AMD on track for launch of Kaveri in February 2014

[Phynaz]

That doesnt make sense at all , OEMs were always first
and retail second when it comes to the initial deliveries ,
it never happened otherwise , so if retail by february
oems will be delivered the first batches in january.

Article says launch though. OEMs may well have product, but they would hold off announcing that product until CeBIT.

This is all based upon the OP's post, which specifically says OEM after channel. Kinda calls the entire slide / article into question though. Oh wait, I did say it was a fake

 

[frozentundra123456]

It kinda pains me to say this because i'm not a fan of APUs really. But, to be fair, you're comparing APUs to the Iris Pro which costs upwards of 650$. It's not like you can just simply go out and buy an Iris Pro for LGA - it's actually BGA only. And it's hella expensive. I think the more apt comparison is HD4600, to be honest. Conversely, APUs are generally sub 150$ LGA parts.

I think the point is better served by comparing apples-to-apples. Not a 650$ Iris Pro to a 100$ LGA APU, you know? I only know of one design using the Iris Pro, and that is the upcoming Macbook Pro retina. I'm sure there are other AIO systems using it, but the price performance isn't good enough to choose it over discrete nvidia graphics - a lot of AIOs are opting for the GT750 instead over the Iris Pro because of this.

Still, the Iris Pro is certainly promising - Broadwell BGA parts will be absolute beasts in terms of graphics performance next year IMO. Certainly intel knows what its doing with integrated graphics, of that there is no doubt.

I agree it is unfair to compare an apu to iris pro. The more apt comparison is to a discrete card like the HD7750 in a desktop or something like the gt650m ddr5 in a laptop. These are pretty much the minimum for a fairly well rounded gaming platform and are available at reasonable cost if one shops carefully. To me these are the levels an apu must reach to even be considered as a gaming platform.

 

[ph2000]

anyone knows what
"Fully validated dual boot SW stack" means ?

 

[monstercameron]

I agree it is unfair to compare an apu to iris pro. The more apt comparison is to a discrete card like the HD7750 in a desktop or something like the gt650m ddr5 in a laptop. These are pretty much the minimum for a fairly well rounded gaming platform and are available at reasonable cost if one shops carefully. To me these are the levels an apu must reach to even be considered as a gaming platform.

have you been inside a bestbuy? the one closest to me only has igp based systems, also do most of the local computer stores down here in miami.
It is completely fair to campare intels iris pro with an apu but for the consumer buying a lownd/cheap/budget/htpc the apu will be better value cuz cheap[er]. As for reaching 7750 lvls of perf, that is good and all but just reaching gt630 or hd6670 levels would be adequate.

 

[ph2000]

anandtech article
http://www.anandtech.com/show/7466/amd-news-a106790k-1311-beta6-drivers-fm2-motherboards-kaveri

Update: AMD sent along an official statement (which they've issued previously) on Kaveri availability: "AMD's ‘Kaveri’ high-performance APU remains on track and will start shipping to customers in Q4 2013, with first public availability in the desktop component channel very early in Q1 2014. ‘Kaveri’ features up to four ‘Steamroller’ x86 cores, major heterogeneous computing enhancements, and a discrete-level Graphics Core Next (GCN) implementation – AMD’s first high-performance APU to offer GCN. ‘Kaveri’ will be initially offered in the FM2+ package for desktop PCs. Mobile ‘Kaveri’ products will be available later in the first half of 2014." If we read "customers" as the large OEMs that make desktops, then we may or may not have actual Kaveri hardware in hand for testing this year, but we'll wait and see.

 

[Idontcare]

So AMD will start mass production of the Kaveri APUs in December 2013, and then launch them 2-3 months later in February 2014.

Doesn't that sound kind of quick?

Intel is supposed to start production of Broadwell in Q1 2014 (see this), yet they will not launch the CPUs until H2 2014 / Q4 2014.

So can AMD really make it in time? If so, how come Intel needs more time from start of production until launch? Maybe Intel is talking about start of production in Q1 2014, and not mass production, i.e. only engineering samples in Q1 2014?

Kaveri is being ramped on a mature production node (28nm) which already has lots and lots of pre-existing capacity installed and operating.

Greenlighting a production run in that environment leaves you with a rather small lead-time to contend with. Basically just the fab cycle-time (~8 weeks) plus ex-fab time (2-3wks) entailing testing, packaging, and distribution.

With Intel's situation, or any other company who is simultaneously ramping both a new product as well as a new node, the timelines are considerably more challenging as it becomes a multi-body problem with critical interdependencies.

The timing differences between these two companies, for these two specific products, is not surprising at all (assuming the information in the OP is correct).

 

[Fjodor2001]

Here, I'll explain to you what you posted.

Your article says OEM sometime after February.

CeBIT is a business IT show that occurs in March. Let's see, an OEM is in the business space and March is after February. Yes, I could see a product launch happening there.

I could also see an OEM product launch happening during the back to school season as ikachu stated. That's also when most OEM's refresh their business notebook lineups in order to catch the fourth quarter business buyers. I leave it to you to research why businesses do most of their buying in the fourth quarter.

Well, first you said that it would be reasonable for OEMs to launch Kaveri in Q3 2014, then in your next post you said that Cebit which is in March 2014 was a likely launch date for them. That is contradictory and confusing. So I don't think it's strange to expect an explanation, which others also requested.

Also, I don't think you can only look at what events are suitable for launching a product when determining when it's likely to actually be launched. You should also take into account e.g. when the product is entering production, and how far from that the events are in time. And if production starts in December 2013, then Cebit in March 2014 (~4 months after production starts) vs The after school season in August 2014 (~9 months after production starts) are two very different time frames.

 

[frozentundra123456]

have you been inside a bestbuy? the one closest to me only has igp based systems, also do most of the local computer stores down here in miami.
It is completely fair to campare intels iris pro with an apu but for the consumer buying a lownd/cheap/budget/htpc the apu will be better value cuz cheap[er]. As for reaching 7750 lvls of perf, that is good and all but just reaching gt630 or hd6670 levels would be adequate.

What does best buy have to do with it? They aren't known for selling gaming systems. If you are talking about basic use, yes, almost any igp is adequate. You dont even need an amd apu. My work laptop is a sandy bridge i5, with optimus and a discrete card. Obviously I dont game on it, and in that usage, I have never even seen the discrete card kick in.

On the desktop, anything less than 7750 levels of performance can be matched for very little more cost with a cheap cpu like the Athlon x4 750k and a discrete card. Why one would want to settle for less is a mystery to me, but if that is your choice, each to his own I guess.

I can understand sacrificing gaming power in a laptop, because it is not possible to add a discrete card. I simply used the GT650M as an example, as that is pretty much the minimum to play current demanding titles at even relatively modest settings at 768p. If one is willing to settle for playing older games or less demanding ones, then you can get by with less.

 

[Fjodor2001]

So you see, when you post random articles without facts (you seem to do that a lot) there is a lot that is left open to interpretation. Now since this is a discussion forum, I can only assume you wanted people to discuss what you have posted which is what we are doing. I even learned something from ikachu.

I honestly don't know what you mean by that the articles don't contain any facts. But if you're serious about that then I think you should direct your complaint to the source, which is VR-Zone and AMD in this case.

As for me posting threads in general, I do so because I think it's subjects that the members of this forum find interesting and would like to discuss. Judging by the number of posts and views in the threads it does seem to be appreciated be a lot of members. And I in turn appreciate the feedback and comments by the members in general, including fair criticism.

However I do not expect to get my head bitten off by some dick swinging dude with an attitude just because I cannot explain what the number-letter combination "2C" means in a leaked slide published in an article on someone else's website. So maybe you for once instead should try to show some appreciation of the ones who post threads on subjects you like to discuss and comment on, instead of just displaying rage and launching personal attacks against them.

 

[Fjodor2001]

Kaveri is being ramped on a mature production node (28nm) which already has lots and lots of pre-existing capacity installed and operating.

Greenlighting a production run in that environment leaves you with a rather small lead-time to contend with. Basically just the fab cycle-time (~8 weeks) plus ex-fab time (2-3wks) entailing testing, packaging, and distribution.

With Intel's situation, or any other company who is simultaneously ramping both a new product as well as a new node, the timelines are considerably more challenging as it becomes a multi-body problem with critical interdependencies.

The timing differences between these two companies, for these two specific products, is not surprising at all (assuming the information in the OP is correct).

Goods points! That makes sense, assuming that production doesn't mean mass production. Because by the time the product actually is in mass production I would assume most production issues to already have been ironed out.

So if that is the case, should we interpret Intel's statement that "Broadwell production will begin in Q1 2014" as that it's production, and not mass production, that will begin then? That seems kind of odd though, because Intel said that they are having yield issues currently (i.e. production issues due to new process node), but expect them to be ironed out soon so that production can start in Q1 2014. I interpreted that as that they plan to enter mass production in Q1 2014. But maybe I am missing something here...? :hmm:

 

[sm625]

I agree it is unfair to compare an apu to iris pro.

But that is what AMD needs to do. Occasionally outperforming intel's $600 chip with a $200 chip is what has kept AMD around all these years. Besides, iris pro is nothing anyway. Look at this:

Granted, that's an i7 that is probably only running at 2.6GHz during that benchmark. But really, cpu speed doesnt have that much impact.

I doubt iris pro @ 100W would score higher than 32 at that benchmark. It is really not a good gpu. I dont think its too much to ask that AMD be able to beat it by 50% or more.

 

[Idontcare]

Goods points! That makes sense, assuming that production doesn't mean mass production. Because by the time the product actually is in mass production I would assume most production issues to already have been ironed out.

So if that is the case, should we interpret Intel's statement that "Broadwell production will begin in Q1 2014" as that it's production, and not mass production, that will begin then? That seems kind of odd though, because Intel said that they are having yield issues currently (i.e. production issues due to new process node), but expect them to be ironed out soon so that production can start in Q1 2014. I interpreted that as that they plan to enter mass production in Q1 2014. But maybe I am missing something here...? :hmm:

You have to decide what you take the definition of "mass production" to mean.

Each company defines it differently for internal usage. It is a not a metric with industry-wide standardization. (akin to node dimensions)

I have not seen any two companies define either "production" or "mass production" to mean the exact same thing between companies.

What that tells us, as laymen, is that the term or phrase itself is largely irrelevant in the larger scheme of how these businesses operate.

There is something else going on, in and around "production", that is far and away more important to these businesses than the specifics of the definition of what "in production" might entail.

To say it differently, this is the small stuff, and big business doesn't sweat the small stuff, so neither should we.

 

[beginner99]

On desktop these don't matter at all. Anyone gaming will go discrete anyway and is better of with intel.

APUs are needed for mobile and at a price that is below intel dual core + discrete which in the past was way faster than APUs especially the gimped ones.

 

[mikk]

Mobile schedule is unknown but knowing the priorities and the early NDA expiration date I'd say that mobile comes first

Desktop comes first. Don't expect mobile before mid-2014.

 

[inf64]

Desktop comes first. Don't expect mobile before mid-2014.

Yeah I saw update from AT article. Odd priority I must say, shouldn't mobile be a higher priority than desktop?

 

[Idontcare]

Yeah I saw update from AT article. Odd priority I must say, shouldn't mobile be a higher priority than desktop?

Mobile is usually a higher priority but the silicon that is exiting the fabs may not be ideal for binning to the mobile SKUs.

Sometimes you have to go to battle with the army you have, not the one you wish you had.

 

[mikk]

Yeah I saw update from AT article. Odd priority I must say, shouldn't mobile be a higher priority than desktop?

Logical explanation below. Furthermore OEMs have to build and validate notebooks unlike a simple CPU channel launch.

AMD expects to start shipping Kaveri in a late Q4 timeframe this year. These parts will be desktop at first and will transition to mobile in 2014. AMD wants (and needs) to get these parts out in a timely manner, and they are pulling in the launch as much as possible. Hence the desktop first release while they refine production to be able to adequately address the mobile space. Achieving good bins and yields at the higher TDP is easier than trying to hit those numbers for a 35 watt and below product line.

 

[Fjodor2001]

You have to decide what you take the definition of "mass production" to mean.

Each company defines it differently for internal usage. It is a not a metric with industry-wide standardization. (akin to node dimensions)

I have not seen any two companies define either "production" or "mass production" to mean the exact same thing between companies.

What that tells us, as laymen, is that the term or phrase itself is largely irrelevant in the larger scheme of how these businesses operate.

There is something else going on, in and around "production", that is far and away more important to these businesses than the specifics of the definition of what "in production" might entail.

To say it differently, this is the small stuff, and big business doesn't sweat the small stuff, so neither should we.

Well, I don't have experience from production of semiconductor chips. But I do have some experience from production of consumer electronic devices. And there it's a pretty clear milestone that you pass. I.e. first you have prototype production, and verify that the production lines work ok, the HW & SW work as expected after leaving the production line, iron out all (or most) known issues, and make sure the yield is above some threshold. Then you take a decision regarding whether to pass the milestone to enter high volume production or not.

I assume it's similar for production of semiconductor chips. E.g. you do not enter high volume production if you have too low yield, simply because the cost per chip would be too high. That decision is of course not clear cut and the threshold may vary, but I find it hard to see that there should not be some decision like that made. So the distinction should be pretty clear. The question falls back to what Intel means by that they will start production of Broadwell chips in Q1 2014. Are they talking about prototype / engineering samples, or high volume production intended for the end consumer?

 

[tweakboy]

I can't wait. NOT! Is it me or is there like 300 diff AMD models. gl

 

[Idontcare]

Well, I don't have experience from production of semiconductor chips. But I do have some experience from production of consumer electronic devices. And there it's a pretty clear milestone that you pass. I.e. first you have prototype production, and verify that the production lines work ok, the HW & SW work as expected after leaving the production line, iron out all (or most) known issues, and make sure the yield is above some threshold. Then you take a decision regarding whether to pass the milestone to enter high volume production or not.

I assume it's similar for production of semiconductor chips. E.g. you do not enter high volume production if you have too low yield, simply because the cost per chip would be too high. That decision is of course not clear cut and the threshold may vary, but I find it hard to see that there should not be some decision like that made. So the distinction should be pretty clear. The question falls back to what Intel means by that they will start production of Broadwell chips in Q1 2014. Are they talking about prototype / engineering samples, or high volume production intended for the end consumer?

You still have yet to define "high volume production".

I'm not arguing that individual companies don't have their own definitions for these phrases, because of course they do and the milestones are all laid out in girded fashion...but the specific phrases we like to think about are not defined in the same ways we want to think about them.

If Company A has a production capacity of 100 units per month and they decide to put widget X into "mass production", is widget X going to command 10% (10 units per month) of Company A's production capacity? 20%? 5%?

And if Company B has a production capacity of 5000 units per month and they decide to put essentially the same widget X into "mass production", is widget X going to be considered to be in "mass production" when it reaches a comparatively paltry 0.2% (10 units per month) capacity loading?

These aren't facetious questions, but they are rhetorical questions because you really need to think about what it is that you think these companies mean when they reference "production" before you can hope to start evaluating and judging their abilities to meet and deliver on those "in production" timelines.

There is a big difference between knowing that a logical procession of events must be occurring, versus knowing exactly how those milestones are internally defined at any given manufacturing company (made all the more complicated when you have a fabless company like AMD in the comparison ).

 

[Enigmoid]

It kinda pains me to say this because i'm not a fan of APUs really. But, to be fair, you're comparing APUs to the Iris Pro which costs upwards of 650$. It's not like you can just simply go out and buy an Iris Pro for LGA - it's actually BGA only. And it's hella expensive. I think the more apt comparison is HD4600, to be honest. Conversely, APUs are generally sub 150$ LGA parts.

I think the point is better served by comparing apples-to-apples. Not a 650$ Iris Pro to a 100$ LGA APU, you know? I only know of one design using the Iris Pro, and that is the upcoming Macbook Pro retina. I'm sure there are other AIO systems using it, but the price performance isn't good enough to choose it over discrete nvidia graphics - a lot of AIOs are opting for the GT750 instead over the Iris Pro because of this.

Still, the Iris Pro is certainly promising - Broadwell BGA parts will be absolute beasts in terms of graphics performance next year IMO. Certainly intel knows what its doing with integrated graphics, of that there is no doubt.

There are much cheaper iris pro parts.

But that is what AMD needs to do. Occasionally outperforming intel's $600 chip with a $200 chip is what has kept AMD around all these years. Besides, iris pro is nothing anyway. Look at this:

Granted, that's an i7 that is probably only running at 2.6GHz during that benchmark. But really, cpu speed doesnt have that much impact.

I doubt iris pro @ 100W would score higher than 32 at that benchmark. It is really not a good gpu. I dont think its too much to ask that AMD be able to beat it by 50% or more.

Intel is held back a lot by drivers. If you look at HD 4000 at launch and at haswell release you can see a nice increase. Its also at intel's Achilles heel, high res and settings.

Granted, its only slightly faster but is using less power and has about twice the CPU perf. I don't think intel HD 5200 is that good, it uses quite a bit of power and a lot of transistors for okay performance.

But considering where intel was 4 years ago with its igps its been impressive.

 

[Fjodor2001]

You still have yet to define "high volume production".

I'm not arguing that individual companies don't have their own definitions for these phrases, because of course they do and the milestones are all laid out in girded fashion...but the specific phrases we like to think about are not defined in the same ways we want to think about them.

If Company A has a production capacity of 100 units per month and they decide to put widget X into "mass production", is widget X going to command 10% (10 units per month) of Company A's production capacity? 20%? 5%?

And if Company B has a production capacity of 5000 units per month and they decide to put essentially the same widget X into "mass production", is widget X going to be considered to be in "mass production" when it reaches a comparatively paltry 0.2% (10 units per month) capacity loading?

These aren't facetious questions, but they are rhetorical questions because you really need to think about what it is that you think these companies mean when they reference "production" before you can hope to start evaluating and judging their abilities to meet and deliver on those "in production" timelines.

There is a big difference between knowing that a logical procession of events must be occurring, versus knowing exactly how those milestones are internally defined at any given manufacturing company (made all the more complicated when you have a fabless company like AMD in the comparison ).

I think you're making it overly complicated. I would say it's at the stage where the criteria (e.g. yield targets) for high volume production is met, and the company starts producing units aimed for the end consumer. I.e. no longer prototypes / engineering samples. The actual volumes may of course differ per company, product, and production capacity available at the time. At least that's the definition I'm used to from working with production of consumer electronics.

Or what do you think Intel means by their statement that they will begin production of Broadwell in Q1 2014?

 

[schmuckley]

All I know is..There have been some produced..
Possibly will see "leaked" stuff in 2 weeks.
I have $$ set aside for FM2+ build :ninja:
Whether it gets allocated to that is a diff story
I wanna see some results, 1st.
Super Pi is @ top of the list, too.

 

[Shivansps]

There are much cheaper iris pro parts.



Intel is held back a lot by drivers. If you look at HD 4000 at launch and at haswell release you can see a nice increase. Its also at intel's Achilles heel, high res and settings.

Granted, its only slightly faster but is using less power and has about twice the CPU perf. I don't think intel HD 5200 is that good, it uses quite a bit of power and a lot of transistors for okay performance.

But considering where intel was 4 years ago with its igps its been impressive.

Iris Pro problem is high quality, thats is where the 128MB runs short.

 

[Enigmoid]

Iris Pro problem is high quality, thats is where the 128MB runs short.

Yep. Kinda shaky for newer games.