I would imagine Nvidia are very uncomfortable with the way their market has been contracting over the last couple of years.
At some point enough x86/x64 patents will expire that Nvidia will be able license the remaining ones and so an x64 chip of their own.
Or alternatively they could sell Arm+GPU SOCs instead - arguably Arm+GPU is a better bet than x64+GPU because the sales of phones and tablets will exceed the sales of x64 PCs. Of course the margins are likely to be thinner because there's a lot of competition in the Arm SOC market - Apple and Samsung have their own in house designs and outside that it looks like Qualcomm have most of the rest of the market.
Still it's not like AMD is doing very well competing with Intel. And the reason Qualcomm do so well is because they design their own Arm microarchitectures - Scorpion and Krait were both designed in house and were higher performance than the best Arm designed microarchitecture. So I guess NVidia could be aimed to compete with Qualcomm since Denver is in house too.
Actually Apple A6 and A7 chips are like this too. Apple have an Arm license but the chips are designed in house. So it seems like of the Arm SOCs that actually sell well only Samsung is using Arm's designs and only in some markets
E.g.
http://en.wikipedia.org/wiki/S...
Galaxy S4 models use of one of two processors, depending on the region and network compatibility. The S4 version for North America, most of Europe, parts of Asia, and other countries contains Qualcomm's Snapdragon 600 system-on-chip, containing a quad-core 1.9 GHz Krait 300 CPU and an Adreno 320 GPU. The chip also contains a modem which supports LTE. Other models include Samsung's Exynos 5 Octa system-on-chip with a heterogeneous CPU. The octa-core CPU comprises a 1.6 GHz quad-core Cortex-A15 cluster and a 1.2 GHz quad-core Cortex-A7 cluster. The chip can dynamically switch between the two clusters of cores based on CPU usage; the chip switches to the A15 cores when more processing power is needed, and stays on the A7 cores to conserve energy on lighter loads
So there are two versions. A Qualcomm Snapdragon one for the US and Europe and an Exynos one for Asia. The Exynos one uses Cortex-A15 and Cortex-A7 in a BIG.little configuration.
Unfortunately they fucked up the big.LITTLE configuration
http://www.anandtech.com/show/...
The Exynos 5410 saw limited use, appearing in some international versions of the Galaxy S 4 and nothing else. Part of the problem with the design was a broken implementation of the CCI-400 coherent bus interface that connect the two CPU islands to the rest of the SoC. In the case of the 5410, the bus was functional but coherency was broken and manually disabled on the Galaxy S 4. The implications are serious from a power consumption (and performance) standpoint. With all caches being flushed out to main memory upon a switch between CPU islands. Neither ARM nor Samsung LSI will talk about the bug publicly, and Samsung didn't fess up to the problem at first either - leaving end users to discover it on their own.
You can see the results here
http://www.gsmarena.com/samsun...
The Qualcomm one has much better talk time - almost twice as much.
You have to wonder what the hell has happened to Arm to be honest. It seems like Apple (A6, A7) and Qualcomm (Scorpion, Krait) do a much better job at Arm core design than Arm/Samsung.
It'll be interesting to see battery life tests on the Snapdragon 801 and Exynos 5422 versions of the S5 to see if Samsung have got big.LITTLE working like it is supposed to. Actually I wonder whether big.LITTLE is even necessary - it seems like it would be much easier to just have the big core and scale the CPU frequency. The S5's CPUs are 2.5 GHz quad-core for the Snapdragon variant 2.1 GHz quad-core Cortex-A15 and 1.5 GHz quad-core Cortex-A7 for the Exynos variant.
That's quite a step up from the S4 so you could probably run them at a much lower clock frequency most of the time. I guess the problem is that a Cortex-A15 uses more juice when run at a low speed than an Krait and big.LITTLE (aka 'use an A7 instead when performance isn't critical') was a sort of band aid for this.
