If they exit server/ws and most of the desktop business in favor of mobile clients and LP core-cluster servers (as I'm predicting) there will be a huge howl from the fanbases at S|A and AMDZone.

November 9 (my birthday, btw) will be pretty interesting. I wonder if a lot of the presentation will be shades of grey or black/white in terms of direction.

gb

I wonder if they'll redirect any gfx activity more toward low power and integration and less toward high performance clients i.e. discrete/add in boards?

gb

Now that would make for some news!

gb

I haven't seen any analysts weigh in yet. Perhaps they're waiting to hear "Project Win" details next week.

gb

If you listen to RR he does sound like a bit of a braggadacio. Lot's of "I" in every statement he made on the Q3CC.

gb

IIRC, the servers (blades was my term) only had 1G of memory so perhaps the ECC and capacity issues aren't really significant. That would also match with my lowend disposable "IP modem" theory.

gb

If focused on shared nothing front end disposable blades it might not require much at all.

gb

In the Q2cc AMD replied to a specific server profitability question with "marginally". That was before the BD launch. It no doubt has degraded from there with the loss of share in spite of presumably aggressive pricing to chase deals.

gb

AMD was listed as a "partner" on the HP press release on that ARM server. Perhaps AMD is considering giving up the high end in favor of LP cores which could be used on the HP server and also in low end clients in emerging markets and low end laptops/tablets.

gb

The press release also mentions LP, emerging markets and cloud. Theoretically all of their existing product lines fit under this. I wonder what "Project Win" will cut of their existing product lines if any?

gb

In the 8K they mention:

AMD 8K just filed says 1400 rather than 1200 employees:

http://biz.yahoo.com/e/111103/amd8-k.html

gb

Since RR will do a "Worldcast" next week, Nov 9, which I presume is for internal consumption, I suspect they will make it public as well somehow.

Leaks would occur anyway.

gb

Maybe they've publicly threating to walk on the GLF agreement since they're not getting progress on extension of their good die pricing?

If so that will get interesting as they would turn into an "ordinary customer" for a boutique process that it's not clear GLF has any other customers for.

The Brazos low end integrated product that TSMC builds seems to be percolating along in lowend mobile. It was likely the source of the share gains vs Intel in Q3. Not clear however that level of revenue and margin can carry the rest of the company if the good die pricing isn't extended and they have to start paying for the whole loaded cost of the SOI 32nm process debacle.

gb

I wonder how many bodies were buried in that Q3 report.

gb

I wonder how they'll hold up compared to Intel chips when the front ends have to start running https for all logins along with what ever other security is necessary?

gb

I'm guessing that these must be cheap disposable front facing web servers for big websites. Think of them as modern day modem banks.

gb

S|A seems to be a hangout for the AbI (anyone but intel) crowd to hang out and root for anyone else. Occasionally one will grudgingly admit that they have a SB system. They seem to have very little microarchitecture experience except as gained by reading websites. And even less semiconductor or general business sense.

Makes the forums pretty useless except for some early warning detection of the latest web rants.

Right now they seem to focus on graphics competition between AMD/Nvidia since there's little competitve CPU AMD news except for some dim hopes for better server benches from BD (curiously absent) and a "new, improved BD" sometime in the spring.

gb

where i live ice and snow "hole a road"

gb

HP and Intel gradually parting ways? I would think both have a lot to lose but perhaps HP moreso. They can scarcely afford to lose their Itanium supply which keeps them nominally in the big iron race. Nor can they afford to lose their likely preferred NDA status on x86 for the main portion of their server line in pursuit of a science experiment.

Seems rather silly.

Didn't their CTO just leave? I wonder if this is one of his grand experiments?

Where is the counter to Oracle's SW stranglehold?

gb

I think as long as the FCC allows cable companies to require a box to access most of the channels (rather than software) the TV will remain for the most part nothing more than a dumb monitor with ports. The current trend to include internet in the TV will likely fall victim to price pressures and the cable boxen will prevail in the provisioning of the 'net.

OCAP was supposed to make it possible to do it all in software but the FCC seems to have lost interest in requiring it to be available. In fact Comcast received a waiver to allow complete elimination of all analog channels and replacement by an encrypted rather than cable card equipped device so as to go all digital.

No wonder Intel gave up on the TV companies. They have no chance in the US vs the cable providers to provide anything computer based. Certainly not much more than an ARM chip provides. Access Netflix on the net via any recent computer then try it on a net equipped TV. Big difference in the interface response times.

That makes the Cable/cable box companies much better targets for compute power. It will be interesting to see if Intel is able to land one of them for more than a "technology demonstration."

gb

I'm not sure what East has to gain by shooting off his mouth. His stock is as high as it's going to get already. If I was him I'd quietly haul my money to the bank rather than taunting Intel. This guy has a personality disorder of some sort.

gb

Forgot to mention: we are line-of-sight to the TV towers 30 miles away and get all the OTA stations just fine here with just rabbit ears. So if we went streaming exclusively we would still get realtime broadcasts. We seem to be hooked on several cable offerings however so I doubt that would occur.

It's only money, maybe someday.

gb

We bought a new TV this year as our old one died. Most of the TVs above $1K have WiFi and/or ethernet internet access and access to Netflix, Pandora, Amazon, YouTube, etc. There's also a browser but it is pathetically slow. And keyboarding via left/right arrows and pressing Enter is a major PITA.

I also have a PC in the stack in a entertainment-equipment styled case. It has a Core i3-2100 and is a rocketship. The bootdrive is an Intel SSD and the data storage drive is a 2T SATA drive. Dead cold power off to Win7MC running in 20 seconds. This runs Win7 Media Center. I have slowly been putting all our CDs on the drive and it has a cablecard tuner capable of recording 4 channels at once. I use a Harmony One remote for it just like I use for the Sony, Denon or TiVo. When I want to enter a bunch of recording schedules I can use the MSFT entertainment BT keyboard which has touchpad and mouse capability built in. It's a whole lot faster to find and schedule shows with this. The CD drive is a combo which also acts as a BluRay player which is nice for watching movies.

The 1080P TV does a great job of displaying the Windows screen as a giant PC unlike the the TVs of old.

It's gotten to the point that we're going to take the TiVo out of the stack and give it to our daughter to use. That will leave me with the Win7MC HTPC and the Denon as the only pieces of gear other than the Sony. Gone is the 5CD changer since we can make huge playlists with a few clicks and without shuffling through a pile of unorganized CDs.

Until MSFT got the copy protect stuff working on Vista and Win7 CableLabs wouldn't approve of cable cards in the vast majority of PCs. Now it's a piece of cake to build one which is what I did from parts from EggHead. Even Comcast got the CableCard install correct on the first try...a miracle!

Win7MC is quite nice. Also since everything is on the harddrive and networked all of the content is accessible thoughout the house vis any Win7 PC that has network access.

gb

One more thing: in the transfer of the fabs to GLF I assume that all the fab knowledgable people were transferred as well. It would make no sense to have a process guy, particularly one with enough knowledge to suggest machine parameter changes, to still be on the AMD payroll.

gb

I don't think that degree of ownership would entitle them to wander around the fabs and mess with the machines.

Given the nature of Rory's responses on the CC I'm not impressed. The board spend 6 months looking for a CEO and this is the best they could do? Says a lot about the prospects for the company. Maybe he's just a placeholder while they shop the company around.

I would characterize him as "all hat and no cattle."

gb

Just hit me:

Did it strike anyone else as odd that Rory would claim that "we" are going through machine-by-machine. I thought AMD was a customer not a fab owner. I'd be surprised if AMD was permitted to run around in GLF's fab. Seems to me to add another thick layer to the BS factor of his execution blabbering.

I tried to listen to the CC replay and it was breaking up a lot. Dunno if it was recorded that way or it was an internet issue. I'll try again tonight.

gb

From reading the announcement it seems like that is aimed at the low power end of the process spectrum.

Perhaps the "club" is pulling too many directions and also perhaps not willing to make the design rule, layout and tool chain changes that Intel described.

Too many cooks...

gb

IIRC in order to mess with the ARM architecture you need a special license which includes giving back anything you do to ARM for potential future incorporation in the base. I'm fuzzy on the details.

Even with a common spec it's typical to have multiple incompatible implementations until the spec is refined enough to ensure that doesn't happen. If ARM is allowing multiple locations indpendently work on 64 bit there's no doubt there will be incompatibilities even if they're overseeing the lot of them.

And you really won't know that until they've all gone to market and the developers have beaten on them. Then it will take another pass by each (at least) to converge. That will add another couple of years.

If ARM ever gets to 64bits in a meaningful volume with performance to match even today's lowend Intel x86-64 it won't be until the latter part of the decade. There however will be grandstanding ppt pronouncements in abundance until then.

gb

Changing from SOI to bulk and doing a half step shrink of features isn't going to make anything better about gate-first. If anything I suspect it may be even worse.

Someone throw a few more shovels down that hole! That will help them dig faster!

gb

At this point you have to wonder what is going on at IBM. They have a bunch of people paying them money for a process at least as I understand it.

They're now woefully late on HK-MG, chose gate-first vs gate-last and are reversing course and likely don't have anything like tri-gate ready to go (understatement).

They have a steady stream of shiny-thing press releases from their research group but nothing seems to make it to production.

Either they're sandbagging and about to come out with something totally killer or they're just taking money for nothing. How much longer will their "club members" hang on?

Or perhaps the club members have no where else to turn?

gb

There's a pony in here somewhere, let's keep digging...

gb

Given some more thought to this unexpected quarter:

AMD has a cost umbrella in the form of the "only good die" agreement with GLF which expires in Q1. If the agreement gets extended in roughly its current form it means that GLF is admitting defeat on gate-first. If agreement doesn't get extended it could mean that gate-first is getting better or they've told AMD to pound sand. It will be interesting to see what transpires.

Of course some maintain that AMD is an anchor tenant which GLF can ill afford to upset. My take is that AMD has no where else to go and GLF may well tell them to live with what they can provide. The SOI process is not mainstream and doesn't seem to be attracting a lot of business otherwise there would likely be press releases trumpeting its success.

I frankly don't expect gate-first to ever get better. So AMD will be waiting for GLF to do gate-last at some node to compete with Intel's highend and midrange. That will leave AMD capable at only competing at the lowend and vulnerable to the ARM encroachment and the Intel response to same.

It's not that Intel will go out of its way to compete with AMD's lowend niche parts, it's just that those niches will get squeezed terribly between Intel and ARM and AMD will be very hard pressed to find a profitable spot to hang out. The ARMy will be blasting parts out from a variety of vendors at a furious pace and Intel will be responding with a more limited set of products but with unmatched process and manufacturing capacity and resultant pricing. AMD can neither match the proliferation of the ARMy or the process, manufacturing and pricing of Intel.

I think this will mostly hinge on the timing of Win8 which will really put this competition in focus. I'm guessing AMD is hoping Win8 is delayed until late 2013 or until GLF can get gate-last ramped to avoid being trampled.

So it will be interesting to see if AMD can continue to post profits or if they start bleeding while waiting for GLF to give them the weapons to fight. Of course they could decide to leave the battlefield but that doesn't seem to be in their genes.

gb

bingo!

gb

What a bunch of babble! That certainly wouldn't give me confidence as an investor.

It will be curious to see how the analysts and markets respond to this.

I never expected them to even break even this quarter. I wonder if they buried a few bodies somewhere.

gb

If that's the case I think they might have the fulcrum a bit too close to their knuckles, perhaps even on the wrong side!

gb

I'd really like to know what the heck they're targetting with these? Are they trying to make edge servers like the Sparc gaggle of idiot chips? They certainly aren't going after big iron.

Of course they'll run Linux and likely Apache or some such.

Seems like a total waste of money. Given their insider knowledge of Intel roadmaps is likely to be as good as it gets outside Intel I wonder why they're not seriously investing in competitive solutions to Oracle SW rather than some dumba** HW thing. There's much more at risk there.

gb

With the number of holes HP seems to be bent on digging it's starting to look like a prairie dog community.

gb

I wonder what's in store for Haswell? IB will help shake out the 22nm process but Haswell will be a new arch and specifically designed to take advantage of 22nm trigate.

gb

cf: Krispy Kreme

gb