AMD product mix by Mercury Research (Q312 numbers)

[mrmt]

http://www.techpowerup.com/175751/APUs-Make-Up-Nearly-75-of-AMD-s-Processor-Sales.html

Techpowerup came with some interesting numbers from Mercury Research and as usual they screwed up when crunching them.

The numbers essentially shows how bad AMD is in the value ladder, with bobcat amounting for 40% of their shipments, but almost 60% of their mobile shipments. Bobcat as their flagship product means that their low gross margins aren't something temporary, but something structural. Low margins are here to stay.

Trinity is a lot more popular than Llano, with the former taking more than 70% of the latter's share in AMD product mix in a single quarter, but aggregated Trinity numbers are 30% smaller than Llano numbers in Q1. The same trend is present in desktops, with desktop Trinity sales amounting for less than half of Llano desktop sales in Q1. This corroborates the structural shift I described in the previous paragraph, AMD is being confined to the very bottom of the market.

Server amounts for a mere 1.3% of their shipments. Nothing to be seen here.

Despite the "Moar Cores!!!™ " hype, 90% of their processors have up to 4 cores and 66% have 2. 8 "cores" processors respond for just 1% of their product mix, with only 127.000 8 Cores bulldozer shipped in Q3.

And here's the surprise: After more than a year of the launch, Bulldozer didn't overtake Deneb/Regor in their product mix as the top product. The most popular product in AMD mix is the Athlon X2 and the fastest growing desktop product is Brazos, with both comprising almost half of AMD desktop shipments. K10 is leading architecture at AMD mix. Good old Phenom X6 still sells more than 8C Bulldozer. This is nothing short of a disaster, as they are using an at the time outdated architecture to compete with Intel's cream.

If anybody had any doubts of the magnitude of the Bulldozer disaster and how hopeless AMD situation is, the numbers are there to everyone to see. AMD fighting an Intel shipping 70%of their products in 22nm node and bleeding edge architecture with a product mix comprised basically of low end chips manufactured in 40nm and 45nm, most of them being old designs.

 

[VirtualLarry]

Are they still mfging Phenom X6 and Athlon X2? I thought that due to reported process cross-contamination issues with 45nm and 32nm, that they had stopped 45nm production months and months ago.

 

[jones377]

Charlie claims that AMD kills off Steamroller, Kaveri and Excavator, leaving only the low power x86 cores (Bobcat+Jaguar) and Seamicro+ARM left. The latter not having reached the market yet.

http://semiaccurate.com/2012/11/19/amd-kills-off-big-cores-kaveri-steamroller-and-excavator/

If true, this means AMD is killing off 60% of their CPU shipments by volume and likely even more by revenue. They are gonna have to lay off ALOT more people than the 15% or so announced. Can they survive this sudden shift? 2 years from now AMD might only pull in half the revenue per quarter if all goes extremely well (Jaguar has to grow the low power x86 line and Seamicro has to be a success.). This has to line up with less than half in expenses too...

 

[sm625]

As many have already said, they were certifiably insane to try releasing a cpu with the single thread performance of a pentium 1. It is just a huge fail and I'm not even sure any company should survive such awful mismanagement. There was just no excuse. I never imagined they could do something so stupid.

 

[Abwx]

As many have already said, they were certifiably insane to try releasing a cpu with the single thread performance of a pentium 1.

Seems that viral marketing did take its toll to the point that some
people dont even realize how ridiculous their claims can be...

 

[ShintaiDK]

39% Bobcat. And some people still wonder why they are the new VIA?

The server segment is utterly destroyed. Everything there is a waste of time and money for AMD. Yet they keep running after it like a headless chicken.

127000 Bulldozer 8 cores shipped. It says it all.

 

[Idontcare]

http://www.techpowerup.com/175751/APUs-Make-Up-Nearly-75-of-AMD-s-Processor-Sales.html

Techpowerup came with some interesting numbers from Mercury Research and as usual they screwed up when crunching them.

The numbers essentially shows how bad AMD is in the value ladder, with bobcat amounting for 40% of their shipments, but almost 60% of their mobile shipments. Bobcat as their flagship product means that their low gross margins aren't something temporary, but something structural. Low margins are here to stay.

Trinity is a lot more popular than Llano, with the former taking more than 70% of the latter's share in AMD product mix in a single quarter, but aggregated Trinity numbers are 30% smaller than Llano numbers in Q1. The same trend is present in desktops, with desktop Trinity sales amounting for less than half of Llano desktop sales in Q1. This corroborates the structural shift I described in the previous paragraph, AMD is being confined to the very bottom of the market.

Server amounts for a mere 1.3% of their shipments. Nothing to be seen here.

Despite the "Moar Cores!!!™ " hype, 90% of their processors have up to 4 cores and 66% have 2. 8 "cores" processors respond for just 1% of their product mix, with only 127.000 8 Cores bulldozer shipped in Q3.

And here's the surprise: After more than a year of the launch, Bulldozer didn't overtake Deneb/Regor in their product mix as the top product. The most popular product in AMD mix is the Athlon X2 and the fastest growing desktop product is Brazos, with both comprising almost half of AMD desktop shipments. K10 is leading architecture at AMD mix. Good old Phenom X6 still sells more than 8C Bulldozer. This is nothing short of a disaster, as they are using an at the time outdated architecture to compete with Intel's cream.

If anybody had any doubts of the magnitude of the Bulldozer disaster and how hopeless AMD situation is, the numbers are there to everyone to see. AMD fighting an Intel shipping 70%of their products in 22nm node and bleeding edge architecture with a product mix comprised basically of low end chips manufactured in 40nm and 45nm, most of them being old designs.

More proof that the BoD actually did the right thing in firing Dirk. His green-lighting of bulldozer is what brought AMD to this.

And with his untouchable credibility, at the time, stemming from being the chief architect of the original K7 Athlon there can be no doubt he was the one who was convinced (and thus convinced the BoD) that bulldozer was the way to go.

I can't say the BoD did anything else right, but at least they kinda did something right by removing Dirk. Their timing sucked, their lack of preparation prior to canning him sucked, their execution after firing him sucked, their choice of an interim CEO sucked, and quite possibly their choice of a replacement CEO sucked...but one thing is for certain and that is while Dirk may have been a microarchitectural genius in creating the Athlon, he was a one-hit wonder and did more harm than good in the end.

 

[Abwx]

It seems that AMD didnt evaluate the consequences of its market positionning
in the low end in respect of the value segment.

Eager to make more profits by selling low costs solutions at values segment prices , unscrupulous OEMs are using Bobcats instead of 2C/4C Llanos as was expected by AMD , hence they saw their value CPUs segment being cannibalized by their own lower margins products with the added downside that this would destroy their reputation since OEMs are using inadequate APUs for AIO and 15" netbooks.

 

[dud]

Healthy competition is good for the marketplace and especially the consumer. This news is not good for consumers. We need a threat to Intel to offset their near-monopoly.

The popularity of their APU line gives one hope.

 

[videogames101]

Healthy competition is good for the marketplace and especially the consumer. This news is not good for consumers. We need a threat to Intel to offset their near-monopoly.

The popularity of their APU line gives one hope.

We have one, it's called ARM. Maybe you've heard of it

 

[guskline]

39% Bobcat. And some people still wonder why they are the new VIA?

The server segment is utterly destroyed. Everything there is a waste of time and money for AMD. Yet they keep running after it like a headless chicken.

127000 Bulldozer 8 cores shipped. It says it all.

Darn. Maybe I'll hold onto my FX8150 instead of trying to sell it. It will be a "collector's item" someday!

 

[mrmt]

I can't say the BoD did anything else right, but at least they kinda did something right by removing Dirk. Their timing sucked, their lack of preparation prior to canning him sucked, their execution after firing him sucked, their choice of an interim CEO sucked...

That's a nice summary of the BoD actions. I think there are a lot of mispredictions too.

The APU strategy, I think they had in mind was to charge a lot more for Trinity than they are charging now, let's say, the same of an equivalent CPU + price premium for the GPU. In reality they can't charge a penny more than Intel for their bigger GPU.

The 28nm debacle is another question too. How could they think that they would have frequency headroom by going from a SOI process for a bulk process? They are a fabless company, they must be very conservative about estimates of their foundry partner, instead it seems that they were gung-ho about frequency and thermals at the planning stage.

There were a lot of f*** ups there. Someone was chasing the golden pot at the end of the rainbow instead of plan according plausible market trends.

...and quite possibly their choice of a replacement CEO sucked

I'm glad to see that someone still thinks that Rory Read is suitable for the task at hand

AMD has an engineering problem, not a commercial one. Rory could be a nice addition to the BoD if they had solid engineering and a nice product mix to sell, then his experience in the OEM world could be a boost for AMD.

As of now, as the company does not have a vision and every engineer out there needs a direction, a non-technical CEO like Rory is a nuisance, it's a clown babbling market buzzwords that won't give a focus for the engineering teams. It's a boss that cannot be challenged by the engineers because he can't even talk like an engineer.

He can't go deep into AMD x86 business and work a solution there, he is just hammering the problem by going ARM full speed ahead.

 

[Idontcare]

It seems that AMD didnt evaluate the consequences of its market positionning
in the low end in respect of the value segment.

Eager to make more profits by selling low costs solutions at values segment prices , unscrupulous OEMs are using Bobcats instead of 2C/4C Llanos as was expected by AMD , hence they saw their value CPUs segment being cannibalized by their own lower margins products with the added downside that this would destroy their reputation since OEMs are using inadequate APUs for AIO and 15" netbooks.

In other words, AMD's management managed to mismanage their own products and the market's reception to them.

A fate that Intel's managers somehow managed to avoid despite fielding Atom.

No matter how you want to try and lay the blame, at the end of the day the point still stands that AMD did it to themselves by not being as good as their counterparts at Intel in carrying out their own jobs at AMD.

 

[mrmt]

Eager to make more profits by selling low costs solutions at values segment prices , unscrupulous OEMs are using Bobcats instead of 2C/4C Llanos as was expected by AMD , hence they saw their value CPUs segment being cannibalized by their own lower margins products with the added downside that this would destroy their reputation since OEMs are using inadequate APUs for AIO and 15" netbooks.

There was a reason Intel priced 45nm Atom so high and only with soldered kits that you could not build anything other than a netbook with it, and you are finally pointing out this now.

That said, don't blame OEM for AMD mess. OEMs prefer Brazos to Trinity because they can get better margins with the former. AMD has only itself to blame for the atrocious design of Bulldozer and for their design of a low-end chip measuring 246mm^2 that cannot get descent margins for OEMs.

 

[pablo87]

Those low BD unit #s are friggin' scary.

 

[Idontcare]

Those low BD unit #s are friggin' scary.

The shock factor from such low unit movements made me immediately think of the Chevy Volt and the 9 units they moved in August (IIRC), there is "poor uptake" and then there is downright "practically no one wants one".

I have to imagine the 8320's and 8350's are doing much better though in terms of units moved.

 

[cytg111]

Charlie claims that AMD kills off Steamroller, Kaveri and Excavator, leaving only the low power x86 cores (Bobcat+Jaguar) and Seamicro+ARM left. The latter not having reached the market yet.

http://semiaccurate.com/2012/11/19/amd-kills-off-big-cores-kaveri-steamroller-and-excavator/

If true, this means AMD is killing off 60% of their CPU shipments by volume and likely even more by revenue. They are gonna have to lay off ALOT more people than the 15% or so announced. Can they survive this sudden shift? 2 years from now AMD might only pull in half the revenue per quarter if all goes extremely well (Jaguar has to grow the low power x86 line and Seamicro has to be a success.). This has to line up with less than half in expenses too...

and updated

"Updated 11/19/2012@10:15am: AMD contacted us with an official denial of the story and stated that Kaveri and the big cores are still on track."

comical ali comes to mind ...

 

[inf64]

Well Charlie covers every possible alternative so eventually "he told us so" . Whatever it turns out to be

 

[Idontcare]

Well Charlie covers every possible alternative so eventually "he told us so" . Whatever it turns out to be

I love it when he goes with the "we knew back then but we couldn't tell you about it because the moles asked us not too :whiste:" after-the-fact-excuse in a "breaking" story.

riiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiight

 

[mrmt]

I have to imagine the 8320's and 8350's are doing much better though in terms of units moved.

There is a reason for OEMs prefer Bobcat to any other AMD chip and this reason is margin.

At 315mm^2 AMD must have a really high starting price, add to that extra cooling, and bigger cases and you have the perfect anti-OEM chip. Worse, the same die

Vishera brings more performance to the table but does not addresss the fundamental flaws of Zambezi, so don't expect something radically different.

 

[MisterMac]

So lemme get this straight

AMD sold less than 500.000 8100 series bulldozer?



Are you ***** ********* me?

Whats the Q4/11 numbers?

 

[Vesku]

I find it hilarious that the market is forcing AMD management kicking and screaming into their best place to position themselves. Lower margin power efficient x86 with as much SoC like features as possible (GPU, memory controller, PCIe on die) is where they needed to go since Intel came to their senses with Core->Core 2 and their aggressive Tick-Tock development.

Their Trinity numbers would be better if they could do a better job of convincing companies to actually produce products around it. The selection of Trinity notebooks is lacking atm. Edit: Which I think also points to management issues, OEMs like reliability in their suppliers.

 

[Idontcare]

At 315mm^2 AMD must have a really high starting price, add to that extra cooling, and bigger cases and you have the perfect anti-OEM chip. Worse, the same die

Yet more irony...the primary argument for going with gate-first HKMG integration (versus gate-last like Intel and TSMC) was that the gate-first process would enable much smaller dies (better scaling) and lower manufacturing costs (simpler process) on top of the lower production cost that would come with having smaller dies and more dies per wafer.

Its been said by a number of people here and I tend to agree with them, half of zambezi's problems were GloFo's 32nm process technology. Even if AMD had stuck with just shrinking the K10, as they did with Llano, the process tech did not deliver in terms of clockspeeds and power usage over that of 45nm K10's.

So it seems fair to conclude bulldozer was hamstrung by the same process tech that hamstrung Llano's performance. But that does not exonerate AMD because they went along with GloFo's plans to pursue gate-first.

Too much of a "not invented here" complex going on there, they just couldn't take Intel's lead in gate-last HKMG and ride on the coattails of billions of dollars of proven R&D. TSMC doesn't have that problem and it shows.

 

[Idontcare]

So lemme get this straight

AMD sold less than 500.000 8100 series bulldozer?



Are you ****** ******** me?

Whats the Q4/11 numbers?

Mind the swearing please, this is supposed to be a safe-for-work technical forum.

 

[pablo87]

The shock factor from such low unit movements made me immediately think of the Chevy Volt and the 9 units they moved in August (IIRC), there is "poor uptake" and then there is downright "practically no one wants one".

I have to imagine the 8320's and 8350's are doing much better though in terms of units moved.

I considered buying one this summer but went with 2500K. The value just wasn't there. The thing that bothered me even more than the module business which I never liked from day one, was how old the 760 chipset was (going to be 4 years old very soon). It felt like we were buying an obsolete product..