POST OF THE DAY
Advanced Micro Devices
In Reply To:
New Alpha Machines

Format for Printing

Format for printing

Request Reprints

Reuse/Reprint

By ChrisRijk
September 4, 2002

Posts selected for this feature rarely stand alone. They are usually a part of an ongoing thread, and are out of context when presented here. The material should be read in that light. How are these posts selected? Click here to find out and nominate a post yourself!

techobserver1 wrote:
One one hand you have a brand new AlphaServer that is fully binary compatible with your existing Alpha applications. And for the long term, HP has the Itanium architecture which should be rock solid in 4-5 years when you are ready for the next generation machine.

In the meantime, why not just postpone the migration decision?


Yes, you need a big overlap to keep existing customers getting too annoyed.

Of course, if all existing HP-UX / Tru64 customers postpone, then Itanium version of those OSs is a dud...

I haven't really studied in detail the short term product plans for the PA-RISC machines. But what I'm wondering is this: What happens if many of the Itanium customers just decide to hold out and wait for their migration efforts. That seems like the lower cost / lower risk thing to do. Then what does Intel do in the meantime? Obviously HP is in this for the long haul, but what about the other Itanium suppliers. How long are they willing to wait for real customers to show up in volume?

Next up for PA-RISC is the 1GHz 8800 "mako" in 2003, which is dual-core, with 32MB external cache. Estimated SPECint of 1000 at 1GHz. Upgrade for existing systems. So, not too bad at all - over 100% (max) throughput increase. After that is a 0.09um version - bit more cache, bit more MHz I guess. Customers with existing production systems will upgrade to this - I think in such cases, Itanium would be ignored 99% of the time.

Maybe I'm missing something. Maybe a migration will happen faster than I would expect. Or maybe Intel has no problem sucking up the costs along the way. It just sounds expensive to me.

I think existing HP customers will only start to seriously consider Itanium when (a) they're planning to do new projects (b) the software they need is available (c) the possible solutions is stable/proven. That's not going to happen quickly at all - the higher end you go, the more paranoid (and slow-moving) customers become. The bigger the change, the more paranoid they become.

A lot of existing HP Unix customers will have production Sun and IBM systems as well.

To existing HP customers, migrating to Itanium is likely to be riskier than migrating to IBM or Sun (until Itanium is very well established - 2005/6 earliest I'd guess). In the early days of Itanium, many customers will choose IBM or Sun for new projects, and as Itanium slowly gets established, more HP customers will start to choose Itanium. Chicken and egg problem though - the longer it takes for Itanium to become established, the more likely it'll "fail". HP's market share (in x86 and RISC) is shrinking though, which could compound their problems, particularly if it continues at current rates - IBM's going to overtake HP in RISC/Unix soon at this rate, and Sun is going to get even further ahead. Many customers like to pick the "market leading solution".

As long as the IT recession continues, customers will be even less likely to take a risk with Itanium.

PS mid-range to high-end RISC/Unix market (systems cost $100k or more) is a big part of the overall server market (revenue wise), though number of CPUs in such systems is probably in the region of 100-200k per year. It'll be a long time before this market is worth more than $100m revenue per year to Intel in CPU sales. This is supposedly the main (new) market Itanium is supposed to go after, too.


Become a Complete Fool
Join the best community on the web! Becoming a full member of the Fool Community is easy, takes just a minute, and is very inexpensive.