Since the summer there was a little debate among analysts of the technology industry, it was known that Apple would depend on two manufacturers for developing their processor mobile applications ARM A9 and A9X. Samsung and TSMC two giants of semiconductor manufacturing
But what was not expected is that the iPhone A9 6s / 6s were also made by the two, given that Samsung usually greater industrial capacity in this task. IPhone 6s two identical two different chips on which you can check a different model numbers and size for APL0898 APL1022 Samsung and TSMC.
The Chipworks group was instructed to find the differences between the two chips. The first APL0898, has an area of 96 square millimeters and is manufactured by Samsung. The second APL1022 is 104.5 square millimeters and manufactured by TSMC. Knowing the manufacturing process, FinFET, it is known that the transistors equal to FinFET Samsung’s 14nm and 16nm TSMC to.
This maneuver creates two very interesting points. From the point of view of industry, Samsung and TSMC have a long history of disputes at various levels (technical and legal) for being the first to get a higher FinFET manufacturing process and therefore capture more demand for manufacturing ARM designs chips.
Samsung was the first to achieve 14nm chips manufactured at a high level with the Exynos 7420 that gave such good results in the Galaxy S6 vs. direct competition with Qualcomm Snapdragon 810. Instead, the first product of this generation TSMC is Apple A9 manufacturing of 16nm FinFET, indicating that, at last, be distributed – and reliably high level. – such chips
More importantly Apple: this is a movement of absolute power. In the past – even in the present, but occasionally – it was relatively common depend on several manufacturers to design the same chip, screen or piece. AMD born from this need many years for example. But on mobile devices, more specifically in processors, is unheard far as I can tell, by how difficult and costly it is to do.
After designing the chip have to agree with the two manufacturers hows and when, requiring large financial and engineering resources by the three parties. Something that, in essence, few besides Apple can do.
The biggest problem is that two different manufacturers to manufacture an chip as identical as possible and meets the requirements exactly. Each manufacturer has its techniques and steps at the time of manufacture. The end result is that for very similar and they meet requirements, it is impossible to be exactly the same.
In this case we can see differences even physically. Although both processes TSMC and Samsung are very similar, the second chip is 8% smaller than the first thanks to a higher transistor density. However, this should not be a problem if the two reach the same speeds, something that Apple has purchased and that the benchmarks that have been used reported. Same speed and performance
But of course, problem comes in the efficiency of the chips, much more difficult to control due to differences in manufacturing and is not something that can be altered from the overnight. The two chips are to be as similar as possible by controlling the voltage using at rest and higher clock speeds. But it was clear that there would be differences, as it is normal. Since it can not be avoided.
Why Apple has decided to invest more money, more resources engineering and create potential problems by hiring two manufacturers to develop their Apple A9?
There various possibilities, but they can only be speculation. The first and most likely is not to count exclusively on the present and the future with a single manufacturer: this would cause problems in the long term volume and potential contracts. It can be dangerous for Apple that only Samsung could make their chips. In addition to TSMC also taking stock of iPhones can grow faster and supply the growing demand every year to enjoy the iPhone. If Apple has decided to invest more resources to do this must be for something.
Differences between TSMC and Samsung
Consumers have been able to make the first test. It all started with this Reddit thread where two different iPhone 6s Plus chips were compared by battery benchmark Geekbench. According to him and screenshots chip manufactured by TSMC was more efficient, giving about 2 hours of battery test
It was expected that Samsung chips -. When 14nm – would be more efficient . What came as a surprise, besides the difference was not as mild as expected. Difference, as I stressed before, you can only minimize, not eliminate.
A single user performing a synthetic test is not enough proof, but mydrivers made a series of tests most enlightening.
Putting the chip to render javascript, with the same brightness and settings in both iPhone, showed the following:
At 45 minutes A9 iPhone with TSMC (APL1022) had 89 % of battery, the iPhone with Samsung chip (APL0898) 82%.
75 minutes the results were 74% for TSMC and 59% for Samsung.
Still, although the difference is quite large, only it is testing the chip in a situation that hardly give so continuously in real life and whose only parameter is a CPU operating at more than 60% of capacity. A slightly higher voltage at that speed could be responsible.
Several teams have made more similar test uses real life, but not be controlled laboratory tests, and found the same and obvious differences between the two chips:
Why this happens
Because they are two different manufacturers, with two different manufacturing processes and manufacturing techniques?. Implement a processor is a very complex task and the goal is to meet the previously stipulated requirements. Within these requirements there are ranges of voltages to the level of absolute rest or higher consumption, for instance.
It is impossible for a chip Samsung and TSMC chip can consistently have the same efficiency at different clock frequencies and performing processing load.
For fans of PC hardware this is nothing new. Even Intel processors, depending on the manufacturing date printed on the box processor can have a voltage identification (VID) different than it left the factory last week. For those who through the BIOS increase the bus frequency and multiplier of processors is very important because if a processor achieves a frequency at lower voltage temperature dissipated so will and, therefore, the margin increased by overclock will be higher.
No comments:
Post a Comment