TI announces OMAP5

As Periander mentioned, NV did license the A15 but they're not one of the lead partners (who are, in order of apparent 'leadness', TI/ST-E/Samsung).
Sorry for OT, but would Apple even be named as an official lead partner, even if they are one (wouldn't fit their release schedule for new SoCs in spring, but who knows)?
 
Sorry for OT, but would Apple even be named as an official lead partner, even if they are one (wouldn't fit their release schedule for new SoCs in spring, but who knows)?

You realise that the topic here is about future 28nm SoC and Apple's coming SoC won't be most likely anywhere smaller than 45nm besides other aspects?
 
You realise that the topic here is about future 28nm SoC and Apple's coming SoC won't be most likely anywhere smaller than 45nm besides other aspects?
Arun listed three lead partners for the A15 and I was just asking if Apple could also be an "unnamed" lead partner for Cortex-A15. Since Apple will likely use around 150 million high-end SoCs (A5) in a year or so, IMHO it must be one of the top licensees for high-end ARM cores. So IMHO it's not too far out there to assume that ARM and Apple would also have an interest in working together on future high-end CPU designs.
 
Last edited by a moderator:
Quite impressive is this race in the mobile market.
I just wonder who will keep up with this tempo (companies that actualy release those products and also software) as the gadgets gets more sophisticated.
 
Sorry for OT, but would Apple even be named as an official lead partner, even if they are one (wouldn't fit their release schedule for new SoCs in spring, but who knows)?
They might not be named, but ARM did explicitly say they had three lead partners iirc, which might imply Apple is only be in the second licensing round (like NVIDIA?) - another possibility is the PA Semi guys were indeed working on a custom ARM core and they have no intention of ever licensing it. We'll see.

Also, I moved the Cortex-A15 discussion to my article's feedback thread since it's really more on topic there than here.
 
Arun listed three lead partners for the A15 and I was just asking if Apple could also be an "unnamed" lead partner for Cortex-A15.

That's besides the point. Again did it possibly strike you that die area isn't infinite for today's conditions under 45nm besides it being a wee bit too early for A15 designs to be ready just yet? I don't think a A15 being by 50% faster clock for clock with a A9 comes for free in terms of transistors for each either.

Since Apple will likely use around 150 million high-end SoCs (A5) in a year or so, IMHO it must be one of the top licensees for high-end ARM cores. So IMHO it's not too far out there to assume that ARM and Apple would also have an interest in working together on future high-end CPU designs.

Is there a link that documents those 150M SoCs even if it's just speculation?

No it's not too far to speculate that ARM will use Eagle eventually but not just yet.
 
That's besides the point. Again did it possibly strike you that die area isn't infinite for today's conditions under 45nm besides it being a wee bit too early for A15 designs to be ready just yet? I don't think a A15 being by 50% faster clock for clock with a A9 comes for free in terms of transistors for each either.



Is there a link that documents those 150M SoCs even if it's just speculation?

No it's not too far to speculate that ARM will use Eagle eventually but not just yet.

No link, but give iphone sold 16M last quarter, has launched on verizon, is extending its reach in other countries, its not too hard to see them sell 70M in a year. Dial in 35M+ itouches (they sold over 10M last quarter), maybe 30M ipads and a 2-3M Apple TVs you aren't a kick in the ass away from 140M.
 
That's besides the point. Again did it possibly strike you that die area isn't infinite for today's conditions under 45nm besides it being a wee bit too early for A15 designs to be ready just yet? I don't think a A15 being by 50% faster clock for clock with a A9 comes for free in terms of transistors for each either.
I'm sorry, but I don't understand why you think I was talking about this year's Apple SoC. I thought (maybe mistakingly) that it was clear that I was talking about a possible Cortex-A15 SoC sometime in 2012 (like the OMAP5). The possibility of 2011 and 45nm didn't even cross my mind, I guess that's why I didn't specify the date and process.

No link, but give iphone sold 16M last quarter, has launched on verizon, is extending its reach in other countries, its not too hard to see them sell 70M in a year. Dial in 35M+ itouches (they sold over 10M last quarter), maybe 30M ipads and a 2-3M Apple TVs you aren't a kick in the ass away from 140M.
I agree. It's highly speculative of course but given Apple's past performance and even just the average smartphone growth rate (plus iPad, iPod touch and Apple TV), around 150 million A5s in 12-15 months (04/2011-iPhone 6) seems reasonable to me. It's also right in the ballpark of (maybe optimistic) analysts' numbers for iPad 2, iPhone 5 and part of iPod touch 5th gen and Apple TV 3 numbers.
 
Last edited by a moderator:
Is there a link that documents those 150M SoCs even if it's just speculation?

It's middle of the road analyst estimation.
More substantial perhaps is that Samsung has officially stated that they will supply Apple with 4 times the number of SoC wafers during 2011 vz. 2010 (20000 wafer vs. 5000). You could do some math : a wafer is 70000mm2 and an A4 SoC is 53mm2 and get some interesting numbers. However, assuming 45nm there is every reason to believe that the size of a portion of the SoCs will grow a bit, nor do we have exact yields.

Nevertheless, when sales projections and material component reservations match up, it's believable. Say 125-200 million iOS devices next year. A wide window, but any way you look at it, that is one hell of a market. And all ship to active users, as compared with administrative/school/et cetera Windows systems. If you compare to the size of the console market... Remarkable.
 
Exactly because the die area of the coming SoCs is likely to increase significantly over the current 53mm2 of the A4 did I ask where those numbers come from. I think but am not sure that the iPhone3GS SoC was somewhere over 70mm2@65nm and I wouldn't be in the least surprised if their next SoC ends up even slightly larger than that.

Of course is it one hell of a market and that's exactly the reason why it becomes slowly clearer why SAMSUNG is making a line of weird decisions for their own product lines probably in order to keep a customer like Apple.

From what I've read the Galaxy SL are going to get TI OMAP3630, Galaxy S2 smart-phones and tablets might employ partially or entirely 3rd party SoCs (I guess we'll have a clearer picture after the MWC in Barcelona and the respective announcement(-s)) and finally Orion and especially their Mali 400MP design choice. If I am to believe ARM's die area quotes for Mali400 (4.7mm2@65LP per core) it sounds quite a bit cheaper than a SGX543 MP2 for example. Of course are there other trade-offs, but that's not important after all. Plus I wouldn't be surprised either if ARM has pushed IP prices down to very low levels; at those quantities and considering how large the employment of Orion could be I don't see why a bundled offer would had been impossible.

I've been puzzled for quite some time considering Samsung's overall decisions/strategy for 2011. Now that a few further details emerged it sounds suspiciously like their doing the best they can to reduce costs in order to serve Apple's manufacturing needs at full capacity.
 
Even though it's incidental and completely natural due to the different platforms they support, it's at least a little funny that Apple is now moving down from a DirectX 9 Feature Level 3 capable mobile GPU to a more GL ES 2.0 positioned one at the same time TI is going in the exact opposite direction.
 
Of course is it one hell of a market and that's exactly the reason why it becomes slowly clearer why SAMSUNG is making a line of weird decisions for their own product lines probably in order to keep a customer like Apple.
I agree that should have a little bit to do with it, but most important of all Samsung is structured very differently from a company like Nokia. For example they've got many different teams that are *competing* with each other with potential products (even on the same OS). I think you'd be hard pressed to find a SoC in the industry that part of Samsung hasn't considered at one point. They nearly certainly do have a preference for in-house SoCs at a given price point if it's as good as a competitor's, but even if they had infinite fab capacity for them they still wouldn't get *any* automatic design wins. And they wouldn't get an overwhelming majority of the business either even if they were by far the best processors in the industry (which they aren't, but they are fairly competitive at least).
 
Even though it's incidental and completely natural due to the different platforms they support, it's at least a little funny that Apple is now moving down from a DirectX 9 Feature Level 3 capable mobile GPU to a more GL ES 2.0 positioned one at the same time TI is going in the exact opposite direction.

Since when is SGX535 DX9.0 L3? Do you see anywhere a higher than 2048*2048 max texture size support for SGX530/5 or even SGX540? Apart from SGX545 there aren't any other SGX cores outside 544 and 554 that are DX9.0/L3.
 
Since when is SGX535 DX9.0 L3? Do you see anywhere a higher than 2048*2048 max texture size support for SGX530/5 or even SGX540? Apart from SGX545 there aren't any other SGX cores outside 544 and 554 that are DX9.0/L3.
I'm not sure if SGX535 is DirectX9 Level 3, but it's definitively DirectX9 at some level or another ;) As for Apple, who knows what their long-term plans for API support are I suppose, heh. For example they haven't looked as enthusiastic about OpenCL lately.
 
Exactly because the die area of the coming SoCs is likely to increase significantly over the current 53mm2 of the A4 did I ask where those numbers come from. I think but am not sure that the iPhone3GS SoC was somewhere over 70mm2@65nm and I wouldn't be in the least surprised if their next SoC ends up even slightly larger than that.
With 20000 wafers per months you end up with enough SoCs for more than 150 million A5, even if the A5 is 80mm2@45nm and some A4s have to be produced as well. Some light number crunching:
http://www.macobserver.com/tmo/forums/viewthread/79535/

Are there any estimates out there of how much TI is going to sell of the OMAP4?
 
The texture size is artificially limited, it supports more than that iirc ;) But either way, as I said I don't remember if SGX535 supports DirectX9 Level 3 (so Lazy8s' post could be wrong) but it does support at least one DirectX9 level/profile where SGX543 does not support any at all. This could just be timeframe thing though (i.e. Apple didn't care about DX9 on SGX535 but it was the most powerful core they could get in the 3GS timeframe or they do care about DX9-level features long-term but not enough to wait for SGX544 on the A5 generation)
 
Sorry, I confused the terminology.

I meant DirectX 9.0c for the 535, yet 5x0 processors also apparently exceed Shader Model 3.0. All are obviously referring to different contexts/features, but it's just too ambiguous here.
 
Maximum texture size for 535 is 4Kx4K, but can obviously be 2Kx2K depending on configuration of other things.

Ok another lesson learned heh. So is the SGX535 DX9.0 L3.0 or not after all?
 
Back
Top