The AMD Execution Thread [2007 - 2017]

That's not a problem, but I'd really like to keep the "AMD is doomed financially" discussions separate from the UVD discussion, because that just encourages trolling too much IMO.

If you want to start a thread on AMD's financial position and subpar execution, feel free as long as the original post is a good enough basis for a mature discussion! If you wish, I can move your above post to a new thread, since it's not a bad starting point (although arguably a bit harsh, and considering exclusively negatives, but I fully understand your point of view).

P.S.: Any PE or stockholder group that reads this and wants to get rid of AMD's current management, Beyond3D will gladly take on their roles with flawless execution and minimal salaries. We also won't ask for as many benefits as Hector got last year, promised! Pretty much just kidding, no hard feelings, but I wish the best of luck to AMD - they'll need it, I fear...

 

I believe overclocked_enthusiasm (oe) has brought up a subject very worthy of further discussion. I don't think oe's statements are overly harsh; I think they are just pointed comments that are being made by shareholders, partners and consumers alike.

At the risk of throwing another log on the fire, here is yet another issue that hasn't exactly been brought up: Brisbane, the next core following Windsor, should have brought many overclocker's back to the AMD fold. While they are not serious competitors to Core 2 in the pure performance arena, once overclocked they are contenders in the bang for the buck category.

What happened? Why aren't loads of people overclocking them and benefiting from lower prices and better performance? I own a Brisbane - it went from 2.1 Ghz to 2.9 Ghz at stock voltage. I paid ~$100 for it. Why aren't people singing its praises, and why am I, a long time fan of AMD processors not singing its praises?

The DTS is broken, so you have absolutely no idea how hot it is really getting! That limits the value as an overclocking processor. I thought it was the motherboard, but upon research, (and I had to dig deep to find this out) AMD stated that there was an issue with the digital thermal sensor. This is yet another misstep that hasn't been adressed, barely confirmed, or even discussed. Will it be fixed? Who knows? You have to recognize that there is a problem first before you can fix it. I don't know if many people at AMD even know that the DTS is not functioning correctly.

UVD happened in a similar way; "everyone" knew it wasn't included with R600, yet "everyone" let slide the misinformation without a correction. It turns out that not "everyone" really did know it was absent.

AMD, with the combined talents that ATI brings to the fold *should* be in a position for market leadership. It seems as though the left hand has no idea what the right hand is doing within the company, though. I believe this lack of focus and communication is what is causing these delays and confusion. It isn't helping AMD any, either to ignore these issues and pray that they go away; unfortunately, rather than addressing the problems with truth and a solution seeking mindset, many of their company spokesmen either deny they have problems or aren't aware of them. As I said, one has to admit/be aware that there is a problem before one can solve it, and that appears to be a large part of the issues currently facing AMD.



EDIT: I will cite my sources for the DTS issue (what I have been able to find, which isn't much)
Here is one, I will look for others: Link
Here is a discussion: Link

 

On that front I'd suggest because the C2D clocks higher in most cases and generally offers higher IPC.

The R600 has become a soft target though. It was late to market and the 512-bit memory interface meant everyone assumed it was going to give the G80 a lesson in GPU performance.
As it is it's late, power hungry and can't really challenge the 8800GTX.

Now add in the fact that it has been advertised, perhaps not by AMD themselves but certainly through their partners and distributors as having a feature that in reality it did not (regardless of whether people would or even could utilise it) and it gives another angle for people to vent their frustrations from.

 

What happened with Brisbane is that it is Socket AM2. The enthusiasts who had overclocked rigs were running AMD on Socket 939. At the time they were looking to upgrade, they were faced with two choices for their upgrade -- AMD or Intel. AMD failed to provide any possible means to upgrade just the CPU. They end-of-lifed a platform that still had plenty of life yet. This irked many enthusiasts. The only option left is to replace all three main components: CPU, MB, and RAM.

At the time Intel had their Core 2 Duo out while AMD had their pre-Brisbane CPUs out. The Core 2 Duo's easily overclocked by 50% if not more and ran cool. The AMD CPUs did not, overclocking was poor and they ran hot. The Intel platform had an upgrade path to Quad-Core. The AMD platform had no such upgrade path or had an uncertain upgrade path. People were not certain if AMD's Quad-Core cpu will work in Socket AM2 or if it will require Socket AM2+ or even Socket AM3.

Did I mention many enthusiasts are irked by AMD abandoning Socket 939? At the time of their switch to Socket AM2, there was no performance benefit of AM2 over 939. In fact, the AM2 platform performed worse than 939 platform.

AMD had a lot of mis-steps. AMD continues to mis-step to this day. Intel is hitting full-stride. ATI continues to mis-step to this day. Nvidia is hitting full-stride. It looks to be a very rocky and uncertain future for AMD/ATI.

 

Actually it seems like AMD is winning market share everywhere. In GPU's last Q according to Peddie. This'll help: http://www.theinquirer.net/default.aspx?article=39906

Much is made of Intel, the company with ten times the R&D budget, actually taking the performance crown back for a few months recently for the first time in years (if they can actually hold it, pending Barcelona). But really, I have to wonder how many people care. Lower cost is probably more paramount, most chips today are going to be the same to most consumers, aka the vast majority of non-gamers. It's actually fairly inexcusable for Intel to ever not have the performance lead, though, let alone get shut out for years.

Losing 600million is absurd, they've got to get that sorted. But beyond that they're in great shape. The fusion of GPU-CPU is really going to put opponets on the defensive (already is). Nvidia is going to have to pair up somewhere it looks like, or they will die. That's become rather clear since the merger.

 

Is that the right link you posted? That shows only IGP and it shows no gain for AMD, just losses by Nvidia?

 

Well it's interesting you mention Barcelona and say that "non-gamers" don't care in the same breath.

Barcelona is, of course, a server chip, and people who buy servers do care about performance, and moreover they also pay high margins for the CPUs. Arguably it's the server market where AMD has made the most strides over the past few years in being taken seriously (arguably it's the reason why Dell finally crumbled). AMD still has a good story to tell in the 4S+ server market by virtue of their superior interconnect, but for 2S it's much more even-stevens these days. Intel has their new platform scheduled for next year (IIRC) so after that who knows what will happen at 4S+. Would be interesting to see market share trends in the 2S and 4S server market ...

 

I dunno I'm just trying to be hardcore AMD okay? LOL. Somebody's got to do it.

But honestly, reading some of the posters around here you'd be amazed that it would ever be possible for AMD to actually have good news and gain share these days, or Nvidia lose same as they indeed just did, and have, as well as Intel until one Q ago, for 12 straight Q's.. Yet it's happening. Not always, but it can happen. And they haven't even begun to really utilize the merger properly of course.

The X2900XT, almost seems like it just proves it almost doesn't matter if ATI puts out a pretty horrible chip.

They're not done yet. And if they can stay afloat short term I think things are definitly in their favor long term, unless Intel buys Nvidia or so, in which case I'd say things are not in their favor again..

But really, it's amazing also the dread with which some posters speak of the fact AMD dared lose the CPU performance crown. You know, honestly, AMD should NEVER have that crown. Anytime they do even for one minute (let alone years) they are outperforming all expectations people have a right to about that imo.

 

I'm not certain that everyone quite grasped my frustrations with Brisbane, nor the significance of the DTS being broken.

The digital thermal sensor is supposed to report the temperature of the processor, an important thing to monitor when overclocking. I have in my box right now a processor that reports itself as at 60C idle through the motherboard, and if I turn on the warning beep, will start beeping under load. Coretemp reports each core at ~25C. Everest reports each core as ~18C, with the CPU being @ 75C! None of these are accurate, and it seems the temperature reported depends upon the position of the stars and phase of the moon (i.e. random).

This is actually a very good performing CPU, and with a bit of voltage increase easily goes over 3Ghz...but I for one am not going to up the voltage on a CPU without being able to monitor the temperature. Brisbane is extremely inexpensive, as well, with the lowest of the line ~$69 and able to easily hit or surpass 3Ghz...if you don't mind the crap shoot of not knowing if you are cooling it adequately or not.

Yes, AMD upset quite a number of their market segment by dropping the 939, I'll agree, but do not forget that the MC is on the processor and 939 depends upon DDR - which is now extremely expensive compared to DDR2. At present, it is cheaper to move to DDR2 + a new motherboard than it is to buy 2GB of DDR. I will refrain from giving them the benefit of the doubt by assuming that they *could* have made the MC work with both DDR and DDR2, and we can all ding them for that, but they are at least making AM2+ backward compatible.

The point of me bringing that up, though, is that is yet another problem with their parts that no one seems to be trying to fix, or even very aware of. I understand the need to do damage control by not fanning the flames when a product has difficulties, but there comes a point in time that a company has to acknowledge that the damage has been done and it is time to work to correct it.

Acknowledging problems, in my opinion, is at the root of many of the problems facing AMD at present. Another issue is just plain bad marketing and communication with the press and public. Had AMD been more upfront with the issues facing R600 rather than the ludicrous stories they fabricated, expectations would have been more in line with reality. They could have marketed the reasons why they switched from Socket 939 to Socket AM2 better, and provided a more transitional upgrade path, or a worst explained the need better.

The falsehoods surrounding R600's "family launch" need to be addressed and accounted for. The shortcomings of their recent products need to be recognized and there needs to be some accountability among the officers of the company. AMD is under a cloud of suspicion at the moment, one that they need to work to dispel in a hurry.

Six months ago, as a mere consumer I looked forward to Barcelona, Agena and all that AMD promised. Now, I am suspicious because of the poor execution of R600 and the flaws within Brisbane which aren't being addressed. I truly want AMD to succeed; they have much to offer besides just providing healthy competition to Intel and NVidia. OE got it right in the initial post; the emperor is not wearing any clothes and it is time for that to be acknowledged and corrected - it is time for AMD to notice the draft and start working to cover their butts.

 

caffeinated, if that is really such a pain for you, you may as well invest about $20 in something like this:

http://heavyweather.de/product_info.php?products_id=550&ref=239

Now on topic:

But they everyone would have bought a G8x and noone would have waited for the R600. So from the strategic POV, it's better that they didn't say it

 

Wow! Sorry to be so blunt, but that's some very real competition for my current signature... Share is only one part of the equation, gross margins and operating margins are the other two. Having 99% of the market for yourself is not very useful when you sell at a loss, and AMD certainly has been very aggressive in terms of pricing lately. Their gross margins in Q1 are down from 58.5% one year earlier, to 31%. My prediction is that they'll be down to 25-27% in Q2.

That is their real problem. It wouldn't matter if they had slower products, if they were so cheap to make that they could sell them at a huge profit and in huge quantities anyway. But obviously, they can't do that. The fact they're bleeding money is not an obscure consequence of unknown cosmic powers. It's a direct consequence of their products and market positioning. Obviously, since their gross margins are at least positive (heh!), they could reduce their operating expenses (massive layoffs) and capital expenditures (byebye fabs) to get back to profitability, but that's not really a solution either IMO.

 

I have a diode with an LCD display for my temps... . I *know* what it is running at - that's not the point. The point is that I can't recommend Brisbane to *most* people, because they can't depend on the things that one normally depends upon to determine if it is working correctly. I'm just one person; how many integrators are also reluctant to build Brisbane boxes because of this?

I argue that plenty of people *would* have waited for R600 anyway, because there are quite a few people that just hate NVidia that much. Their position would have been a lot more defensible, too. It is easy to support the underdogs with a solid, but underperforming part. It isn't so easy to support an obviously underperforming part though, when you have been mislead for months that it would be superior. The perception of AMD has been tainted.

Frankly, the number one problem at AMD isn't their technical efforts, it is their leadership, marketing and public relations that have the biggest problem. That was my point with bringing up Brisbane.

Perhaps I'm off track, and their current situation is completely due to technical reasons and plain bad luck. As an outsider, though, it just doesn't seem to me that AMD is playing to their strengths.

 

Sorry..somehow I did a double post, my apologies for the mess.

 

To be flippant, it seems like combining the 2nd place CPU maker and the 2nd place GPU maker yields a 4th place IHV.

AMD is stuck in a nasty situation. It has shown to be a laggard in its ability to design, manufacture, and market its products.

I've never been impressed with the pre-buyout AMD's marketing.
(Does anyone remember the AMDme campaign?)

I hoped that the ATI acquisition would lead to some life in that department, but so far I don't see any major change, and I think ATI may have suffered more for it.

Not that it was the best:
ATI has a history of being unable to create a great impression for a reasonably competitive or highly competitive product several GPU refreshes in a row.

Perhaps my memory is flawed, but I've always been more impressed by Nvidia's marketing anyway.
The NV30 fiasco alone showed that Nvidia's marketers can really polish a turd.

Both Intel and Nvidia seem to have stronger benchmark, dev relations, and software initiatives in their respective spheres.

The new AMD seems to have worse luck than if it were still two separate companies.


For manufacturing:
AMD lags in CPU fab capability. There is no dispute on that.
It will most likely now lag even further.
Costs for new fabs have not gotten lower, and AMD's new products have large dies.
AMD's process agreements don't yield optimal results readily, and Intel is capable of fielding a process design effort that a consortium of competitors has thus far been unable to match.

AMD's slow process ramps are evidence of limitations to the less specialized fab tech, limited engineering resources, limited cash, or a mix of all three.

ATI theoretically is on a level playing field with Nvidia.
It still has managed to misplay its hand on several occassions.
R600 is a very blatant example of how relying on the next (or half to the next) process node to beat your opponent's products on today's process is not a winning proposition for GPUs.

For design:
Delays and cancelled products happen all the time. AMD and ATI have more widely known examples.
R400+whichever next R(n)xx core that was to be the "New R400".
AMD and K8,K9,K10 when those cores were supposed to be significantly different.

Intel and Nvidia have doubtless had their share of cancelled products. I can't speak to much on Nvidia, but Intel has shown to be far more capable of juggling multiple core designs in the pipeline.
AMD is simply too small to do the same, so even the slightest mistake is disproportionately costly in terms of money and especially time.

The original successor to K8 was to be a super-wide processor, but it was determined to be too hot for too little performance early on. Even so, it was cancelled 6 months into the process.

AMD's Griffin core is an example of the slowed design progress. The company's future mobile product that is to compete against a new version of Intel's Conroe is a chip that some at AMD have characterized as being almost identical to K8, a core from several years ago.

Griffin also shows a possible direction for AMD in the laptop and desktop markets. The core itself is small, and it should run relatively cool. While it's not on a cutting-edge process, it should be within AMD's manufacturing sweet spot.
It will likely pull in decent margins, even if marketed as a value or price/performance proposition.
At least against Penryn's derivatives.


About leadership:
I don't trust Ruiz, at least no more than I'd trust the head of a company that is losing.

Orton:
Can anyone find a quote where Orton or somebody else stated that he didn't see there being a power constraint for GPUs like there was for CPUs?
I sort of remember someone saying that.

I think that alone is a sign of a massive strategic blunder years in the making.

 

Gee, OE, don't hold back on the doom and gloom..tell us what you really think

 

3dilettante: Nice post on AMD's recent execution, I guess that's an aspect I don't tend to focus on much, heh. I certainly agree with most of what you said, with perhaps one or two slight primary exceptions: I'd trust Ruiz much less than the CEO of another company which is losing, because not every loser is incompetent at predicting the future (some just know they're doomed), and not every loser feels it is their duty to hype themselves up to the point that people would think they plan to completely annihilate the competition. I also don't completely agree with what you said wrt the various versions of R4xx; I actually think ATI managed that juggling quite nicely. It certainly didn't turn out as good as if they never had to change things around at all, but that's pretty much always the case; as least they managed to decdent backup plan. There certainly were some problems with ATI, I wouldn't think that was the most significant one though.

Well, I would seriously tend to believe that AMD will regain some marketshare in Q2, partially due to OEMs/SIs no longer having tons of excess inventory. The more important reason, however, is that AMD's price cuts in April were more drastic than Intel's.

AMD's guidance for Q2 is "revenue flat to slight up". What happens when your margins go down at the same time, though? And that you are not able to cut operating expenses substantially? Heh... As you say several times, AMD seems to be getting some high-profile design wins, but nearly only because they are willing to be more aggressive for marketshare than Intel.

That's desperation for marketshare, to say the least, and it's not going to make their financial results any better. In fact, it's going to make them worse, most likely. What happens when you reduce your margins by a substantial amount to increase your marketshare by a smaller amount? Because your cuts affect most of your sales, and not exclusively your new ones, your gross profits tend to go down.

In terms of margins, one major difficulty for AMD also will be that they have been delaying Fab30's conversion to Fab38 to reduce capital expenditures in 2007, which implies that they will have 90nm products until even later than previously expected. That's obviously not good for gross margins.

Yeah, what I'm actually expecting is gross profit to be lower than Q1 (!!!), although probably not horribly so. I don't know if they'll make a $500M+ loss though. I mean, how are you counting that? Net loss? Maybe then. I'm not sure that number even matters though.

In AMD's case, I'd look at EBITDA and add in interest income & interest expenses. Or alternatively, I'd ignore the income statements completely and just look at the balance sheet. The highlighted income figures are interesting for a company that is making money or doesn't have any cash issue, as they are often the most important measures of profitability and thus dictate the valuation of the company on the stock market. On the other hand, when the question is 'How long can the company stand without further financing?' then I'm not convinced those numbers are the most important.

I'd tend to agree with that, while emphasizing that it is the entire Barcelona product family that matters, not just Barcelona itself. And arguably Griffin/Puma too, because Griffin is not even a Barcelona derivative: it's a K8 derivative.

As for PE. I'm sure that might be interesting today - AMD's mid-term and long-term prospects don't look bad at all, in theory. It will likely still be interesting in Q3 and maybe even in Q4. Beyond that, I'm not convinced the company's disclosed roadmap against Intel's (Nehalem, Fusion clone, Larrabee...) will be very enticing, and unless they manage a miracle (how they'd turn back to profitability in 2007 is completely beyond me), their balance sheet can only get worse going into next year.

Why any PE would want to even bother with AMD in Q2 2008 or so is completely beyond me, and in fact I just can't see that happening. Either AMD manages to fix their financial situation rapidly (that'd be rather miraculous!), or they turn to PE sooner than some might be expecting them to, or they should find some other way to get cash. Selling Fab30 (aka Fab38) before it's even fully converted to 65nm would certinly be very interesting, and it'd even make a lot of sense - assuming they could find an interested buyer, of course.

 

What guidance did they give for Q2? As in, should we expect a warning in a couple of weeks?