Windows XP 2002 for the Itanium: Unbridled rage
virtuallyfun.com129 points by jandeboevrie 3 days ago
129 points by jandeboevrie 3 days ago
A historical note: Windows XP 64-bit Edition that was mentioned in the article, was Itanium-specific. It was based on XP kernel while Windows XP x64 Edition (for AMD64 architecture) was based on Windows Server 2003 kernel. Because of that, Windows XP x64 Edition had some significantly different performance characteristics compared to other Windows XP editions. If anything, it was closer to Vista than XP in some aspects.
I would love to read the Microsoft PM spec for this product. Who in the world were they making it for?
Itanium never made it into the consumer-class hardware that was XP’s audience. AFAIK, Intel never even published a roadmap for that to happen!
Maybe a proof of concept they shipped as a demonstration of loyalty to Intel?!
In approx. 2000 to 2002, 2003 or around then, there was a theoretical dream that Itanium based workstations would become a real mass market thing, but it obviously never caught on.
Intel never made Itanium CPUs cheap enough to realistically purchase and Itanium motherboards never became a thing from the top-10 sized Taiwanese motherboard makers (the same companies that were making really nice boards for the 1 GHz to 1.2 GHz socketed Pentium 3 with 512KB cache and the first generations of Pentium 4).
Also the performance sucked. This was before AMD64 existed on the corporate desktop or home desktop.
It totally worked! Took out like four decent risc competitors with just a little FOMO
Paper-Itanium is unfairly blamed / credited with that, but it's not exactly true. The vertically integrated RISC/UNIX minicomputer companies were not on economically viable trajectories. x86 systems were undercutting their high margin workstation and server business, subsidized by enormous volumes in PC market needed to sustain increasing silicon manufacturing and design costs of new generations. The RISCs were forced to keep retreating to higher margin, lower volume, more specialized, higher performance markets which was really their only option by that point but it was also exactly the wrong direction to go in for long-term viability in the face of these prevailing trends.
Intel had their foundries and CPU cores paid for by video gamers and receptionists and students typing up their assignments and people looking at porn at 14.4kbps. Servers and workstations were pretty much gravy and every year there was less their systems couldn't do compared to the big iron.
Those computers pretty much sold themselves, contrast with every $20k workstation or $100k server that the unix guys had to explain why they were worth 5x the grey box on the desk.
SGI was struggling to make their MIPS chips and were pretty quickly finished off when nvidia started making GPUs. Itanium originated at HP as a PA-RISC replacement. They owned DEC Alpha when they killed it but arguably DEC didn't have a big base of legacy software at that point. IBM and Sun stood for a lot longer, mainly I would say largely due to locked in customers/data on Oracle and DB2 databases and proprietary software, rather than the merits of their hardware.
If Itanium never happened, 64-bit x86 server chips would have done the same thing to MIPS, Alpha, PA-RISC, SPARC, PowerPC. Might have happened even sooner from Intel since they allegedly sandbagged efforts in that direction to protect Itanium.
Alpha cancellation happened before HP/Compaq merger, when it was purely Compaq decision.
At the time some partnerships were slowly reducing the base costs of making Alpha machines (with K7 Athlon motherboards and EV6 Alpha being essentially signal compatible but not pinout compatible - you could reuse the chips and most of the design).
Compaq however... was not interested, for various reasons.
Oh you must be right, I mix up my time lines.
In any case the economics were more or less the same as Alpha as for the others. They couldn't support integrated silicon manufacturing. It wasn't even just the manufacturing (which could be outsourced although arguably not as competitively as integrated shops like Intel until TSMC really hit its stride). Or the logic and physical design. But the software development and ecosystem support costs were ballooning too. Alpha like the rest of them just never made sense unless they could make competitive cheap systems that were compatible, and by the time efforts like Windows ports and x86 emulation came along, it was already too late.
I think you're right overall that Itanium is used too often as the reason why all the RISC platforms died/were cancelled, as if they would have had healthy futures had Itanium simply not existed, but I do wonder about Alpha.
I wasn't around at the time, so I don't know what availability and pricing actually looked like, but unlike the other RISC platforms, Alpha was actually available on PC-like systems, with PCI and ISA busses, and x86 emulation at the firmware layer, allowing regular PC peripherals like VGA cards to work. Windows and, significantly, Linux, had pretty superb support for Alpha.
It's not enough to suggest that it had a chance "in any reality", but it seems like a much more potent competitor than PowerPC or SPARC or MIPS, it only because of the platform and OS support. Both Windows and Linux had pretty capable x86 emulation software layers too, and not just at the very end!
> I think you're right overall that Itanium is used too often as the reason why all the RISC platforms died/were cancelled, as if they would have had healthy futures had Itanium simply not existed, but I do wonder about Alpha.
Alpha had no fundamental differences that would have enabled it to avoid the same problem as the others IMO.
> I wasn't around at the time, so I don't know what availability and pricing actually looked like, but unlike the other RISC platforms,
No that's not true, other RISCs had similar, even before Alpha was on the designboard. They all did RISC workstations, Sun, SGI, even HP and IBM. IBM most famously went into PowerPC with Apple and Motorola early on, and then also did CPUs for several games consoles.
IBM probably of all of them understood the commoditization fight and they tried to put up a fight, but even they couldn't keep ahead of it. Alpha couldn't have.
> Alpha was actually available on PC-like systems, with PCI and ISA busses, and x86 emulation at the firmware layer, allowing regular PC peripherals like VGA cards to work. Windows and, significantly, Linux, had pretty superb support for Alpha.
Just too expensive to produce for what it did. Why would you use it for a Windows or Linux machine if you could do the same with a Pentium or K6 that actually ran your games and word processors and other stuff.
Windows on Alpha and FX!32 were reasonably solid options, as was the general health of the Alpha ecosystem. The platform also was not overextended in dot-com boom as much as let's say Sun. Windows port was also pretty early, even if it was 32bit limited.
The big benefit for EV6 was that it had actual parts commonality with PC, and I do not mean few I/O chips like some of the others (SuperIO, network, etc.) but northbridge - anyone making a K7 motherboard could make an Alpha motherboard, and Samsung became first vendor selling both CPUs and complete systems with non-Compaq chips.
FX!32 was not good. DEC's own paper only provided results for "Byte Benchmark" (nbench) which is a few trivial compute kernels. Even that most favorable test they could find showed 500MHz 21164A performed about the same as a Pentium Pro 200, the core that would the next year be in consumer CPUs. It was also NT only, at the time not used for PCs. It's just not enough to be "as good" in a scenario like that where not only are you are massively disadvantaged by costs, but you have to break into existing markets.
EV7 and EV8 development did not go well either, costs and complexity was blowing out, EV7 slipped 18 months and planned 130nm shrink was not able to hit performance targets. EV8 was enormous and expensive totally unsuited to mass market commodity bottoms-up path of Intel, and exemplifies their retreat into high end niche rather than accept and embrace commoditization.
AMD licensed the EV6 bus, but of course there was no socket/pin compatibility between AMD and Alpha parts. So they still didn't have commodity motherboards, though that would only be one small part of the puzzle if they did have.