AMD shows off their "Barcelona" quad-core chip designs

If you're not super excited by technical chip jargon like "advanced branch prediction" and "sideband stack optimizer," you might want to skip this one over, but for you chipheads out there, AMD has just unveiled the nitty gritty specs behind their forthcoming Barcelona quad-core processors. Built on a 65nm SOI process, the new chips will take the place of AMD's Opteron line, and will power workstations and servers sometime mid-2007 before the technology trickles down to consumer versions of the chips. AMD couldn't help but take a little pot-shot at Intel for their quad-core systems, which they claim are just two dual-core CPUs packed together, but we're pretty sure most consumers are going to be more concerned with performance, price and performance per watt than semantics. AMD hasn't provided any benchmarks yet, but we'll all be watching closely, since they're currently playing catchup to Intel on most of those fronts.

[Via Slashdot]

[Revrant2394 @ Oct 11th 2006 3:47PM]

Well if you weren't off in Intel Fancamp maybe you'd realize AMD has been delivering what they promise since they hit us with a fine mainstream processor, the Athlon XP.

Accusing them of nothing but "vapor" is entirely too ignorant, possibly even stupid, when given the fact that Intel is on the brink of doing it themselves. I recall Intel was the sluggish one dragging their feet when AMD was pumping out 64-Bit processors and Dual Core, and for so many years leading up to the superb C2D, Intel produced a hotter, pricier, lower performing counter to AMD, futhering their market gains.

Vapor, just, Bah.

[Buzzsaw @ Oct 11th 2006 3:15PM]

A little "pot shot" at Intel Quadcore ... LOL.. Only problem for AMD is that the quadcore benchmarks that various websites have already done completely back up Intel's performance claims.

Now for a "pot shot" at AMD "Quadcore"- what am I supposed to do with their cute quadcore design diagram - can I stuff that piece of paper into my PC next month to power my PC like I'll be able to with Intel's Quadro chips as of November 1? I can't? Oh nevermind

So in a nutshell, AMD's answer to Intel's quadro muscle coming out next month is "wee - look at our drawing!"

[Jonathan Sundy @ Oct 11th 2006 3:18PM]

There IS a performance difference between having a "real" quad core setup and having 2 dual cores on the same package.

One case, and the one I've most heard brought up in this type of situation is memory bandwith.

To simplify we'll talk about dual cores. Done properly each core should have full memmory bandwith, thrown together, each core is sharing the same memmory pipeline, cutting it in half. It was a big discussion when AMD introduced the opteron with it's on chip memmory controller vs intels multiprocessor setups. I assume the same generally applies here. If Intel takes 2 dual core chips and drops them on the same package, each set is going to be fighting for the same set of bandwith. (Imagine two 1-lane hi-ways merging into a 1-lane hiway.) This of course is much simpler to do, and lets them get it out faster, and won't always affect the performance gain from the 2 additional cores. When done properly,and probably around the same time AMD will be coming out with their quad-core setup I assume, it would be more like having two 1-lane hiways merging to form a 2-lane hiway, giving you the full benefit of the 2 additional cores.


Atleast I assume this is one of the things AMD is poking at.

[Jonathan Sundy @ Oct 11th 2006 3:19PM]

Proper cache sharing is probably a more likely problem than memmory bandwith now-adays.

[Benson Leung @ Oct 11th 2006 4:00PM]

It should be noted that the Kentsfield design of two Conroe dies on the same package is a move that Intel expects to get them to market faster. Indeed, they will have Quad processors out by the end of this year...

Intel has plans for a quad core on a single die design when they shrink to 45nm next year, codenamed Bloomsfield.

So yes, Kentsfield is a non-optimal quad solution designed to get to market quickly by 2006 and 2007, but then again so is AMD's 4x4.

AMD's strategy is much like Intel's in that they are going to start marketing to high end customers a dual-dual solution, with two sockets with dual core Athlons, around the same timeframe as Kentsfield. It will have some of the same disadvantages as Kentsfield, namely some core to core communication will be slower, but it will be out to market sooner than K8L.

4X4 will likely cost more for consumers than Kentsfield since 4X4 will require 2 socket motherboards and require the purchase of two processors.

Either way, AMD's K8L won't be competing with Kentsfield when it debuts in 2007. It will be compared with Intel's next generation 45nm quad-core-single-silicon processor also due out next year..

[Andir3.0 @ Oct 11th 2006 4:16PM]

Isn't it obvious! By putting the arrow between the Res and grouping the ALU and AGU together you simplify the design of the chip far more than drawing a parallelogram for OPs. But taking the FMUL and FADD boxes and completely ignoring what comes out of them is sheer brilliance!!

[Akira @ Oct 11th 2006 5:39PM]

Nice name! I hope these have no heat problems, it gets really hot here in summer. Greetings from Barcelona. Akira

[fredouil @ Oct 11th 2006 9:16PM]

AMD does not have to hurry, they are quite smart :
they are leaving Intel spending on marketing/energy to educate the market about quad-cores, and when the market will be ready they will sell their powerful solution.

[Sean @ Oct 12th 2006 2:17AM]

Whee! Look at our pretty pictures! No, wait, don't look at the benchmarks Intel has already posted that wipe the floor with everything we produce! No, no. Look at the pretty pictures...

[Darkskypoet @ Nov 6th 2006 1:43AM]

True, 2 Core 2 Duos are better then one. Yay, you get a cookie :)

two prevailing streams of thought seem to run through the comments, Intel rocks with 4 cores (True), AMD has some decent silicone coming soon but isn't here. (True)

The important thing to realize about AMD, is that at this point in time there really is no need to push Quad core out tommorow. There isn't.

AMD currently sells almost every single processor it makes. Has anyone noticed the side effect to intel's ASP with having to sell all 90nm P4(there's lots in channel), all 65nm P4 (lots of those too), all celerons (oh yeah tons) at a loss because C2D is so blazingly excellent?

See AMD has the midrange market sweetly tucked away in it's A64, A64 x2, Turion, Sempron goodness. We all know C2D is faster and better then A64, same as A64 family has been better then the P4 family for quite some time.

So Intel brings out C2D and it's amazing! Blows A64 out of the water, which makes it that much better then any P4. an e6300 can be had for around $200 CDN, What then can you sell a P4 for?

AMD slashes it's prices and continues to sell all it produces, intel sits on billions in inventory of chips it can't sell.

So as AMD gobbles up mainstream market share for all those not buying a $200 cpu (most PC's selling right now). Intel Lauds it's C2D performance wins; while it's ASP, and stock price continue to fall.

Conroe is an awesome Architecture! Finally Intel got the picture, "Wider is better". And, as I handed a cookie out before, kudos to those that agree 4 good cores are better then 2.

But AMD is not in Danger, in fact, I believe they are purposefully taking their time, as the only threat to their carved out market share and much improved reputation is if they fail to respond, and fail to execute.

Rushing is the surest way to misstep, and lose all they have gained by Pawning intel for the last couple years.