ST-Ericsson Nova A9600: dual-core ARM A15, PowerVR Series 6

[anexanhume]

Especially sad because this was one of the first product lines to use GloFo's 28nm FD-SOI process. I'm half curious if either of the forthcoming console SoCs will use it.

 

[Exophase]

Especially sad because this was one of the first product lines to use GloFo's 28nm FD-SOI process. I'm half curious if either of the forthcoming console SoCs will use it.

I can say with almost absolute certainty that they will not. Expect them to be on the same 28nm TSMC process the standard Jaguar SoCs will be on.

 

[tangey]

What a waste. Always sad to see promising hardware that was almost complete get axed. At least OMAP5 wasn't cancelled :/

Interestingly however not a single omap5 design win/product announcement at either CES or MWC. I would have expected some folks who were committed to omap5 before the TI "refocus", to have had product available.

It's looking increasingly likely that 5430/5432 has little if any design wins, unless amazon has a committed product coming.

 

[anexanhume]

I can say with almost absolute certainty that they will not. Expect them to be on the same 28nm TSMC process the standard Jaguar SoCs will be on.

Why is this certain? GloFo advertises FD-SOI capable of taking bulk designs as-is. AMD has a demonstrated need to keep its orders up with GloFo.

 

[Exophase]

Why is this certain? GloFo advertises FD-SOI capable of taking bulk designs as-is. AMD has a demonstrated need to keep its orders up with GloFo.

Because Jaguar and GCN were already designed for TSMC 28nm which is going to be much more mature than GF's FD-SOI (design considerations notwithstanding, as it relates to yield). Sure, GCN is going to be ported to GF and Jaguar is supposed to be very portable in the first place. But AMD isn't ready for the former or else they wouldn't be releasing Richland with VLIW4, and for the latter we already know that Jaguar will be coming on TSMC later. And I'm sure AMD expects volumes for Kabini/Temash to at least similar to those for consoles. Based on everything we've seen thus far TSMC's 28nm is probably a lot denser than GF's 28nm (Bulk or FD-SOI). These are not small chips, for PS4's at least it's probably around 250-300mm^2.. you'd really want to keep that down.

They'd be paying an extra expense for improved power consumption when it probably isn't critical enough to be worth it.

Interestingly however not a single omap5 design win/product announcement at either CES or MWC. I would have expected some folks who were committed to omap5 before the TI "refocus", to have had product available.

It's looking increasingly likely that 5430/5432 has little if any design wins, unless amazon has a committed product coming.

It doesn't need the design wins you're thinking of to be viable as a product though.. several companies are making OMAP5 modules, making it an option for lower volume and embedded devices that you'll never read about design wins for. But they'll be around. Most other companies are not making high end ARM SoCs available in this fashion so they're serving a useful market, if one that may not be large enough to justify the investment.

To be fair, it's not like you hear that much about other product wins for SoCs that aren't really out yet (what has been confirmed to contain a Tegra 4?), and it's not like Samsung where you can count on them to ship it in one of their own products. OMAP5 only stands out because it was announced so early. OMAP4 was the same way; no one knew what it'd be in until shortly before the first products shipped.

 

[anexanhume]

Because Jaguar and GCN were already designed for TSMC 28nm which is going to be much more mature than GF's FD-SOI (design considerations notwithstanding, as it relates to yield). Sure, GCN is going to be ported to GF and Jaguar is supposed to be very portable in the first place. But AMD isn't ready for the former or else they wouldn't be releasing Richland with VLIW4, and for the latter we already know that Jaguar will be coming on TSMC later. And I'm sure AMD expects volumes for Kabini/Temash to at least similar to those for consoles. Based on everything we've seen thus far TSMC's 28nm is probably a lot denser than GF's 28nm (Bulk or FD-SOI). These are not small chips, for PS4's at least it's probably around 250-300mm^2.. you'd really want to keep that down.

They'd be paying an extra expense for improved power consumption when it probably isn't critical enough to be worth it.

Everything you say makes sense; I knew TSMC wasn't one of the 28nm partners in the Samsung, Glofo etc. 28nm alliance. Why do we think their process is so much denser though?

 

[Exophase]

Everything you say makes sense; I knew TSMC wasn't one of the 28nm partners in the Samsung, Glofo etc. 28nm alliance. Why do we think their process is so much denser though?

You can make some rough density comparisons between some Samsung die shots and TSMC ones, like Apple's SoCs vs nVidia's, TI's, and Freescale's. For instance Cortex-A9s are much smaller on TSMC for the same node generation. And Samsung's process should be similar to GF's.

For another comparison, look at the VLIW5 arrays on Brazos vs Llano. One unit of 40 SPs takes about 2.08mm^2 on Brazos (TSMC 40nm) vs about 2.28mm^2 on Llano (GF 32nm). Now imagine what the difference would be look for a process both are giving the same feature size designation.

 

[anexanhume]

Thanks.

 

[tangey]

It doesn't need the design wins you're thinking of to be viable as a product though.. several companies are making OMAP5 modules, making it an option for lower volume and embedded devices that you'll never read about design wins for.

Let me rephrase then.

I would have expected some smartphone/tablet designs that were at the committed stage prior to the TI refocus, to have been announced at CES or MWC.

It would suggest that if omap5 did have any meaningful design wins in that space, they were subsequently pulled. Seems hard to believe that no one had selected omap5 for a volume tablet, I supposed there is still an outside chance for kindle ?