Sunday, June 25, 2006

Henri Richard proposed a second life for Itanium

AMD's EVP said in this interview:

"Maybe the salvation for Itanium is to make it socket-compatible with Opteron. That would make it an excellent candidate for a Torrenza co-processing opportunity. That’s probably the only chance it’s got!"

Very interesting idea indeed. But I wonder what is good about Itanium besides its 24MB cache.

21 Comments:

Anonymous Anonymous said...

Most people have a lot of trouble understanding Itanium.

Itanium's mission was never to succeed as a processor.

On the contrary, Itanium's mission was to leverage Intel's power and dominance of the processor business to drive all other processor designs out of the market.

Thus we see Itanium's successes:

ALPHA - Dead

PA-RISC - Dead

SGI/MIPS - Dead

PowerPC - dead outside of IBM and slowly dying even inside of IBM

SPARC - took a big hit on the high end. SPARC now evolving towards "Cool Threads" machines and traditional SPARC development is stagnant.

Intel killed off many competitors with nothing more than a crappy chip and lots of payoffs.

Intel can pat themselves on the back for putting the entire processor industry 10 years or more back of where it could have been.

If anything, AMD should be looking at Itanium for clear evidence of Intel's illegal antitrust activities.

12:00 AM, June 26, 2006  
Blogger Mad Mod Mike said...

Intels move to Quad-Core might be interesting for the Server-Front as far as Itanium is concerned. AFAIK, "Tulsa", the Quad-Core Itanium 2 set to debut whenever Intel can decide they're losing by too much, can't have all 4 cores running 100% at a time to stay within its thermal limit. Tulsa is also supposed to use CSI and have an IMC, yet they're putting 6MB of L2 Cache per CPU on it...kinda defeats the purpose of CSI and IMC....Intel is too dumb.

4:45 AM, June 26, 2006  
Anonymous Anonymous said...

To the 1st poster... Great commments on the state of Intel and to expose what they really are.

You'd think a technology/market leader would do more to further the industry...

In reality though, since they are a monopoly, they are content (You can see ran by Marketing folks) to just drive all other companies and any contending uP, and call it success.

I have yet seen the must respected Intel of the 80/90's, to a company that does not innovate anymore, and have no interest in serving the tech community at all.

6:42 AM, June 26, 2006  
Anonymous Anonymous said...

This is rather strange conslusion. Intel didn't kill anything because number of Itaniums sold is very very low. This is all about Linux and better price/performance ratio of x86 machines. That's why Solaris is working on x86 and SUN or HP sells more and more PC servers.

7:02 AM, June 26, 2006  
Anonymous Anonymous said...

I think the chances of Itanic ending up on Torrenza as a co-processor are likened unto a snowball's chance in Hell. Who would want that overrated POS processor as a Torrenza plug-in when there is great likely-hood that another company will develop one instead that will far surpass it's specialized capabilities, perhaps 10-fold? Intel will make sure an Itanium NEVER ends up on a AMD platform in any form, I'm pretty confident of that speculation judging by past history. Ain't gonna happen Henri!

8:29 AM, June 26, 2006  
Anonymous Anonymous said...

QUOTE:
"This is rather strange conslusion. Intel didn't kill anything because number of Itaniums sold is very very low. This is all about Linux and better price/performance ratio of x86 machines. That's why Solaris is working on x86 and SUN or HP sells more and more PC servers."

Yep, that's the point. One single architecture, you can use in servers AND desktops without pain: no need to develop specific platform apps and programs, just one takes it all

9:16 AM, June 26, 2006  
Anonymous Anonymous said...

i must say, henri has gotten a lot more arrogant lately...
maybe he needed to be reminded his AMD is going to have a hard time battling conroe.

9:31 AM, June 26, 2006  
Anonymous Anonymous said...

To Mad Mod Mike:
"AFAIK, "Tulsa", the Quad-Core Itanium 2 set to debut whenever Intel can decide they're losing by too much, can't have all 4 cores running 100% at a time to stay within its thermal limit."

Tulsa is dual-core processor built Intel’s 65-nanometer process and is the last remnants of the NetBurst architecture. It has nothing to do with Itanium. I strongly suggest you to check your statements before you call any company or individual stupid. In fact I don't think calling others stupid is a good thing at all. Try to behave yourself, at least don't make stupid mistakes while calling others stupid.

9:49 AM, June 26, 2006  
Anonymous Anonymous said...

In response to the first post I believe HP got into intels' trap and played a big role in killing the ALPHA,PA-RISC,SGI/MIPS(Tandem/HP NonStop was a big MIPS shop). Or Is that their big plans together got wary? HP bought into idea of single hardware platform and have gone deep into it would be in trouble if cannot come up with an alternate stratergy fast.

Donot know much about PowerPc.

But definitly should watch out for the new generation SPARCs.

10:16 AM, June 26, 2006  
Blogger Sharikou, Ph. D said...

henri has gotten a lot more arrogant lately...

Or maybe this:
http://www.theinquirer.net/?article=32628

while Intel is spending money hyping chips off development FABs, AMD is busy selling chips

10:25 AM, June 26, 2006  
Blogger TheKhalif said...

Thus we see Itanium's successes:

ALPHA - Dead

PA-RISC - Dead

SGI/MIPS - Dead

PowerPC - dead outside of IBM and slowly dying even inside of IBM

SPARC - took a big hit on the high end. SPARC now evolving towards "Cool Threads" machines and traditional SPARC development is stagnant.



Sorry but Opteron did all of that. Itanium is a good idea, but it didn't do any of that. HP cancelled Alpha because Windows stopped supporting it.
PA-RISC was going anyway.

SGI/MIPS is still alive. Well MIPs is.

Opteron has made Itanic OBSOLETE. it wasn't AMDs idea. they knew they couldn't leave x86 but needed 64bit. So here comes AMD64. So great it is now the platform of choice for all MS server apps.

I mean if Itanium is so sucessful why did MS cancel XP for it and stop updating Server? Why is MS using all HP DL 585s for ALL WEBSITES AND MAYBE EVEN Domain Controllers?

I thought next week was kiss Otellini's ass week?

10:39 AM, June 26, 2006  
Anonymous Anonymous said...

I love it how almost every one of Henri Richard's interviews become generally about Intel and their products; and not about AMD and AMD's products...

10:46 AM, June 26, 2006  
Anonymous Anonymous said...

To put things in perspective:-

http://forumz.tomshardware.com/hardware/modules.php?name=Forums&file=viewtopic&all=1&t=189184&postdays=0&postorder=asc

10:55 AM, June 26, 2006  
Anonymous Anonymous said...

This discusses the future of Intel:-

http://forumz.tomshardware.com/hardware/modules.php?name=Forums&file=viewtopic&all=1&t=189217&postdays=0&postorder=asc

Be sure to read all the comments...

11:15 AM, June 26, 2006  
Anonymous Anonymous said...

Thus we see Itanium's successes:

ALPHA - Dead

PA-RISC - Dead

SGI/MIPS - Dead

PowerPC - dead outside of IBM and slowly dying even inside of IBM

SPARC - took a big hit on the high end. SPARC now evolving towards "Cool Threads" machines and traditional SPARC development is stagnant.


alpha died the minute it was sold to compaq

hp, the producer of pa-risc is working with intel to replace it with intanium

mips is still very alive in the embedded market

although apple dropped the power architecture, it's still very happy and prospering from emedded market through all major console gaming platforms (xbox360, ps3 & nintendo) up to super computing.

sparc seems to be alive to - in strange way.

what made the market decline, is the high performance of x86 based processors and the low prices for this machines...

11:31 AM, June 26, 2006  
Blogger Mad Mod Mike said...

"ry to behave yourself, at least don't make stupid mistakes while calling others stupid."

Thanks retarded Intel fanboy for correcting me, I lose track of worthless products and their names, excuse me.

2:23 PM, June 26, 2006  
Anonymous Anonymous said...

It was indeed Itanium/EPIC "mother of all new architecture chips" that killed off the RISC market.

Opteron had no mindshare at the time Itanium was being pumped by Intel. At the beginning of Itanium, Opteron did not even exist outside of AMD's labs.

Instead, Intel used its monopoly power to force EPIC on the market and all the other designs suffered.

There were entire IDF conferences focused on Itanium.

AMD at this point was "Athlon" and "Athlon MP". Intel was NetBurst and Pentium 4.

What really happened is that not too long after Itanium killed all its competitors, Opteron entered the scene and exposed the Itanium for the sham that it was.

As Opteron was designed intelligently to solve a wide variety of real world computing problems, AMD's Opteron easily torpedoed the Itanic and the rest is history.

But let us not forget that Itanium did indeed fulfill its mission before it was exposed as a fraud.

3:16 PM, June 26, 2006  
Anonymous Anonymous said...

Any one could like to comment on why SPARC is still alive and kicking? ANy thing special with this chip ?

3:25 PM, June 26, 2006  
Blogger Eddie said...

todospara1/chicagraf0 here.

Do you really want to know what's good about Itanium?

It is half of the architecture. The half that is not absurdly and unnecesarily complicated.

In Itanium you have interesting ideas for parallelism:

* Instruction Predication: Instead of using branches, you can predicate the instructions with a predicate register and the instruction runs only if the predicate is "true"

* Speculative execution: Instead of doing it implicitly in the µ-processor, you explicitly state where to put the check for the speculative code and the the procedure to retry the failed speculative segment

* The huge register file (128 GPRs)

* The rotating registers that work like a sliding window and allow advanced techniques such as software pipelining

There are some more.

Interested?

6:15 PM, June 26, 2006  
Anonymous Anonymous said...

"Thanks retarded Intel fanboy for correcting me, I lose track of worthless products and their names, excuse me."

I will excuse you, more retarded AMD fanboy

6:22 PM, June 26, 2006  
Blogger Sharikou, Ph. D said...

Instruction Predication: Instead of using branches, you can predicate the instructions with a predicate register and the instruction runs only if the predicate is "true"

All these features needs compiler support. How can they help in a x86 computing environment? You issue a bunch of EPIC instructions via ccHT to the Itanium? For what? Emulate Itanic on Opteron so HP's SuperDome customers can keep their software assets with some residual value?

2:42 AM, June 27, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home