A chip of real beauty died so that Intel could sell a few hundred Itanics a year...

Solaris/SPARC apps? On Itanium?

What gives?

(Not that anybody really cares about Itanium, but still… it's a talking point)

That Jon Stokes book on processor architectures does indeed look excellent.

But it's $33 in the US (amazon.com) and £34 in the UK (amazon.co.uk).

Amazon, surely, is having a laugh.

That word "remaining" bothers me. I can read it as "still left", implying that RISC is on the way out industry-wide, or I can it as "other" implying that Intel reckon Itanium is a RISC chip. Both positions seem totally insane, but the latter probably has the edge.

That was the DEC UK Alpha Marketing slogan on one year's Christmas cards and freebies. It was, if I remember rightly, dreamt up by a chap called Richard. Could just be coincidence of course.

Anyway, the industry no longer has to imagine being the one *without* Alpha AXP, soon it will have to imagine being the one *with* an Alpha (you can still buy Alpha systems at the moment if you know the right handshakes).

Hopefully the inevitable port of VMS to AMD64 will go relatively smoothly, but the HP-UX people will have the "byte order" problem to solve when their turn comes, and whether there'll be enough paying customers left by then to actually make the port(s) worthwhile is arguable.

VMS on a decent Proliant might be an interesting prospect for anyone who cares about virus-free high-availability servers.

Until Intel gets QuickPath up and running and there's a common Xeon/Itanium motherboard then they're just treading water.

Whether it will help or just remove some excuses for the price/performance gap is another matter.

You can have my Alpha boxen when you pry them from my cold dead hands.

And yes, they'll probably still be running!

The SPARC Solaris bit is about using Transitive's emulator software to run old SPARC Solaris code unmodified on top of Linux on Itanium. Basically, it offers those with older Solaris apps that could be too tough to port to more modern kit a pretty risk-free option to just carry on running the same code on modern systems. There was a demo at the last IDF where Intel hailed an Itanium system with Transitive software as the fastest Solaris system available.

OpenVMS on Itanium has been available from HP for a while and it is fully supported on their Integrity servers from the blades right through to Superdomes. Alpha is still available but only via HP Renew or brokers, but raw performance of Itanium now makes even the fastest Alphas look slow.

As for the 3Ghz vs 1.66GHz bit, everyone knows there is more to chip performance other than just clockspeeds. With healthy Integrity sales I'm sure HP is qute happy to be "saddled" with Montvale, and even happier at the idea of Tukwila. Maybe even SUN will give it a go - after all, they've started selling Windows on Xeon, what else have they got left to do they always promised they wouldn't?

The fact that Alpha development effectively stopped over 6 years ago (though they did put the final polish on EV7 after that) could help explain why today's Itanics can actually out-perform it. Took 'em a fair amount of time, though: HP was still fielding large Alpha systems rather than the next-generation, huge-cache Madison Itanics against POWER in some benchmarks until Madison IIs finally managed to sneak by the aging Alpha scores.

Nice try at spinning away the 1.66 GHz vs. 3 GHz observation, but that comparison was not to *other* 3 GHz processors (e.g., x86-64s, which indeed do run circles around today's Itanics) but to the performance that Intel was once projecting for Montvale. I.e., Montvale is many days late and many dollars short, just as Itanic has been since its inception in 1994 (when Intel was at first projecting a 1997 - 1998 ship date for a competitor-killing part, rather than the 2001 ship date that the unusable Merced part achieved or the 2002 ship date that the more respectable but still hardly world-beating McKinley achieved).

Yes, HP does manage to sell Itanics, but mostly to existing customers who need to upgrade previous Itanic purchases or replace aging PA-RISC or (more rarely) Alpha systems. Last I knew, HP's *total* non-x86 enterprise system revenue was still declining, suggesting that the demand for Itanics by any *new* customers (as contrasted with the replacement market) is, shall we say, less than clamorous.

It was clear to people not taken in by Compaq's frantic spinning immediately after the Alphacide that 1) Itanic would advance at nowhere nearly the rate they were predicting, 2) Alpha could have easily continued to out-pace it had development continued, 3) the extension of x86 to x86-64 would soon exacerabate the erosion of Itanic's potential base from the low end, and 4) POWER (and even to some degree SPARC variants, especially Fujitsu's) stood a good chance of holding the high ground against Itanic - leaving Itanic wallowing through heavy seas despite the power of the corporate and marketing engines driving it. Had HP not initiated take-over proceedings very soon thereafter, there was some chance that the incompetent Compaq management responsible for that fiasco could have been ousted and Alpha development resurrected; as it was, we had to be satisfied just saying "We told you so..." as we watched Compaq/HP's enterprise business swirl slowly but inexorably down the toilet.

Too bad about VMS, though: it coulda been a contenda had it been ported to something more palatable and given more than life-support-level development funding. Oh, well: as the updated marketing slogan goes, "Digital had it then."

- bill

It seems absolutely hilarious to me that people are willing to give up half of Itanic's already dismal performance to get absolute certainty that their hardware even works in the first place. As opposed to the vastly more common software errors, which it will happily compute wrongly twice. What would you do if it failed regularly, anyway? Turn it off for five minutes every hour to cool down?