Texas Instrument made a choice not long ago to make their OMAP platforms available to markets outside the Mobile phone one , simply because that could give them the money needed to keep developing new designs and most importantly since at most other markets , products like those have longer life circles , ones that can return the original investment made on their design and production .
SoC designing is starting to become an expensive "Game" not just because the technology used is pushing the limits even further every time which costs money by itself but also because each new design usually stays on the top segment for just a while .
Most SoC's require two different things to return the investment made in their design and production . High volumes in the order made and some time on the market . High volumes mean that more than one manufacturers have to choose your solution for their phones and that they also must sell many devices based on it . Long time on the market can be achieved more easily since each platform can move down to cheaper devices as time passes and they stop being considered high end solutions .
A combination of both those requirements can not be easuly achieved and at least Texas Instruments was never particularly successful at that .
Qualcomm on the other hand has the advantage of controlling the North American market (not always by playing fair but that is another matter completely ) , Samsung is pushing everyone else even further with each new design since they always seem to make the best SoC's available on their launch , and Nvidia is determined to stay on this market regardless of the money they have to spend in the meantime .
Analynizing even further this notion is interesting .
Samsung can easilly cover the costs of the development by just using each new SoC on their own devices since they can achieve very large volumes , and at the same time , they also have Apple as a customer to bring the extra cash needed to always improve their production technology and facilities .
At the same time , they always choose to end the production of each SoC shortly after the device based on the next one is released and choose Qualcomm's solutions to power the new mid range devices since their own last year's high end ones cost more to keep in production than making a large order to Qualcomm (which by the way is now also a customer to Samsung's foundries )
This is how Samsung was able to complete first the development of the next generation SoC's that are based on the Cortex A-15 core along of course with newer GPU's and peripherals .
Sadly for most end users me included , this was not made on time so they could use the Exynos 5250 on the Galaxy S III but there is always the next Galaxy device to wait ....
As it was expected , even a dual core configuration of A-15's is faster than any available A-9 design , even the ones using four cores . At the same time , Qualcomm's Krait core was never designed to compete with the A-15 , but to become available sooner by leaving out some of it's features and after all Qualcomm intends to use it for more than another year as a high end and gradually a mid range solution .
Samsung was able to kick-start the 5250's production by securing Google as a client both for the new chromebook and the new Nexus Tablet and that is by no means a small victory for a SoC maker .
In the coming months , Samsung will have their production volumes of the 5250 sold to Google , and by the time Nvidia announces the next Tegra SoC based on the A-15 they will already have a significant advantage .
Cortex A-15 however are not without their drawbacks .
At the manufacturing processes available now , both at 32nm by Samsung and 28nm by TSMC (in Nvidia's case...) each core is quire big in size . This means that creating a SoC containing more than two of them along with a capable GPU and other peripherals , in a reasonable size is almost impossible and certainly outside the space available inside a mobile phone .
The new architecture does provide exceptional performance per clock and overall , it also requires more power than an A-9 to operate . Since mostly marketing reasons require each SoC based on the A-15 to come with a GPU that is at least 1.5 times more powerfull than the ones used on SoC's based on the A-9 , the power consumption is increased even further .
Apple's creation of the A5x to be able to drive the third generation's Ipad high resolution screen showed everyone that capable GPU's require both space and energy and they also produce significant amounts of heat .
That last one is the reason that for the time being , A-15 based designs will be used inside tablets ,. since the space provided there for heat dissipation is much larger that any mobile phone .
Some technology leaps forward will have to be made in order to properly fit a 5250 inside a mobile phone and this is probably what Samsung is trying to do right now .
Nvidia is also facing the same problems , since they also have to pair the A-15 with an even more capapble GPU , which is by no mean an easy task . After all , most of the power consumed by the Tegra 3 and the heat generated was from the GPU and not the 4 A-9 cores that most of the time were not running at maximum speed all at once . Being able to deliver their solution on the market later than Samsung , will give them the time needed to figure out how to solve most of these problems and this is probably the only advantage of being late at a highly competitive market . At the same time , they are making efforts to combined the Network technology they aquired from Icera on the same chip with their Tegra solutions , but that will probably happen on some later version of their next Tegra SoC the same way that Qualcomm can not fit the network part along with four Krait cores and everything else inside the same chip .
Qualcomm and Nvidia are at different stages of creating SoC's containing the Network stack along with everything else . Actually Qualcomm has achieved this goal many years ago , but they do face the same space problems with each new Core design as everyone else . Nvidia is willing to spend the money needed for achieving this and at the meanwhile sell the Tegra chips as stand alone SoC's and cover the losses made if their products are not selected by many phone makers . Actually the Tegra 3 SoC managed to find it's way to just as many devices as the Tegra 2 did , which should at least make it not a damaging product for the company money-wise .
Samsung on the other hand has the advantage of high volumes and is actually designing devices based on Nvidia's and Qualcomm's solutions at the same time in order to fully cover their needs for different devices and price segments . They too are designing their own network parts but for the time being they are unable to produce them at large volumes . Besides that , the techical know-how they have by using Intel's XGold solutions on most european phones that are not based on a Qualcomm SoC , along with the one they have gained by using Qualcomm's solutions means that for the time being they can keep using them and invest smaller amounts of time and money on their own designs .
They are however known to want to be free from other companies where ever they can , so they will probably end up using their own network components at some point in the near future .
This battlefield is filled with companies that have both the money and the clients to secure their involvement on this market . Texas Instruments on the other hand , could not get the orders needed to both keep developing new SoC's at the same pace with the other ones let alone the money needed to start designing and producing their own network components for UMTS and LTE networks .
This company was a key player on this market long before Nvidia and Samsung even started producing their own chips of any kind (for mobile phones that is ) . At the same time they have given us many products that were ahead of their time and most of us have at some point used a phone based on an OMAP SoC .
Along with NXP that started as a branch of Phillips they were from the first companies that provided the mobile phone market with 32-bit ARM designs and DSP's capable of delivering multimedia content .
Sadly the modern day high end technology market has no room for nostalgic feelings and Texas Instruments will probably never be able to gain the edge it once had on this market .
No comments:
Post a Comment