savantu (savantu@email.ro) on 2/5/09 wrote:
---------------------------
>Linus Torvalds (torvalds@linux-foundation.org) on 2/5/09 wrote:
>---------------------------
>>David Kanter (dkanter@realworldtech.com) on 2/5/09 wrote:
>>>
>>>This article talks about the catalyst for the delays, which
>>>was primarily a change in the platform, and the history and
>>>decisions that ultimately lead to the delay.
>>
>>The much more interesting issue is if they fixed anything
>>else than the platform?
>>
>>You compare to (very optimistic) POWER6 platform numbers,
>>and try to make it out to be the achilles heel of Itanium,
>>but look at Spec numbers, and realize that the weakest
>>point of Itanium was never the platform - it's the core.
>>
>>SpecInt runs basically close-to-100% out of the cache, and
>>of all the architectures out there, Itanium is way way
>>behind the competition. Improving on the platform will do
>>absolutely nothing to that, since it isn't even
>>relevant.
>>
>>Look at the SpecInt2006 numbers:
>>- Core i7-965: 36
>>- p570 (4.7Ghz): 21.7
>>- Itanium (1.66GHz/18MB): 17.0
>>
>>Now, the Core i7 number is obviously much more recent than
>>the other two, but hey, that's what Tukwila will be facing.
>>(And quite frankly, you can find old Core 2 numbers that
>>are just under 30, so even if you want to go back to the
>>previous generation, you're really not changing the argument
>>at all).
>>
>>And if Tukwila only improves on the platform, not the
>>core (and frequency), then it will be stuck at roughly
>>the same performance. Don't tell me that the platform
>>matters that much for SpecInt when you have a 18MB L3. I
>>just don't believe you.
>>
>>Btw, the numbers aren't that different on the SpecFP side
>>either. Itanium does fractionally better, but not by much.
>>
>>Yes, by comparing to only POWER6 you can make the race
>>look much closer, and can make the argument that POWER
>>has the better interconnect, but that's almost as absurd
>>as comparing Itanium to Sparc (which you also did).
>>
>>What's the real competition again?
>>
>>Oh yeah - it's those Core i7 Xeon chips with eight cores
>>and HT. They're not out yet, but neither is Tukwila.
>>
>>Linus
>
>Let's take another POV : does Tukwilla really need to beat Nehalem and X86 in general ?
>
>I'd say the answer is clearly no.In fact Tukwilla needs to equal Power and handily beat Sparc ( which will happen ).
>Intel can happily push Tukwilla as a direct competitor to other high end server/mainframe
>architectures while using Nehalem as a hammer to squeeze RISC/mainframes from above and below.
>
>It doesn't matter to Intel if somebody picks Nehalem over Tukwilla, still money
>in their pocket.

Excuse me?

When was the last time you ordered hardware?
What price would you want to pay for Tukwila versus Nehalem?

>It is a problem however if money goes to Sparc/Power even if Nehalem
>has 2x the performance per socket and per thread.Sometimes that isn't enough and
>that's where the whole Itanium philosophy comes into play.

Software gets upgraded in lots of companies. All kind of software from the past already has been replaced long time ago. All that software already scales pretty ok. It is NO problem to let single socket or dual socket machines do the job, as long as they got ECC.

Very few softwareproducts can profit major league from 4 socket machines over 2 socket machines, pricewise seen.

Basically we're speaking about AI engines such as my chessprogram where a 4 socket box is real cool to have.

Note for 2009 even that isn't relevant much, as some total idiots decided single handed to have a 8 core (logical cores) limit there.

Yet there is a lot of other tournaments out there. Thing is, that is software that doesn't run crucial services for companies now.

Vincent