Earlier today on Twitter, AMD engineer Phil Park identified a curious nugget of PC architectural history. Bob Colwell (Pentium Pro chief architect) has been posting on Quora. Pentium 4 had a version of x86-64 that was fused off. This design path was axed by the higher-ups at Intel, who clearly feared it would eat into or harm Itanium (profits).
The aim of the game was Intel getting a 64 bit CPU without having to share the instruction set with AMD.
Intel was crossing many lines in terms of the anti-competitive practices they were willing to engage in, of which hobbling their own internal development was probably the least damaging to the market overall.
This article is filled with misinformation that any journalist should be ashamed of getting wrong. For example:
> where "64-bit" is immediately understood to mean x86-64 — because who wants a computer that can't natively run most applications made for a PC?
Huh? Sure, when people hear "64 bit", they don't think about Alpha, UltraSPARC or Power, but I don't think "64 bit" universally makes people think about x86-64, considering the ubiquity of 64 bit ARM CPUs in practically everyone's pockets. And "most applications made for a PC" has very little to do with whether something is 64 bits or not.
> the pure 64-bit architecture of Intel Itanium did not allow 32-bit (x86) applications to run natively
They were totally different instruction sets, so this had almost nothing to do with 32 versus 64 bit. The Itanic even had support meant to accelerate emulation of 32 bit x86, so if anything, Itanic would've performed dismally running 64 bit x86 code compared with 32 bit x86 or native Itanic code.
> Itanium landed with a thud in the market despite being among the first to the 64-bit punch
Itanic was late in the 64 bit world. Alpha, UltraSPARC, MIPS and even Power had 64 bit long before the 2001 introduction of the Itanic. AMD had already announced their spec and had development environments and toolchain support for x86-64 before Intel's Itanic release.
> However, the story of Intel Itanium doesn't quite end at the launch of AMD64/x86-64 in 2003 despite the obvious chilling effect. Itanium still kept support and was iterated on until February 2017, and shipments didn't completely halt until July 2021. For some fringe scenarios, Itanium was very performant and even ideal.
No. No, it wasn't. This tries to make Itanic seem like it was good enough to hold on until 2017, which is definitely not the case - Intel threw literally billions of dollars at the Itanic, but it really sucked. It never had any real edge over any of its contemporaries - not in raw performance, not in price per FLOPS or MIPS, not in aggregate performance. It was Intel proving that they're a one trick pony, which they still haven't figured out.
Aside from ethernet chips and x86, Intel has failed at everything - at embedded processors (Atoms have typically been woefully underpowered and have had reliability issues, like bricked Cisco routers), at broadband chips (see the Puma chipset debacle), at CPUs for embedded devices (even their modern "e" core devices like the N200 can't compete in performance per watt compared with ARM, and are too power hungry for anything you'd want to actually put in your pocket), at cellular chipsets, and so on.
Don't take my word for it. The Register does an excellent job of calling Intel out for their lackluster products and inability to compete outside of x86 (and problems competing inside of x86).
When I hear “64-bit” without any additional context, I absolutely think x86-64. ARM64 devices aren’t really that widespread in people’s mindshare, I would wager.
That doesn't mean that's universal, particularly for those of us who remember 64 bit CPUs before x86-64. I'd even wager that very young people stopped associating it with x86 at all since all x86 CPUs, save some Intel holdouts, have been 64 bits for at least the last fifteen years.
It is just the context, those of us who came all the way from 16bit 80286, 32bit 80386 to 64bit will read the 64bit a little differently.
Especially in today's world. Some level of Hardware understanding used to be a requirement. Now a days most developers grown up with abstractions they dont have to care about any of these any more.
Earlier today on Twitter, AMD engineer Phil Park identified a curious nugget of PC architectural history. Bob Colwell (Pentium Pro chief architect) has been posting on Quora. Pentium 4 had a version of x86-64 that was fused off. This design path was axed by the higher-ups at Intel, who clearly feared it would eat into or harm Itanium (profits).
The aim of the game was Intel getting a 64 bit CPU without having to share the instruction set with AMD.
Intel was crossing many lines in terms of the anti-competitive practices they were willing to engage in, of which hobbling their own internal development was probably the least damaging to the market overall.
This article is filled with misinformation that any journalist should be ashamed of getting wrong. For example:
> where "64-bit" is immediately understood to mean x86-64 — because who wants a computer that can't natively run most applications made for a PC?
Huh? Sure, when people hear "64 bit", they don't think about Alpha, UltraSPARC or Power, but I don't think "64 bit" universally makes people think about x86-64, considering the ubiquity of 64 bit ARM CPUs in practically everyone's pockets. And "most applications made for a PC" has very little to do with whether something is 64 bits or not.
> the pure 64-bit architecture of Intel Itanium did not allow 32-bit (x86) applications to run natively
They were totally different instruction sets, so this had almost nothing to do with 32 versus 64 bit. The Itanic even had support meant to accelerate emulation of 32 bit x86, so if anything, Itanic would've performed dismally running 64 bit x86 code compared with 32 bit x86 or native Itanic code.
> Itanium landed with a thud in the market despite being among the first to the 64-bit punch
Itanic was late in the 64 bit world. Alpha, UltraSPARC, MIPS and even Power had 64 bit long before the 2001 introduction of the Itanic. AMD had already announced their spec and had development environments and toolchain support for x86-64 before Intel's Itanic release.
> However, the story of Intel Itanium doesn't quite end at the launch of AMD64/x86-64 in 2003 despite the obvious chilling effect. Itanium still kept support and was iterated on until February 2017, and shipments didn't completely halt until July 2021. For some fringe scenarios, Itanium was very performant and even ideal.
No. No, it wasn't. This tries to make Itanic seem like it was good enough to hold on until 2017, which is definitely not the case - Intel threw literally billions of dollars at the Itanic, but it really sucked. It never had any real edge over any of its contemporaries - not in raw performance, not in price per FLOPS or MIPS, not in aggregate performance. It was Intel proving that they're a one trick pony, which they still haven't figured out.
Aside from ethernet chips and x86, Intel has failed at everything - at embedded processors (Atoms have typically been woefully underpowered and have had reliability issues, like bricked Cisco routers), at broadband chips (see the Puma chipset debacle), at CPUs for embedded devices (even their modern "e" core devices like the N200 can't compete in performance per watt compared with ARM, and are too power hungry for anything you'd want to actually put in your pocket), at cellular chipsets, and so on.
Don't take my word for it. The Register does an excellent job of calling Intel out for their lackluster products and inability to compete outside of x86 (and problems competing inside of x86).
When I hear “64-bit” without any additional context, I absolutely think x86-64. ARM64 devices aren’t really that widespread in people’s mindshare, I would wager.
That doesn't mean that's universal, particularly for those of us who remember 64 bit CPUs before x86-64. I'd even wager that very young people stopped associating it with x86 at all since all x86 CPUs, save some Intel holdouts, have been 64 bits for at least the last fifteen years.
It is just the context, those of us who came all the way from 16bit 80286, 32bit 80386 to 64bit will read the 64bit a little differently.
Especially in today's world. Some level of Hardware understanding used to be a requirement. Now a days most developers grown up with abstractions they dont have to care about any of these any more.
As the parent poster said:
> Alpha, UltraSPARC, MIPS and even Power had 64 bit long before the 2001 introduction of the Itanic.
And even PA-RISC had 64 bits. Itanium was just a processor trying to disrupt the RISC market.