their own 64 bit processor to favour itanium.

As far as I know, Intel's plan is to move to a smaller die and scale up the P4 to 4 GHz + - I'd be impressed if the x86-64 can keep pace with it, considering the current breed of Athlons cannot.

Movie editing is still hideously complex - my dad bought a PC for that intention, and he hasnt been limited by the hardware - its the software he cant cope with!

[Laze]

Originally posted by MT
Ah, it must have been Hewpaq that decided to drop their own 64 bit processor to favour itanium.

Yes indeed. I bet they are worried.

As far as I know, Intel's plan is to move to a smaller die and scale up the P4 to 4 GHz + - I'd be impressed if the x86-64 can keep pace with it, considering the current breed of Athlons cannot.

The Athlon is an old chip. It started selling in August 1999 with a maximum clock speed of 650 Mhz. Generally a chip design runs out of steam when it is clocked to about 5* the release speed
e.g.
Pentium - 66 Mhz launch 300-500 ish when axed (cant remember tbh)
Pentium Pro 200Mhz launched retired as PIII at about 1.3 Ghz
So as you can see the Athlon is comming to the end of its life.

A new chip design will have far more scope for having the clock speed boosted, so unless AMD have totally bolloxed up the design they should be able to keep up.
btw PIV was launched at 1.4 Ghz so we can expect it to reash 7 Ghz.

Movie editing is still hideously complex - my dad bought a PC for that intention, and he hasnt been limited by the hardware - its the software he cant cope with!

I know people working on consumer level movie editing tools - so hopefully that will change.

[Luckeh!!!!]

Originally posted by Laze
Yes indeed. I bet they are worried.

The Athlon is an old chip. It started selling in August 1999 with a maximum clock speed of 650 Mhz. Generally a chip design runs out of steam when it is clocked to about 5* the release speed
e.g.
Pentium - 66 Mhz launch 300-500 ish when axed (cant remember tbh)
Pentium Pro 200Mhz launched retired as PIII at about 1.3 Ghz
So as you can see the Athlon is comming to the end of its life.

btw PIV was launched at 1.4 Ghz so we can expect it to reash 7 Ghz..

the above is quite wrong i'm afraid, there are certain physical limitations with CPU core Mhz scaling.

P4 will reach about MAX 4ghz, possibly only upto 3.5ghz, before Intel will need to again reduce the core size (in microns), remember, the P4 has already been reduced in core size to allow it to scale higher in speed.

[Laze]

Originally posted by Luckeh!!!!
the above is quite wrong i'm afraid, there are certain physical limitations with CPU core Mhz scaling.

P4 will reach about MAX 4ghz, possibly only upto 3.5ghz, before Intel will need to again reduce the core size (in microns), remember, the P4 has already been reduced in core size to allow it to scale higher in speed.

I dont recall claiming that intel would not have to scale the PIV again. If you read the list of chips I gave as examples they all went through a couple of core shrinks. I was talking about the usefull lifetime of the basic core design.

[Luckeh!!!!]

I still think your deductions are incorrect though, the old Slot A Athlons had off-die cache running @ half speed and I'm sure they different to the Athlon XPs of today.

They also initially went as low as 500mhz.

The original pentium went as low as 50Mhz and as high as 450Mhz (P1 50-233, P2 233-450) that is actually 9 times.

Pentium Pro 200Mhz launched retired as PIII at about 1.3 Ghz

Actually ended up as high as 1.4ghz, 7 times.

Also, 650*5 = 3250Mhz, I do hope you realise that the Athlon isn't likely to reach that high, it is on the Barton PR3000+ right now, but even then the REAL Mhz is lower than that of the Tbred PR2600+.

[Laze]

Originally posted by Luckeh!!!!
I still think your deductions are incorrect though, the old Slot A Athlons had off-die cache running @ half speed and I'm sure they different to the Athlon XPs of today.

They also initially went as low as 500mhz.

The basic core design on the Athlon is the same. There have been large tweeks such as putting the cache on the same silicon, but the basic design remains.

It is more interesting to compair the highest launch speed, not the lowest, as this is what the chips were capable of (frequently chips sold at lower speeds can do a higher speed)

The original pentium went as low as 50Mhz and as high as 450Mhz (P1 50-233, P2 233-450) that is actually 9 times.

The Pentium Pro chip was a total different core from the Pentium. The Pentium II core was based on the Pentium Pro, not the Pentium. The Pentium III was based on the Pentium II (dont you just love intel naming convensions !)
The Pentium MMX was based on the Pentium core.

Also, 650*5 = 3250Mhz, I do hope you realise that the Athlon isn't likely to reach that high, it is on the Barton PR3000+ right now, but even then the REAL Mhz is lower than that of the Tbred PR2600+.

The speed the Athlon chip reaches probably depends more on how AMD splits its development resources than what it could be pushed to.
The PIII shows how far a chip can be pushed when it has to be.

er

i want to get a clawhammer as soon as possible

cos like 64bit linux on the desktop

phw00000000000000000ar

youll be prying me off the seiling with a rake, and it wont just be because of all the acid ill be on!!!

It will rock. All you people saying 'we dont need 64bit' just think back to your most favourite person and hes most famous quote ever..

'64k is enough for anyone!'

or whatever it is that that ms twat said!

[Gayle29uk]

Originally posted by Idi
'64k is enough for anyone!'

or whatever it is that that ms twat said!

"640K of memory should be enough for anybody"

Originally posted by Gayle28uk
"640K of memory should be enough for anybody"

yeah that.

toms hardware reviewed an early sledgehammer ages ago and they had nothing but good things to say, although iir the agreement they signed to review it probably had 'say only good things' as a clause.

but i still get wet thinking about them.