Two Perspectives...By Ed Stronglio

two perspectives

I liked Ed's article on viewing AM and INTEL from both an overclockers and investors standpoint and being an ex daytrader I could get really carried about writing about stock performance for both companies both past and future. But rather, I'm a little curious about the comment:

"Despite its expected performance, Conroe/Merom is still essentially a quick fix. Intel needs a new design, and sooner rather than later."

This is a brand new design, built from the ground up for both today's power and performance needs. Netburst is gone, the lengthly pipes that serve no use in real world computing, gone. Extra lanes allowing multiple instructions to be executed simultaneously, power is down performance is up and a basic platform is now in place for multiple cores as well as being on the edge of 45nm.

If you ask me, Presler and Cedar mill was the quick fix, simply scaling the Prescott architecture down to 65nm. maybe dual purpose marketing.engineering ploy to get the bugs out of 65nm and set the stage for the leap ahead that CORE duo will bring to computing.

So like, uh, what's up with that comment Ed.?

I think he was making reference to the fact that the architecture draws much of its character from past Intel procs as illustrated in this Ars article:

A question of breeding?

Before I get into the more technical discussion of Core's features, I want to quickly spell out how I view Core's relationship to its predecessors. As Intel has repeatedly claimed, Core is a new microarchitecture that was designed from scratch with today's performance and power consumption needs in mind. Nonetheless, Core does draw heavily on its predecessors, taking the best of the Pentium 4 and the Pentium M (Banias) and rolling them into a design that looks much more like the latter than the former.

Because the Pentium M itself is a new design that draws heavily on the P6 microarchitecture, I've chosen to place Core very generally within the P6 "lineage." However, I ask the reader not to read too much into this loosely applied biological metaphor, because my comparing Core to its P6 predecessors and talking about its development in terms of the "evolution" of the "P6 lineage" is really nothing more than an way to organize the discussion for ease of comprehension.

Click to expand...

http://arstechnica.com/articles/paedia/cpu/core.ars

Yes, it's a new architecture, but the similarities to some of the old ones is much more pronounced than say the transition of XP to A64.

Actiually K7 to K8 wasn't a huge change. The execution caors are very similar. Mostly just tweaked a bit (widened to 64-bit, pipeline length grew slightly, SSE2/3 added) and a memory controller/hypertransport interface added. Actually I think "core" is more different from P6 than K8 is from K7.

"Despite its expected performance, Conroe/Merom is still essentially a quick fix. Intel needs a new design, and sooner rather than later."

Click to expand...

Hmm. While I wouldn't necessarily agree that Core is a quick fix, it's a step they needed to take...a step backwards. Maybe it's P3 on steroids for 2006, but they really messed up after P3 and now is their chance to go back in time and take the other road

The next fix they need (and maybe what Ed is eluding to?) is to stop relying on the external bus for each chip to get data. With quad-core procs coming up, their plans of stacking more and more cores in procs is going to be a waste of time if there are more cores all fighting for the same limited bandwidth and have to go the "scenic route" to work at that. From what I've seen, multi-core server procs will start using an "interconnect" between chips in '07 that's akin to AMD's Hypertransport. I haven't heard anything about it on desktops/mobiles and there's just not a ton of info on it yet at all really.

If you ask me, Presler and Cedar mill was the quick fix,

Click to expand...

If you ask me, Presler and CedarMill were not even a fix (maybe for heat), so much as a reduction in costs (you can get a lot more 65nm chips out of a 300mm wafer than 90nm chips) and them getting the process fine-tuned for Core CM/Presler really were nothing except a Prescott die shrink with some extra cache tossed in.

don't forget to post a link to the article in question..... see the stickies for more details.....

Two Perspectives

Interesting read...loooks like the stockholders could be screwed coming up

PPro, PII, PIII, Pentium M, Conroe
Are all based on the P6 core, justed tweaked

Gnufsh said:

Actiually K7 to K8 wasn't a huge change. The execution caors are very similar. Mostly just tweaked a bit (widened to 64-bit, pipeline length grew slightly, SSE2/3 added) and a memory controller/hypertransport interface added. Actually I think "core" is more different from P6 than K8 is from K7.

Click to expand...

My bad. Try to help the guy out and only end up confusing him more

JaY_III said:

PPro, PII, PIII, Pentium M, Conroe
Are all based on the P6 core, justed tweaked

Click to expand...

Well if they can get this much of a performance increase with a tweek, imagine what can happen if they start from scratch.

Gnufsh said:

Actiually K7 to K8 wasn't a huge change. The execution caors are very similar. Mostly just tweaked a bit (widened to 64-bit, pipeline length grew slightly, SSE2/3 added) and a memory controller/hypertransport interface added. Actually I think "core" is more different from P6 than K8 is from K7.

Click to expand...

That isn't really a fair comparison, Conroe is more a development of Banias, Dothan, and Yonah rather than some largely done from scratch redesign of the original P6, a better analogy would be between P6 -> Conroe and the original Athlon (or maybe even the K6-X) and current generation Athlon 64s, and there is a lot of difference on both sides. I know that Conroe isn't just a tweaked Pentium Pro by any means, but you can trace it's design back to the earliest P6 chips, just as the current AMD cores can be traced back to the Nx686 at NexGen.

proth said:

Well if they can get this much of a performance increase with a tweek, imagine what can happen if they start from scratch.

Click to expand...

Given Intels recent track record not a heck of alot much
P68 (Netburst)
P7 (Itanium)
P8 (Itanium II)

but only time will tell

>HyperlogiK< said:

just as the current AMD cores can be traced back to the Nx686 at NexGen.

Click to expand...

Are you sure about that?

k5 was the 1st original AMD (everything before was clones more or less)
K6 as you stated was bought from NexGen know as the Nx686

However the K7 sometimes called the x86 Alpha is not based on the Nx686, as far as i know its made by a different set of engineers that came from DEC Alpha/Compact ect. Based on the Ev6 bus of the alpha.

Do correct me if i am wrong here

It is hardly gospel, but Wikipedia describes the original Athlon as 'a major reworking of the K6 core designed for compatibility with the EV6 bus'

They list the changes as including:

A heavily improved floating point unit
Big L1 Cache
Off die L2 Cache

I read something along these lines somewhere else ('PC Plus' i think) so either it is a widespread myth, or it is true, gnufsh?

>HyperlogiK< said:

It is hardly gospel, but Wikipedia describes the original Athlon as 'a major reworking of the K6 core designed for compatibility with the EV6 bus'

They list the changes as including:

A heavily improved floating point unit
Big L1 Cache
Off die L2 Cache

I read something along these lines somewhere else ('PC Plus' i think) so either it is a widespread myth, or it is true, gnufsh?

Click to expand...

You probly are correct then
K6 had an awsome ALU but weak FPU either way, as we all know K7 did wonders for AMD

JaY_III said:

You probly are correct then
K6 had an awsome ALU but weak FPU either way, as we all know K7 did wonders for AMD

Click to expand...

i can vouch for the k6, (and k6-2, and k6-3) fpu being completely atrocious.