64-bit desktop computing unnecessary, says Intel CTO

[pm]

There seems to be some debate over clockspeeds, but I don't see a lot of people talking about why Mr. Gelsinger is mistaken.

As someone who has spent at least a small amount of time porting 32-bit apps (like SPICE) to 64-bits, I don't see the switch to 64-bits as something that is likely to be a slam dunk for speeding up software. Any program with a fair amount of pointer operations is likely to slow down since you replace 32-bit pointers with larger versions that take up more cache space and more bus bandwidth. Same information, twice as much space. That said, there are plenty of applications do speed up when you reoptimize for 64-bits, but not by a lot. And there aren't that many that I have seen. Anyone who thinks that recompiling a 32-bit program to 64-bits will result in it running twice as fast, needs to think about this some more.

So, if performance isn't the primary motivator for moving to 64-bits, then the other obvious reason is the 4GB memory limitation. And I tend to agree with our CTO's theory on memory. I just upgraded my home computer to 1GB. I don't see any reason why I need more memory than this right now on my home desktop.

I don't see anyone actually refuting what he said. Do people really think that they need 64-bits on their desktops right now? And if so, why?

Yes, I work for Intel so no doubt plenty will say, "well of course pm's going to say that 64-bits don't matter" and that's fair enough I guess. But if anyone accuses me of bias, I'd like to see exactly why they think that I'm mistaken.

 

[NFS4]

I don't see anyone actually refuting what he said. Do people really think that they need 64-bits on their desktops right now? And if so, why?

Question...

What was the FIRST 32-bit processor?
How many 32-bit applications were available at its release?
How long did it take before 32-bit apps became commonplace after its release?


THEN, we can take a look at the CTO's theory

 

[BoberFett]

So pm, no disrespect as you're obviously one of the most knowledgeable people on this board as far as CPU technology goes. But let's be honest here. Is Intel going to wait until the world does need a 64bit desktop? If so, they're going to take in in the keister, lagging that far behind. If we want to see 64 bit in the mainstream in a few years, the desktops need to start hitting soon in order for there to be an appreciable user base for software developers to bother with 64 bit software.

Or did the 386 only get released when the world absolutely needed a 32 bit CPU? I don't remember having a huge selection of 32 bit software in the mid 80s.

 

[MDE]

Ooh ohh, me first.

Intel 386
Almost none (a guess)
A friggin long time (Win 95 was largely 16 bit still)

 

[NFS4]

Originally posted by: MonkeyDriveExpress
Ooh ohh, me first.

Intel 386
Almost none (a guess)
A friggin long time (Win 95 was largely 16 bit still)

I wonder what Intel's CTO would have said in regards to the i386 at launch 🙂

"It's not necessary, but here you go anyway!" 😛


You've got to start somewhere. AMD has thrown their hat in the ring. Intel doesn't have a counter response...simple as that.

 

[pm]

What was the FIRST 32-bit processor?

There's some dispute to this as it depends a little on what "bitness" is. Some might say the 16/32 bit DEC PDP-11 in 1970. Some people might say that it was the National Semiconductor 32016 microprocessor in 1978(?). But if you think that a 32-bit processor requires a 32-bit bus, then the answer becomes more complex. I'd probably argue that the first 32-bit microprocessor of any consequence was the Motorola 68000 released in 1979. It had 32-bit applications made immediately available along with the CPU and most people immediately compiled for mixed 16-bit/32-bit mode. Gaining more widespread support for the computing platform took years.

But I presume you want was what was the first 32-bit x86/IA32 compatible microprocessor. That would be the Intel i386 DX which was released in 1985.

How many 32-bit applications were available at its release?

I have no idea, but I would imagine that it was very few.

How long did it take before 32-bit apps became commonplace after its release?

I personally don't think that 32-bit computing on the IA32 platform started to become widespread until the release of Windows 95 in 1995. So, I'd answer that it was about a decade later.

 

[Pariah]

For those of us that remember the launch of the Pentium Pro, it makes perfect sense that Intel is not jumping on the mainstream 64bit bandwagon yet. Let someone else create the market, and only step in when the market dictates it. There's no reason for Intel to worry about the 64bit ability of the Athlon64 until there are applications that actually show the benefits of 64bit computing.

 

[pm]

So pm, no disrespect as you're obviously one of the most knowledgeable people on this board as far as CPU technology goes. But let's be honest here. Is Intel going to wait until the world does need a 64bit desktop? If so, they're going to take in in the keister, lagging that far behind. If we want to see 64 bit in the mainstream in a few years, the desktops need to start hitting soon in order for there to be an appreciable user base for software developers to bother with 64 bit software.

So if I might paraphrase, and correct me if I'm wrong, but you seem to be saying "Ok, maybe you are right, the world doesn't need 64-bits right now, but they will soon and so we should start releasing 64-bit microprocessors now". Which, if I have paraphrased correctly seems to indicate that you don't think that my CTO is completely wrong in his statement, but that you think he's being shortsighted. You may well have a point, although I have a good deal of respect for both Intel's management, and for Patrick Gelsinger himself.

I might counter that although the Itanium is not currently a desktop microprocessor, it has the capability to be one. And that therefore there already is a 64-bit microprocessor for Intel that is creating new 64-bit applications in preparation for the 64-bit future that we all agree is coming.

 

[Excelsior]

Originally posted by: dexvx

Originally posted by: Excelsior

Originally posted by: dexvx

Originally posted by: BoberFett

Originally posted by: dexvx

Originally posted by: BoberFett
Intel loves to play the numbers game. If they can sacrifice a few IPCs for a higher clockspeed, they're more than happy to because they know that Joe Schmoe can tell the Intel is faster. I mean, the Pentium's got 3 giggihertz and the Athalon only has 2. That's half again as fast!

Then explain to me why Centrino Pentium-M laptops are so damn popular and why the Itanium II 1.5/6MB is the highest single processor in the world.

Your logic is flawed.

Uh oh, little boy wants to play with the big kids. Sigh.

... blah blah blah....

Come back with some ammo, fanboy.

Lets get to the point of the arguement instead of straying, little kid.

You said Intel would sacrifice high IPC for high clock speed. I pointed out 2 products in Intel's line that has low clockspeed but high IPC. I have disproven your lovely theory about Intel loving high clock speeds.

So come back with some ammo, fanboi and prove me wrong.

No you haven't disproven the theory that Intel loves high clock speeds. Its a fact. We are talking about desktop chips here, not the Itanium. Now, in desktop chips, does Intel favor high mhz and low IPC or high IPC and lower mhz? You lose.

Yes I have disproven it. Pentium-M and Itanium are low clock high IPC. Do you understand that?

Note the original poster said "Intel loves to play the numbers game. If they can sacrifice a few IPCs for a higher clockspeed, they're more than happy to..." He did not say Intel loves to play blah blah blah with desktops. So next time how about taking some reading classes before you stumble again?

And don't you fvcking understand that neither the Pentium-M or Itanium are in the desktop machines sold to consumers? Dipsh!t. Don't tell me to take reading classes when you obviously don't understand what CPUs are used in the consumer desktops right now. Not servers, not laptops, but the OEM machines sold at best buy or from dell. He was OBVIOUSLY talking about Desktops because he said P4 @ 3Ghz and Athlon @ 2Ghz, and those two CPUs are in desktops, simple as that.

 

[zephyrprime]

 

[pm]

One comment that I'll make on the IPC versus clock speed debate is that what you guys are arguing about has been the subject of equally fervent debates at microprocessor conferences for decades. Yes, decades.

In the 'history' of microprocessors, it's called the Brainiac vs. Speed Demon debate and it's really a matter of personal preference which arguement you accept. It has been the subject of many heated debates and each side always brings their favorite architectures that have their implementation of choice. The brainaic approach is a high IPC approach with a relatively short pipeline, and the speed demon is the fairly obvious opposite which is a long pipelined microprocessor. There's an interesting article on the subject at Microprocessor Report's website here. You can read more about the debate by typing both "speed demon" and brainiac into Google.

 

[TerryMathews]

Originally posted by: pm
So, if performance isn't the primary motivator for moving to 64-bits, then the other obvious reason is the 4GB memory limitation. And I tend to agree with our CTO's theory on memory. I just upgraded my home computer to 1GB. I don't see any reason why I need more memory than this right now on my home desktop.

I strongly disagree with you. Two years ago, my desktop had 128MB RAM. Now, it's got 1GB. Extapolate out and you get my memory doubling every year, on avg. Meaning, two years from now, give or take, I'm gonna want 4GB on my desktop.

If Intel and AMD don't start getting 64bit processors out soon, they won't have cheap processors for everyone within two years. I certainly don't want to pay $700/processor for my desktop.

 

[BoberFett]

Originally posted by: pm
So if I might paraphrase, and correct me if I'm wrong, but you seem to be saying "Ok, maybe you are right, the world doesn't need 64-bits right now, but they will soon and so we should start releasing 64-bit microprocessors now". Which, if I have paraphrased correctly seems to indicate that you don't think that my CTO is completely wrong in his statement, but that you think he's being shortsighted. You may well have a point, although I have a good deal of respect for both Intel's management, and for Patrick Gelsinger himself.

Yep, you pretty much understood my meaning. I don't know if I'd say the world will need 64 bits soon, but eventually. In the world of computers, I think of "soon" as within a year. I don't think we'll see significant desktop usage of a 64 bit code for at least a couple years. Part of the reason for will be the slow migration to 64 bit, and part of it will be the lack of necessity. You know as well as I do though, that this industry has to plan ahead, and if 64 bit starts hitting desktops now, then in a few years we'll see a respectable number of home users that have a 64 bit CPU. If AMD had waited a year or two until Intel said "Go!" to 64 bit, that would push the software back even further and probably beyond the point where it'll be required.

I might counter that although the Itanium is not currently a desktop microprocessor, it has the capability to be one. And that therefore there already is a 64-bit microprocessor for Intel that is creating new 64-bit applications in preparation for the 64-bit future that we all agree is coming.

I do believe it was you I was reading about that has an Itanium on the desktop, and that it is a capable 32 bit CPU as well. This means that modifying Itanium for the home user is a very real possibility, but the words of your CTO seem to say they're going to push 64 bit off for quite some time. If they had plans to release a modified, lower cost Itanium as a contender to Athlon 64 anytime soon, you'd think they'd say something about it to muffle AMDs thunder. As it is I just see Intel missing a market segment in which they'll have to do a lot of work to catch up to AMD when they do decide to finally enter the 64 bit home market. I do wonder what they know that I don't, because obviously Intel isn't going to hire morons for upper management. I guess we'll see what the next couple years bring.

 

[BoberFett]

pm

As for the Brainiac vs. Speed Demon issue, I don't think your average techie cares how the work gets done; just that it gets done. But your employer does realize that your average home user goes to Best Buy or calls Dell and hears about how the competition only this many megahertz while we have double that number. This is nothing new, it's been going on since the home PC market started.

So regardless of what the chip designers think, I'm willing to bet that Intel's marketing department has at least some say in which way the technology moves. Obviously they can't do anything detrimental to the company's future, but you can rest assured that they love clock speed over IPC. It make their job easy. 🙂

 

[pm]

I strongly disagree with you. Two years ago, my desktop had 128MB RAM. Now, it's got 1GB. Extapolate out and you get my memory doubling every year, on avg. Meaning, two years from now, give or take, I'm gonna want 4GB on my desktop.

You may be disagreeing with me, but I'm not disagreeing with you. I agree in a couple of years we will need 64-bits on the desktop. I just don't think it happens to be necessary right now. I just upgraded to 1GB right now. I think I'm good for a while.

If Intel and AMD don't start getting 64bit processors out soon, they won't have cheap processors for everyone within two years. I certainly don't want to pay $700/processor for my desktop.

Both Intel and AMD both have 64-bit microprocessors out right now.

 

[pm]

So regardless of what the chip designers think, I'm willing to bet that Intel's marketing department has at least some say in which way the technology moves. Obviously they can't do anything detrimental to the company's future, but you can rest assured that they love clock speed over IPC. It make their job easy. 🙂

I agree that marketing a "Speed Demon" is easier than a Brainiac, but I agree with the others who mention the Itanium and the Pentium M as examples that Intel doesn't always apply the Speed Demon style of design to its entire product line. I personally think that they both have their place. For raw integer performance, it's hard to beat a Speed Demon design with a Brainaic version. But for low-power designs, a Speed Demon is probably not the optimal choice. They each have their areas where one excels over the other..

 

[Adul]

Originally posted by: MonkeyDriveExpress
Ooh ohh, me first.

Intel 386
Almost none (a guess)
A friggin long time (Win 95 was largely 16 bit still)

Windows NT 3.1

 

[BoberFett]

Originally posted by: Adul

Originally posted by: MonkeyDriveExpress
Ooh ohh, me first.

Intel 386
Almost none (a guess)
A friggin long time (Win 95 was largely 16 bit still)

Windows NT 3.1

Again, yes that was available. But I don't know anybody that ran NT at home. I didn't even run NT at work until we abandoned Novell and moved to an NT 3.51 server

 

[Budman]

64-bit desktop computing unnecessary

So was AGP 8X when it came out,so was ATA133.

But it's got to come out sometime,we probably dont need it but it will still good to have it.

 

[imgod2u]

One thing to note is that one 64-bit ISA isn't neccessarily another. That is, two 64-bit ISA's could be completely different an incompatible. Those arguing that Intel should release their own 64-bit ISA should think of what this would implicate. Either the will:

1. Release an x86-64 compatible chip (which I think is highly unlikely given their adamant stance against this and their willingness to move to IA-64)
2. Release an IA-64 chip with an adequate, if not stellar IA-32 compatibility.

I'd say 2 is more likely and if that were to happen, it would be absolute *chaos* in terms of desktop software. We'd have an i386-compatible version, an x86-64 compatible version *and* an IA-64 compatible version. You'd have to port *three* different platforms worth of software.

Saying we'll eventually need to move on to 64-bit is definitely true, however, x86-64 may or may not be laying down the "clear path" as hopefully as others think. If Intel doesn't adopt x86-64 in the future, the software base it builds up will simply cause problems later on that could be as bad as the Mac vs PC compatibility issues.

As for the point of buying 64-bit processors now in order to be able to upgrade to 4 GB of RAM later on.....by the time you actually would need 4 GB of RAM, wouldn't the processor you buy now be horrifically slow? I'd think you'd have moved on and upgrade to a new processor/motherboard/memory solution by then.

 

[NFS4]

I personally don't think that 32-bit computing on the IA32 platform started to become widespread until the release of Windows 95 in 1995. So, I'd answer that it was about a decade later.

So how is the A64 any different? AMD is just pioneering the technology now just as Intel did back in the day. It just gives AMD a better grip at getting x86-64 support by having an earlier showing.

 

[dmw16]

I'd say that too if my big competitor was putting 64bit into the hands of desktop users before I was.
-doug

 

[TerryMathews]

Originally posted by: pm

I strongly disagree with you. Two years ago, my desktop had 128MB RAM. Now, it's got 1GB. Extapolate out and you get my memory doubling every year, on avg. Meaning, two years from now, give or take, I'm gonna want 4GB on my desktop.

You may be disagreeing with me, but I'm not disagreeing with you. I agree in a couple of years we will need 64-bits on the desktop. I just don't think it happens to be necessary right now. I just upgraded to 1GB right now. I think I'm good for a while.

If Intel and AMD don't start getting 64bit processors out soon, they won't have cheap processors for everyone within two years. I certainly don't want to pay $700/processor for my desktop.

Both Intel and AMD both have 64-bit microprocessors out right now.

Good points. 🙂 I should have stated my points more clearly. When I said that Intel and AMD need to get 64 bit processors out soon, I was meaning 'relatively' affordable destined-for-desktop CPUs since this whole thread was in reference to home computing.

Neither Itanium nor Opteron is a desktop processor as is obvious by the features available and not available on their respective motherboards. Some Opteron boards are close, though. Entirely too pricey though.

 

[BoberFett]

Re-reading the article, I'm questioning Gelsinger more than I originally did. I first jumped on "desktop" to mean the home market. There is a corporate and technical desktop market as well. As a programmer, it's quite easy to chew through gigs of RAM while developing. And the scientific user can definitely use the higher speed, high precision floats that a 64 bit CPU can provide.

So what is he referring to exactly? So far I've argued in this thread that, yes, 64 bit isn't necessary now but it's a good idea to start presenting it to the home user. But now I'm wondering if he think 64 bit only has a place in enterprise servers. If that's the case, then I think he's WAY off base.

 

[magomago]

Originally posted by: NFS4

I personally don't think that 32-bit computing on the IA32 platform started to become widespread until the release of Windows 95 in 1995. So, I'd answer that it was about a decade later.

So how is the A64 any different? AMD is just pioneering the technology now just as Intel did back in the day. It just gives AMD a better grip at getting x86-64 support by having an earlier showing.

So if AMD could get a decent userbase in - or atleaset something that can't be ignored then ini order to be compatable Intel would have to liscned their technology?