Talk:Itanic

Misc

Merge? Why won't it be deletd? A derisive nickname has no place in an encyclopedia aside from a footnote. If there really is pressure to document the downside of the Itanium, then let's make it a section of the main topic.Mikeblas 01:40, 25 December 2005 (UTC)

Itanic is a derisive nickname referring to Intel's IA-64 semiconductor architecture, comprising the Itanium and Itanium 2 series of processors. The name is a reference to Titanic, a highly-touted and expensive ocean liner which sank in 1912. The IA-64 architecture is considered by many to be a white elephant, as it cost Intel and partner Hewlett-Packard billions of dollars, and has achieved unit sales far below expectations, primarily due to cost and performance concerns. IA-64 was further hampered by competition from AMD's AMD64 architecture on the low end, and IBM's POWER architecture on the high end. Many early IA-64 vendors, such as Dell and IBM have dropped or drastically scaled back support, and even Hewlett-Packard now only offers limited IA-64 products.

I put the wrong comment in my revert edit. What I meant to say was, "How can you argue that Itanium has not been a disappointment to Intel?" Point to one thing in the statement above that is not true. —Joseph/N328KF (Talk) 14:30, 17 November 2005 (UTC)

Hi Joseph, thanks. I'll get to the statements that are not true in a moment, but there's a lot of FUD here. Anyway, let's start at the top. You're absolutely right about the nickname and what it refers to, but IA-64 (the architecture) does not "comprise" two series of processors that implement it. The article has gotta remain up to this point. Beyond that, it looks like your brain wasn't completely turned on. ;) Rough day at work, maybe, but I can't tell you how silly it is to hear you write that the Titanic was expensive, as if this were somehow a point against it. While I also dream of a future where my children can buy ocean-liners for $3, I think for the time being it is quite reasonable for ocean liners to cost rather substantial amounts of money.

Anyway. Moving on, things gets a bit murky.

Vague statements like "considered by many" are certainly disputable and so don't *really* have a place in an encylopedia just yet; you go on to say that Itanium development cost Intel and HP "billions of dollars" - but how many billions? Do you know? I wish I did, but this kind of operating information is usually tightly guarded. (How many billions did Xeon cost to develop? Pentium 4? Opteron? POWER5? I don't know. Do you?) What you can convince yourself of, though, is that some analyst's claims are outrageous; you can do this because Intel disclose total R&D expenditure in their financial statements. Some "analysts" (and other, typically British web sites) have claimed that Itanium development has cost Intel more than $10 billion, without bothering to notice that Intel's R&D expenditure is only around $4 billion a year. It seems unlikely that Intel would have cancelled all non-Itanium-related projects for 10 years running, and I do say 10 years because the vast majority of Intel R&D expenditure goes into their fabrication processes, which is money that must be spent regardless of whether those fabs build x86, Itanium, or any other processors. You can dig up this kind of information by sifting through Intel's SEC statements. Here's one to get you started:

(see e.g. http://ccbn.10kwizard.com/xml/download.php?repo=tenk&ipage=3754578&format=PDF )

- you can find more if you poke around Intel's website for a bit.

Anyway, it would be great if you could point to exact R&D figures, but until you can, your statement seems a bit bogus, and only gets worse when you cite "cost and performance concerns". Which performance concerns are these? The last time I checked, Itanium was at or near #1 in all industry standard processor benchmarks (e.g. SPECcpu2000.)

Speaking of Itanium sales generally, it comes as a surprise to some to learn that Itanium sales are actually on the increase. For example, see:

http://informationweek.com/story/showArticle.jhtml?articleID=166403335 or http://www.hp.com/hpinfo/investor/financials/quarters/2005/q3.pdf

Where 2005 Itanium CPU sales grew 170% over 2004, while HP's Itanium server sales grew 113%.

Best estimates are that Itanium processor sales are currently on the order of 8,000 per week. Whether or not this is large enough to cover Intel's development costs is hard to determine, but it is certainly far too early to refer to the thing as a white elephant, particularly when growth is at around 170% per year. It will be interesting to see where and when the figure levels out, but it sure won't be this year, and is unlikely to be next year due to the release of the Montecito processor.

Moving on:

you write that Itanium's market presence has been limited by x86-64 processors, first from AMD (and now also from Intel) and this is absolutely spot on. But you go on to write that POWER processors have been pressuring Itanium at the high-end. This is not quite so black-and-white. If you want to convince yourself of that, all you need to do is hit the IBM website and look carefully at the pricing of POWER5 systems. You will find that smaller systems are being sold for less than the price of additional *processors* in larger systems. This means that IBM *must be* doing one of the following two things:

- Selling low-end POWER5 systems at a loss, in response to cheaper Itanium systems that are making real inroads at the "mid-range": 4 to 32 processor systems, and making up for this by taxing high-end customers, who are unlikely to notice an additional $12,000 per processor since the entire system is already approaching $1m+.

- Not selling low-end POWER5 systems at a loss, but making their high-end systems disproportionately more expensive. This could certainly be seen as a "successful" attack from Itanium at IBM's high-end systems, which is where most of their profits are made.

Anyway, now we get to the stuff that is simply false or misleading. You write:

"Many early IA-64 vendors, such as Dell and IBM have dropped or drastically scaled back support, and even Hewlett-Packard now only offers limited IA-64 products."

but this is a bit misleading. There's Dell and IBM, but I can't think of any others. (Can you?) Two, three, or five - whatever the number is, it's certainly not "many", certainly not compared to the number of current Itanium vendors. As for HP, your statement that they now only offer limited IA-64 products is absolutely false. They offer more than they ever have, and the timing of your comment is particularly unfortunate: just a couple of weeks ago HP introduced an entirely new line of Itanium blade servers:

http://www.hp.com/hpinfo/newsroom/press/2005/051101b.html

I'm going to try editing the article to keep it as close as possible to yours, while removing the stuff I think is either highly misleading or outright false. If you think it's OK, let's keep it at that. Here goes!

AFD result - merge

Yeah, merge sounds good to me. The term 'Itanic' clearly gained widespread notoriety in the industry. And let us not forget, this was supposed to be Intel's next big thing. They lavished billions on it. has to go down as one of the biggest screw ups in chip history. Timharwoodx 13:35, 26 December 2005 (UTC)