Tricky. It'll have a very high level of integration, but it'll be built on TSMC's 28nm (don't know if it's LP or HPM) process, so all of the benefits of 22nm FinFEts go away.
I think the micro-architecture is still very good, and I'm sure the SoC guys are porting over all the power management goodness from the Silvermont products on 22nm, but it won't be the knockout that SoFIA on 14nm will be (and I suspect this will be a MAJOR problem for QCOM).
Did you miss the headline?
"OCZ Filing for Bankruptcy, Announces Offer From Toshiba to Purchase Assets"
Stop pumping. It's over.
Well, I'm just saying that GPU is a focus for just about everybody in this industry. NVIDIA is exceptionally good at GPUs, no question there, but to suggest that others aren't *trying* seems a bit off. That's all :-)
" Yet no other SoC supplier thinks entertainment is a good area on which to differentiate. "
Tell that to the Snapdragon 805. Or Cherry Trail.
I think you grossly underestimate what the other guys are doing in this space.
They didn't include it because mgmt does not expect material sales from Merrifield. This was supposed to be the "knock out, kill ARM" part, but the market moved too fast. Moorefield will be Intel's next chance at bat, but I suspect the real traction doesn't come until Broxton.
Just as we both predicted - ARMH is now back above $50. Does the market realize, however, that despite the "we'll fab for anybody" bit that Intel could do the exact same thing that it's doing with tablets to phones? ;-)
Hi uppereastboy, sorry. Here is the response:
Basically, Intel has all of the right IP/technology to succeed in this space, but the company miscalculated with some of its products. For example, Bay Trail - which is a great performer, btw - was designed for the high end of the tablet market AND the low end of the PC market. See, when you're selling a $50+ chip into the low end PC market or a $30-$40 chip into the high end tablet market, you can afford to have a die size of 102mm^2 (this is very close to a PC "Ivy Bridge", btw) for the entire SoC.
On top of that, the Atom was designed for a "rich" bill of materials. In other words, all of the components that surround the chip on the PCB (i.e. the board on which all of the chips go) were fairly expensive. Further, since the Atom didn't have things like Wi-Fi, Bluetooth, GPS, etc. built onto the die, anybody building a Bay Trail system would need to buy a separate combo chip from either QCOM, BRCM, or MRVL to get that functionality. These are about $4-6, so it's a non-trivial adder for a $199 tablet.
But that's where SoFIA comes in. It integrates Wi-Fi, Bluetooth, GPS, and cellular onto one relatively small chip. It's not as fast as Bay Trail, but it's more highly integrated which makes life easy for the system vendors as well as Intel (the die size of SoFIA is smaller than Bay Trail IIRC).
The company also hinted that Broxton would be more highly integrated. It won't have cellular, but I highly suspect that it'll have Wi-Fi, Bluetooth, and GPS (i.e. connectivity).
Basically, anytime somebody says "DERP INTEL NEEDS ARM" they're just clueless. This is all about picking the right design points for the products you're building, and it seems under Perlmutter, the company underestimated QCOM/NVDA at the high end and it missed the mark on the low end, leading to the situation we have today .
But the bottom line is that this will be corrected by 2015.
Sure - I'd be happy to give you a detailed explanation. I'll be back in about 20 minutes with a reply (just taking care of something at the moment). Don't want you to think that I'm leaving you hanging/ignoring your question.
Also, 100% agree that I have no effect on the share price. That is not my intention and I'd be a delusional fool if I thought that I could "move" Intel.
I am just an investor who loves to share his thoughts with the investment community.