The AMD Execution Thread [2007 - 2017]

Status
Not open for further replies.
There's no reason why specific Kabini SKUs that don't particularly push the perf/W envelope should be delayed, unless:

1) it's a marketing decision;
2) they have real power management with Turbo, unlike most current Kabini/Temash SKUs, as this probably requires much more testing/qualification.

Turbo would be much more useful in notebook so that would seem to only leave one choice, which imo, is the obvious reason for this "delay", which is actually a reschedule.

It might not be the interesting or apocalyptic reason that many people hoped for, but the reason for the "delay" is almost certainly that Kabini would cannibalise Richland if it were made available on desktop right now.
 
AMD isn't specifically producing these low-end desktop chips - they exist already - but they are not releasing them into the market right now.

Is there anything remotely official mentioning them? I know that AMD always has all the chips it needs but always strategically decides not to release them, but some official acknowledgement that these are not actually mid-life bumps (for which Feb 2014 mass production would make sense) wouldn't hurt. Also, is it not interesting how AMD is always on a path to inventory cleaning that never gets there, because the next great thing that would take the market by storm ends up over-provisioned too?

EDIT: to be very clear, I don't think the article that got the ADF so riled up actually says what you think it does - "OMAGAD, Kabini not being mass-produced / shipping yet" - but rather that some new SKUs, which might be a mid-life bump or something, will come in 2014 / with mass production happening in early 2014. Which is not only reasonable, but would also make sense, even from AMDs standpoint, as opposed to them harvesting and stockpiling stuff for the lulz...pardon...fear of culling the trickle of Richland sales.
 
Is there anything remotely official mentioning them? I know that AMD always has all the chips it needs but always strategically decides not to release them, but some official acknowledgement that these are not actually mid-life bumps (for which Feb 2014 mass production would make sense) wouldn't hurt. Also, is it not interesting how AMD is always on a path to inventory cleaning that never gets there, because the next great thing that would take the market by storm ends up over-provisioned too?

The mentioned chips have been known about in drivers since June at least - http://www.cpu-world.com/news_2013/2013062701_AMD_leaks_model_numbers_of_future_Kabini_APUs.html

EDIT: to be very clear, I don't think the article that got the ADF so riled up actually says what you think it does - "OMAGAD, Kabini not being mass-produced / shipping yet" - but rather that some new SKUs, which might be a mid-life bump or something, will come in 2014 / with mass production happening in early 2014. Which is not only reasonable, but would also make sense, even from AMDs standpoint, as opposed to them harvesting and stockpiling stuff for the lulz...pardon...fear of culling the trickle of Richland sales.

It's not a trickle of Richland sales they are worried about, it's the majority of Richland sales, probably talking millions per month. Those 65W ~3GHz A4's and A6's with 128 SP's would be unsellable vs a 25W Kabini.
 
Nevertheless the desktop, ie itx, Kabinis (which I'm waiting for to upgrade the mediacenter) are still nowhere to be seen. You can get laptops and even some cheap integrated systems, but still no motherboards.

A theory attempting to explain that scarcity is that most of the Kabinis they're getting end up either in the next XBOX or in the next Playstation, with both of those currently building launch stockpiles. Do note that it takes (this is handwavey, before the obligatory "it's a totally custom ausumsauce different chip" retort comes from one source or another) 2 Kabinis to get a single SoC for either of those, and that as far as we know both MS and Sony are using relatively close to the median specs, so they might be gorging the majority of valid chips coming from the fab (based on the sort of capacity AMD has allocated).
 
A theory attempting to explain that scarcity is that most of the Kabinis they're getting end up either in the next XBOX or in the next Playstation, with both of those currently building launch stockpiles. Do note that it takes (this is handwavey, before the obligatory "it's a totally custom ausumsauce different chip" retort comes from one source or another) 2 Kabinis to get a single SoC for either of those, and that as far as we know both MS and Sony are using relatively close to the median specs, so they might be gorging the majority of valid chips coming from the fab (based on the sort of capacity AMD has allocated).

I was under the impression that 20% of TSMC's 28nm capacity was available. At least I think I've read something to that effect recently. If true, this shouldn't be an issue.
 
I was under the impression that 20% of TSMC's 28nm capacity was available. At least I think I've read something to that effect recently. If true, this shouldn't be an issue.

If AMD had to get any spare capacity they would still probably put more into consoles though, or mobile Kabini. Desktop Kabini must be about as far down the list as it gets for them, would literally be suiciding Richland for no gain.

Anyway, this is clearly not the execution problem that is being reported in some quarters.
 
A theory attempting to explain that scarcity is that most of the Kabinis they're getting end up either in the next XBOX or in the next Playstation, with both of those currently building launch stockpiles. Do note that it takes (this is handwavey, before the obligatory "it's a totally custom ausumsauce different chip" retort comes from one source or another) 2 Kabinis to get a single SoC for either of those, and that as far as we know both MS and Sony are using relatively close to the median specs, so they might be gorging the majority of valid chips coming from the fab (based on the sort of capacity AMD has allocated).

Huh? How on earth could MS and Sony SoCs have anything to do with Kabini availability? It's not like TSMC manufactures tons of Kabinis first and then you pick some of them and make MS/Sony SoCs out of them :???:
 
Huh? How on earth could MS and Sony SoCs have anything to do with Kabini availability? It's not like TSMC manufactures tons of Kabinis first and then you pick some of them and make MS/Sony SoCs out of them :???:

I don't think I follow. This would only be true if AMD could get as many Kabini chips from the fab as they want. Practically, I imagine they can only get a set number. That number may also be hard to change after the fact.

Not saying this is the case, but certainly plausible. I assume Sony/MS have the highest priority for Kabini chips.
 
I don't think I follow. This would only be true if AMD could get as many Kabini chips from the fab as they want. Practically, I imagine they can only get a set number. That number may also be hard to change after the fact.

Not saying this is the case, but certainly plausible. I assume Sony/MS have the highest priority for Kabini chips.

Think of it in terms of product lines. AMD has 3 different Jaguar-based chips at TSMC.

The normal Kabini chip is the first one. The laptops, tablets and desktops are all from the same die and are binned depending on whatever characteristics they have - but that is their limit, they can't be used in any of the consoles.

Then you have the PS4 die, with 8 Jaguar cores and however many SP's, and all the rest of the SoC. It can only ever be a PS4 chip.

Finally you have the Xbox die, with 8 Jaguar cores and however many SP's and all the rest of the SoC. It can only ever be an Xbox chip.

What you are getting at is that AMD might have to increase console chip output first - and you'd probably be right with that suggestion. It's unlikely that they'd be forced to do that as per any agreement, but they certainly might choose to do so. In effect they would have to give up the normal Kabini wafer starts to boost console wafer starts, which would certainly mean they have less desktop Kabini parts as well.
 
Last edited by a moderator:
I don't think I follow. This would only be true if AMD could get as many Kabini chips from the fab as they want. Practically, I imagine they can only get a set number. That number may also be hard to change after the fact.

Not saying this is the case, but certainly plausible. I assume Sony/MS have the highest priority for Kabini chips.

My point was that they're not building Kabini-chips to the SoCs, they're building whole SoCs which have absolutely nothing to do with how many Kabinis TSMC makes for AMD
 
But AMD reserves an amount of capacity on a given process for a given fab. It's unlikely they reserve individual capacities for each and every SKU.

So the point Alex made stands.
 
But AMD reserves an amount of capacity on a given process for a given fab. It's unlikely they reserve individual capacities for each and every SKU.

So the point Alex made stands.
Unless it's MS and Sony who are contracting with TSMC for their consoles. That doesn't seem all that unlikely?
 
Wouldn't AMD's margins be higher if their customers took on the contracts at the foundry?
Once the custom APU division was disclosed, the margins dropped more than some analysts expected.

Is there a particular set of hoops they can jump through to allow some kind of pass-through ownership of the produced chips to include AMD if Microsoft or Sony are paying for the wafer starts and taking the output?
 
Wouldn't AMD's margins be higher if their customers took on the contracts at the foundry?

They might not be able to do that due to cross-license restrictions with Intel. Of course I have no information one way or another, so just throwing that out there as one possible reason why they could not do that.
 
They might not be able to do that due to cross-license restrictions with Intel. Of course I have no information one way or another, so just throwing that out there as one possible reason why they could not do that.

I know why this probably wouldn't happen, but in the hypothetical case where it were permissible AMD's results would reflect that it wasn't responsible for paying for the wafer starts. That the released numbers showed a greater margin hit than expected would hint that it is in actuality not happening.
 
Unless my recollection of the latest licensing agreement is incorrect (and it might well be), whilst AMD was granted the right to go about fabbing x86 wherever they chose, there was no loophole through which AMD's partners could go about doing the same thing, even if AMD wrote the RTL and gave the design all of its love.
 
Status
Not open for further replies.
Back
Top