Intel Q313 Results

[mrmt]

But the mobile section and other future sections is doing well slow. The are using a lot of money on mobile using windows 8 mobile as a way of pushing into tablets. I can not see that losing is going to change.

Given that almost half of BT design wins are Android, I don't think they are really keen on W8 anymore. And they shouldn't, W8 is awful.

 

[SammichPG]

You're so full of anti-Intel propaganda that it's ridiculous. You do realize that you have to invest BEFORE your products show up, right?

The massive R&D you're seeing is preparing the products that will be competitive beginning with Bay Trail and extending into next year. Until then, all other IA has to actually sell today are a bunch of OK 3G modems and some old Atoms. As the LTE/LTE-A solutions, Bay Trail-T/I, Merrifield, and the like ramp, you will see what we call "operating leverage".

Say, what happens to Intel's bottom line if that $2B operating loss just hits breakeven? Or - gasp! - what if Intel actually makes money at it?

"just wait for the next year!"

Intel on every mobile product release.

 

[lefty2]

So, if PC sales dropped 8% YoY (according to Gartner), and Intel PC sales only dropped 3.5%, where is the rest of the sales drop?

I can only think of one other x86 manufacturer that had to of had a massive drop is sales to make up the difference.

That is a very good point. Either Gartner got it badly wrong, or AMD lost a massive amount of marketshare this quarter (I calculate they would have to lose 5% to account for the shortfall)

 

[liahos1]

That is a very good point. Either Gartner got it badly wrong, or AMD lost a massive amount of marketshare this quarter (I calculate they would have to lose 5% to account for the shortfall)

gartner doesnt really account for the difference between Sell-in and Sell through. It also doesnt account accurately for the whitebox DIY market in china and other asia pacific countries.

 

[Phynaz]

"just wait for the next year!"

AMD on every product release.

Lol, FTFY

 

[mrmt]

Intel on every mobile product release.

I don't think you realize the shift here. Intel said that netbooks are giving single digits revenues, so the bulk of sales on that division is already mobile, phones and tablets, some 500MM here. That said, it is still very small by Intel standards but it's safe to say they already have a foothold on the mobile market.

 

[lefty2]

gartner doesnt really account for the difference between Sell-in and Sell through. It also doesnt account accurately for the whitebox DIY market in china and other asia pacific countries.

but is that enough to account for a 4% shortfall? I wonder has Gartner ever been this wrong before?

 

[SammichPG]

Lol, FTFY

Completely true as well.

I might be out of my league but Intel is playing a DEC or an IBM, trying to compete in a market that changed too fast for them.

Am I the only one seeing similarities to mainframes vs microcomputers 80s market and the mobile market now?

 

[24601]

Completely true as well.

I might be out of my league but Intel is playing a DEC or an IBM, trying to compete in a market that changed too fast for them.

Am I the only one seeing similarities to mainframes vs microcomputers 80s market and the mobile market now?

The use cases are what drive technology in the long term. Unless mobile OS's improve as fast as Windows did, everything on the hardware side is pointless.

Wintel was Wintel for a reason.

I'm still hoping for a Wintel phone with 100% backwards compatibility all the way back to Dos.

One can dream

I'm not getting a smartphone until that happens.

 

[teejee]

Well, seems like Intel will get some competition in server business from Broadcom. I guess this chip won't be available until 2015.
But server chip to become a highly competitive business is a nightmare for Intel.
And the CPU cores are optimized for server use, this could easily compensate for using less advanced process than Intel.
I think we'll see more dedicated server CPU's.

http://www.broadcom.com/press/release.php?id=s797235

 

[Arachnotronic]

Well, seems like Intel will get some competition in server business from Broadcom. I guess this chip won't be available until 2015.
But server chip to become a highly competitive business is a nightmare for Intel.
And the CPU cores are optimized for server use, this could easily compensate for using less advanced process than Intel.
I think we'll see more dedicated server CPU's.

http://www.broadcom.com/press/release.php?id=s797235

Not really...this is just a follow on to the current XLP line of quad issue/quad threaded MIPS server chips. This isn't anything new.

 

[Arachnotronic]

"just wait for the next year!"

Intel on every mobile product release.

I'm pretty sure Bay Trail is ramping this year, and it is competitive.

 

[bullzz]

@teejee
from ur link - "quad-threaded 64-bit ARMv8-A core"

isnt this still bay trail IPC level. they still have years to go till they are anywhere close to Xeon
and its not always only about the chip right..

 

[krumme]

I don't think you realize the shift here. Intel said that netbooks are giving single digits revenues, so the bulk of sales on that division is already mobile, phones and tablets, some 500MM here. That said, it is still very small by Intel standards but it's safe to say they already have a foothold on the mobile market.

No. They have foothold the day they can do that without losing loads of money. Right now they are nearly giving them away at low end arm cost and probably paying oem to design for them on top of that. At least they must do something like that to lose all those money on mobile. I dont know where the money is going if thats not the situation.

 

[Arachnotronic]

No. They have foothold the day they can do that without losing loads of money. Right now they are nearly giving them away at low end arm cost and probably paying oem to design for them on top of that. At least they must do something like that to lose all those money on mobile. I dont know where the money is going if thats not the situation.

It's going into R&D. You know. The spending that makes competitive chips.

 

[liahos1]

Well, seems like Intel will get some competition in server business from Broadcom. I guess this chip won't be available until 2015.
But server chip to become a highly competitive business is a nightmare for Intel.
And the CPU cores are optimized for server use, this could easily compensate for using less advanced process than Intel.
I think we'll see more dedicated server CPU's.

http://www.broadcom.com/press/release.php?id=s797235

by the time this comes out (2015) Intel will have interated its xeons and atoms 2x and have stepped down to a new node.

You realize that intel is an ARM architectural licensee. They can see whats coming for years before it actually comes. One would hope they would act sensibly on it. I'm talking to Mark Henninger at Intel tomorrow afternoon. If anyone wants me to ask him some questions let me know.

 

[mrmt]

No. They have foothold the day they can do that without losing loads of money. Right now they are nearly giving them away at low end arm cost and probably paying oem to design for them on top of that. At least they must do something like that to lose all those money on mobile. I dont know where the money is going if thats not the situation.

Lots of wild assumptions here, Krumme. We can't really say how much is R&D and how much is marketing money to grease OEM hands. FWIW doing two die shrinks, three architectural updates, developing a digital radio *and* LTE modem costs a lot of money and that's what Intel is poised to do in the next three years.

I especially doubt things like "Intel is selling chips at loss" or "Intel is paying OEMs big money to make them adopt Atom", because this will build a relationship where Intel won't be able to make money, ever. There is a huge gulf between giving some incentives to new customers, like free samples or marketing money, and actually paying someone to use your product. The first one will yield you money once the customer come back, the second one will not. But even if Intel were dumb enough to pursue that strategy, they wouldn't do that on the mobile market. Do this from a dominant position and you prevent your competitor from entering your market and then you can remove your subsidy afterwards. Do this when entering a market and by the time you think about removing the subsidy your partner will roll back to the previous supplier.

 

[teejee]

@teejee
from ur link - "quad-threaded 64-bit ARMv8-A core"

isnt this still bay trail IPC level. they still have years to go till they are anywhere close to Xeon
and its not always only about the chip right..

This is a quad-issue design. That indicates IPC far above bay trail. Probably similar to Haswell.

 

[teejee]

by the time this comes out (2015) Intel will have interated its xeons and atoms 2x and have stepped down to a new node.

You realize that intel is an ARM architectural licensee. They can see whats coming for years before it actually comes. One would hope they would act sensibly on it. I'm talking to Mark Henninger at Intel tomorrow afternoon. If anyone wants me to ask him some questions let me know.

No, Intel has no idea what the ISA licensees are doing except for the instruction set.
Intel will only learn about the ARM designs (cortex A57 etc).
Broadcom is doing their own design, just like Apple, qualcomm , Nvidia (soon), Samsung (soon) etc.

 

[Exophase]

@teejee
from ur link - "quad-threaded 64-bit ARMv8-A core"

isnt this still bay trail IPC level. they still have years to go till they are anywhere close to Xeon
and its not always only about the chip right..

Why would it be BayTrail perf/MHz level (IPC is the wrong term)? How do you extract that from what little information is given? The word "issue" isn't even used consistently by CPU designers, so we don't really know what it means here. I could call Haswell "quad-issue and dual-threaded", and this would be consistent with Intel's own terminology. I think you're just making assumptions because this is ARM, even though that really says very little about uarch.

And why are people so fixated on perf/MHz to begin with?

The quad-threaded part can give them some extra performance opportunities over dual-threaded SMT in the right conditions. Same reason SPARC, POWER, Larrabee, and some others have gone quad-threaded in various implementations. This is interesting since it's the first multi-threaded ARM design I'm aware of.

 

[liahos1]

No, Intel has no idea what the ISA licensees are doing except for the instruction set.
Intel will only learn about the ARM designs (cortex A57 etc).
Broadcom is doing their own design, just like Apple, qualcomm , Nvidia (soon), Samsung (soon) etc.

sure thats a fair point. but having knowledge of the core ISA gives you a pretty good range of expectations around it.

 

[Exophase]

sure thats a fair point. but having knowledge of the core ISA gives you a pretty good range of expectations around it.

No it doesn't. Why do you think so?

It's not like you need an architectural license to understand the ISA. Most of the relevant details of ARMv8 have been common knowledge for a long time, and I think the real specifications are available now (just like they are for ARMv7a)

 

[liahos1]

No it doesn't. Why do you think so?

It's not like you need an architectural license to understand the ISA. Most of the relevant details of ARMv8 have been common knowledge for a long time, and I think the real specifications are available now (just like they are for ARMv7a)

Id just assume that having knowledge on paper vs being able to build the chip yourself and test it were two different things. although i have ZERO first hand experience with this. i'd defer to u to let me know how it works.

 

[Arachnotronic]

Id just assume that having knowledge on paper vs being able to build the chip yourself and test it were two different things.

Not quite. ARMv8, like X86, is just an instruction set - a specification, if you will. It just tells you which "commands" the processor needs to be able to execute (this is a VERY simplified explanation). The implementation (i.e. building the machine that can actually perform those commands) can vary greatly. Intel and AMD both implement X86, but Intel's chips are better than AMD's (even on the same class of process node).

So, having an ARM ISA license doesn't help much in figuring out what the competitors are going to do.

Hope this helps.

 

[liahos1]

Not quite. ARMv8, like X86, is just an instruction set - a specification, if you will. It just tells you which "commands" the processor needs to be able to execute (this is a VERY simplified explanation). The implementation (i.e. building the machine that can actually perform those commands) can vary greatly. Intel and AMD both implement X86, but Intel's chips are better than AMD's (even on the same class of process node).

So, having an ARM ISA license doesn't help much in figuring out what the competitors are going to do.

Hope this helps.

that makes a lot of sense but the difference in intel vs amd performance (even at same node) would be due to microarchitechture differences correct? Vanilla a57 cores should be the same irrespective of how they are implemented into an SOC (at a given node). e.g. V8 is the ISA but a57 is the uarch on that isa? or is that wrong?