Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Actually I think this is the most remarkable achievement of Intel in the recent years. Their Android team did a great job at porting to x86. This was not an easy task and it works on Medfield just great with only minor incompatibilities.
Joining Intels Android workforce together with Google's may help them release a 64 bit Android port much sooner than they could without Intel.
Anyway, Broxton will be critical for Intel to dominate the high end and, at the same time, gaining performance/watt and cost leadership. Probably the hardest work in front of them is integration. Intel is a threat to Qualcomm, since almost every mobile processor sold by Intel means also a sold modem by Intel.
This whole BOM issue really is bothering me. What the heck was Intel thinking years ago about what degree of integration they would need in order to be competitive in this market? Many years ago, Qualcomm was already integrating as much as was possible and the trend was pretty clear to see. Is Intel really that blind to not being able to forecast that integration is key? Something that has been ongoing for so long in the PC space already? Or is it a technological issue with their process technology so that they are not capable to integrate certain parts that are needed for further integration (e.g. some analog stuff, even though I don't see much of it needed here). But again, even taking that into account, they would have been able to see this issue coming many years ago. That is the biggest fail from Intel management that I can imagine. It also is something very hard to see from outside as an investor, making it very hard to judge Intel's claims of the past (and future).
I mean, how hard can it be to integrate existing IP like GPS, LTE and so on for a company that is capable of setting up a 14 nm process and which designs the best CPUs on earth? There is just nothing I can see missing on the technological side.
How is planar 20nm giving a power advantage? Was Intels move to Finfet not a necessity? They'll lose on leakage what they gain on dynamic power. Have fun power gating I'd say. Remember how even TSMC says that 20nm doesn't make much sense but customer contracts demand it?
What happened to your insider selling?
Ok, I think that this is really worth to discuss on this board. Would be good to have some more opinions on that.
My thesis is that Intel needs volume to fill their fabs. CapEx for new fabs increases heavily and each shrink brings more and more capacity to the fabs. If Intel has an 80% utilization of their fabs today, they'll probably have less than 40% at 10nm, which doesn't make any sense economically. There is a reason why only a few foundries today can produce what in the past has been produced by many fab owning semicos. Intel is about the last in this business who can afford leading edge fabs.
So, since the PC and server market isn't anywhere growing as fast as the technology shrinks, they need something else to fill their fabs. Mobile is about the only option despite the foundry business Intel has got. Mobile SoCs are pretty much standard components where on contrast, the IoT or more generically, the controller business is highly specific with many custom designs. That's also the reason why I don't believe in IoT for Intel. That's ARMs domain where the licensing model is key and service for small players is important. Intel can't do that.
What is left is the foundry business. There I think Intel can't compete with TSMC and Co. Again, this is a very service oriented business and also Intel is specialized on digital ICs, not analog ones (I guess that's what they have their trouble with integrating their modem).
Mobile SoCs are pretty similar to the PC business though. You need standard interfaces and customization is not necessary. That's just fine and Intel has a high chance to win that market due to their process lead (yes, it does matter) and their processor know how. Still, currently their execution just stinks and unkept promises are not what people are used from Intel. I mean they can introduce a 14 nm process almost in time but they can't deliver BayTrail in time on Android as promised some months ago? WTF?
Regarding time frame I am not quite sure and that can be debated. I guess 10 years from now on we need significant increase in volume, likely from mobile and some key foundry customers (like Apple, Altera etc). Intel is not a secure or conservative investment at all. It may look range bound right now but there is high potential for futere up and downside, largely depending on their success in mobile. I am seeking for major advances in mobile and foundry business next year, otherwise I'm out.
It's not really new information but I don't know if this has already been posted on this board:
Intel is obviously trying to port Android to Haswell successors.
I hope this will be the case with Broadwell or its successor. If Intel is clever, they combine the integration level of Atom/Goldmont (hopefully finally with integrated baseband and all the other stuff needed in mobile) with the power of Haswell/Core I. First for a tablet platform and maybe later (e.g. at 10 nm) for smartphones, both as extension towards the higher end of Atom. That is definitely something the ARM guys will have a very, very hard time to beat. Remember how that core is still outperforming most of what exists today, from AMD to server RISC etc? That's Intels most valuable asset and they should use it also in mobile to grab the high end (and I definitely believe there is a high end in mobile just as everywhere). Nevertheless, the same degree of integration is needed for these Core-I variants for mobile applications - no BayTrail like mistakes to be made anymore. A modular architecture is a must there.
What Intel also needs to work on is the performance improvements of their Core-I processors. The 10% improvement steps as it was the case with Haswell won't be enough to keep the distance between the high end mobile cores (like Apple's A7) and Intels high end core. People need a reason to pay a premium. I know that this is very hard to achieve and performance improvements in the future are much harder to get than it was in the past but Intel's business highly depends on that. I hope they take this challenge as serious as they can. Skylake will give an indication, which I think is very important.
Also Intel should consider stripping Haswell a bit for mobile applications. Stuff like shorting the pipeline and improve the power saving, e.g. by removing some of the most complex/power consuming circuits without sacrificing too much performance (the shorter pipeline will compensate nicely for that) could be good to have for mobile applications. Something like a Core-I light. Also taking advantage of very sophisticated clock gating would help the desktop version of Core-I to turbo even more aggressively. I am for example thinking about an even wider design for out of order execution where there is a sophisticated circuit which can determine when the additional execution units can be clock gated since their result will likely not be needed.
[/quote]So basically, the Snapdragon is designed for high performance over short periods of time (e.g. handling a touch interface with reasonable latency), but more moderate performance over long, continuous loads. A great mobile processor but lackluster as a web server? [/quote]
One trick pony. They can't do that next time around anymore. By the way, remember how Apple's A7 is going +8 Watt when going from idle to full load? That's stressing a mobile power budget about as much as it can. There's not much headroom left. Remember how BayTrail consumed around 2,5 Watt while running a benchmark? It is still outperforming Apple's A7 in benchmark. Same goes for Snapdragon 800. I declare performance/watt leadership for Intel already. This will improve further with the 14nm transition. The others have already eaten their lunch, Intel can still eat it and outperform them. Does anyone realize how Snapdragon 805 is just a small performance increase. I bet it will be just like that for everyone else.
Just consider that you can get a Haswell with 5 Watt TDP already which is in the league of Snapdragon or Apple's A7 and definitely outperforms them. Broadwell will improve further on that, making high end Core I class SoCs possible for mobile applications - something Intel hopefully is working on and something they can charge a nice premium for (e.g. high end tablets or even phablets).
All the low hanging fruits have been picked by the ARM crowd, things like going wider, clocking higher, adding cores and going out of order has been done (stuff that has been done with PC processors a very long time ago). From here, performance improvements get really difficult and complex, while they can't stress their power budget any further. We are in Intel territory now, where there is a lot of knowledge and patents hold by Intel and the others have to learn stuff that Intel learned many years ago while fighting AMD and the server processors.
Having said that, Intel really needs to execute better, I am completely with Ashraf there. If they think they can stick to their multi year tick-tock model of developing new cores, they are misplaced in the mobile business. The dinosaur Intel needs to adapt to a very agile market, otherwise they'll lose. If they'll lose the mobile battle, they will not be able to finance their new high end fabs just from the PC and server business and finally will be overtaken by foundry based cores, simply due to their process advantage (which currently is still hold by Intel, not to forget).
Merrifield fills a gap. It can do very well in compact phones with sizes up to 4.7". I think there is some tendency in the market that phablets aren't for everyone and impractical in every day life. There are almost no devices with smaller size that feature a Qualcomm Snapdragon 800. The Xperia Z1 compact is the only exception I know of and that uses a very large battery. Is there any dual core Krait 400 in the market? Not that I am aware of and Intel may fill that gap.
Concerning real life performance, Intel benefits from its know how in high level optimization. Stuff like branch prediction, cache controlling etc. is very complex and Intel is ahead of them all in this regard. This is something you can only see in very complex benchmarks and Silvermont doing very good in SPEC int is a clear indication where this good real-life performance actually comes from. This is the thing where Intel has a big advantage over ARM wimpy cores and these won't learn that quickly nor will they have an easy time working around Intels patents in that space.
Now it is time for the customers to realize that an 8 core Mediatek with Cortex A7 just doesn't perform anywhere as good as the benchmarks suggest. Apple already guided into that direction with the Apple A7 dualcore which has very high single thread performance. Nvidia is going the same way with Denver and I believe others will follow that route. Intel is basically first here in the Android space and they should play their 64 bit card for marketing reasons so people believe the performance benefit would come from that, just like it was the case with Apple's A7 (which it wasn't as we all know).
Besides: Remember that video where Samsung Galaxy 3 real life performance was compared to Motorola Razr I with its Medfield? Medfield lost all benchmarks but its good single thread performance helped it to perform better in many applications than the Galaxy 3 (not in games though, which may be due to its weak graphics and emulation of the native compiled games code).
The Merrifield video shows web browsing as strong point and Medfield also delivered very good results there. The reason may be highly optimized web engines for x86 together with the best compilers that exist. Since web browsing is important for Smartphones and still an issue with non mobile websites, Intel delivers a big advantage. My Nexus 7 2012 edition sucks so much when web surfing compared to my Razr I.
I still believe Intel should hurry to bring Haswell or its successors down to mobile power figures. A single core Haswell with some energy tweaks and hyperthreading would beat any ARM or Atom core in real life easily. It just doesn't bring the marketing GHz and many core stuff.
What do you base your investment on if not first hand information from an investor conference?
Well, in that same conference he claimed density leadership in the SoC field if I recall correctly. Misspeaking is one thing, making such an important claim at an investor meeting is not something you can just slide away. To me this was very important about how I positioned myself since I expected major design wins at CES. This is not some small cap CEO but the CEO of the largest Semico in the world.
Having said that, I don't judge Krzanich or Intel on such single fails but if they add up, I would sell my shares just due to invalid information from the company. I mean, what else can you trust if not even public information at shareholder meetings? What I don't get: Why are those lousy analysts not asking such questions at the Q&A? I mean, what are they paid for? Instead they suck with questiona about inventory and such ...
By the way, you guys should be reading the transcripts more carefully it seems