AT conroe review updated

[Ycon]

Itanium = Expensive Failure
Prescott = Failure but at least it made money
Conroe != P6 based
and just for wusy
ASUS A8N SLI Deluxe=Prescott

I usually agree with your opinion, but like everything is wrong in that post.
Itanium = Best processor on this planet, dunno how you think its a failure
Prescott = beats its predecessor and delivers great computing experience thanks to HT. Not as much as real Dual-Core but its still OK.
Conroe = in a certain way it is P6 based cause it (Merom to be precise) follows the same idea
ASUS A8N SLI Deluxe != Prescott, it does neither beat competitors (not to mention DFI) nor does it offer any great computing experience

 

[Action_Man]

Itanium = Best processor on this planet, dunno how you think its a failure

I don't believe they've made any money on it and if they have it'd be so little it wouldn't justify the cost.

Its way too expensive and it costs heaps to make. Its a niche CPU at best.

Prescott = beats its predecessor and delivers great computing experience thanks to HT. Not as much as real Dual-Core but its still OK.

Way way too hot and slow. The K8 hands its ass to it whilst being cooler and cheaper.

Conroe = in a certain way it is P6 based cause it (Merom to be precise) follows the same idea

It borrows technologies from the PM but it isnt based on it.

ASUS A8N SLI Deluxe != Prescott, it does neither beat competitors (not to mention DFI) nor does it offer any great computing experience

I threw it in for wusy. I'm sure he'll get laugh out of it, well he better!

 

[Whizzard9992]

Itanium is a good processor, no doubt about it, if you can afford it. The architecture is antiquated now because the market is going multi-core.

There's a lot of speculation that the last Itanium 2 release is indeed the last Itanium release.

P6 = Pentium Pro. Just stop.

 

[Action_Man]

P6 = Pentium Pro. Just stop.

And Merom/Conroe/Woodcrest = Built from the ground up.

 

[apache_lives]

Conroe - based off P6 - take note the efficency like all other P6 chips!!!

 

[Action_Man]

Intel says otherwise and I agree. The amount of changes from the PM it'd make more sense to start fresh, which they did.

 

[ltcommander_data]

Whether the Merom architecture is based off the P6 is just a technicality. In truth, it is designed around the Pentium III architecture. However, there are sufficient changes to warrant it being classified as a new architecture in it's own right. Features like micro and macro-ops fusion were added completely new and required changes in how data information is stored. The changes weren't drastic, at least in the case of micro-ops fusion which we've heard more about, but the concept is different. The decoders have also been completely reworked with 4 full complex decoders instead of 1 complex and 2 simple. The number of ALUs and ports has increased by 1 while the FPU has been completely redesigned and now supports all the latest SSE instruction sets including a supposed SSE4. Data prefetch mechanisms have also been vastly improved along with branch prediction. The FSB is now borrowed from Netburst and both caches have increased in size with the associativity changed. As well, even the parts of the architecture that haven't been redesigned are still different from the PIII days since the various parts can now shut themselves down to save power.

If you either look really hard, or drastically generalize you will still see the P6. However, the perform levels and the reasons for those levels is quite different.

 

[Action_Man]

Just because you see the P6 doesn't mean its based off it.

 

[joset]

Now if woodcrest is indeed 20% faster that will eat further into their itanium sales, unless they cripple woodcrest such that it can't work in more than 2/4 socket variations (i.e. upto 16 way with the release of quad core).

First off, and once and for all:



There you have it. Architecture, Microarchitecture & Processors (from Intel, of course.).

No more reasons for speculation & ill judgements.

Edit: to avoid coherency issues, I underlined changes:

Aside skunk works-like projects, Intel has two strong architectures (counting out IXA...), within three different spaces: x86 IA-32/EMT64 and EPIC; AMD has two strong architectures, within two different spaces: x86 (32-bit) & AMD64 (no need to go through IBM, Sun, MIPS, ...).


Now, for the myth:

Itanium = Itanic. False.

Itanium & EPIC are an entirely different species (like IBM Cell & the Octopiler are), in the computing space: either will it succeed or not, it'll set (again, just like Cell) a computing paradigm for the future. Many factors are at stake, in this project, on which its success might depend, but... do you think that a highly sophisticated & successful computing platform company would bet its reputation (& best resources) in something only dreamers dream?
Intel took over HP's project; recently, it was decided to drop off IA-32 hardware support. So, Itanium was going more & more depleted, right? Then, why this, now:

http://arstechnica.com/news.ars/post/20060307-6335.html

We, really, don't have a clue about most of what's going on in those "rooms", do we?

Striking, was also Steve Jobs move towards Intel; it was weird, unexpected, offensive & almost anachronic. In a way, it was understandable, profit-wise, since Intel is... well, Intel. And, the intent was to gradually replace its mobile space, right? Wrong! Steve clearly sated - a few years ago - that he intended to replace its entire 'fleet', from laptops to servers. Why? Well, the reasons put forward was that, IBM (...) couldn't reach the 3.0GHz plateau within schedule and was in short supply... or, was Steve already aware of Intel's next big jump, in all computing fronts?

As for the collection of disparate, uneducated (& often disgusting) statements & speculations about who's who in the podium's first place, let me join my different approach to it:

Do any of you think that, both the most verbalized chip manufacturers, here in these threads, are a mere bunch of idiots, cooking strategies & devising schemes just for the next "IDFs", "CeBits" and every other fair?
Billions & billions are invested by each one, each year. And that's not just to look 'pretty'; Intel's came up with a very promising computing platform and is worth for what it's accomplished. AMD has amazed everyone - being 7 orders of magnitude smaller - and will keep doing so. It's their job, after all!

(Don't like the style, but...):

http://www.theinquirer.net/?article=30042,

as a kind of balance between the 'now' Intel & the 'briefly' AMD.

Post scriptum: sorry stimpy... had to post to someone!

Cheers!

 

[Whizzard9992]

Wow, that is a humbling post, to say the least.

I'll start with a hearty thanks.

It's amazing to take a step back and see how these boards are really the result of marketing, and how everyone really just gets sucked into it an consumed by it, myself included.

So questions: what is the Niche market Itanium is milking? I've seen it applied to SQL Server with much success, but that's just my world. Doesn't x86 offer a more inexpensive alternative to IA64? What exactly is the benefit of IA64 over x86? What about multi-core on IA64? I know Intel is pushing Itanium, but I've seen press lately saying that the Itanium 2 is the last of IA64.

 

[joset]

So questions: what is the Niche market Itanium is milking? I've seen it applied to SQL Server with much success, but that's just my world. Doesn't x86 offer a more inexpensive alternative to IA64? What exactly is the benefit of IA64 over x86? What about multi-core on IA64? I know Intel is pushing Itanium, but I've seen press lately saying that the Itanium 2 is the last of IA64.

IF Itanium succeeds (that's a big IF: like IBM's Cell, these are spearhead advances into new Design/Architecture/Microarchitecture & Compiler territories, with only DEC's Alpha serving as a [post-mortem] contextual reference.) As a note, even Cell's giving a lot of headaches to programmers & OEMs and many foresee its 'extinction' beyond the console space...
Intel & HP's aim towards Itanium, has always been the 64-bit arena, from the workstation up to supercomputing. Dynamic internet-content traffic, complex [scientific] computation, rich-media & object-oriented purposes, were & are Itanium's main target (Although HP has already abandoned the 'boat', I still refer to them on everything Itanium-related, 'cause they've had more than a decisive role in its development...). Both Intel's Xeon & AMD's Opteron lines, won an astounding share of Itanium's market (the later, keeps gaining both Sun's UltraSPARC T1 & IBM's POWER series market share...), due to costs (evidently) but, especially, to the large entreprise software inertia (backwards compatibility; application re-compiling & generalized implementation, are some of the 'heavy-loads'...).
The recent transition to the 64-bit processing space through the [obsolete] x86 ISA and the use of highly 'transparent' network interconnects (allowing large MP scalability) is, certainly, cheaper and uses more off-the-shelf technologies to achieve current performance requirements.
"Montecito" is a 90nm, dual-core part (IBM's POWER5+ is a MCM with four dual-core chips in it and Sun's T1 "Niagara" is a [very simplified, comparatively] 8-core on-die processor); aside production costs & low demand, Intel would have no major problem to go with a multi-core, smaller node Itanium.
IA-64 might go for good (and Cell, according to some...). But, most certainly, x86 ISA will decay, within the next 10 years. And, being so different in their approach to computing, I'm quite sure that both - IA-64 & Cell - will leave their mark as computing platform paradigms, just like Alpha once did...


Thanks for the compliment. That's what the THG forumz should be about, in the first place: help & enlighten. And, a good laugh is always welcome.


Cheers!

 

[ecosoft]

joset Dude

You presented a well reasoned analysis of "what's so" in the IT world ... what's your background, and how is it that you are located in Portugal?

I take singular exception with your comment "x86 ISA will decay, within the next 10 years". How do you define "decay"? On first reading, my interpretation was "becoming replaced with ??? tech." While anything is possible (cheap quantum CPU's), the notion that the massive PC User world will begin throwing out their accumulation of application programs AND their experience for some new paradigm appears ludicrous.

I say this within the context of your comments for IA64's failure to achieve significantly profitable market share because of (your quote) "especially, to the large entreprise software inertia (backwards compatibility; application re-compiling & generalized implementation, are some of the 'heavy-loads'...)"

What say ye, Sir? 😀

 

[joset]

Hello there!


For a start, the obsolete x86 ISA has been decaying for a long time (it was meant to have already disappeared, were it not for «the massive PC User world will begin throwing out their accumulation of application programs AND their experience», i.e., all software-related & platform implementation issues.
Well, it happens that, for the same reasons x86 still kicks, x86 will vanish. Not because it's old (and it is); rather, because it's complex (in a way that complexity isn't needed to perform better or differently, but because it's inherited & hardwired), because it's very limiting, programmig & compiling wise (the programmer/compiler has a narrow, if any, access to the hardware units) and, because it hampers hardware performance (it wasn't by whim that, from the P6 & K7, more & more RISC-like features have been implemented in both architectures; and no, I'm not mixing up x86 ISA, CISC & RISC.).

I believe that software (application programming & compiling) will drive the next ISA change: it's already happening, with this [apparent] shift towards performance-per-watt & increasing thread parallelism (x86 doesn't like it much, to say the least!), instead of full-speed ILP.
Actually, both RISC & CISC architectures have long departed from its original blue-prints, being a mix of both plus a bunch of sophisticated microarchitectural improvements, by now ("NetBurst" is/was a microarchitecture...). How gradually the transition will occur, only time will tell.
Finally, it may happen that the next architecture will be neither based upon IA-64, EMT64 nor AMD64 (although based in 64-bit memory space address & processing, most probably.); Cell...

Another apparent paradox (which addresses your last paragraph, in particular) is that, coming high-end technologies from the mainframe arena to the desktop as they come, it's common sense to state that we - end users - will be the first probes to use whatever comes out, instead of the large inertia-entreprise business space; and we - end users - are a very large space...

(For the sake of correctedness, "Quantum Computing" is a side-shift in microarchitectural processes but not, necessarily, in architectural paradigms...).

This is my opinion and it is what I meant by "x86 ISA decay" (it has something to do with my educational background, certainly), sir! ; )

Sorry, but I didn't quite grasp the meaning of «how is it that you are located in Portugal?»; well, I'm portuguese...


Cheers!