Aceshardware

That last five quarters and the continuing rise of integrated graphics by both amd and intel probably explains why nvidia is so focused on anything but pc addin graphics.

of course jhh doesn't come out and state that since shareholders would flee.

nvidia will be left fighting for tv, handheld and other embedded graphic sockets soon enough. larrabee will likely take the hpc market not because it has the highest theoretical performance but because of the work intel is doing on the software tools it can be programmed by mere mortals...and on time.

gb

Im not saying nVidia is in great shape, just wanted to point out the potential profits in this market for "ATI" are more than $200m per quarter (especially if it would take the entire market.. somehow as suggested above).

Your first sentence for some reason made my brain recall SUN's Millennium chip as well as USV chip...both had plenty of time allotted for their development and design phases (nearly 5 yrs) and both failed to see the light of day (tapeout) due to anticipated lackluster performance metrics.

Not saying AMD is SUN, SUN had way more money. But look where AMD's past *major* microarchitecture advances came from...pre-K5 was Intel borrowed microarchitecture, K5 was all AMD, then K6 was bought by buying up NexGen, and K7 simply would not have happened were it not for DEC's implosion and AMD being there to hire up the Alpha team members (including Dirk Meyer) to build the K7 (of which we've really only seen AMD field micro-iterations of, add on IMC, 64-bit, etc...no new *major* architecture ever since).

Then they bought ATI, and their next product vision is to bolt ATI's architecture onto the existing K10 (K7 derivative) cpu architecture. Nothing has really changed, Bulldozer will be the first truly in-house home-grown CPU microarchitecture from AMD since the K5, they've had oodles of time to "get it right" so let's hope it isn't another Millennium project.

Also AMD stated in the last cc that BD will not be 1H2011, it will be shipping to customers after 1H2011, and the first 32nm product will be Q4/2010 which will be 32nm deneb-shrink + GPU monolithic fusion deal. (there are plenty of articles on this already, I won't dig up the links, just google if you hadn't seen them yet)

You're not really giving a lot of credit to the other hundreds of AMDers that worked on K7 who weren't part of the DEC hiring. I mean, if we're going to go down that road, you can pretty much cut out credit to anyone that worked for a particular company before joining AMD. Lots of Intel folks joined AMD, lots of SUN folks joined AMD.

True I'm not explicitly going out of my way to give them credit but don't think for a moment that I don't think there is credit to be given there...obviously my post was meant to be a "big picture" generalization viewpoint which inherently is going to lack the kind of granularity of details necessary to attribute kudos at the headcount level.

Just looking at the trend, a multi-decade one at that, and saying first we must convincingly discount the trend's trajectory before giving credence to prospective alternative futures of AMD that won't merely be evolutions of its past.

At the moment I haven't seen anything in the big picture to materially suggest the fundamentals that created the trend have changed...corporate culture's are like that, regimes change but if you left AMD 10yrs ago and returned today I bet you'd recognize a lot of the old work environment and corporate culture that made/makes AMD be AMD and not Intel or IBM. There is a slow evolution of the culture of course, but old habits are hard to kick.

This is kinda self-evident though, I would think, for example we generally expect different things of Intel's Hafia team versus their Portland team. We know the sub-culture of those design teams are different in ways that permeates the resultant product. If you can recognize and accept the presence of these differences that exist within Intel then surely you can see how the history of AMD's architecture is a thumbprint of the interjection of the differing teams that occurred over time at AMD as well.

AMD's successes to date can be in large part attributed to injections of talent and IP from what was external resources at one point in time, from Nex-gen to DEC to ATI. I don't think this is a bad thing, clearly AMD are finding efficiencies in this model that enable them to remain performance competitive despite the gap in R&D resources they have to their competition.

I mention this (big picture trend of microarchitecture inspirations and sources) only because I feel it is relevant, naturally.

We don't expect 10GHz quad-thread/core 500W behemoth architectures from Hafia for a reason, and likewise when I personally look to AMD and ask myself "given the history here, what can I reasonably expect of this team for BD and what can I reasonably rule out?"...well I don't know what to expect of AMD's BD design team, just as is the case with the Hafia team, but them having lots of time to develop BD is not a metric that elevates my confidence in their forthcoming product, long development timelines in this industry have more tended to be the harbinger of bad things not good things.

since we're giving AMD shout outs you should not forget about AMD friends at IBM that supplied the green company with all things FAB and HP which contributed a lot to the CPU itself.

Nobody knows what Bulldozer will look like. AMD didn't provide any public info on the topic. Thinking like "truly in-house home-grown", "brand new", "excellent, "SMT/CMT" etc are simly speculations.

Why wouldn't I?