The Intel Atom Thread

[Abwx]

factory OCed Trinity (oops, Richland).

Oced within the same 35W TDP......

 

[Khato]

They will sell them at a loss that s for sure , if their production
cost was in the sub 10$ zone and considering their ASP they
should have at least 80% gross margin even with all other
activities included.

Let's have a show of hands. Given Intel's agreements with regulatory bodies around the world, who thinks that they'd be so stupid as to engage in below-cost predatory pricing? Anyone? Especially when the legal precedent last I checked is that above-cost pricing gets a free pass regardless of whether or not it's technically predatory pricing intended to drive competition out of a market. (See section I.C.2 - http://www.justice.gov/atr/public/reports/236681_chapter4.htm )

 

[Phynaz]

just find
a translator.

http://tel.archives-ouvertes.fr/docs/00/63/12/46/PDF/thA_se_avec_perspectives_3.pdf

I will - my wife is French.

You know this is going to backfire on you, right?

Edit:
And it did. My wife says this is an old paper, that is mostly speculating on the challenges of 45nm and 32nm manufacturing. She says it has no cost numbers in it at all, and doesn't even mention anything past 32nm. It contains no inside knowledge and barely mentions Intel's manufacturing by name.

 

[Phynaz]

So you're not an engineer and you get all of your information from marketing slides. Phynaz isn't competent because he's not an engineer (your words), but you're competent despite not being an engineer. Makes perfect sense!

He's an Engineer who is still posting from a phone, and refuses to state any of his PC specs. Makes you think, doesn't it.

But he is correct, I'm not an Engineer. I'm no good at math.
That's OK, because the Engineers work for me

 

[Vesku]

It s likely that they are even cheaper than Temash
by a sizeable margin wich point us to the purpose ,
that is that Intel see ARM as an existencial threat
that must be countered.

In short better selling thoses at a loss now than
risking to sell nothing later and worse,see the ARM
brigade starting to dent elsewhere than in their usual
bottom of the barrel market.

If BT is cheaper by any appreciable amount then they are barely breaking even or actually selling at a loss. Unless they get creative with their gross margin calculations, shifting R&D and such.

 

[liahos1]

I will - my wife is French.

You know this is going to backfire on you, right?

Edit:
And it did. My wife says this is an old paper, that is mostly speculating on the challenges of 45nm and 32nm manufacturing. She says it has no cost numbers in it at all, and doesn't even mention anything past 32nm. It contains no inside knowledge and barely mentions Intel's manufacturing by name.

amazing. abwx where you at?

 

[Arachnotronic]

I will - my wife is French.

You know this is going to backfire on you, right?

Edit:
And it did. My wife says this is an old paper, that is mostly speculating on the challenges of 45nm and 32nm manufacturing. She says it has no cost numbers in it at all, and doesn't even mention anything past 32nm. It contains no inside knowledge and barely mentions Intel's manufacturing by name.

LOL Owned.

 

[mrmt]

If BT is cheaper by any appreciable amount then they are barely breaking even or actually selling at a loss. Unless they get creative with their gross margin calculations, shifting R&D and such.

You must first understand what selling at loss means.

One thing is to spend 10 to manufacturing something and sell it at 8. Then I'm selling at loss. But I cannot sell this straight. I have to book 2 per unit as loss in my balance sheet. This is what AMD did with their old inventory, and the procedure is pretty nasty. If the volume is significant or recurrent you must disclose it to your fillings in the end of the quarter and the analysts will kill you for that.

But let's say that I spend the same 10, but I sell for 12. But I'll sell only 10 units of those and I spent 30 in R&D to develop my product. I won't be selling at loss, but the project itself will be a loss for the company as it won't be able to recoup all the money spent to bring a product to the market.

I'm sure Intel isn't selling Atom at loss, because that would be dumping and regulators do not like dumping. On top of that you better not do that in a high volume, price sensitive product because even smaller losses per unit become really big numbers and it's hard to rebuild pricing once you crater it, especially on price sensitive markets like this one. And in the end there isn't much you can do to mask negative gross margins that this kind of practice would bring, you would have to report it on the end of the quarter and that would be bad, really bad. When you reach that point it's better to shut down production, sell fabs and try to find something better with your time and money.

That leaves us with the second option, that project Atom ROI might be negative. Any of the trolls claiming that is pulling numbers from god-knows-where, because the only people that actually have this information are the Intel guys working on the project management and in their financial planning and they can't disclose that information to no one. To claim to know this information is akin to claim fairy tales like "I bet against Goldman Sachs official position".

 

[Nothingness]

Let's have a show of hands. Given Intel's agreements with regulatory bodies around the world, who thinks that they'd be so stupid as to engage in below-cost predatory pricing? Anyone? Especially when the legal precedent last I checked is that above-cost pricing gets a free pass regardless of whether or not it's technically predatory pricing intended to drive competition out of a market. (See section I.C.2 - http://www.justice.gov/atr/public/reports/236681_chapter4.htm )

I have no doubt Intel is pricing extremely low, probably not below production cost, but most probably below R&D + fab capex + prod cost. This is based on pricing some of our customers reported.

Is that predatory pricing? Can't say, I don't understand your link, and most probably very few people do

 

[mrmt]

I have no doubt Intel is pricing extremely low, probably not below production cost, but most probably below R&D + fab capex + prod cost. This is based on pricing some of our customers reported.

Do you know that fab CAPEX is included in the production costs, don't you?

 

[Khato]

I have no doubt Intel is pricing extremely low, probably not below production cost, but most probably below R&D + fab capex + prod cost. This is based on pricing some of our customers reported.

Is that predatory pricing? Can't say, I don't understand your link, and most probably very few people do

It'd definitely be in a more gray area, since they'd basically have to write off/apply those costs elsewhere. Which they can probably get away for the time being. I expect that their pricing includes both manufacturing and product design costs, but not the fixed manufacturing costs which are likely absorbed entirely by another group. That would put them more or less in the clear.

And yeah, legal opinions tend to be quite interesting to read and properly interpret. I'll usually come to a slightly different conclusion than is legally correct, but luckily I have a few lawyers around to set me straight

Do you know that fab CAPEX is included in the production costs, don't you?

Normally yes, but for these purposes they could be playing some slight accounting games.

 

[liahos1]

Do you know that fab CAPEX is included in the production costs, don't you?

fab capex runs through the cash flow statement. deprecation on that capex runs through the income statement. so indirectly its included.

 

[Vesku]

You must first understand what selling at loss means.

One thing is to spend 10 to manufacturing something and sell it at 8. Then I'm selling at loss. But I cannot sell this straight. I have to book 2 per unit as loss in my balance sheet. This is what AMD did with their old inventory, and the procedure is pretty nasty. If the volume is significant or recurrent you must disclose it to your fillings in the end of the quarter and the analysts will kill you for that.

But let's say that I spend the same 10, but I sell for 12. But I'll sell only 10 units of those and I spent 30 in R&D to develop my product. I won't be selling at loss, but the project itself will be a loss for the company as it won't be able to recoup all the money spent to bring a product to the market.

I'm sure Intel isn't selling Atom at loss, because that would be dumping and regulators do not like dumping. On top of that you better not do that in a high volume, price sensitive product because even smaller losses per unit become really big numbers and it's hard to rebuild pricing once you crater it, especially on price sensitive markets like this one. And in the end there isn't much you can do to mask negative gross margins that this kind of practice would bring, you would have to report it on the end of the quarter and that would be bad, really bad. When you reach that point it's better to shut down production, sell fabs and try to find something better with your time and money.

That leaves us with the second option, that project Atom ROI might be negative. Any of the trolls claiming that is pulling numbers from god-knows-where, because the only people that actually have this information are the Intel guys working on the project management and in their financial planning and they can't disclose that information to no one. To claim to know this information is akin to claim fairy tales like "I bet against Goldman Sachs official position".

Your first half basically covers what it means when I state things like "unless they play with how they calculate gross margins". Gross margins would normally include direct R&D but there are plenty of clever ways to massage business numbers when management wants something to look a certain way. R&D that was really directed at Bay Trail becomes some other "general process R&D" booked to Haswell or Broadwell, etc.

Even with some careful massaging I don't think Intel is selling any current Baytrail-T for $10 unless for a one off "get in this particular device" and I'd need to see some serious evidence to believe it's done any other way than that price being delivered to the OEM not by selling them at $10 but providing enough additional discounts on device related parts and marketing to hit seal the deal.

 

[mrmt]

Normally yes, but for these purposes they could be playing some slight accounting games.

Do you mind describing some of the accounting games here?

 

[mrmt]

Your first half basically covers what it means when I state things like "unless they play with how they calculate gross margins". Gross margins would normally include direct R&D but there are plenty of clever ways to massage business numbers when management wants something to look a certain way. R&D that was really directed at Bay Trail becomes some other "general process R&D" booked to Haswell or Broadwell, etc.

Gross margin does not include R&D. R&D is OPEX, it goes straight to the bottom line and that's why every bleeding edge MPU company needs a fat gross margin in their income statement, because they need to pay the R&D bill.

 

[Khato]

Do you mind describing some of the accounting games here?

Basically as stated in the first portion of my previous post. I make no claim of being familiar with the specifics, just that it's possible to separate out a number of the fixed costs from being included in the 'cost' that they need to sell at or above in order to effectively shield themselves from future antitrust litigation. Which is quite important when it certainly appears to be the case that they intend to use their manufacturing advantage (that advantage primarily being that they actually have fabs) in order to start and win a price war... which would basically be the definition of predatory pricing if they succeed in driving the competition from the market.

 

[Abwx]

I will - my wife is French.

You know this is going to backfire on you, right?

Edit:
And it did. My wife says this is an old paper, that is mostly speculating on the challenges of 45nm and 32nm manufacturing. She says it has no cost numbers in it at all, and doesn't even mention anything past 32nm. It contains no inside knowledge and barely mentions Intel's manufacturing by name.

Lol...

The paper is of 2008 and explain how a given transistor is synthesized
with explicit mathematical formulation of the physical shape and doping
influence on the mosfet transistors performances and so on , it will
be the same for any process , i gave this link as exemple of where
i get my information about current technoilogies and at wich level ,
this wasnt meant as a exemple of what intel is really using but whatever
their real implementation it will follow the said mathematical modelisations
of the devices caracteristics and said modelisations say that an actual
product can only be a compromise but i let this for another post...

 

[Abwx]

He's an Engineer who is still posting from a phone, and refuses to state any of his PC specs. Makes you think, doesn't it.

But he is correct, I'm not an Engineer. I'm no good at math.
That's OK, because the Engineers work for me

No pun intended from me on my previous posts.

Posted from a core2 duo laptop that i got for free....

 

[mrmt]

Basically as stated in the first portion of my previous post. I make no claim of being familiar with the specifics, just that it's possible to separate out a number of the fixed costs from being included in the 'cost' that they need to sell at or above in order to effectively shield themselves from future antitrust litigation.

Funny. You don't know the specifics but yet you claim to know it is possible to make something like that. Let me be more explicit: You can't do what you are saying, because it is abjectly stupid. Any company that is selling something at loss has to go out of business. Selling at loss means that it is cheaper to leave the factory idle than manufacture anything. But if you want to engage in unfair commercial practices and buy market share, you won't mess with your costs in order to offer lower prices to your customers, because:

a) It's easier for regulators to notice

b) It's easier for your audit company to notice

c) It's easier for financial analysts to compare and notice something fishy

d) you still have to shove the costs somewhere else

What you are going to do is offer rebates, marketing money, bonus and so on. This goes straight to a blackbox called SG&A, something that everybody accepts as wildly different from company to company.

 

[AtenRa]

Wrong again. Core i5 4288U is a 2C/4T Haswell CPU, at 2.6-3.1GHz it will beat any 35W Trinity/Richland in MT taks and smash it in ST tasks. Poor 35W Richland has yet to match last-year's 17W Ivy Bridge ULV MT CPU performance according to Anand. Also, this Core i5 has the same GPU frequency as a Core i7.

Quick CB11.5 comparison using slightly a higher-clocked 2.8-3.3GHz Core i7
A10-5750M: 2.32 Multi / 0.85 Single.
Core i7 4558U: 3.48 Multi / 1.45 Single.

You make it sound like there is a big difference between Trinity and Richland, but in fact Richland is nothing more than a slightly higher-clocked Trinity (performance-wise).
If 28W ULT Haswell manages to match AMD's best 2012 mobile APU's GPU performance (and smash it in CPU tasks), then a hypothetical 35W Haswell with GT3 + more clock headroom would probably do fine (if not beat) factory OCed Trinity (oops, Richland).

We were talking about iGPUs not CPUs, and the fact remains that at the same TDP and same price points Intel has nothing to directly compete against 35W Ritchland in Graphics performance. And this is with a full node process advantage(22nm) for Intel and an older VLIW4 architecture from AMD on 32nm, wait to see what will happen with Kaveri.

 

[liahos1]

We were talking about iGPUs not CPUs, and the fact remains that at the same TDP and same price points Intel has nothing to directly compete against 35W Ritchland in Graphics performance. And this is with a full node process advantage(22nm) for Intel and an older VLIW4 architecture from AMD on 32nm, wait to see what will happen with Kaveri.

let me take a stab. kaveri will come out have god awful cpu perf, good but ultimately unplayable gpu perform (relative to a discrete) all in a TDP uncompetitive with intel. amd will talk talk up HSA as the next coming of Christ and claim many many many design wins. and ultimately we may see kaveri in a few budget notebooks on the bargain / special sales section of best buy. sort of like what happened with trinity

 

[Vesku]

Gross margin does not include R&D. R&D is OPEX, it goes straight to the bottom line and that's why every bleeding edge MPU company needs a fat gross margin in their income statement, because they need to pay the R&D bill.

Even when the engineering team is working specifically on getting a chip like Bay Trail through to production, not just on the design phase?

 

[mrmt]

Even when the engineering team is working specifically on getting a chip like Bay Trail through to production, not just on the design phase?

Yes. Only the final qualification run is accrued as COGS, everything else is OPEX.

 

[AtenRa]

let me take a stab. kaveri will come out have god awful cpu perf, good but ultimately unplayable gpu perform (relative to a discrete) all in a TDP uncompetitive with intel. amd will talk talk up HSA as the next coming of Christ and claim many many many design wins. and ultimately we may see kaveri in a few budget notebooks on the bargain / special sales section of best buy. sort of like what happened with trinity

Let me save this to quote you on Kaveri release

 

[liahos1]

Let me save this to quote you on Kaveri release

please do. if they build something halfway decent i would be happy to buy it. i'm already awaiting 290x reviews to see if i sell my 2x gtx 780s. seems unlikely at this point though...will have to wait for 20nm product.