Qualcomm does not use "pure" ARM designs in their solutions , they prefer to take them , redesign them but keeping them compatible with the ARM architecture and set of commands in order to make them compatible with ARM's solutions .
This is possible because the company purchased this kind of license from ARM holdings . the company responsible for ARM's development and licensing .
This policy reached it's peak with the Scorpion cores in the first Snapdragon chips , ones of the APQ range since at first , manufacturing at that time was not able to fit the CPU core , all the other modules and the Radio Logic in one chip .
Scorpion Cores were a redesigned Cortex A-8 ARM core one that in theory could outperform most A-8 implementations , depending on the optional features that ARM provided to each A-8 maker .
Qualcomm's cores are designed to outlast the ARM's generation on which they are based and usually make it untill the next one reaches the end of it's life .
Scorpion cores were designed to live theoretically up until the time A-8 successor the A-9 would be close to the end of it's life .
For the most part Qualcomm was successful on this , not without any drawbacks however .
Scoprion cores appeared on the market sooner that most A-8's and were far more efficient than ARM-11 the cores that the A-8 later replaced . Once A-8's reached the market however it became clear that most carefully designed A-8 SoC's were much better than their Qualcomm counterparts based on the Scoprion core .
However , when A-9's started to appear , the situation changed somewhat in favour of Qualcomm .
Most A-9 SoC's came in Dual Core designs , and Qualcomm simply created Dual Core Scorpion MSM's and APQ's . Since in the A-9 the NEON module was optional unlike the A-8 , and some SoC makers like NVIDIA chose not to include it on their design , Scorpion based solutions managed to outperform these SoC's . Since they were based on the older A-8 core , even though they were an improved version of them , they could not outperform A-9 designs that included the NEON engine and had carefully designed FPU's .
This brought us to the Krait era , with the Krait cores being redesigned Cortex A-15 cores , and came to market a few months before the first A-15 based solutions .
This time around , Qualcomm bet on the fact that A-15's were a far different design from both A-8 and A-9 ones and along with their many advantages they carried many disadvantages like increased power consumption and increased die size inside the SoC . They also made a bet that most other SoC manufacturers would first jump to Quad Core solutions based on the A-9 design in order to both buy some time to get their A-15 solutions ready and because marketing reasons were shifting the market at that direction .
With all that in mind , this time Qualcomm designed the Krait core as an improved version of the A-9 core , based on the A-15 . In essence they improved every aspect of the A-9 that could be improved and left out the features of the A-15 that caused "problems" .
The end result was of course as we all know an amazingly fast core compared to most A-9 designs , one that in dual core configuration could match the CPU performance of most Quad Core A-9 SoC's taking advantage of the fact that most software is not optimized for Quad Core solutions , including Android itself .
Qualcomm hopes that this way , by the time the first A-15 cores reach the market in Dual Core designs they will already have gained a substantial market share with both their Dual and their Quad Core Krait solutions .
And now let us see how the APQ's got a foothold on the mobile phone market starting at the time the first Krait MSM's reached the market .
Qualcomm is not actually manufacturing it's own chips . They outsource this task to foundries like TSMC and make use of their manufacturing technology and subsequent advances in manufacturing technology .
Qualcomm was the first to take advantage of TSMC's 28nm manufacturing process by building their first Krait based MSM's based on it . At this time , only the Dual Core design was ready , and the one with the slightly updated Adreano 225 GPU , with GPU's being a seperate article of their own .
This was done in order to achieve the goal of creating an MSM solution containing everything an MSM normally does along with the new cores that are of course larger in die size than the A-9 while at the same time being smaller than the standard A-15 ones .
The first MSM's created were of two different kinds . One of them supported each imaginable radio interface including LTE and the other one lacking just the LTE capabilities .
MDM's suppporting LTE were already available , and although they also supported UMTS and GSM network interfaces up untill this point they lacked the capability of supporting voice calls , a capability that the first Krait MSM's of course had .
This was done in purpose from Qualcomm's part in order to guarantee sales of their dual core krait designs to phone makers that required the simplicity of the MSM design in order to make LTE capable mobile phones .
Qualcomm knew from the start that at some point they would have to make MDM's that support voice functions available and being one of Apple's key suppliers they knew before hand that this had to happen before Apple made the iPhone 5 available which would support LTE networks but given that Apple would use their own A6 design as the phone's SoC they would require from Qualcomm an MDM capable of voice calls .
They also knew that by the time MDM's with voice support would start being sampled to Apple , the same should be done to all other manufacturers that were up untill then left only with the dual core Krait MSM's to support LTE connectivity along with voice calls .
They also made sure that by the time that happened they should also have some kind of Quad Core Krait solutions available to keep the phone makers from choosing their new voice capable MDM but at the same time choosing someone else's SoC since regardless of the real life performance of Krait cores , the market was demanding Quad Core solutions as a result of manufacturer's marketing strategies .
At this point it became clear that four Krait cores along with everything else including and upgraded GPU would not fit inside a normally sized MSM solution and Qualcomm choose to create the first Quad Core designs as APQ's , lacking the Radio Logic .
Qualcomm's effort to beat the A-15's to the market with their Krait cores and provide them as Quad Core solutions along with voice capable MDM's proved to be successful for the most part since for the time being A-15 designs are not on the final stages of actual phone designs and are only destined to live inside tablets for the next few months that provide more space to fit the increased SoC size .
For the most part , even though someone could argue that this was a monopoly strategy , one that could be considered hostile to other SoC makers , it proved to be successful . Given that most high end phones containing Quad Core APQ's with or without LTE capable MDM's will stay on the market long enough to give Qualcomm a good reputation for their Krait cores before the first A-15 SoC appear on mobile phones everything worked as Qualcomm wanted and they also indirectly made the APQ series a successful one .
Finally my prediction is that the Scorpion history will repeat itself once again and when the A-15 reach the market , they will of course prove to be much more capable that the Krait core .
Given that the optional features of the A-15 core are limited and most crucial ones are mandatory for all SoC makers that will make use of this core , even Dual Core A-15 solutions will outperform Quad Core Krait ones and Qualcomm will once again find itself trailing behind the competition , with their monopoly of the LTE market for the coming months being their only stronghold up until capable solutions from Icera appear inside Nvidia's Tegra SoC's and Samsung starts using their own LTE solution combined with their Exynos SoC's outside their home Korean market .
The fact that A-15 cores were actually delayed on the market , also provided SoC makers with the option to combine them with even better and more capable GPU's , that we will discuss later on in the coming days .
No comments:
Post a Comment