AMD Breaking up

if intel wanted to, it could....

it is not illegal to be the only supplier of something in america..... capitalism is a beauitful thing....

edit: and besides, there many are other microprocessor makers out there... so i really do not understand the idiotic statements frmo so many people saying that for some reason Intel would be sued if AMD died.... its just stupid....


IMHO.... the foreign stake in AMD makes them way less attractive to me as a consumer. Recently, I've bought: a 780G Motherboard, Athlon X2, Athlon BE, Sempron 3000, 2x HD 2400, and Radeon 4850 from AMD... really, i've "done my part" to help them out... but with a middle eastern company owning so much of them and their fabs I'd say that 4850 will end up being the last thing I ever directly or intentionally purchase from AMD....

Intel all the way now....
FALSE

Read up on the Sherman act( http://en.wikipedia.org/wiki/Sherman_Act ) Monopolies are only allowed in certain cases, and even then its highly regulated(Example of a legal monopoly: the garbage company, having 2 companies picking up trash every week would be wasteful). Having a monopoly on X86 cpu's would not be legal(given that they power almost all consumer pc's)

As for foreign investment... welcome to the global economy... where pontiac vibe's are really toyota's(well not exactly, but close), fords are made in mexico, your tech support calls go to india, and china is loaning our government obscene amounts of money. Yeah you can vote for with your money, but don't be surprised when other countries vote against you with theirs.
 
That's impossible at this point. x86 is far too entrenched to go away. MS + the world is not going to accept a ISA that can't run binaries going past 15+ years without big emulation overhead for a consumer cpu, and x64 is also too far into wide adoption for Intel to give up.

You can still get away with dropping x64, I think. Most people don't yet use it, or need it.
As for the overhead... Server software is mostly available in native Itanium versions aswell, so they won't be affected.
Desktops can take some overhead, CPUs are spinning their wheels most of the time anyway... even in games you start to see that, because we're GPU-limited.
It wouldn't be such a big deal, really. In the end we get faster processors (Itanium core logic is simpler and smaller, allowing for more cores in the same die-space, and better IPC), and quite possibly finally people realize how important it is to write portable software.
So if we don't like Itanium, we can jump over to some other architecture. That would be REAL competition (and it would open a window for AMD to stay in the CPU market, they'd just need a non-x86 CPU).
 
Actually, your point doesn't stand. On April 21st, of 2008 Intel cut their prices of the Q6700 CPU in half. This of course was just about 3 weeks after AMD started shipping Phenom B3s. This is an example of how competition benefited the consumer.

Not a good example though.
Mid 90s? lol, how about the 80s? There was no AMD, yet we saw the same CPU price-cutting as we did now. It's just the business model of the CPU world. You start out with immature production, and as yields on your product increase, you can drop prices. And as I said, you want to cash in on the early adopters.
I've paid the same for my 386 machine as I did for my 8088 machine. How is that possible if Intel never dropped prices?
 
Not a good example though.
Mid 90s? lol, how about the 80s? There was no AMD, yet we saw the same CPU price-cutting as we did now.
Prices for computers gone down very very much. Just check how the prices for video cards has gone down when amd released 48xx. Computers at the early 90's or end of 80's was much much higher than now.

Anyone who says that competition isn't good for prices and the consumer is just plain stupid.

I remember when Compaq released 386 20 MHz for about 16 000 $, add inflation to that to get the value today
 
Not a good example though.
Mid 90s? lol, how about the 80s? There was no AMD, yet we saw the same CPU price-cutting as we did now.

First of all, there absolutely was an AMD. I was ordering 286-20Ns from Dell for a NY bank circa 1988.

It's just the business model of the CPU world. You start out with immature production, and as yields on your product increase, you can drop prices. And as I said, you want to cash in on the early adopters.
Your general premise that CPU prices drop is correct. The timing and magnitude of the drop, however, are the substance of the point I'm making.

I've paid the same for my 386 machine as I did for my 8088 machine. How is that possible if Intel never dropped prices?

There are many, many holes in this reasoning:

1. AMD did exist during that timeframe, and did have a competetive product. AMD was getting faster clockspeeds on their 286 part and matched them on their 386 part. That's at least partially why Intel got pissy later on and tried to freeze them out on the licensing.

2. As for what you paid, precisely what kind of systems did you buy in each case? What were the components? It just so happens I sold computers from 1985-1987. I am quite familiar with the prices of the PCs and their components. I even remember what the Seagate ST-225 hard drive was selling for, and how the price changed over time.

3. What portion of the system price was the CPU vs. the system? Since you've been around a while, you probably know that the margins on PCs really started to nosedive as clones started to proliferate the market. How much of your system price consistency was due to the erosion of the margins at Computerland and IBM vs. processor price.
 
You can still get away with dropping x64, I think. Most people don't yet use it, or need it.
As for the overhead... Server software is mostly available in native Itanium versions aswell, so they won't be affected.

Um, a 1.5Ghz Itanium is $1375. A Q6600 is $189. You're not exactly helping your case that a lack of competition is good for the consumer.

Desktops can take some overhead, CPUs are spinning their wheels most of the time anyway... even in games you start to see that, because we're GPU-limited.
We're mostly GPU limited, but that's largely because there aren't many multi-threaded games out there, but that's going to be changing.

It wouldn't be such a big deal, really. In the end we get faster processors (Itanium core logic is simpler and smaller, allowing for more cores in the same die-space, and better IPC),
VLIW software is even harder to write than multi-threaded software until you get decent compiler support (and therein lies the problem).

and quite possibly finally people realize how important it is to write portable software.

Microsoft tried the portable OS thing, and everyone pretty much ignored it. During the early-mid 90s there was this huge movement towards a micro-kernel OS. Both IBM and Microsoft decided it was a good idea. Dave Cutler went off to do his version and it became Windows NT. IBM went off to do their version and it was OS/2 for the PPC (internally codenamed Falcon). NT would eventually be released to run on Mips, Alpha, PowerPC, and Intel. Only the Intel version made it to the next version. Portable software in general is very hard to do well. At an operating system level, Linux is the first one to get it "right", and I use the term "right" loosely. As a general sidenote, it is puzzling why people decided to continue to buy Intel during the early 90s when Alpha was around. Yes, it was that good.

At the application software level, in most cases you end up with a bastardized version of your app which doesn't look and feel at home on any of the OSes it's running on. The notable exceptions I can think of are Firefox and anything built on Java or .NET.
 
1. AMD did exist during that timeframe, and did have a competetive product. AMD was getting faster clockspeeds on their 286 part and matched them on their 386 part. That's at least partially why Intel got pissy later on and tried to freeze them out on the licensing.

Wrong.
Firstly, Intel was outsourcing production of some of its CPUs to AMD. What you're saying is the same as saying that Chartered is competing with AMD (AMD has outsourced production of Athlons to Chartered).
"Their 286 part" was Intel's part, manufactured by AMD, sold by Intel. It actually had (c) Intel on the package.
If you want to argue about actual independent clones, you should mention something like the NEC V20/V30.

Secondly, at the time that I bought my 386, there was no AMD alternative on the market. Intel introduced the 386 in 1985. AMD introduced their 386 in 1991, by which time I already used a 486 (which again cost no more than the 386 a few years earlier)

2. As for what you paid, precisely what kind of systems did you buy in each case? What were the components? It just so happens I sold computers from 1985-1987. I am quite familiar with the prices of the PCs and their components. I even remember what the Seagate ST-225 hard drive was selling for, and how the price changed over time.

Then you'll also know that the CPU was one of the most expensive components back in those days.
Both systems were fairly similar anyway (basic PC + colour graphics + monitor, keyboard, mouse etc), biggest difference being that the 386 had VGA instead of CGA, and the 386 had a harddisk on board.

3. What portion of the system price was the CPU vs. the system? Since you've been around a while, you probably know that the margins on PCs really started to nosedive as clones started to proliferate the market.

Both my PCs were clones (both Commodore, in fact), so they were both after the nosedive.
The only logical conclusion is that CPU prices came down along with the prices of all other components.
Else I'd never have bought a 386, because the price of those processors at introduction was of the same level as Extreme Editions today. The CPU alone would have cost more than what I paid for my complete system.
 
Um, a 1.5Ghz Itanium is $1375. A Q6600 is $189. You're not exactly helping your case that a lack of competition is good for the consumer.

Did you look at the prices of Dunnington? They go up to nearly $3000.
You seem to be confusing server processors with desktop processors, which is not helping your case.

We're mostly GPU limited, but that's largely because there aren't many multi-threaded games out there, but that's going to be changing.

How will multi-threading affect games that are GPU-limited?
It will only make the games even MORE GPU-limited, because the CPU will be used more efficiently.

VLIW software is even harder to write than multi-threaded software until you get decent compiler support (and therein lies the problem).

That doesn't make sense. What exactly is "VLIW software"? Assembly? 99% of all software doesn't require any assembly, just a HLL like C++ is fine. C++ compilers for Itanium have been around for ages. So for most people, nothing changes.
Also, multithreaded software is much harder than anything else. Assembly isn't much harder than C++, it still follows the same basic rules, it just requires some more manual labour. Multithreading on the other hand requires you to completely change the way you think about software.

Microsoft tried the portable OS thing, and everyone pretty much ignored it. During the early-mid 90s there was this huge movement towards a micro-kernel OS. Both IBM and Microsoft decided it was a good idea. Dave Cutler went off to do his version and it became Windows NT. IBM went off to do their version and it was OS/2 for the PPC (internally codenamed Falcon). NT would eventually be released to run on Mips, Alpha, PowerPC, and Intel. Only the Intel version made it to the next version. Portable software in general is very hard to do well. At an operating system level, Linux is the first one to get it "right", and I use the term "right" loosely. As a general sidenote, it is puzzling why people decided to continue to buy Intel during the early 90s when Alpha was around. Yes, it was that good.

Windows was and is portable, and as a result, any well-written straight C++ code for Windows is also portable (Visual Studio warns about potential pointer arithmetic problems. Fix those and in many cases your code can compile as-is on x86, x64 and IA64. I know all my code does, I've tried). Problem is that a lot of code isn't that well-written, and requires fixing.
Linux the first one to get it right? A laughable notion. A big part of the code that turns the linux kernel into an actual OS is taken from the GNU project, which was started years before the linux kernel. The GNU code is (mostly) portable, the linux kernel itself is not.
You will also see that many applications used on linux are not actually developed on/for linux, but rather for any POSIX-compliant system (often with a history starting on commercial *nix systems). Software that is specifically written for linux is often quite hard to get running on other architectures or POSIX-compliant OSes because people made the same mistakes of painting themselves into a corner as is common in the Windows-world.

Portability is not something that linux brought to the table. It was just common in the *nix world, because it allowed them to run applications developed on one system to also be compiled on another system. If anything, you have the ANSI C/C++ and POSIX standards to thank for it. Linux contributed to neither.

At the application software level, in most cases you end up with a bastardized version of your app which doesn't look and feel at home on any of the OSes it's running on. The notable exceptions I can think of are Firefox and anything built on Java or .NET.

Windows doesn't have that problem. Windows on Itanium is still Windows.
I've used NT4 on various architectures (yes, including Alpha, and yes, native Alpha applications were insanely fast at the time... but it had a pricetag, x86 machines were just much cheaper, and affordable enough to put one on every desk), and UI-wise you couldn't tell the difference. Since you could also run x86-executables on other architectures (the Win32 binary format isn't called Portable Executable for no reason you know), it was almost a seamless transition from the user's point-of-view.
I've also used the x64 version of both Windows XP and Vista, and ported some of my applications over.
 
eehhh do you think less competition will lower the price?

I could quote Opteron prices if you like (http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_609,00.html?redir=CPPR01).
You'll see that AMD too sells many processors that are more expensive than the $1375 Itanium mentioned earlier.

Conclusion: high-end server processors are more expensive than high-volume mid-end desktop parts such as the Q6600.
Doesn't say a whole lot about Itanium really, other than that there is no desktop-range for Itanium currently, because the x86 desktop-range is in the way.
As far as server processors go, Itanium pricing is nothing unusual.
 
I'm guessing most of intel based server are run on 64bit OS/Software, which means they would take a huge hit in the server market if they were barred from selling x86-64bit cpu, but then again amd would too from not being able to sell x86 based cpus =p
 
I'm guessing most of intel based server are run on 64bit OS/Software, which means they would take a huge hit in the server market if they were barred from selling x86-64bit cpu, but then again amd would too from not being able to sell x86 based cpus =p

Thing is that most server-oriented Windows/*nix software is also available for Itanium.
Let's assume that most servers these days are either database or web/mail/related internet servers.
You can solve most of that with linux or FreeBSD or such, Windows Server, Apache, IIS, MSSQL, mysql, postgresql, Exchange etc, which are all available for Itanium.

So for the majority of servers the transition would be quite smooth.
 
Conclusion: high-end server processors are more expensive than high-volume mid-end desktop parts such as the Q6600.

You must be joking? Do you really not understand that competition will lower price? Two competitors is better than monopoly. But two competitors are not that many if you know what I mean. 20 competitors would be a better situation. Companies isn't like charity organizations.

Do you know why there exists antitrust laws?
http://en.wikipedia.org/wiki/Antitrust

EU slaps Intel with formal antitrust charges
 
Did you look at the prices of Dunnington? They go up to nearly $3000.
You seem to be confusing server processors with desktop processors, which is not helping your case.

Okay, how much did you pay for your desktop Itanium solution? Yeah, that's what I thought. In any case, there isn't a helluva lot of difference at the silicon level between server and desktop parts.


How will multi-threading affect games that are GPU-limited?
It will only make the games even MORE GPU-limited, because the CPU will be used more efficiently.
You are assuming that all games are GPU limited. On my rig, I don't see that problem.


That doesn't make sense.
It does to an engineer.

What exactly is "VLIW software"?
VLIW stands for very long instruction word, and is one of the key features of Merced. It requires very intelligent compiler and software design to get the benefit of the hardware.

Assembly? 99% of all software doesn't require any assembly, just a HLL like C++ is fine. C++ compilers for Itanium have been around for ages. So for most people, nothing changes.
But how good are those compilers? And if the software were that much more performant, the Itanium that much more economical, and it overall just a happier place to be, then why are we not all using Itanium based computers?

Also, multithreaded software is much harder than anything else. Assembly isn't much harder than C++, it still follows the same basic rules, it just requires some more manual labour. Multithreading on the other hand requires you to completely change the way you think about software.
I've written quite a bit of multi-threaded software, and agree it's hard. It's not harder than VLIW optimization.

Windows was and is portable, and as a result, any well-written straight C++ code for Windows is also portable (Visual Studio warns about potential pointer arithmetic problems. Fix those and in many cases your code can compile as-is on x86, x64 and IA64. I know all my code does, I've tried). Problem is that a lot of code isn't that well-written, and requires fixing.
Well I agree that many don't write portable code. I also agree that Visual Studio will warn you about most problems (I've been using Microsoft C++ since it was Lattice). Actually we pretty much agree on that whole paragraph. But my point is that I never claimed that Windows wasn't portable (in fact, I do just the opposite by mentioning the other ports). My point is that nobody cares about anything about Windows on anything other than on x86 and x64 machines. Itanium is as relevant now as the 80186 was in 1986. Let it go. The best thing Intel could do to salvage some of their investment would be to incorporate aspects of VLIW into their future x86 processors.

Linux the first one to get it right? A laughable notion. A big part of the code that turns the linux kernel into an actual OS is taken from the GNU project, which was started years before the linux kernel. The GNU code is (mostly) portable, the linux kernel itself is not.
How many other OSs can you name run on as many different CPUs? Off the top of my head, I can't think of any. Show me HpUX running on a Cell, for instance...

You will also see that many applications used on linux are not actually developed on/for linux, but rather for any POSIX-compliant system (often with a history starting on commercial *nix systems).
[/quiote]
We agree on that point..

Software that is specifically written for linux is often quite hard to get running on other architectures or POSIX-compliant OSes because people made the same mistakes of painting themselves into a corner as is common in the Windows-world.
We don't agree there. I work with the GNU compiler everyday and it's pretty friggin' anal; moreso than VC.

Portability is not something that linux brought to the table. It was just common in the *nix world, because it allowed them to run applications developed on one system to also be compiled on another system. If anything, you have the ANSI C/C++ and POSIX standards to thank for it. Linux contributed to neither.
Okay, so aside from BSD, which of the other *nix implementations operate on such a wide variety of CPU hardware?

Windows doesn't have that problem. Windows on Itanium is still Windows.
Well there is that whole device driver thing.. In any case, there are far more people out there running Vista 64 on x64 than on Itanium.

x86 machines were just much cheaper, and affordable enough to put one on every desk),
Which is still the case today when you compare x86 and Itanic.

and UI-wise you couldn't tell the difference. Since you could also run x86-executables on other architectures (the Win32 binary format isn't called Portable Executable for no reason you know), it was almost a seamless transition from the user's point-of-view.
I've also used the x64 version of both Windows XP and Vista, and ported some of my applications over.

Been there done that, bought the t-shirt. We're way off-topic in any case.
 
Do you really not understand that competition will lower price?

I don't see how that has anything to do with the example of Itanium vs x86 given earlier.

Other than that you've fallen for the same trap as many others: you are so conditioned into thinking that competition lowers prices as a rule, that you take this as absolute truth and base all your reasoning on that.
You'll first have to prove that competition actually DOES lower prices (in *all* cases), before you can base any further reasoning on it. Else it's all just a big fallacy.

See, I am not as convinced as some people are that competition ALWAYS lowers prices and benefits the consumer, as if it's some kind of law of physics.
I can see *some* cases where competition could be good, but I can also see cases where competition doesn't make any difference, or actually has an adverse effect on the companies and/or consumers in the end.
Therefore I don't see it as a rule that competition is good by default.
 
Other than that you've fallen for the same trap as many others: you are so conditioned into thinking that competition lowers prices as a rule, that you take this as absolute truth and base all your reasoning on that.
You'll first have to prove that competition actually DOES lower prices (in *all* cases), before you can base any further reasoning on it. Else it's all just a big fallacy.

Ohh my...

Do you have any example where competition hasen't enhanced development and lowered price?
I think that 99,99% understand this
 
In any case, there isn't a helluva lot of difference at the silicon level between server and desktop parts.

No, but that doesn't explain why Xeon and Opteron processors are much more expensive than their desktop counterparts either. So your argument is simply flawed.

You are assuming that all games are GPU limited. On my rig, I don't see that problem.

No I'm not, but I think we can both agree that a lot of them are, in which case the best possible CPU performance is not required for the optimal gaming experience. Which means that CPU emulation might be acceptable in those cases.

It does to an engineer.

I am an engineer, and it doesn't make sense to me. I've not programmed in assembly in years (outside of tinkering with hobby projects that is). Therefore I don't "program x86", nor would I "program VLIW", if you mean what I think you mean.

VLIW stands for very long instruction word, and is one of the key features of Merced. It requires very intelligent compiler and software design to get the benefit of the hardware.

Yes, but that doesn't affect the average application developer. They just use the compiler, the rest is out of their hands.

But how good are those compilers? And if the software were that much more performant, the Itanium that much more economical, and it overall just a happier place to be, then why are we not all using Itanium based computers?

They're probably pretty damn good, because Itanium-based systems managed to break some records in TPC-benchmarks with Windows Server + MSSQL. Nearly all of that is written in C/C++ and therefore compiled code.

I've written quite a bit of multi-threaded software, and agree it's hard. It's not harder than VLIW optimization.

You don't have to do VLIW optimization, the compiler does that for you. And I still disagree, VLIW optimization is no different from puzzling with the optimal set of instructions in any other type of assembly language. x86 isn't exactly easy either. There are so many pitfalls and legacy issues in a modern x86.

My point is that nobody cares about anything about Windows on anything other than Windows on x86 and x64 machines. Itanium is as relevant now as the 80186 was in 1986.

But my point was that code written for x64 will compile pretty much as-is for IA64, so people don't HAVE to care.

How many other OSs can you name run on as many different CPUs? Off the top of my head, I can't think of any.

I already covered this earlier. The number of architectures that an OS runs on is not a reflection of the portability of its codebase.
Aside from that, NetBSD is the obvious answer. I believe it holds the record for number of architectures that a single OS runs on. This partly has to do with the very portable nature of its codebase, by the way. It was one of their design goals.

We don't agree there. I work with the GNU compiler everyday and it's pretty friggin' anal; moreso than VC.

It's not about the compiler, but more about things beyond the scope of the compiler.
For example, people using hardcoded include paths, which differ on other *nix systems. Or people using libraries/extensions of APIs that are only available in the linux implementation.
Or people using gmake, breaking compatibility with the standard BSD make found on most commercial *nix/BSD OSes.
I've encountered many such problems trying to build code on non-linux systems.
In fact, you seem to assume that all *nix OSes use the GNU compiler, which is also not true. Therefore, gcc extensions to the C++ language will not compile on other ANSI C/C++ compilers. Which is another issue I ran into.

Okay, so aside from BSD, which of the other *nix implementations operate on such a wide variety of CPU hardware?

Beside the point.

Well there is that whole device driver thing..

Itanium is a blessing in disguise in that sense. It doesn't have such a legacy or diversity as x86 has. Support for basic chipsets, videocards and things is plenty available. Consumer hardware may not be supported yet, but it's not THAT hard to port Windows-drivers over to IA64, since Windows was designed with cross-architecture portability in mind, and most of the driver is developed in C/C++ with a common kernel API interface across all CPU architectures.

Which is still the case today when you compare x86 and Itanic.

Obviously, since there are no desktop Itaniums available (current Itaniums are all aimed at high-end multi-CPU servers... if we looked solely at Xeon MP or Opteron MP processors, x86 would look very expensive aswell). Doesn't mean Intel can't make those.
That would mean two things:
1) Rules of supply and demand dictate that desktop parts can be much cheaper because they are sold in much higher volumes.
2) If Intel removes some of the server-only features such as MP logic, L3 cache, ECC cache etc and the extra validation required for server-grade processors, and builds them on their current 45 nm manufacturing line, they end up with a much smaller and cheaper to manufacture CPU. I'd say it could be done cheaper than x86, because the base core logic of IA64 is more compact by default (same reason why an ARM will always be more power-efficient than an Atom, all other things equal).
 
Do you have any example where competition hasen't enhanced development and lowered price?
I think that 99,99% understand this

Ironically enough x86 itself is a fine example. It killed off most competing architectures, which were often far more modern, powerful, advanced and/or efficient.
Competition does not guarantee that the better technology/development always wins.
 
Ironically enough x86 itself is a fine example. It killed off most competing architectures, which were often far more modern, powerful, advanced and/or efficient.
Competition does not guarantee that the better technology/development always wins.
So you are saying that if Intel had been alone, no AMD and no other company had competed with them. Then we would have better and cheaper processors?
 
So you are saying that if Intel had been alone, no AMD and no other company had competed with them. Then we would have better and cheaper processors?

No. I'm just saying that it wouldn't necessarily be worse, or even make a difference at all.
However, I do think that Itanium would have been better than x86 yes. But I've been thinking that about 68000, Alpha, PowerPC, PA-RISC, ARM and MIPS aswell, so that doesn't say much about Intel really, more about x86.

If it wasn't for AMD, we probably wouldn't have x86 today.
 
Scali, a few things are clear to me.

You don't like AMD. That's fine, I don't like Intel. When they were so arrogant that they initially refused to replace their flawed P54C processors, I made a conscious decision to avoid them. They were simply so successful that they got cocky, and their arrogance ticked me off. Here they were making billions in profit per quarter and they basically said, "Screw you consumer." Will I avoid them in the future? Well not necessarily. When it's time for me to upgrade, which is about every three years or so, then I'll see what's out there and decide then. But I'll be heavily biased against them, because I believe that AMD and Intel co-existing in the marketplace was the sole reason I could buy a quad core processor for about $200.

Now am I altruistic or some kind of religious zealot? No. Honestly I don't think that AMD would have necessarily acted any more honorably during the P54C fiasco if they were the ones with 98 percent market share, or whatever it was at the time. AMD certainly hasn't been consumer-pricing-friendly when they didn't have to be. I paid $1000 for a 4800x2 when it was new, and that aint cheap. When they didn't have any competition, they did what Intel or any other company with no competition does, they gouge you for as much as they can. It's been proven time and time again, and it's part of the economics curriculum in any institute of higher learning in the country.

You also seem particularly enamored with Itanic. Good, bad, or otherwise, it's a non-entity in the market. The best thing Intel can do with Itanic is what Microsoft is doing with Vista. Vista got a bad rap (unfairly so, in my opinion), so now they're trimming it and it'll get relaunched as Windows 7. Sure, they'll fix some things, snazz it up, etc, but I'd guess that the device drivers for Vista will work just fine on Windows 7, which is a pretty good indication that underneath it's the same beast. But once there's the perception that something is dead, then don't try to marry it.

  1. x86-64 is here to stay.
  2. Neither Intel nor AMD is saint nor demon. They are merely companies looking out for their own best interest, which isn't necessarily the interest of the consumer
  3. Two companies providing viable alternatives in the marketplace usually means good things for the consumer
  4. If the roles were reversed, I wouldn't want to see Intel go away either
  5. Nobody in their right minds is going to switch out their production servers to run IA-64. And not many are going to put those in a datacenter, either.
  6. Device drivers are important. The fact that device drivers are not as well supported on a platform isn't doing that platform any favors, despite how you'd like to spin it.
x86-64 isn't going away, not in the next ten years anyway.
 
You don't like AMD. That's fine, I don't like Intel.

I don't like Intel either, I don't like anything x86. But at least Itanium is a more modern and more efficient architecture than x86. It's the lesser of two evils. Pretty much every other nice architecture is all but gone these days (68k, PPC, Alpha, to name a few great architectures back in the day, that were annihilated by x86).

When they were so arrogant that they initially refused to replace their flawed P54C processors, I made a conscious decision to avoid them. They were simply so successful that they got cocky, and their arrogance ticked me off. Here they were making billions in profit per quarter and they basically said, "Screw you consumer." Will I avoid them in the future? Well not necessarily. When it's time for me to upgrade, which is about every three years or so, then I'll see what's out there and decide then. But I'll be heavily biased against them, because I believe that AMD and Intel co-existing in the marketplace was the sole reason I could buy a quad core processor for about $200.

I don't allow myself to get that emotional about any company.
I'd also like to note that AMD didn't exactly handle the TLB bug in the Phenom very gracefully. The user only got a silly patch that hampered performance. At least Intel actually DID replace all the flawed Pentium processors in the end. I think it's only logical that they were reluctant at first, considering the financial loss such an operation would incur.
Another lesser-known bug which annoyed the heck out of me personally is the Athlon's handling of pages larger than 4 KB. Windows used to use large pages for the AGP aperture for obvious performance benefits. This crashed Athlon processors, so you had to use a registry patch that forced the pages to 4 KB (and hamper performance). Throughout the lifetime of the K7 architecture, this bug was never fixed. Most people aren't even aware of it, because the patch is installed automatically on XP systems.

You also seem particularly enamored with Itanic.

Not really, but it's pretty much the only 'realistic' alternative we have to x86. Intel could make it in large enough volumes for it to be a successful desktop part, and Windows already runs on it. All other architectures are even more far-fetched for desktop/Windows usage at this point.

[*]x86-64 is here to stay.

That remains to be seen. If it turns out that the current AMD action has invalidated the x86-license then technically neither Intel nor AMD are allowed to manufacture x86-64 chips.

You [*]Nobody in their right minds is going to switch out their production servers to run IA-64. And not many are going to put those in a datacenter, either.

Why would you switch them out? Even if the x86-64 license got invalidated, I don't think that would affect any processors that are already in use.
I also don't see why you wouldn't put Itaniums in datacenters?

[*]Device drivers are important. The fact that device drivers are not as well supported on a platform isn't doing that platform any favors, despite how you'd like to spin it.

Yes, but that doesn't mean Itanium can't be successful. The best possible incentive for Itanium to become successful is killing off x86-64. Less than 2 years ago, x86-64 driver support wasn't exactly hot either, and even today there are many devices that don't work in Vista and/or x64 OSes, so I wouldn't put too much value in all of that.

x86-64 isn't going away, not in the next ten years anyway.

Again, that remains to be seen.
If it doesn't, we have AMD to blame for it. Keeping us in the dark ages for 10 more years. I thought Intel was bad, but Intel has at least tried to replace x86 with a more modern architecture twice.
 
No. I'm just saying that it wouldn't necessarily be worse, or even make a difference at all.
So you don't know, you are just guessing? Well if that's the case it is one advancement from the lack of knowledge that competition is ALLWAYS good for product advancement and the consumer.
 
You don't like AMD. That's fine, I don't like Intel. When they were so arrogant that they initially refused to replace their flawed P54C processors, I made a conscious decision to avoid them.
Well, at least your standards are consistent. It's not like AMD failed to replace buggy AM386, K6/K7 (3DNow/FPU failures due to poor validation) or BA/B2 Barcelona/Agena chips. Wait... AMD didn't. :p It slides in AMD's cases because AMD can't afford to do recalls.

But at least Intel did replace the affected Pentium chips.
 
So you don't know, you are just guessing? Well if that's the case it is one advancement from the lack of knowledge that competition is ALLWAYS good for product advancement and the consumer.

Your logic is backwards, as I already said before.
I've already given Betamax and x86 as examples where competition crushed the more advanced products, and made the consumer get stuck with inferior technology for years.
Then how can you still argue as if "competition is always good" is a fact, and base all your reasoning on that premise, which was already proven to be false in the first place?
 
Well, at least your standards are consistent. It's not like AMD failed to replace buggy AM386,
Wait, so according to Scali above, AMD's AM386 implementation was really just Intel's. So if it was buggy then does that mean Intel's design was buggy?

K6/K7 (3DNow/FPU failures due to poor validation)
This is the first I've heard of this. Have any supporting documentation? I never had a problem affecting any of my apps, but that's probably because nobody ever used 3DNow.

or BA/B2 Barcelona/Agena chips.
Wait... AMD didn't. :p It slides in AMD's cases because AMD can't afford to do recalls.
But at least Intel did replace the affected Pentium chips.

So what if it does slide? Which is worse, making a couple of billion in profit a quarter and being able to afford to do the right thing but initially choosing not to because you're too greedy, or not being able to afford to do it?

So what about you genius? Where do you stand? Just out of curiosity, when did you buy that Q6600 you have in your sig? What did you pay for it?
 
Wait, so according to Scali above, AMD's AM386 implementation was really just Intel's. So if it was buggy then does that mean Intel's design was buggy?

No, rather that their reverse-engineering of Intel's chips wasn't 100% accurate.
I had an Am486DX2-66 that generally worked okay, but for some reason I couldn't get OS/2 to boot in anything but safe mode.
I tried a real Intel 486DX2-66 in the same PC, and suddenly OS/2 worked. I guess that was another bug resulting from a not 100% perfect copy of Intel's architecture.
 
Wait, so according to Scali above, AMD's AM386 implementation was really just Intel's. So if it was buggy then does that mean Intel's design was buggy?

This is the first I've heard of this.
AMD used parts of the licensed Intel 286 design in their AM386 chip, but the unofficial clone was buggy*. Does that clear it up?

I would point you to AMD's errata, but AMD hides bugs that have been fixed. You need to request older bugs under NDA. AMD released an official statement that it was a problem with validation that slipped through the cracks. You can find theinq's apology for this in the archive and in older news archives. AMD quietly fixed the bug in later steppings and that was the last of it. No recall, no replacements. Zippo.

I'm not the one holding the companies to different standards, especially since you hold something against Intel that they did offer to replace at great expense. I'm just pointing out the irony, that's all. It's hard to be a man of principles when they're not applied consistently. :rolleyes: Intel still doesn't make a couple of billion a quarter even now with record revenues, much less profit by that much 14-15 years ago. IIRC, the few hundred million cost of the replacement program was worth more than a quarter's profit, but the cost was spread out of course.

* In historical context, AMD felt that it was entitled to future Intel technology as part of the cross-license agreement, Intel felt that AMD had not lived up to its terms. The long lawsuit that followed in the early 1990s allowed AMD to clone the 386 in exchange for payment to Intel since AMD in fact had not lived up to the cross-license obligations (supply technology back to Intel).
 
* In historical context, AMD felt that it was entitled to future Intel technology as part of the cross-license agreement, Intel felt that AMD had not lived up to its terms. The long lawsuit that followed in the early 1990s allowed AMD to clone the 386 in exchange for payment to Intel since AMD in fact had not lived up to the cross-license obligations (supply technology back to Intel).

The debate was on how far the second-source manufacturing license was stretched.
AMD claimed it applied to ALL x86 processors, but Intel claimed it only applied to the 8088 and 80286 processors that it had produced by AMD.
It mainly revolved around the microcode implementation. AMD was not allowed to use Intel's microcode in their own products, so they had to do clean-room reverse engineering.

In the end Intel and AMD came up with the first x86 cross-license agreement, allowing AMD the right to produce x86-compatible processors, but it didn't get access to Intel's designs, microcode or other implementation details. So AMD had to clone the CPUs without Intel's help, or design its own architecture (which it did with the K5).
 
Scali: I think that you should take a brake and socialize with other people, just a tip
 
AMD used parts of the licensed Intel 286 design in their AM386 chip, but the unofficial clone was buggy*. Does that clear it up?
No, not really. Honestly I'm not a process or chip engineer.

I would point you to AMD's errata, but AMD hides bugs that have been fixed. You need to request older bugs under NDA. AMD released an official statement that it was a problem with validation that slipped through the cracks. You can find theinq's apology for this in the archive and in older news archives. AMD quietly fixed the bug in later steppings and that was the last of it. No recall, no replacements. Zippo.
So I should just trust you, right? :rolleyes:

I'm not the one holding the companies to different standards, especially since you hold something against Intel that they did offer to replace at great expense. I'm just pointing out the irony, that's all. It's hard to be a man of principles when they're not applied consistently. :rolleyes: Intel still doesn't make a couple of billion a quarter even now with record revenues, much less profit by that much 14-15 years ago.

Oh, my bad, okay then.. They made 1.6Billion then for Q2 2008. They've certainly made more than $2 billion in the past (2.11 Billion to be exact).

IIRC, the few hundred million cost of the replacement program was worth more than a quarter's profit, but the cost was spread out of course.
Incorrect. They took a charge of $475 million in 1994, which is less than the 659 Million they made in the second quarter of 1994.

So how much did you pay for that Q6600? And when did you buy it?
 
Scali, I can understand your dislike for x86 in one sense, but as a developer, does it really affect us? I mean if we really want the extra registers then we can just use Intrinsic variables in x64 mode. We all know x86 is clunky at the op level, but most CPUs today rearrange all the instructions anyway to microops. Like you, it's been a while since I've had to open up Michael Abash's assembly optimization book because everything I do is in a HLL. Honestly I spend more time optimizing performance across these "seamless" distributed services.

It just seems odd that on one hand you talk about how compilers hide the complexity of VLIW but then have such a dislike for x86. I admit it's not elegant, and there are certainly better alternatives out there.. But does it really affect us? And if the end result is that bad, why did Apple dump PPC for x86? I'm not attacking PPC here.. just mentioning it for arguement's sake.
 
So I should just trust you, right? :rolleyes:
Of course not. http://babelfish.yahoo.com/translat...n--/meldung/19599&lp=de_en&btnTrUrl=Translate It's hard to dig up items that are 8-10 years old. That's one of the surviving links.
AMD confirms that some processors work at the representation of JPEG and MPEG files incorrectly. Are concerned all three families, the K6-2, in production, the Duron and the Athlon.
That was the 3DNow/FPU validation problem. It affected more than a "handful" of processors and even I had one of the buggy K6 CPUs. It was a fairly widespread problem. And it wasn't just a problem 5-6 places to the right of the decimal, it gave completely large magnitude wrong answers.

Incorrect. They took a charge of $475 million in 1994, which is less than the 659 Million they made in the second quarter of 1994.
I remembered the amounts incorrectly. That's still bad, but not as bad as I remembered.

So how much did you pay for that Q6600? And when did you buy it?
I paid $170 for my Q6600 last year from Fry's on black Friday, but what does that have to do with anything?

You still fail to explain your double-standards. I can take one guess to the reason why. :p
 
Scali, I can understand your dislike for x86 in one sense, but as a developer, does it really affect us? I mean if we really want the extra registers then we can just use Intrinsic variables in x64 mode. We all know x86 is clunky at the op level, but most CPUs today rearrange all the instructions anyway to microops. Like you, it's been a while since I've had to open up Michael Abash's assembly optimization book because everything I do is in a HLL. Honestly I spend more time optimizing performance across these "seamless" distributed services.

It just seems odd that on one hand you talk about how compilers hide the complexity of VLIW but then have such a dislike for x86. I admit it's not elegant, and there are certainly better alternatives out there.. But does it really affect us? And if the end result is that bad, why did Apple dump PPC for x86? I'm not attacking PPC here.. just mentioning it for arguement's sake.

Well, firstly, my dislike for x86 stems from the days when we still used assembly. Secondly, it doesn't affect most of us directly, but we do know that all this extra x86 translation hardware takes up valuable die space, and makes the chip less efficient.
HyperThreading is a good example of that. The backend of a modern x86 core can perform far more instructions than what you can feed through a single x86 instruction stream (thread).
So if we had a more modern chip design, built on the cutting-edge technology of current x86 processors, we'd end up with more powerful processors at the same price.

Apple dumped PPC for a simple reason: Originally the desktop variations of the PPC architecture were developed and manufactured by Motorola. Motorola eventually gave up on the desktop market, because Apple was the only customer and was not big enough to warrant further development of the architecture. So Motorola concentrated mainly on the embedded/DSP market.
Apple then went to IBM, which still had suitable PPC processors... but these too didn't develop quickly enough for Apple to remain competitive.
So eventually Apple made the sensible choice to move to x86, which had by then surpassed the PPC in performance and power efficiency (a very large market of Apple is the notebook market, so they need performance-per-watt mostly. PPC was always ahead of x86 there... until Intel introduced the Pentium M/Core architecture).

So technically PPC had an advantage over x86... but due to slower development, x86 eventually surpassed it. If PPC had developed at the same rate as x86, it'd have an even bigger advantage over x86 today.
Even Itanium suffers from that, its architecture isn't updated that often, and it's still manufactured on an older process. Intel plans to pull Itanium ahead though.
 
So technically PPC had an advantage over x86... but due to slower development, x86 eventually surpassed it. If PPC had developed at the same rate as x86, it'd have an even bigger advantage over x86 today.
Even Itanium suffers from that, its architecture isn't updated that often, and it's still manufactured on an older process. Intel plans to pull Itanium ahead though.

Less competition -> inferior products -> worse development -> fewer customers
 
Less competition -> inferior products -> worse development -> fewer customers

You have it backwards. x86/Windows was actually PPC/Apple's biggest competitor.
It's the inferior product (x86/Windows) that won, ironically enough because the inferior product had the most customers.

Basically what happened to PPC was this:
Not enough customers -> less development -> inferior products -> more competition from x86 -> PPC loses its last big customer to x86.

The same story applies to many other architectures, such as 68000 and Alpha. They started out with an advantage, but eventually x86 caught up with them because Intel just had more money to throw around on manufacturing and beating more efficient architectures with brute force and/or low prices.
One example today is that x86 is a very successful server processor. Not because it's so good, but because it's cheap enough to throw a ton of boxes in a server room.
 
I paid $170 for my Q6600 last year from Fry's on black Friday, but what does that have to do with anything?
Do you honestly think you would have paid $170(even though the ad says $198 *cough* Correction, that's with a motherboard; my bad) for that CPU if AMD didn't exist?

You still fail to explain your double-standards. I can take one guess to the reason why. :p
I don't need to explain anything. Why don't you explain why you're in an AMD forum attacking them?

In any case, I've restate it:

  1. Competition is good for the consumer
  2. You probably wouldn't have paid that low a price for that CPU, nor would I have paid $237 for mine, if either of these companies didn't exist.
  3. Neither AMD or Intel is a Saint
  4. I'll always support a second source provider if it helps insure that there will always be a second source provider.
 
You have it backwards. x86/Windows was actually PPC/Apple's biggest competitor.
It's the inferior product (x86/Windows) that won, ironically enough because the inferior product had the most customers.
More customers and more competition will win. If apple didn't protect it's market they would have done better. Markets with no competition will die
 
Objection, your honour. Conjecture.

Seriously, what do you hope to gain with questions like these? Fact of the matter is that nobody knows exactly what would have happened if AMD didn't exist.
What all knows (exept you?) is that the price would have been higher and the processor would have been worse
 
Back
Top