Low-cost emerging market SoC/phone discussion

[Arun]

Skipping L2 cache even? Now that sounds like real cheap .

600MHz Cortex-A5 with 32/32KB of L1 cache and NEON takes ~0.75mm² on 40LP iirc. The 1MB of L2 cache on Tegra 2 takes ~3.8mm². So adding just 256KB of L2 cache would more than double the die area, and I'm not sure 128KB would be worth the trouble. Maybe with 16/16KB L1...

Do you have any numbers for that? I've always wondered about die sizes of these SOCs (and how large the invidiual parts of it are) but no die shots tend to get published

I do, but that's another story for another day Just two quick numbers off the top of the my head: Infineon XMM2130 is ~25mm² including RF/FM/PMU while the Qualcomm MSM6290 is 42mm² for the baseband alone (both on 65nm).

Also maybe that's different in the US but I'm not sure you could really sell a 2G smartphone (even if it's cheap) in Europe. The only 2G smartphone I even know of was the original iPhone...

I expect the same EDGE chips to be viable for both feature phones and smartphones. There are plenty of 2G feature phones out there, do you expect no manufacturer to switch to Android if they can do so at zero extra cost? It won't be for western markets as much as developing ones, but even here I can see a small opportunity.

 

[Deleted member 13524]

So there won't be any Cortex A5 in smartphones?
That's strange, the performance/power shown in ARM's website would suggest otherwise.

Also maybe that's different in the US but I'm not sure you could really sell a 2G smartphone (even if it's cheap) in Europe. The only 2G smartphone I even know of was the original iPhone...

There are some "recent" low-end Symbian^1 models without 3G. Nokia 5530, 5228 and 5250, for example.

Of course, they all made sense over a year ago. That price point has now been fully replaced with MSM7227/7225 Android phones from ZTE and Huawei, with carrier branding.

 

[Arun]

So there won't be any Cortex A5 in smartphones?
That's strange, the performance/power shown in ARM's website would suggest otherwise.

Heh, the Cortex-A5 marketing has been weird to say the least. We might see it in HSPA smartphones, but based on my understanding not from: Broadcom (1xA9), ST-Ericsson (1xA9), Infineon (Atom), NVIDIA (high-end is all we need), Texas Instruments (boohoo we lost Nokia), or Qualcomm (cheaper would mean less money for us). So that only leaves Mediatek in the HSPA market.

But as I said, I do expect to see it in EDGE and TD-SCDMA smartphones down the line. I don't think GSM/GPRS will shrink below 65nm, but it makes a lot of sense to shrink EDGE to 40nm (probably not 28nm) and it makes sense to upgrade from ARM9 to Cortex-A5 even if they didn't target ultra-low-cost smartphones (which I expect they will).

 

[Erinyes]

We're not talking $250 phones here. We are potentially talking as low as $100 unsubsidised with a BoM of less than $75. Every cent counts. The main differences with the MSM7227 are the CPU for faster applications and browsing (the most important use cases by far) and, even more importantly, lower cost.

I agree, however, that a MSM7230A on 28LP would be a very big deal. 1.4GHz Krait, 720p encode/decode, 14.4Mbps HSPA+ and all at a cost very similar to these 40nm chips. It remains to be seen whether that will happen and in what timeframe.

While i agree with your point i would be very surprised if Cortex A9 based phones manage to hit the $100 pricepoint by next year. They'll definitely go down to the $200-250 pricepoint by next year though

Would a MSM 7230A on 28nm be cheap enough? 28nm wafers will be quite expensive initially so would the reduced die area be worth the extra wafer cost or are they better off sticking to 45/40nm?

Strangely, the MSM7227 seems to be in just about any midrange smartphone these days (so is the chip really expensive? In any case looks to be the fastest ARM11 chip sold in smart phones, might not even be far behind some of the slower-clocked Cortex-A8), and roughly half of them only support video recording with QVGA/15 fps - way below what the chip is supposed to be able to handle...

Yep, pretty much every midrange phone seems to be using the MSM 7227, these days (I had a ZTE blade with an MSM 7227 myself ). Baseband integration means lower cost and a cheaper PCB. And its been in production for what two years now so it should be fairly cheap as well.

Afaik it supports VGA at 30 fps, donno why some manufacturers only enable QVGA at 15 fps

 

[mczak]

600MHz Cortex-A5 with 32/32KB of L1 cache and NEON takes ~0.75mm² on 40LP iirc. The 1MB of L2 cache on Tegra 2 takes ~3.8mm². So adding just 256KB of L2 cache would more than double the die area, and I'm not sure 128KB would be worth the trouble. Maybe with 16/16KB L1...

Though if you say 25mm² for some SOC 1 mm² for 256KB L2 cache doesn't sound like that much even if it doubles cpu area (granted going A9 might only add another 2 mm² or so) . Though I don't know how much it helps performance, it's difficult to get comparable numbers for different SOCs (but the L2 cache definitely seems to help the ever-so-popular arm11 msm7227).

I do, but that's another story for another day Just two quick numbers off the top of the my head: Infineon XMM2130 is ~25mm² including RF/FM/PMU while the Qualcomm MSM6290 is 42mm² for the baseband alone (both on 65nm).

Wow that's huge for just the baseband. Does it have to be that large?

I expect the same EDGE chips to be viable for both feature phones and smartphones. There are plenty of 2G feature phones out there, do you expect no manufacturer to switch to Android if they can do so at zero extra cost? It won't be for western markets as much as developing ones, but even here I can see a small opportunity.

I'm not sure switching to android might incur other costs. For instance, those low-end nokia smartphones mentioned come with 128MB ram, 256MB rom. That might not be enough to run android "reasonably" (and a 500Mhz A5 without L2 wouldn't be much of an upgrade for these anyway, they come with a 430Mhz arm11), not to mention even cheaper feature phones probably have even less memory today.

 

[Arun]

While i agree with your point i would be very surprised if Cortex A9 based phones manage to hit the $100 pricepoint by next year. They'll definitely go down to the $200-250 pricepoint by next year though

Sigh, for the last time...

MSM7227: 65nm, 7.2Mbps baseband, ARM11 with FPU, 256KB of L2 cache, WVGA encode/decode, 1xTMU OpenGL ES 2.0 Adreno 200 GPU, 8MPix camera
U4500: 40nm, 7.2Mbps baseband, Cortex-A9 with NEON, 256KB of L2 cache, WVGA encode/decode, 1xTMU OpenGL ES 2.0 Mali-400 GPU, 8MPix camera

Everything is similar except for the process technology and the Cortex-A9 with NEON which will add less than 1.5mm² of silicon. That's ~10x less than the 7.2Mbps baseband alone! There is absolutely no reason to associate the inclusion of a single-core Cortex-A9 with a higher price point. The process shrink will make this a cheaper chip with a significant mass market opportunity (same for the Broadcom equivalent).

Would a MSM 7230A on 28nm be cheap enough? 28nm wafers will be quite expensive initially so would the reduced die area be worth the extra wafer cost or are they better off sticking to 45/40nm?

Qualcomm is focusing on 28LP which is still SiON, not High-K. The premium in the early 2012 timeframe should be fairly negligible compared to the die area savings on a chip with so much digital logic. Also, MSM7230 has a rather pointless 64-bit LPDDR1/2 memory bus which could be replaced by a 32-bit LPDDR2 for cost savings in that timeframe.

Yep, pretty much every midrange phone seems to be using the MSM 7227, these days (I had a ZTE blade with an MSM 7227 myself ). Baseband integration means lower cost and a cheaper PCB. And its been in production for what two years now so it should be fairly cheap as well.:

Oh, it IS fairly cheap, but it has nothing to do with production costs. It's still a pretty big chip and there's no magical reason why it should be cheaper than other 65nm chips simply because it has been in production longer. Consider this instead: http://www.intomobile.com/2011/03/0...ling-our-chips-cheap-drive-smartphone-market/ - I assume it refers to the MSM7230 to some extent as well.

 

[Arun]

Though if you say 25mm² for some SOC 1 mm² for 256KB L2 cache doesn't sound like that much even if it doubles cpu area (granted going A9 might only add another 2 mm² or so).

It's ~25mm² for the SoC on 65nm and the extra 1mm² is on 40nm. I'd certainly expect less than 20mm² on 40nm so this would increase die size by >5% (and probably ~10% if only considering the digital part). That's much more than the cost of upgrading from ARM9 to Cortex-A5.

Though I don't know how much it helps performance, it's difficult to get comparable numbers for different SOCs (but the L2 cache definitely seems to help the ever-so-popular arm11 msm7227).

The MSM7201A did not just lack L2 cache - it didn't even have a FPU! All FP operations were done in software as if we were still in the 1980s. I'm not sure how significant this was for most handheld applications which are very integer-centric, but it does make it hard to judge the real benefit of the 256KB L2. I'm also not sure how much L1 cache the MSM7201A had - since they were penny pinching on everything else, I wouldn't be surprised if it was only 16/16KB.

Wow that's huge for just the baseband. Does it have to be that large?

Qualcomm has the fastest basebands, but they're also slightly larger than many competitors. I don't have precise numbers but I think Infineon's XMM6160 in the iPhone 4 is closer to 30mm². On the other hand, the Qualcomm MSM8200 for 21/28Mbps HSPA+ is about 100mm² on 65nm. Yes, 100mm², I'm not kidding. It's probably the largest baseband ever though (excluding ones with integrated application processors) and even the Qualcomm MDM9600 for 100Mbps LTE is slightly smaller on 45nm (I don't think I'm allowed to give the precise number, but it's fair to say ]80;100[ mm²)

I'm not sure switching to android might incur other costs. For instance, those low-end nokia smartphones mentioned come with 128MB ram, 256MB rom. That might not be enough to run android "reasonably" (and a 500Mhz A5 without L2 wouldn't be much of an upgrade for these anyway, they come with a 430Mhz arm11), not to mention even cheaper feature phones probably have even less memory today.

Yeah, zero extra cost isn't entirely fair, but it IS practically zero extra cost for the SoC manufacturer. So it makes a lot of sense for them to allow OEMs to do it if they want.

Also 430MHz ARM11 is for Nokia's lowest-end 3G smartphones. I promise you their S40 EDGE phones have significantly slower processors, and obviously any Cortex-A5 used on a 3G SoC would have L2 cache (and maybe higher speed since ARM's own TSMC 40LP hard macro reaches 600MHz).

 

[Entropy]

Everything is similar except for the process technology and the Cortex-A9 with NEON which will add less than 1.5mm² of silicon. That's ~10x less than the 7.2Mbps baseband alone! There is absolutely no reason to associate the inclusion of a single-core Cortex-A9 with a higher price point.

There is one piece of the puzzle missing, for me at least. Are the licensing costs of the A9 and the A5 identical?

 

[Arun]

There is one piece of the puzzle missing, for me at least. Are the licensing costs of the A9 and the A5 identical?

A9 would be higher, but keep in mind nearly everyone will take a multi-use license, and they'll need the A9 for other markets anyway. So the real question is royalties which would certainly be higher, but the difference in percentage terms will also nearly certainly be smaller than the difference in die area.

I think ARM expected companies to favour dual-core A5 over single-core A9 for power efficiency and marketing reasons, but I don't think that's going to happen much if at all. I expect most implementations to be single-core (Samsung's old roadmap had a single-core variant in 2010/2011 and a dual-core in 2012/2013 - we'll see what happen to them). Obviously this is all still speculative (who knows what these companies will do in the end) but you get the idea.

 

[mczak]

It's ~25mm² for the SoC on 65nm and the extra 1mm² is on 40nm. I'd certainly expect less than 20mm² on 40nm so this would increase die size by >5% (and probably ~10% if only considering the digital part). That's much more than the cost of upgrading from ARM9 to Cortex-A5.

Ok but even if it's 10% of the chip it's still not that much in absolute terms. How much does 1 additional mm² cost on 40nm? Granted maybe it does make a difference if you want to end up with a <100$ phone...

The MSM7201A did not just lack L2 cache - it didn't even have a FPU! All FP operations were done in software as if we were still in the 1980s. I'm not sure how significant this was for most handheld applications which are very integer-centric, but it does make it hard to judge the real benefit of the 256KB L2. I'm also not sure how much L1 cache the MSM7201A had - since they were penny pinching on everything else, I wouldn't be surprised if it was only 16/16KB.

Qualcomm has the fastest basebands, but they're also slightly larger than many competitors. I don't have precise numbers but I think Infineon's XMM6160 in the iPhone 4 is closer to 30mm². On the other hand, the Qualcomm MSM8200 for 21/28Mbps HSPA+ is about 100mm² on 65nm. Yes, 100mm², I'm not kidding. It's probably the largest baseband ever though (excluding ones with integrated application processors) and even the Qualcomm MDM9600 for 100Mbps LTE is slightly smaller on 45nm (I don't think I'm allowed to give the precise number, but it's fair to say ]80;100[ mm²)

So Qualcomm has ~40mm² basebands but skimps on the ~0.2mm² FPU? (Scratching head...).
Is the area of the baseband somewhat linear to the number of channels supported? So a chip supporting 3.6mbit hsdpa is only a bit more than half the size of a 7.2mbit one? Strange then that you'd go penny-saving on the cpu side - especially since I think there's almost nil practical difference between 3.6mbit and 7.2mbit hsdpa anyway... Would a 3G baseband (but no hspa) be much larger than a 2G GPRS/EDGE (class 12 one)?

Also 430MHz ARM11 is for Nokia's lowest-end 3G smartphones. I promise you their S40 EDGE phones have significantly slower processors, and obviously any Cortex-A5 used on a 3G SoC would have L2 cache (and maybe higher speed since ARM's own TSMC 40LP hard macro reaches 600MHz).

I was referring to the Nokia 2G smartphones mentioned by ToTTenTranz (they are more like 150$ though) which have 430Mhz ARM11 (not sure which chip).
I actually got a S40 3G phone (just umts, no hspa) from nokia (n6267) and I believe you if you say it has a slower cpu (might be arm9 not sure) . Though with S40 it doesn't really matter, except for the half-minute hangs sometimes which aren't the cpu's fault. This is certainly old though.
I agree though if the 3G basebands are that large (btw do they shrink well?) Cortex-A5 might not make that much sense in such chips. Would probably be a much better balance to skimp on HSUPA/HSDPA modes a bit and go for a A9 instead if you really want to save costs.

 

[Arun]

Ok but even if it's 10% of the chip it's still not that much in absolute terms. How much does 1 additional mm² cost on 40nm? Granted maybe it does make a difference if you want to end up with a <100$ phone...

Oh, we're probably talking closer to $40 for a feature phone without touchscreen/WiFi/GPS and $65 for a capacitive touchscreen/WiFi/GPS smartphone with no subsidy by late 2012. $100 is for 1xA9/HSPA ala U4500

Look at this for inspiration: http://www.gsmarena.com/lg_cookie_lite_t300-3412.php and http://www.gsmarena.com/vodafone_547-3297.php (the former is 60 euro unsubsidised while the latter is 40 euro prepaid but SIM locked so probably a small subsidy)

So Qualcomm has ~40mm² basebands but skimps on the ~0.2mm² FPU? (Scratching head...).

Ironic isn't it? Of course, they fixed that in the MSM7227, but I don't know what they were thinking. Maybe they thought it would make Snadragon look better...

Is the area of the baseband somewhat linear to the number of channels supported?

Not really, it's definitely less than that. The MSM8200 is the exception, not the rule. I don't have numbers, but I'm pretty sure Infineon's 21Mbps baseband on 40nm is smaller than their 7.2Mbps one on 65nm (and its spectrum efficiency is better too).

Would a 3G baseband (but no hspa) be much larger than a 2G GPRS/EDGE (class 12 one)?

Keep in mind you need to support multimode so many implementations are really just a WCDMA baseband next to a separate EDGE baseband. I think the difference between EGDE->WCDMA is smaller than WCDMA->7.2Mbps HSPA, but I don't have hard numbers.

I was referring to the Nokia 2G smartphones mentioned by ToTTenTranz (they are more like 150$ though) which have 430Mhz ARM11

Hmmm, could be WCDMA ones with only the EGDE baseband enabled. I'm not sure - there are plenty of custom TI baseband chips at Nokia with no public info. Anyway that's the exception, not the rule.

I agree though if the 3G basebands are that large (btw do they shrink well?)

They shrink perfectly, it's all digital nowadays. The RF is separate but surprisingly enough it's also mostly digital and shrinks relatively well with a lot of design effort (it's shipping on 130nm today, sampling on 65nm, and in development on 40/28nm).

Would probably be a much better balance to skimp on HSUPA/HSDPA modes a bit and go for a A9 instead if you really want to save costs.

Yeah, although I don't think anyone will go lower than 7.2Mbps HSDPA/2Mbps HSUPA in future products. HSUPA is valuable and the you won't save that much silicon going from 7.2 to 3.6Mbps. We'll see.

---

Anyhow, enough time spent on this thread, I think I'll shut up now

 

[Deleted member 13524]

Hmmm, could be WCDMA ones with only the EGDE baseband enabled. I'm not sure - there are plenty of custom TI baseband chips at Nokia with no public info. Anyway that's the exception, not the rule.

Interesting. If true, I wonder if a hacked firmware could actually enable 3G in those models, as they're all quite similar to the 3G-enabled 5230 (actually, the 5228 is exactly the same as the 5230 but without 3G)..

 

[wco81]

So ARM has a product called A5, like Apple does?

Confusing.

Wouldn't A8 cores with older GPUs become cheap enough for this market?

 

[Deleted member 13524]

Apple\Intrinsity designed a SoC called A5. ARM has a CPU core called Cortex A5, which can be part of a SoC.

A I see it, the Cortex A5 is part of the 2nd-gen of ARMv7 cores, while the Cortex A8 is first-gen. The Cortex A5 has higher performance/watt and performance/area than the Cortex A8 while using the same process and clocks.
I don't know if it still makes sense to create new SoCs with a Cortex A8 anymore, as it seems that the single Cortex A9 + NEON is quite a bit faster and not much more expensive at ~1GHz, both in IP licensing and area costs. And at ~600MHz the Cortex A5+NEON seems to offer lower power consumption while providing approximately the same performance.

 

[mczak]

Oh, we're probably talking closer to $40 for a feature phone without touchscreen/WiFi/GPS and $65 for a capacitive touchscreen/WiFi/GPS smartphone with no subsidy by late 2012. $100 is for 1xA9/HSPA ala U4500

Look at this for inspiration: http://www.gsmarena.com/lg_cookie_lite_t300-3412.php and http://www.gsmarena.com/vodafone_547-3297.php (the former is 60 euro unsubsidised while the latter is 40 euro prepaid but SIM locked so probably a small subsidy).

Well that 60 euro phone still lacks quite a few things from your supposed 65$ phone (while being more expensive). Only resistive touchscreen, no WiFi, no GPS, not to mention it would need more ram (along with that cpu upgrade). Though you're saying late 2012 and I was thinking more early 2012 for new generation...
Also, a $100 3G phone with Cortex-A9 would be seriously impressive as well, though the U4500 is said to be for phones for 100-200$ so if that's late 2012 for a 100$ version (or 100$ earlier but missing some key features) it sounds a bit less impressive. I've yet to see a "decent" android smartphone (meaning at least HVGA capacitive screen, GPS, WiFi, 3G) below 200$, and Cortex-A8 based devices typically don't start below 300$ (though it's quite possible I missed some in either category). There are exceptions to this like the ZTE Blade but those are more a special deal (not available everywhere for instance).

Keep in mind you need to support multimode so many implementations are really just a WCDMA baseband next to a separate EDGE baseband. I think the difference between EGDE->WCDMA is smaller than WCDMA->7.2Mbps HSPA, but I don't have hard numbers.

Hmm ok. Still seems like it would be a more substantial difference than going from a A5 to a A9.

Yeah, although I don't think anyone will go lower than 7.2Mbps HSDPA/2Mbps HSUPA in future products. HSUPA is valuable and the you won't save that much silicon going from 7.2 to 3.6Mbps. We'll see.

Well if it shrinks perfectly it indeed might make sense to always support 7.2Mbps. Even that large 42mm² Qualcomm baseband would only be 20mm² on 40nm which might be tolerable even for quite cheap phones, especially if that would only be a 10% or so difference in die size going to 3.6Mbps.

Anyhow, enough time spent on this thread, I think I'll shut up now

Noooo, lots of interesting information here

So ARM has a product called A5, like Apple does?
Confusing.

Well it's called Cortex-A5 in full name .

Wouldn't A8 cores with older GPUs become cheap enough for this market?

Dunno, but if you'd want to do a 2G phone you'll need a new chip anyway since the Cortex-A8 SOCs all are 3G (I believe). And I don't think you'd really want to do a new design with A8, both A9 and A5 look like better designs to me (more efficient). And if you want to shrink it to 28nm for cost savings and lower power consumption (maybe), there's also less incentive to hang on to an old chip design.

 

[metafor]

Wouldn't A8 cores with older GPUs become cheap enough for this market?

With the metrics of the A9 in terms of performance, perf/W or perf/area, it almost never makes sense to use an A8 nowadays unless you're that worried about licensing cost per core (I imagine A8's cost less).

 

[mczak]

A I see it, the Cortex A5 is part of the 2nd-gen of ARMv7 cores, while the Cortex A8 is first-gen. The Cortex A5 has higher performance/watt and performance/area than the Cortex A8 while using the same process and clocks.

I believe that is not quite true, the Cortex A8 should be able to reach quite a bit higher clocks. Though a lot of smartphones have it only clocked around 600Mhz, I suspect that's due to power reasons (which the A5 won't suffer from to the same degree).

I don't know if it still makes sense to create new SoCs with a Cortex A8 anymore, as it seems that the single Cortex A9 + NEON is quite a bit faster and not much more expensive at ~1GHz, both in IP licensing and area costs. And at ~600MHz the Cortex A5+NEON seems to offer lower power consumption while providing approximately the same performance.

I agree except that I don't think a 600Mhz Cortex A5+NEON will quite reach the same performance as a Cortex-A8 (clocked at 600Mhz), even when using same size L2 cache. Granted it wouldn't have that crappy VFP unit (which doesn't matter as long as you use NEON), but it's still single-issue vs. dual-issue chip (granted it was in-order dual-issue so probably not especially efficient but still I see no way how the A5 could reach the same performance - unless maybe with NEON-heavy code if the NEON unit was significantly improved).

 

[metafor]

I agree except that I don't think a 600Mhz Cortex A5+NEON will quite reach the same performance as a Cortex-A8 (clocked at 600Mhz), even when using same size L2 cache. Granted it wouldn't have that crappy VFP unit (which doesn't matter as long as you use NEON), but it's still single-issue vs. dual-issue chip (granted it was in-order dual-issue so probably not especially efficient but still I see no way how the A5 could reach the same performance - unless maybe with NEON-heavy code if the NEON unit was significantly improved).

For the same area as a Cortex A8, you can fit 3-4 Cortex A5's with NEON. I'd venture that for a smartphone application, having a dual Cortex A5 is preferable to a single Cortex A8 in just about every metric save purely single-threaded performance.

 

[mczak]

For the same area as a Cortex A8, you can fit 3-4 Cortex A5's with NEON. I'd venture that for a smartphone application, having a dual Cortex A5 is preferable to a single Cortex A8 in just about every metric save purely single-threaded performance.

Well that's why I was initially suggesting dual-core A5 for next-gen cheap phone, but it looks like it's going to be single-core A9 instead, except for REALLY cheap phone where it might be single-core A5...

 

[Arun]

Well that 60 euro phone still lacks quite a few things from your supposed 65$ phone (while being more expensive). Only resistive touchscreen, no WiFi, no GPS, not to mention it would need more ram (along with that cpu upgrade). Though you're saying late 2012 and I was thinking more early 2012 for new generation...
[...]
Also, a $100 3G phone with Cortex-A9 would be seriously impressive as well, though the U4500 is said to be for phones for 100-200$ so if that's late 2012 for a 100$ version (or 100$ earlier but missing some key features) it sounds a bit less impressive.

Oh, I'd say early 2012 too, but maybe it'd be more than $65 then
Even if you had a Cortex-A5 SoC-based phone in that timeframe, cost savings for the touchscreen (both incremental and disruptive), bluetooth/wifi/gps (CSR/Broadcom will only start sampling on 40nm shortly, see SiRFStarV and CSR9800), and memory/display (natural cost reduction as they will still be a big part of the BoM).

Noooo, lots of interesting information here

Bah, and now you've made me reply again

But we're really just arguing about the same things again and again here. So let's try adding some juicier info to the mix ARM has said that they will announce a new core later this year codename 'Kingfisher' and that it is "aimed at lower end smartphones and feature phones and cost-sensitive digital TV applications. It's a small A-class processor" (source: EETimes)

So far, so good. The Cortex-A5 is a nice core, but it needs a performance upgrade. The Cortex-A8 has nice performance, but surprisingly bad perf/mm² and perf/watt, so I bet they're working on a cheap in-order dual-issue CPU that still achieves higher performance than the A8. So far nothing super exciting, right? Okay... now watch this video from 5:30 to 5:45 or so: http://www.youtube.com/watch?v=_4Eo84Uia0Y

This guy is the Director of Strategic Processor Technology for Dell - not just for handhelds, but for the entire company, including servers. So why would he say: "And we're also very excited about Kingfisher. I think if you guys can get software support for that, that's going to be a game changer." - why in the world would he say that? The answer is rather obvious if you think about it...

I suspect Kingfisher will be the first ARMv8 (64-bit) core despite being reasonably low-end. A strange choice perhaps, but it's easier to start with an in-order architecture, and it would certainly remove any advantage MIPS might have had with the upcoming 64-bit Prodigy. I could be wrong, but it'd make quite a bit of sense.