Constantine, Re: "purchasers won't go near it [Itanium] with an 8-foot pole unless it can run their legacy software faster and more reliably than their existing hardware."
Ummm... duh.
Why would you ever think that Intel had plans otherwise?
That's what cracks me up about a lot of the Itanium nay-sayers. They point out blatantly obvious things like, "businesses won't spend money unless it can benefit them." Well, thanks for the words of wisdom, but people already know that. And they also know that IT spending is based on lowering TCO of their current systems.
As systems age, maintenance costs increase, the performance to workload ratio decreases, and newer options come up across the whole industry offering better solutions. Of course, they are not going to spend money, and then find themselves in a worse situation than when they started. But the environment of a lot of IT centers is pretty complicated.
There are costs involved in power and cooling, costs involved with administration and hiring of technicians, costs involved with setup and new software installation, costs involved with networking, service, support, and a host of other areas. Performance is *not* just the deciding factor in upgrading a computer center, even though it directly relates to productivity. The fact is that even commodity x86 servers exceed the TCO of highly scalable RISC mainframes in many situations.
The decision of a company to abandon their whole infrastructure in favor something else is not an easy one to make. Clearly, Intel knows it has to not only offer the most premium performance of any other choice, but that it also has to enable OEMs to introduce lower levels of total cost for their clients.
I'm not sure what ways Itanium lowers cost, or how Intel intends for it to be a realistic player in the high end space. I know that long time businesses like Sun and IBM probably have much more experience in offering businesses what they want. But I also know that Intel has the manufacturing to be the low cost provider over any other option that exists. I'm hoping that Itanium 2 does as well on other performance metrics as it does on SPEC, because if it does, then Intel will have both low cost and high performance to drive new solutions.
Intel has gained a lot of experience in the enterprise with Xeon and Pentium Pro over the last 5 years or so, but I realize they have a long way to go to catch up to the RISC experts. Nevertheless, acquiring engineers from the Alpha and Superdome teams will no doubt give them valuable experience, and with trial and error, Intel will eventually get it right. Itanium is a work in progress, but it's built with an architecture that was meant to deal with the trend towards more parallelized data, which will make it grow stronger with each iteration. That's why Intel is eager to show so many new cores on their roadmap, because it ensures the industry that much more work is being done on Itanium for the future. Not even Intel's money making commodity products go into so much detail.
What I don't know is how much Intel has learned so far, or whether they are prepared to enter the high end space with McKinley this year. What I do know, however, is that you are not experienced in this matter, which is clear from how you speak of it. Since that puts you far from being the authority, I think I'll wait and see for myself whether Intel can execute with Itanium. As I see it, they still have much of an opportunity to do well with it.
wbmw |
