Intel launched its ‘Montecito’ Itanium, doubling performance compared with its predecessor. Each Montecito chip has two processor cores. Of the six Montecito models introduced, the top-end 9050 has 1.7 billion transistors and 24MB of high-speed cache memory. The models range in clock speed from 1.4GHz to 1.6GHz.
It would not be as fast as sun new processors and the version II of their new processor that they currently test.
Itanium must pass through Israel design team to pump the performance they need, much like what happened to Pentium 4>Core Due transformation.
It would not be as fast as sun new processors and the version II of their new processor that they currently test.
To say it softly, current SPARCs are even not close to be as fast as Montecito and Power5+ in specfp and specint. Sure Niagara II will be faster than T1, but it will compete with multicore next-gen Itanium w/ CSI interconnection technology.
> It would not be as fast as sun new processors and the version II of their new processor that they currently test.
To say it softly, current SPARCs are even not close to be as fast as Montecito and Power5+ in specfp and specint. Sure Niagara II will be faster than T1, but it will compete with multicore next-gen Itanium w/ CSI interconnection technology.
Ok, benchmarks at 20 paces….. It all depends what you use as a benchmark. If you want HPC style performance then pick SPECint/fp. Maybe you should also include AMD in there. This is not the Niagara’s teritory. There is something else coming from Sun to fill that role.
If you are looking at Web/Application serving, the Niagara is clearly the leader (at the moment by a fair amount). So far the crystal ball has not said otherwise
There are many benchmarks around……
If you want HPC style performance then pick SPECint/fp. Maybe you should also include AMD in there.
Yes but AMD is not Sun. A lot of constructor sell Opteron.
If you are looking at Web/Application serving, the Niagara is clearly the leader (at the moment by a fair amount). So far the crystal ball has not said otherwise
There are many benchmarks around……
Such as this one on Anandtech:
http://www.anandtech.com/IT/showdoc.aspx?i=2772
I don’t see the supremacy of the T1 on this benchmark:
http://www.anandtech.com/IT/showdoc.aspx?i=2772&p=7
nor in this one:
http://www.anandtech.com/IT/showdoc.aspx?i=2772&p=6
If you want HPC style performance then pick SPECint/fp. Maybe you should also include AMD in there.
Yes but AMD is not Sun. A lot of constructor sell Opteron.
Maybe you missed the point. I did not relate AMD = Sun, but since you brought the point up. Here is some more current testing figures
http://blogs.sun.com/roller/page/bmseer?entry=woodcrest_doesn_t_giv…
lol, you give me a bench prooving that Sun is better than Intel made by Sun.
I am waiting the day a Sun employe will say “our products sucks”.
“To say it softly, current SPARCs are even not close to be as fast as Montecito and Power5+ in specfp and specint.”
SUN is not the only supplier of SPARC parts, Fujitsu’s SPARC64 is currently pretty close to Power5+ and montecito in SPEC numbers. So SPARC is still quite competitive IMHO. Plus it has far larger software and usre base than either POWER or IA64.
I’m not sure the source of your assertions, but something like a link or an explanation may be in order, a belief is not fact, as much as we would all like it to be so
hmmm…interesting what do u mean “Itanium must pass through Israel design team to pump the performance they need” care to elaborate?
That’s definitely a brute force attempt to gain more performance. But I guess it’s working out for Intell. I’m curious to the powerconsumption and heatproduction of this puppy.
I’m curious to the powerconsumption and heatproduction of this puppy.
Thanks to Foxton technology its only 104W.
http://www.intel.com/products/processor/itanium2/specs.htm
Foxton is not enabled in Montecito. It was broken, so they shipped with the functionality disabled.
URL?
Cache uses little power and spending transistors on cache instead of other stuff is a legitimate architectural tradeoff. Customers care about compatibility, performance, price, and power, not elegant design.
>Customers care about compatibility,
Congratulation, you said it yourself why Itanium is failing.
Yep, Itanium is failing based on its own merits; there’s no need for us to critique its design. 🙂
Is this the long-awaited Itanium stepping that brings new OEMs onboard and rekindles discontinued product lines at others, or is this where Intel starts to think about cutting its losses?
In a high-end UNIX space that moves glacially slow, is it possible that Intel can effectively compete with Sun, let alone IBM? Will they also ship a platform that offers the kind of dynamic resource modification/partitioning and workload management features that customers have grown to expect? Can they rely on independent OEMs to provide a highly reliable and well-integrated server product?
I find it hard to believe that Intel has any confidence that Itanium can compete as a platform against their competitor’s offerings, let alone their notoriously conservative prospective customers. It has less to do with the processor architecture and more to do with the overall server product and the decidedly negative perception that they must overcome.
I find it hard to believe that Intel has any confidence that Itanium can compete as a platform against their competitor’s offerings
Sure they don’t have any confidence in Itanium, thats why they plan to spend $10 billion on IA64 R&D during the next 4 years.
> I don’t see the supremacy of the T1 on this benchmark:
> http://www.anandtech.com/IT/showdoc.aspx?i=2772&p=7
> nor in this one:
> http://www.anandtech.com/IT/showdoc.aspx?i=2772&p=6
If performance was the only thing that mattered to make Itanic relevant, Itanium still would not survive against Sparc and Power. Availability of applications and platform momentum are much more important factors than just performance alone. An excellent take home example is Alpha, which shined on performance but still failed miserably. And even if we go to the subject of performance, it is not that stellar even on the latest Montecito. Sun Niagara absolutely stomped the previous generation Itanium 2 on throughput oriented workloads. Below is a good real life example where Niagara delivers about two times better performance than dual processor Itanium2 (requests per second):
http://www.stdlib.net/~colmmacc/2006/03/23/niagara-vs-ftpheanetie-s…
So Niagara will still beat Montecito on a chip per chip basis quite handily given the right workload. And I don’t think there is a chance Itanium would be able to hold a candle to the upcoming Niagara II. Niagara II will deliver more than twice the performance of current Niagara, so it will absolutely destroy Montecito on most of the througput oriented benchmarks. The current UltraSparcIV+ in the high end Sun machines also holds well against Montecito being slower by maybe 20% realistically. But keeping in mind that processors are a leapfrog game, I would expect Sun leapfrogging Montecito in performance with the next revision of higher clocked UltraSparc IV+ processors, which are supposedly due soon.
Overall it seems like there is little chance for Itanium survival, the industry pretty much written off Itanium as a viable platform. A simple fact that HP controls 80% of Itanium market should give you pretty good picture on how “popular” it is. Even though HP is drumming up the revenue numbers, the unit shipments for Itanium are *flat* year over year — HP did not ship more Itanium systems compared with the last year. On the other hand unit shipments for both Sparc and Power are up by a good margin. So all of the above is saying one thing Itanium is being relegated to a very small niche of very high end systems that command high premiums. Given the cost cutting situation at Intel I will not be surprised if Itanium will be sold off to either HP or some Japanese IT giant to slowly wither away.
Edited 2006-07-20 10:11
An excellent take home example is Alpha, which shined on performance but still failed miserably.
How is Alpha failed?! Alpha was killed expressly by HP.
Sun Niagara absolutely stomped the previous generation Itanium 2 on throughput oriented workloads. Below is a good real life example where Niagara delivers about two times better performance than dual processor Itanium2 (requests per second):
Such kind of workload is where Niagara would stomp not only Itanium but Power and Opteron as well.
A simple fact that HP controls 80% of Itanium market should give you pretty good picture on how “popular” it is.
That fact gives you insight that other big OEMs (Sun, IBM) have they own cpu architectures to sell.
Even though HP is drumming up the revenue numbers, the unit shipments for Itanium are *flat* year over year — HP did not ship more Itanium systems compared with the last year.
HP shipped 7200 Itanium servers in 2006q1, comparing to 5400 in 2005q1.
Overall it seems like there is little chance for Itanium survival, the industry pretty much written off Itanium as a viable platform.
Overall it seems that you are following the same theinq-style “itanic is sinking” “analysis” based on the years of constant arrogant Itanium bashing.
Iwould rather see the kentsfield comming sooner:-)