Note that smartphones these days all have 2 or at most 3 Gigabytes of DRAM. Why? because with 32 bits, 4 Gig is all you get, and you need at least some set aside for IO.
Yes, I know with virtual addressing and (shudder) page registers, you can use more memory than this but if you want software to have the capability to address a single block of memory that's more than 4 Gig, you are out of luck with a 32 bit processor.
With lower power DDR4 coming, phone manufacturers are going to want bragging rights for having lots of DRAM, and so 64 bits is coming. Apple is already there, though they haven't increased their DRAM yet.
Junko, I agree that there are capacity concerns with TSMC and competitive concerns with other ARM partners. If they are willing to consider Samsung, then there will be 3 more 14nm fabs, plus one from Globalfoundries. However, I'm not convinced that this will improve Rockchip's competitive position.
Hi, JimMcGregor, I would agree to what you mentioned here.
However, what I found extremely interesting inMiya Kong's piece is the fact that Chinese fabless chip companies are concerned about the potential capacity constraint in the future, and more importantly, at least Rockchip seems to believe the fact that being one of the many Chinese fabless companies plaing in the ARM eco-system gives them neither advantage nor differentiation.
That anixiety -- reported on the ground in China -- is newsworthy.
Agreed. I don't see how this is a "big deal". Intel's chips aren't that competitive in both performance and price even at 22nm with Trigate. How are these chips going to be competitive on 28nm in mid-2015, without Trigate? And only 3G?
Also, this is a sign of both desperation and defeat from Intel - the fact that they are recognizing they can't use their own brand (Intel) and their manufacturing advantage to crack the mobile market, and that they have to license their architecture away to do that (only as wishful thinking).
It's a sign of desperation, just like it was for Nvidia to license Kepler (which nobody seems to want). It's definitely a "plan B" type of strategy.
And Rockchip will get get 14nm from Intel in the "near future"? Give me a break. Even Intel is delaying 14nm until 2015 for mainstream. Rockhip will be lucky to get it in 2017. ARM chips will be at 10nm FinFET by then.
I do agree that the cortex A12 and Cortex A17 make NO SENSE, and I really don't understand why ARM is pushing them. They should've been pushing an ARMv8 Cortex A55 at the same time with A53 and A57. Instead they are forcing the market for remain on 32-bit for a lot longer than necessary. But this is of no consequence in terms of competition with Intel or Rockchip's x86 chips.
This is definitely a move by Intel to break into the Chinese market, but the products still need to be competitive, which has been the issue with Intel's mobile strategy thus far, especially on Android. Also note that this announcement is only for 3G, while Rockchips'a competitors are coming in with 4G technology and a broad range of ARM-based products.
I would agree that this may be a good opportunity for Rockchip, especially since they will gain Intel support. However, I am still skeptical that it can overcome the momentum and performance of the competition.
It's great to get some analysis of the broad trends from someone inside China.
I would note the deal appears to be limited at the moment to work on a single SoC. There was no mention on the conference call with Krzanich Tuesday of Rockchip getting a license to Intel's x86 or 3G technology.
I think that only RockChip is the real winner of this agreement. Now RockChip, a chinese company, can study the best x86 architectures, the best manufacturing processes and a good modem. Intel has just the hope to enter in the low and mid market. The worst agreement ever made.
We've been saying all along that those Chinese fabless guys who make apps processors are in need of the modem technology. When I was travelling in China a few years ago, Rockchip was rumored to merge with Spreadtrum -- China's modem chip king.
While that never happened, I didn't see Intel-Rockchip hookup coming. It is brilliant.
What's driving the two companies to the agreements, are, however, different. Intel needs to fill the 14nm capacity and get into the China market pronto in a big way; Rockchip hopes to secure future capacity and possibly establish the plan B to the ARM-base ecosystem.
But if the X86 ecosystem doesn't develop fast enough, well, what happens then? All bets are off?