No matter how hard I try, I am unable to go through an AMD earnings report without holding some puts. Must be the gambler in me. My order went through today for a (very) small number of Jan09 $10 puts at $1.05 apiece. If the stock gets above $15, I am looking to double down.

Is it possible that the "channel checks" that these analysists supposedly make aren't able to ascertain much detail? ie, when they do these "checks" and get feedback that inventories are up, could it be that they are simply assuming that Intel is impacted just as much as AMD?

When I pair Intel's spectacular earnings report with all of the separate warnings about trouble in the channel, I'm left wondering if AMD's Q3 report is going to be especially bad.

The only thing that gives me pause is that AMD never warned. Still, I am probably going to pull the trigger on puts tomorrow or Thursday if the price tops $14.50.

They wouldn't have done this without good reason.

When Intel was requesting an extension a week or two ago, I read in the one of the numerous news articles (can't remember which one, and I can't find it now), that Intel was basing its request for an extension on the fact that the EU Appeals court ruled against Microsoft only 1 month earlier. The writer of that article said that some of the language in that court's decision went beyond just restricting monopolies, and put some unexpected burdens on "dominant suppliers" (or something like that). Supposedly Intel's lawyers wanted more time to frame their response in a way that took into account the EU's thinking in that decision. Makes sense to me, anyway. Especially if Intel was planning to argue that it isn't a monopoly (in which case, the EU's response is apparently, "So? You're successful so you must suffer").

Personally, I have a really bad feeling about this whole thing. The EU looks like it's on a witch hunt, and wants blood. Big government run amok, as far as I'm concerned...

Thanks for the pointer, Saturn.

Since Intel hasn't commented on the Nehalem floorplan at all yet, then I think it's probably inappropriate to comment further. I will say you are right to be skeptical; there are inaccuracies there in Hans' analysis (more like guess, I think).

To address one of your guesses at an error, I can say that I am honestly not sure if the L2 cache is inside or outside the core. So I will speculate there. You will notice that Hans says the core size has increased in size from 22 mm2 to 29.6 mm2. It seems more logical to me that this is due to the L2 being inside the core. But like I said, I don't know that for sure.

My little teaser comment about there being "clues" in the floorplan was meant to be taken at a very high level, and point out the significance of lining up the cores in a row like this. What I always thought was interesting was how many boxes you could draw around various subsets of that die, and have all the necessary ingredients for a product in a particular market segment. I don't think that Intel is throwing around the term 'scalability' as just an empty buzzword these days...

The way I've always internalized tape-in is that it's the point where all design work is frozen, and the design team itself is basically done. From there, a small central team runs all the fullchip verification checks, fractures the database, and creates the masks, which of course is tapeout.

I suspect Intel started doing this because so many of the team members would focus on the tapeout goal, and it must have got tiresome for management to always remind everybody that they actually need to have their block wrapped up a couple weeks before that tapeout goal. So during the design process, tape-in is what is emphasized to the ground troops to meet the project goals.

I think it's just an Intel thing, and is just an internal milestone that is more meaningful to most designers than tapeout (over which they generally have little control).

In my experience, when it comes to milestones, Intel typically sets an external commit target, and also an earlier internal target that they strive for as a stretch goal. When it comes to getting a part production-worthy after tapeout, the external commit target usually corresponds to somewhere around a C-1 stepping, while the internal target usually corresponds to the timeline of a B-0 or B-1 stepping (ie the B0 stepping will tapeout with all known problems fixed, and the design team will start right away on further timing fixes and whatnot for C0. If B0 comes back and has no further problems, then it can get qualified for production, otherwise C0 will be the production candidate). This is why you sometimes see product pull-ins from Intel, because they didn't need to iterate as much as their contingency plans called for.

I think the fact that Penryn was pulled in from Q108 tells us that things have gone just fine with regards to this project. I would guess they are falling right between their internal and external targets. It's easy to get spoiled when we see a lot of B0 parts go into production, and therefore have their release date pulled in. But I imagine that Penryn's new power capabilities require a lot more platform validation than usual.

The observant reader will notice the contrast in project management between Intel and AMD. A promised release date of Q2 for Barcelona implied that an A-dash stepping would be production-worthy, which is just complete and utter fantasy. Most of the regular folks here had this pegged from the start...

At the time of Intel's Q2 earnings conference, Nehalem had just taped out. I'm sure that's where much of Paul's excitement stemmed from. Maybe not a big deal to the average joe, but anytime the initial design is wrapped up on a 3-year project involving hundreds of engineers, and it's right on schedule, it's kind of a big deal for those involved.

At the very least, Paul knew he could show off wafers of both the new architecture on 45nm, as well as the test wafers from 32nm. I think the demo was a nice bonus.

(Incidentally, I am hearing scuttlebutt that there may have been some internal debates within Intel this last week whether to even do a Nehalem demo at this IDF. Not because of anything that is wrong with Nehalem, but because this IDF is really Penryn's coming-out party, and now that Barcelona details are known many Intel folks aren't as worried about Penryn's ability to compete. So the question may have been asked whether Intel should put all its focus and PR behind Penryn at the moment. I predict we won't see any more details about Nehalem until Spring IDF because of Penryn's positive competitive position).

It is not clear if the 4 core chip has 1 or 2 CSI channels[20 bit wide]. I think that the chip has two independent memory paths,one on the right and one on the left.

Performance Scalability: Eight core chip will be obtained by mirroring the chip vertically and doubling the CSI and memory channels.Two core chip will be obtained by cutting it in half, with half the memory buffers, and half the CSI resources. Obviously the control logic in the central vertical path cannot be neatly sliced, and most of it will be retained. So external memory and interconnect bandwidth scale with the number of cores. So performance scales linearly with number of cores.

Nice analysis. Very nice indeed...

Pat Gelsinger demo'd a 2S/8C/16T Nehalem system, with all 16 threads active. Picture is at the bottom of this story:

http://www.theinquirer.net/?article=42440

(it does look like one thread is having a little trouble, LOL).

Just in case it hasn't sunk in for anybody yet: This is A0 silicon, with an A0 chipset, with the very first implementation of a brand-new bus architecture. I know Intel is famous for quick and successful A0 boots, but this really is an outstanding achievement all around...

Beamer, I'll just say this. Notice that whenever Intel mentions Nehalem, they include the term 'scalability' in the same breath. The layout of the cores plays a big part in the development of the complete product line, however as Intel isn't saying anything yet then I probably shouldn't either. Keep staring at that die, though; there are clues there for the observant...

Yeah, Hinton isn't much of a speaker (fairly horrible, in fact), but he's a very smart guy and very well respected within Intel.

Interesting factoid - Pentium 4's multi-threading capability was Glenn Hinton's baby. He advanced the initial proposal, and it was evaluated after Willamette had already started development. After it was elevated to POR for the project, the design team was fiercely discouraged from referring to the feature as "MT", because management didn't want to risk even a hint of this feature getting out into the world. Instead, the official acronym was "GH2", referring of course to Glenn Hinton. This feature-naming problem caused Intel to later standardize around code-names for features, which is where we get all those *T acronyms (the most famous would probably be YT, or Yamhill Technology, which referred to the 64-bitness of Prescott).

OK, so maybe it's a not-so-interesting factoid... (shrug)

From a live blog at this link:

http://blogs.business2.com/utilitybelt/2007/09/live-paul-otell.html

Glenn Hinton, Nehalem Architecht, is now on stage. "Our team in Oregon has been working on Nehalem for a long time, and it's great to see it finally here," he says.

Hinton says Nehalem was designed to power scalable, multi-core systems, but also to deal with power efficiency. He's getting very geeky right now about the chip's capability, and he's clearly a bit nervous. He talks about the ability for the chip to communicate quickly.

He walks Paul over to a computer that says, "Hello Paul; I am Nehalem. I am only three weeks old, and I am already talking." He says that this computer is running XP, but that the team has already gotten it to boot Mac OS X.

I'm not so quick to judge the Fool harshly because of what they say about AMD, or any company for that matter. I subscribed to one of their newsletters for a year, and I made alot more in that one year than what the subscription cost. The Fool really stresses that no one person speaks for anybody else. There is no 'company line', and you will often see different 'TMF' folks having polar opposite opinions about any particular company.

The problem as it pertains to AMD is that the Fool seems to have employed a full-on 'droid in Anders Bylund. This guy is a die-hard AMD stockholder, and is always quick to paint his comments as favorable to AMD as possible, even as his AMD stock has plummeted in value over the last year. Every time I see a Fool article related to AMD linked on a news page, the auther is Anders Bylund. Always. I believe this guy also played a big role in the Fool naming AMD as one of the top tech companies of 2007. Yeah, great call there.

The guy is a fanboi who happened to fall into the dream gig of getting paid to write positive drivel about his favorite company. At least some of the stalwarts over on the SI AMD thread can admit that things will be tough for AMD for the foreseeable future, but Bylund is always sure that a turnaround is imminent and just around the corner. Complete ignorance and idiocy...

Yeah, these pre-arranged selling plans seem to be all the rage these days with company execs. I see them all the time in other companies I follow.

One point of interest: The $19.00 options have an expiration date of 4/14/08. I'm guessing we'll see more blocks of these particular options exercised at regular intervals between now and then, both from this exec as well as others.

The days of exercising options for $3 a share are over; now the big money will only be made if we get substantial stock appreciation from here on out. Fine by me! That's how stock options are supposed to work.

(Oh, and as far as I'm concerned Dadi Perlmutter deserves every penny he gets, and more. He leads the design team in Israel and fought hard to drive the parallel engineering effort focusing on power efficiency back in the P4 days. I shudder to think where Intel would be without him).

Intel has put their new TV ads up on youtube:

Looks like everyone on the SI AMD board has obtained their law degree and are convinced that offering rebates based on percentage of units is blatantly illegal.

I suppose Intel has no choice but to shut its doors in the face of such a highly regarded and unanimous legal decision...

I took advantage of the recent market action and sold all of my AMD puts today when the stock was right at $14.

The Jan08 $12.5 puts I bought 3+ months ago just prior to the Q1 earnings announcement when the stock was around $14.50, I sold for about a 10% gain. Just last week, however, I was down 45% on these suckers, so I'll take my meager winnings and be happy.

The Jan08 $15 puts I bought a couple weeks ago, a few days before the Q2 earnings when the stock was around $15.50, I sold for a 50% gain. My double-down worked...

My investment goal here was to hit a home-run when the stock dropped precipitously (was hoping for <$10), but it is clear that too many people and institutions are still true believers in the AMD faith that this probably won't happen for another 2-3 quarters.

It's possible AMD could go lower (especially in this market), but the chart over the last 4 months is showing a clear up-trend, with local bottoms at $12.6, $13, $13.3, and today $13.9. Support at $14 was looking pretty solid today, so I pulled the trigger. If the market continues tanking, at least I have some fresh gunpowder to add to a few other positions in my portfolio.

I will be looking for the next upswing past $16 to open a position in Jan09 puts. Hopefully before September so I can get in before any potential Q3 warnings (but I've promised myself I'm not going to chase this thing...the stock just has too much irrational momentum behind it).

I'm still very, very long INTC. Good luck to all..

Whaddaya know, a Barc derivative can actually run something besides Task Manager. And apparently it will 'change the game'. Link was posted on SI AMD thread:

http://www.hothardware.com/Articles/AMD_Consumer_Electronic_Event_in_NYC/

Perhaps the most interesting item on display - at least to us anyway - was AMD's Phenom powered demo system. We weren't able to ascertain the clock speed or run any benchmarks, but we were still able to glean some good information about the system and CPU. The Phenom processor in this rig is installed in an AMD reference RD790 chipset-based motherboard. This motherboard features an AM2+ socket (support for higher HT link speeds and split power planes) and true, dual PCI Express x16 graphics slots. Affixed to the processor was what looked like your standard AMD PIB heatsink with an 80mm lighted fan. Throughout the entire demo, we never heard the fan spin up and it remained very quiet throughout. It wasn't throwing off a lot of heat either, so whatever the clock speed this Phenom processor was running at, it couldn't have been running all that hot.

To showcase the machine, AMD was running a beta version of John Woo's Stranglehold game, featuring Chow Yun Fat; well his likeness anyway. The game uses a modified version of the Unreal 3.0 engine in coperation with the Havok physics system. The engine is multi-threaded and designed to take advantage of multiple processor cores. As you can see in one of the pics above, all four of the cores are being utilized while the game is running. During the demo, we noticed on average about 70% of the processor's resources were being utilized with higher (or lower) spikes depending on how many objects were on screen. We say objects because most of the items in the game world, like signs, widows, propane tanks, and almost everything else for that matter, can be manipulated or damaged. Shoot a propane tank, for example, and it'll explode causing a ton of damage and throwing debris in all directions. After watching the game for a while it looked like a cross between Kingpin and Max Payne, with much better graphics than either. It had plenty of blood and gore, like Kingpin, with the third-person perspective and slo-mo bullet-time of Max Payne. Although they call it 'Tequila Time' in this game.

Over the course of the demo, representatives from AMD said a top-to-bottom line-up of desktop Phenom processors, priced competitively in the marketplace, will be available in the fourth quarter of this year. Motherboards based on the RD790-chipset will arrive sooner, from AMD / ATI's traditional board partners. Although AMD was intentionally tight lipped and wouldn't disclose specific details regarding the processor and chipset just yet, reps seemed confident in Phenom and very excided about the RD790, especially in its overclocking capabilities. One AMD rep even went so far as to say the Phenom / RD790 platform is going to "change the game". We should know that means in the next few months, so stay tuned.

It's obvious I doubled down on my Q1 puts too soon. I figured the massive ASP dropoff that you and Buggi have been witnessing in your price checks would be devastating and overcome any unit increase, but I never foresaw a mammoth 38% unit increase. This is undoubtedly what is moving the stock AH.

I will now keep a close eye on any price weakness and look to get out of all my AMD puts, and then look for a good entry into Jan'09 puts. The short-term positives of this quarter are masking the devastating long-term prospects of this company, IMO.

The good news is that my entire put investment is still just a portion of the profits I made last summer and fall riding AMD down, so all of this is just playing with house money for me...

Nice investigative work, Alan. Those items combined with the $500M they burned through last quarter would indeed add up to about a billion dollars spent during the quarter.

It just boggles my mind that anybody could burn through 1 billion dollars in about 2 months. I am truly in awe of this company.

So they took out a huge $2.2 billion loan, and all it bought them was a measly 6 months? That can't be right, can it?

Or can it?

This is what sticks out to me:

 
                                    June 30,    March 31,  December 31, 
                                       2007        2007       2006(a) 
                                   (Unaudited) (Unaudited) 
 ---------------------------------------------------------------------- 
   Cash, cash equivalents and 
    marketable securities          $     1,594 $     1,167 $      1,541 
  
  
 Long-term debt and capital lease 
  obligations, less current 
  portion                                5,318       3,659        3,672 
  
 

Note that Intel is claiming some more one-time tax benefits that are adding 3 cents per share to earnings. If you back this out, then Intel hit the consensus earnings estimates on the nose. So revenue was higher, but margins lower, with the end result being right where everybody was expecting.

I think alot of folks were expecting a mild upside surprise, which is why the stock shot past $26, and just meeting estimates is probably why it will fall right back down below it.

Boy, it will sure be nice to get rid of that Flash albatross...

Gross Margin only 46.9%, primarily due to NOR flash. That could hurt, although Q3 guidance is 52% and full-year guidance is still unchanged at 51% (implying a pretty strong Q4).

No clue what the stock will do (probably go down, if history is any indication).

Actually it's 17.6%. 15/85 not 15/100.

Not only that, but if it's rewarded every 6 months, the annual yield on your investment is actually double that. And if you really want to get fancy about the APR yield, you can consider that each paycheck deduction has a smaller amount of time before that 17.6% can be realized, increasing the annual yield of each deduction. That last paycheck deduction will be worth 17.6% more a mere 2 weeks later, for an obscene annualized yield. Somebody I knew once calculated the blended annual yield over the 6-month period assuming you got paid twice monthly, and it came out to something like a ~60% return annualized.

And that is worst-case, ie if the stock is flat or goes down. If the stock rises even a little bit, the yield numbers get obscene very quickly. I remember once or twice in the late 90s Intel stock doubled within the time period of the employee stock purchase plan, and I think the blended APR yield on my SPP shares came out to something like 350 or 400%.

Granted, those numbers don't compound, since you start over at the next period, but there is simply no better short-term investment for your money if you have access to such a stock purcase plan. For me, there was never a question of whether to participate, but rather what to do with the stock once I bought it. One thing I quickly learned was to never sell the shares acquired in August; if I wanted to sell for a big purchase or just to diversify, I would always wait at least until Q3 and Q4 were in the books. That doesn't seem to have much significance the last 5 years or so, but during the high-growth days there were some massive movements towards the end of the year.

I am somewhat hopeful that we might see a similar run-up this fall...

I just doubled down on my AMD Jan08 puts. I still have the position I bought just before Q1 earnings were announced when AMD was at ~$14.40, and those have fluctuated between +/- 20% gain/loss over the last 3 months. With today's jump, those are now down 37% from my purchase price, and I couldn't resist putting a little more coin on the ride back down.

This will be purely a short-term play, and I will lighten my position when AMD gets back down closer to ~$14. In fact, if there is no significant downside after they report Q2, I will be looking to unload my entire AMD put position and roll it into Jan09 puts. I have no doubt AMD will see single digits, and I intend to profit from that, but the market has shown a lot of faith in AMD, and it's looking like it will take a couple more quarters for the last institutional holdouts to see the writing on the wall...

The resistance level has got to be tremendous.

I believe tecate is already on record here as saying that she would sell a big chunk of her shares at $25. I have also had this same thought, and may have even expressed this sentiment on these boards. Sarmad sold 15% of his holdings to clear out his margin shares.

I can picture bunches of people who were watching their Intel shares languish at $18, convinced that the Intel growth story is over, and thinking, "If only it can get to $25 again, I can dump this POS stock." I know that thought crossed my mind once or twice the last couple years. I don't think we will see anything north of $25 unless a significant upside surprise is announced by Intel.

Me, I'm holding, and not selling any shares. Intel began a new chapter in its history one year ago with the release of benchmarks on Conroe, and I believe this chapter will last until at least mid-2009 when Nehalem gets fully ramped up. We have only completed 1 year of this 3-year story, and I'm sticking around to see how this particular story ends, regardless of the share price in the interim. I am hopeful we will all be well rewarded by then...

Re: AMD/ATI HDTV/DVR product.

This product will increase the demand for all PCs.

Nah, there's not much new here. Note the key word in this sentence:

"Extending the ATI TV Theater family, the ATI TV Wonder 650 Combo PCIe features ClearQAM tuning, allowing PC users to receive unencrypted digital content available from their local cable television provider."

The only signals that are unencrypted in the typical cable package are the local channels, and those HDTV broadcasts can already be received over-the-air using a UHF antenna. My media PC already has an HDTV tuner that allows me to receive all OTA HDTV broadcasts in my area.

I've owned a media PC for almost 2 years now, and have yet to join the HDTV age, even though I have a 55" HD-capable TV. I've resigned myself to a lack of HDTV premium content until a satellite provider finally makes one of their receivers as a PC add-in card (DirecTV has been hinting at this for some time, but nothing has come of it), or until the cable industry relaxes the restrictions on the Cable Card to the point where it becomes usable.

It's not all bad, since my DVD software upconverts to a higher resolution and I get a beautiful picture when watching movies. But when it comes to watching TV, media PC users are pretty much stuck with standard 4:3 low-res pictures (not that I watch much TV anyway, and besides, recording HDTV can take up much more hard drive space. Yes, I'm rationalizing. ;p)

But assuming the X-step came following all the previous letters of the alphabet means that Intel not only had at least 24 full mask steppings, but probably a lot more metal layer steppings.

wbmw, did you misread my post? I couldn't tell from the smiley in your reply if you were jokingly agreeing with me, or were trying to correct me on something (even though you and I are saying the same things).

A few folks expressed confusion at the jump in alphabet of Conroe steppings, so I made my post to try and explain why a jump in alphabet letters does NOT mean that there were a bunch of steppings in between, but rather one big one. So nobody is assuming that an X-step came following all the previous letters. In fact I said exactly the oppositte.

I mentioned that Prescott had major design steppings at A0, B0, F0, K0, and M0, and then said that Northwood might have gone all the way up to X0 (meaning that they could have done a couple more big steppings like an R0 and then X0, for a total of 6 or 7 major broad-based design steppings....sorry if I didn't make that clear).

And I agree with chipguy that even though there is no way to compare such things, I have to believe that Northwood must have been the most heavily-tuned processor in history. The extremes those guys went to squeeze out just a few more picoseconds towards the end was mind-boggling, and never would have met any kind of normal engineering ROI requirement if Intel's enormous cash cow didn't depend on it. Let's hope they are never faced with that situation again....

Whenever you see a gap in stepping designations, this is a reflection of the parallel design effort that is going on between a simulation-based design effort and a silicon-based design effort.

Imagine you just taped out a B-step part which you expect will be the first production-worthy design, and silicon won't be back for several weeks. Usually the design team gets right to work on the next revision that is based on all the same timing and power models that they've been using. Maybe they'll try to shave a bit more off all the max paths, or do some re-sizing to improve leakage (or in the case of Prescott, maybe they will implement much better clock-gating than was originally designed ;). These design steppings usually take 3-6 months, and are a fairly broad-based effort involving a big chunk of the design team.

Now imagine that the B-step part comes back from the fab, and some showstopper problem is found with it in the silicon-debug labs. To minimize time-to-market, a new stepping will be done in parallel to this major stepping that has only those changes necessary to enable production. This will be called a C-step, and if God-forbid something is found wrong with THAT, then you might need a D-step. In order to allow for this possibility and not have naming confusion with the fab guys, the major stepping that the design team is working on will skip several letters and be called F-step or G-step from the very start.

My recollection on Prescott was that they went from A0 to B0 to F0 to K0 to M0 (Prescott needed alot of work ;). Now that I think about it, I think Northwood may have gotten all the way up to an X-step before they were EOL'd (they really squeezed every ounce of juice from that puppy to make up for the Prescott delays).

Bottom line, you cannot imply how many revisions have been done based on a stepping designation. But what you can be sure of is that if a product is being updated from B-x to G0, then there is likely alot of goodness that's been added to the entire design in terms of frequency, power, health, etc.

Found this link on SI:

http://www.uberpulse.com/us/2007/05/intel_responds_to_amd_barcelona_has_a_problem_you_better_fix_it_...

Intel's 8 core faster than AMD's 16 core/Barcelona!

At the Microprocessor Forum, Intel demoed its V8 workstation running the POV-Ray benchmark. The machine is equipped with Intel's Workstation Board S5000XVN, 2 quad-core Xeons 5365, clocked at 3GHz and 16GB RAM. And the results are simply impressive: Intel scored over 4,900 pixels per seconds versus a little bit over 4,000 for AMD's 4 sockets quad-core (Barcelona) system. Again, this is an AMD 16 cores system versus Intel's 8 cores V8 machine.

Yeah, yeah, questions remain about clock frequency and thermal limits, but still... kind of a nice poke in AMD's eye, IMO...

I just don't think there's enough evidence to conclude either way.

Common sense tells me that Intel probably needed to sweeten the pot a bit and likely included their phase-change IP. The whole reason this sale has taken so long to happen is that it's hard to find a buyer for a consistent money-loser. One thing that would make this an attractive investment is the hope of new technology coming down the pipe.

Even if the new JV has the phase-change IP, what's to stop them from signing a cross-license agreement with Intel, with both companies continuing to do research in parallel from their common baseline today? Seems like that would make sense for both companies.

That Anand kid has got a head on his shoulders, and does a bit of critical thinking. It's a shame this trait is in such short supply when it comes to AMD.

As a side note, you may remember that AMD seemed particularly cautious about sharing too much about its plans for Barcelona/Phenom. AMD stated that it didn't want to clue Intel in on what it had up its proverbial sleeve. However, with Griffin not scheduled for release as late as the middle of 2008, we can't help but wonder if all arms of AMD believe the same things, or if AMD is simply giving us excuses for why we don't get more information today.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2992&p=6

Very interesting marketing tactic by Intel. And very smart, I think.

A game developer is building up hype for its product's release, and invited about 25 journalists and fan site webmasters to their corporate headquarters for a sneak peek at the game. This part of the report on the visit caught my eye:

Wondering about the swag? Get this. In addition to a custom Community Day t-shirt, and a selection of posters, everyone who attended received, courtesy of Intel, a brand new quad-core processor and matching motherboard. I don't know the model numbers, but the Intel guy told us they were brand new, and not even on the market yet. We even got nifty Community Day name tags, displayed in an collectible EA Games lanyard!

http://hellgate.incgamers.com/forums/showthread.php?p=12079

Ya gotta like how Intel intends to capture and retain the hearts and minds of computer enthusiasts.

Some interesting observations on AMD pricing from SI:

http://www.siliconinvestor.com/readmsg.aspx?msgid=23551087

http://www.siliconinvestor.com/readmsg.aspx?msgid=23551491

Coupled with the rumors about increased orders from Dell, my guess is that we are going to see a reverse in unit/ASP action compared to last quarter. In Q1, management said that ASPs stayed fairly steady, but units got wolloped in a huge way (down 37% according to their SEC filing).

My expectation for Q2 is somewhat higher units, but with a significantly lower ASP. Whether the increased volume is enough to make up for these lower prices is the big question. AMD guided for flat-to-slightly-higher revenue in Q2 compared to Q1, but I am still betting that they will miss this and warn. Hitting this revenue target would mean that they overcame much lower prices, strong Intel momentum, and the poor seasonality of Q2. I think that is way too much to ask, and I expect the losses to deepen compared to Q1.

My puts are down 24% from when I bought them just before the Q1 earnings announcement (after being up 30% last week), but I'm holding on to them through at least the quarterly report.

OK, I wasn't going to put much stock in this "B0" nonsense, but if we have two supposedly independent rumors about this (I know, that's a big 'if'), then I start to wonder some things.

First of all, the reality is that it will take a minimum of 6 months before any new silicon stepping out of a fab will find itself in consumers' hands. There is a ton of validation to do at the electrical level, the pin level, as well as the system level, and customers aren't going to ship anything until they've had some time to qualify some production samples. If this new material came back from the fab in early April, we shouldn't expect to see it in quantity until September at the earliest.

Does this mean that Barc won't launch until September? Not necessarily.

My first thought upon reading this rumor is how odd that this would be called a B0 stepping considering the timing. Didn't they get first A0 silicon back around August, over 8 months ago? Surely they must have done a full layer stepping well before now? I seem to remember a Register article where they mapped out timelines and surmised that the infamous task manager demo in December was done on either an A-2 or a B0 stepping, implying a couple cycles of debug and new steppings.

I believe there are 2 possibilities here:

1) This really was the first full-layer stepping since they got A-step silicon back last year. If this is the case, then you can bet that they've got nothing production-worthy ready to go, as the health of that A-step silicon must have really been suffering if they took so long fix any issues. So if this rumor is true, I peg the Barc launch date no sooner than September.

2) The other possibility is that the source of this rumor is mistaken about what stepping this really is, and this is actually a C0 or something like that. If this is the case, my first contention would be that we don't know how many other details this source might have gotten wrong (ie, this might have been a skew lot with high leakage, or one part out of many that is set aside for the debug lab, etc). But if we give them the benefit of the doubt and surmise that everything else about the rumor is true except for which stepping it is, then it's possible that there could be a relatively healthy B-step that is going through the production validation and qualification now. This could mean a launch prior to September at slower speeds, with a nice speed boost in Q4.

If I were an AMD fanboi, I would hope that this rumor is false in some way, because if it's true that they just now have healthy production-worthy silicon, then this provides verification that AMD will continue bleeding money through both Q2 and Q3 (much like we've all presumed). If anything, the lack of any actual Barc test systems or benchmarks would support the theory that they only just now have something that is healthy enough to produce. So I actually wouldn't be surprised if this rumor was completely true. What's funny is that the AMD folks are totally misreading this as good news, when in reality having silicon that is finally healthy a whole 8 months after tapeout is a monumental failure. Think about it....this rumor would imply a level of silicon health in April that Penryn already achieved back in January.

If this rumor is true, hope may spring eternal about the better-than-expected performance, but AMD will need to resolve themselves to another billion dollars in losses before Barc can even *begin* to ramp in Q4. And my hunch is, if there really are 8 months' worth of design changes in this stepping, that more issues may be lingering under the initial onion they have peeled. I wouldn't be surprised to see another dash stepping or two before production, pushing a launch well into Q4 or even later.

That's my doom-n-gloom scenario. For AMD's sake, this rumor had better be false...

I've always thought that Intel and AMD know what the performance is going to be long before the part is issued.

You got that right. High-level performance simulations are done very early in the design, as in 3 or more years ago. Typically, 90% of the RTL is completed about a year before tapeout, and then the architects can start running performance traces to see if actual peformance correlates with their simulation models. AMD had to know at least 1.5 or maybe even 2 years ago exactly what sort of performance they are getting with Barcelona. And you're right, there is little that anybody can do to significantly improve on a CPU design that is less than a year from tapeout.

On the other hand, there is something to be said for maintaining the fog of war for as long as possible so as to cloud any future planning that your competitor might be doing. For example, I could envision AMD frantically updating their future project goals and timelines in February of 2006 when Intel released benchmarks for Conroe at IDF. This effectively gave AMD an extra 6 months to plan and start their long-term response compared to what might have been had Intel waited until Conroe's launch in June/July. But that is an effect that won't be felt for 3 or 4 years down the line, and given the dire straits Intel was in within the marketplace, it was absolutely the right thing to do.

Just like it'd be the right thing to do for AMD if they had benchmarks showing a clear leadership with Barcelona. Then again, if their benchmarks don't show clear leadership, or heaven forbid they don't even have actual benchmarks yet, then the right thing to do would be to shut the hell up. AMD management may be sleezy and deceitful, but they aren't dumb. I really believe they are doing what is in the best interests of their company right now, namely delaying what will likely be very mediocre results until the last possible moment. They already know they can provide excuses or future promises to improve from there, and the market will eat it up. They always do...

Ya know, I understand that Intel needed to at least give some insight into some of their future plans, and I'm sure their presentation on Silverthorn was quite good.

But I think they should just shut up about it now and go out and get design wins, build relationships, and basically get some results in this space before they talk any more about it.

The analysts are completely justified in being skeptical, as the hype and promises of Intel playing in the mobile CE space has been unending for about the last 10 years now, and their execution has been horrible. Promises won't mean anything, we need to see Intel silicon in some real devices with real applications that get consumers excited. If they're going to sell anything in the volumes they need to be profitable, the positive buzz needs to be deafening. And that is not going to happen even if they do 1,000 slideshows...

I think Silverthorn and it's follow-ons have great potential that are cause for optimism, but if Intel really is finally serious about this market, they need to stop the talk and start walking the walk. Until then, I will file this under future R&D that may or may not pan out.

Personally, I think they could do alot worse than simply giving about a million Silverthorns to Apple for free and say, "Here, see what you can do with it..." If anybody knows how to get a new market off the ground, it's those guys...

Hector is in fine form at today's AMD shareholder's meeting:

Another shareholder said that the conference call AMD did in December painted a very rosy picture, but Intel had been executing much better since.

He wanted to know why AMD is being so quiet about new products it might have.

Ruiz said that nothing AMD said in December had changed. He said AMD did have a lull of activity because the next generation begins with Barcelona in the second half of this year.

http://www.theinquirer.net/default.aspx?article=39385

I wouldn't be surprised if this new debt/stock issuance causes significant downward pressure on the stock price. Such an unexciting move (not to mention expensive and dilutive) shows they are going to be pretty much left to their own devices to get out of this mess, and should put to rest most of the wild rumors of some white knight swooping in with boatloads of cash and rescuing them from their predicament.

(For some reason I suddenly have an image of Richard Gere driving up and rescuing Julia Roberts, but unfortunately for AMD they are NOT a pretty woman...)