I don't know about you gentlemen, but I'd personally prefer a quad Swift, Krait or A9r4 nowadays in a smartphone then A15.
No way imho. Might match a krait 200 though with some luck (granted it's not _that_ much slower than a 400), and it's clocked higher, so if it's any good or not entirely depends on the power it needs for that.I actually agree with you for once.
That A9 r4 seems really interesting to me, I would thing at least in integer ops it would match krait 400 at the same clock speed...give or take a few %.
Yes, and Cortex-A9r4 isn't much faster than older ones neither (though I wonder how much actually, the other revisions might also have brought some increases - just about the only thing I know of is that r4 is listed as having potentially more TLB entries (but the number isn't fixed so an actual implementation might not have more). I guess it is very difficult to isolate such changes performance-wise as the respective implementations will have other differences too (e.g. memory system).
I don't think it could quite be the "ultimate" smartphone SoC, a bit too late for that for the performance and featureset (on the gpu side) it offers. You'd think though it is decent enough (particularly thanks to its high level of integration including the baseband) to get some design wins.
http://www.nvidia.com/docs/IO/116757/NVIDIA_Quad_a15_whitepaper_FINALv2.pdfYes, and Cortex-A9r4 isn't much faster than older ones neither (though I wonder how much actually, the other revisions might also have brought some increases - just about the only thing I know of is that r4 is listed as having potentially more TLB entries (but the number isn't fixed so an actual implementation might not have more). I guess it is very difficult to isolate such changes performance-wise as the respective implementations will have other differences too (e.g. memory system).
- TLB entries increased to 512
- branch prediction history buffer and counters increased
- L1 data prefetcher
http://www.nvidia.com/docs/IO/116757/NVIDIA_Quad_a15_whitepaper_FINALv2.pdf
Some more information here: http://www.slideshare.net/caulfield2600/tegra-4i-expands-the-market
Basically dhrystone speed won't change, but larger programs that matter to end-users (browsers for instance) will have a nice boost.
I dunno I think the Cortex-A9 itself wasn't all _that_ great neither. Sure it was much better than the Cortex-A8 in perf/w but a A7 is miles ahead there still (of course, not a fair comparison since the absolute performance reached is lower). My guess is the A12 could be the ideal smartphone cpu but it's very late (and in 20nm those A7+A15 combos might be a lot more viable). So I'm not convinced the A9r4 really is any better smartphone cpu than Krait (say, 300) is, while appearing later.No I was talking more about the cortex a9 r4 itself...not tegra 4i...I think tegra 4i wilk make an impressive midrange SOC, but it wouldnt make the perfect all round solution, only single channel memory ala tegra 3...its more like a tegra 3++
Ah nice link. You can get that information out of the arm documentation but it's a very annoying search...
Yes they'd be terrific for non-highend smartphones, but unfortunately there's not really anything out there using them (ok some chinese companies use them). Granted there's a qualcomm chip coming (supposedly with 1.4Ghz) which I guess might get popular for cheaper smartphones, but it can't really qualify as ultimate smartphone chip since its absolute peak performance is too low (I guess roughly half that of those high clocked Cortex-A9) - certainly would be ways faster though than my aging single-core arm11 .On a sidenote A7s (by themselves) are nice but I'd have them clocked at least at 1.2GHz. At 1.5GHz their perf/mW ratio is nearly ideal for today's =/<mainstream smartphones.
Well there's always the A12 right?
Wow, a tablet with 2560x1600 and a low latency pen seems damn desireable. You could do math all day on it (the kinds where you need to be able to write and draw anything in free form)
I can imagine being easy to learn drawing as well, or why not an app that helps you train your handwriting into looking better for someone else that would need to read it. Why not some software that analyzes your handwriting and suggest you a slightly better but similar way to draw those 'f', 'g' and 'e'. And let me log in with a handwritten (invisible) password.
Maybe I'm falling for it because Toshiba used that "Excite" word but the concept of usable pen computing is interesting to me. Pen and paper is still superior to laptops and classic tablets in many ways.
I would only worry about cost, battery life and fear of getting it broken.