OT: Where do I find...

The internal micro-architectures are somewhat different, such as the number and type of execution units. But the ISA - the "programming level" - is almost identical. It is only special-purpose instructions such as those related to virtualisation, and some SIMD instructions that are different. For general code, except when using the vector processing units (which always change from generation to generation), the instruction set is identical. Anything else would have been crazy.

Intel tried to make their own new independent 64-bit architecture with the Itanium. The idea was nice, since x86 is such a horrendous architecture that it would be good if it could be replaced. But the implementation was terrible, and the Itanium flopped. But Intel couldn't make a 64-bit x86 ISA without losing all credibility for the Intanium while they still had hope for it. AMD, on the other hand, /could/ make x86-64, and saw it as a chance for leadership in some areas. So they made the "amd64" or "x86-64" ISA, and did a pretty good job of it. Intel had no choice but to follow, and copied the ISA entirely except for differences in the SIMD and a few system instructions.

Yes, and the entire rest of the industry is behind Intel, aren't they? So AMD will be so perpetually, right?

What is your point?

prosper.

In what way does it make sense? AMD could easily be on the auction block in a few more years. I see ARM in too many products made by too many different people for AMD to have any impact. What exactly does AMD do that everybody else can't??? Is AMD even selling any ARM parts? I guess you are trying to tell them how to run their business?

--

Rick

What you say is true, but it ignores the fact that almost no one was willing to write code for the AMD 64 bit chips until Intel came out withe their new 64 bit instruction set. Then it got momentum. Why would anyone write code for < 20% of the market?

--

Rick

is

Yup, the computing world is changing dramatically as the market shifts to entertainment and away from the original purpose of computers, science and engineering.

Intel is not stupid. They can feel the winds changing just like everyone else. They also know a lot about making teeny tiny transistors and how to make them do big things. They will come out of the transition just fine by making both portable devices and still selling big iron to those who still need it.

--

Rick

Mostly that Intel is shrinking. x86 is their only trick, it's ancient, and the future is hard to predict.

prosper.

Gosh but you're crabby today.

--

John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
http://www.highlandtechnology.com 

Precision electronic instrumentation 
Picosecond-resolution Digital Delay and Pulse generators 
Custom laser drivers and controllers 
Photonics and fiberoptic TTL data links 
VME thermocouple, LVDT, synchro   acquisition and simulation

You really don't know anything about it do you?

--

Rick

The future? Not much. But I'm pretty sure things will change a lot.

--

John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
http://www.highlandtechnology.com 

Precision electronic instrumentation 
Picosecond-resolution Digital Delay and Pulse generators 
Custom laser drivers and controllers 
Photonics and fiberoptic TTL data links 
VME thermocouple, LVDT, synchro   acquisition and simulation

The Linux kernel and most of the basic Linux system were ported and working on AMD's simulator before they had silicon available. By the time you could buy the chips, several major distros had full ports available (it helped that Linux was already available for four other

64-bit architectures, IIRC).

So for people using Linux, amd64 code was common from very early on. Considering the usage that Linux had at the time, that meant AMD had a big advantage in the "technical enthusiast" market (which was small), the high-power, clustering and supercomputing market (which was growing

- and vital to a processor manufacturer's reputation), and in the growing market of Linux servers (which encroached on Intel's biggest margin market).

/Windows/ was not 64-bit, so /windows/ users (and developers) had no use of x86-64. So for the huge majority of desktop users, it had no relevance - people bought AMD cpus because they gave better value for money for running 32-bit code.

So 64-bit got visible momentum in the mass market only after 64-bit Windows became common (on the desktop side there was 64-bit XP, but it wasn't until Vista that 64-bit versions were in common use). MS would not bother making 64-bit Windows until Intel had the chips. But Intel had to make x86-64 chips long /before/ this point - they had to stop AMD from encroaching the server market, and they had to stop AMD's reputation as the technical leader.

And please don't write things like "until Intel came out with their new

64 bit instruction set". Intel did not come up with a "new 64-bit instruction set" (except the Intanium, of course). Intel copied AMD, and paid dearly for the privilege. I can only guess what they paid in terms of licences, copyrights, trademarks, patents, etc. - maybe it was already covered in their technology cross-licensing arrangements. But they certainly paid in reputation.

Wasn't that Pournelle's Law?

Somebody I used to work with back in the 80's claimed that the personal computer setup you really wanted always cost $5K. In 1980 that was Z80, 64K of RAM, nice serial terminal, dual DSDD 8" floppies,

300 baud modem, and a 9-pin dot-matrix printer. 20 years later, the specs had changed, but the price was still pretty close to $5K for everything.

In the past 10 years or so, it actually seems to have dropped. I'd be pretty happy with $2500...

--
Grant Edwards               grant.b.edwards        Yow! UH-OH!!  I put on 
                                  at               "GREAT HEAD-ON TRAIN 
                              gmail.com            COLLISIONS of the 50's" 
                                                   by mistake!!!

And how many 64 bit apps were there?

I believe you are agreeing with me, right? The 64 bit market did not develop until Intel and Microsoft got behind it. End of story. Even if they had the better idea first, AMD was largely irrelevant in this regard.

My point in using the word "new" was to differentiate it from the "old" Itanium instruction set.

--

Rick

Itanium.

Nope. HP developed the Itanium as a successor for PA-RISC. HP realised they couldn't fab it, so got into bed with Intel. HP was interested in the future of HP-UX systems, not silicon, so they didn't mind giving the development team to Intel in exchange for certain considerations.

End of story.

Nope. One part of the market didn't develop, others did.

Itanium instruction set.

Que?

Pretty much all apps were 64-bit. That's what it means to compile a Linux distribution for a 64-bit processor.

There are a few things, like Wine or binary-only programs, that are

32-bit. Of course, 64-bit Linux distros run 32-bit code (for the same architecture) without problem.

Or do you mean how many applications were there that could really take advantage of 64 bits? Well, 64-bit had been in use in the workstation, server and high-performance (supercomputing) markets for a good while - at least in the serious market rather than the Windows market. Most such systems were Unix rather than Linux at that time, but Linux was already popular amongst more technical people. So you found 64-bit code on database servers, big CAD systems (Intel probably designed their processors on systems with 64-bit SPARC or PPC chips), video processing, animation, simulations, etc.

I think you have totally and utterly missed the point.

It is correct that the mass-market for 64-bit cpus came when Intel and Microsoft joined in. But the real money-making market for processor manufacturers is not the low-margin home/work PC world - it is the high-margin world of big servers, and performance workstations. Super computing hardware has mid-level margins, but very high visibility - you want everyone to know that the fastest computers use /your/ chips. These markets were dominated by 64-bit processors. Intel had tried to compete with the Itanium - and then AMD outclassed it with 64-bit x86 chips that were a fraction of the price and power, much easier to work with, but nearly as fast for most jobs (and faster for some tasks).

And in the home/work PC market, AMD had 64-bit processors while Intel was stuck with old-fashioned 32-bit chips. It did not really matter that the majority of people couldn't take advantage of the 64-bit power

- AMD's chips were also fast for 32-bit. But for the marketing numbers game, AMD had a huge leap over Intel.

Intel played catch-up by copying/licensing AMD's amd64 ISA because they had no choice. AMD was the driving force behind the 64-bit processor in the mass market.

That's just silly. x86-64 may have been new to Intel, but it was already established by AMD - it was not new. Nor was the Itanium particularly old - it was in heavy development at that time (though it was clear to the world that it was a doomed architecture).

I know that Intel worked closely with HP on the Itanium, and that it was to be a replacement for PA-RISC (and the Alpha), but I thought it was mainly an Intel design. Did the PA-RISC also use EPIC?

Ok, so AMD was hot for a couple of years in the server market. I think you miss the point. That is not what drives the profits of a CPU maker. Profit is in the numbers, the mass numbers. Otherwise, why would AMD even bother with the home/business computing market? The server market supplements their profits, but without the mass market they can't run the fabs.

No, I don't miss the point. AMD is the tail and Intel is the dog, always has and always will be. The "old fashion" 32 bit chips which Intel dominated the market, with did just that, dominate the market.

*That* is the point. Linux may have had 64 bit support, but that could only run on 20% of the machines, so that means it was likely installed on what, maybe 1%,... less?

Being the "driving force" means AMD designed the obvious extension to the 32 bit instructions which most likely anyone would have done, but otherwise means nothing. In fact, I'm pretty sure you can't patent typical instructions and you can't copyright hardware, so the only handle AMD even had on Intel were the mnemonics. I seem to recall that is why the Z80 instruction set is slightly different from the 8080 instruction set, but the opcodes are the same... with extensions.

The only instruction patent I know of is some feature of an ARM instruction that *requires* a particular piece of hardware to implement it. Otherwise anyone can duplicate the entire ARM instruction set other than that one interrupt feature. Or so I was told when someone had designed an ARM7 for FPGA. He did such a good job of it that ARM contacted him. He never said exactly what the discussion contained, but he pulled his design and went to work for ARM.

What's silly? I was trying to differentiate my statement from referring to the Itanium 64 bit instruction set. I was making no statement about the Intel x86-64 instructions vs. the AMD instructions. The Itanium instructions were the "old" 64 bit instructions, the x86 compatible instructions were the "new" 64 bit instructions.

--

Rick

Many years ago there was a Scientific American article which analyzed Intel profits. They showed a *very* clear cycle and analyzed the reason behind it. I don't recall the details, but it had a lot to do with the relationships between making capital investments and reaping the rewards. I seem to recall that the profits increased with each cycle.

Yes, things will change and Intel has a lot more capability to change with the market. AMD has been in running the red for more than 60% or more of its life, at least in the last 15 years. I was never in a position to make any money off of it, but I advised my friends to buy their stock on three separate occasions when they were in a major slump losing huge amounts each quarter, but with a huge prospect in the wings. The release of the Athlon was one of those times. Each time they swung into the black in 6 to 12 months and stayed there for a couple of years while they had a technical lead on Intel pushing market share up 1 to 2%.

But they could never hold the fort against the dreadnought and would slip back into the red for a few years until they could find another way to gain a foothold.

Those days are gone. First AMD started loosing ground in the fab race. They just couldn't afford to make the ever larger investments in capital that Intel could and started falling behind in process technology. They would reach the next process node 6 months after Intel and have to deal with the lower ASPs in the meantime. Then Intel's lead increased to a year. Seeing no way out of this progression, AMD decided to dump fabs and got into all sorts of hot water.

I see they are currently selling for $4 a share with $1 per share loss. That says in four years they are worth nothing...

I wouldn't advise anyone to bet on AMD still being AMD in two years.

--

Rick

Wrong way round.

Directly.

Not really. Alpha was developed by DEC around the same time as HP developed EPIC. Compaq bought HP. HP bought Compaq. Not much interest about preserving Alpha within HP.

Other way round. If you look at the Itanium ISA, each Itanium instruction is more-or-less three PA-RISC instructions in parallel. That's a gross over-simplification of course, but there's more than a grain of truth in it.

I ought to expand.

HP originated the concepts and architecture for the Itanium, using ideas based on earlier VLIW machines.

The first processors were implemented by HP in conjunction with Intel, since Intel knew about its semiconductor processes.

Later, HP's Itanium team transferred to Intel, and that's the source of your misunderstanding.