Qualcomm ’s SOC performance must catch up with Apple’s several hurdles:
①. Big-core performance
②. Small-core performance _p1span_span.
Judging from these three items, Qualcomm has lost very thoroughly, especially in the comparison of small core performance, which is most likely to catch up with the peak performance of the GPU.
From the perspective of Apple's design model, stronger peak performance will bring stronger real-time performance test performance, and at the same time, it can also burst out of good strength under low power consumption. For example, this time iPhone 13 The A15 on the Pro MAX uses much lower power consumption in the power-saving mode to outperform the Snapdragon 888 with full firepower
large-core performance, I think it will be useless for Snapdragon to catch up next year. The gap can be shortened on IPC, because the IPC of A15’s Avalanche has basically remained unchanged, but the single-core performance will still be. There is a very big gap, because Apple has increased the frequency
In addition to the powerful single-core performance of Apple, it also has a large and efficient micro-architecture, which is impossible for Qualcomm to achieve in one step or two steps. To catch up with the distance, even with the acquisition of NUVIA, whether it is possible to create an architecture stronger than Apple with limited R&D resources is a question mark
A15's SPEC score has not yet come out ,So taking A14's Firestorm architecture to compare, Firetorm's INT IPC is 46% higher than the Cortex X1 on Snapdragon 888, which means that when the frequency gap is not large, the products corresponding to these two architectures will have huge performance. The gap,
, and even according to the INT IPC that the Cortex X2 written on ARM’s PPT is 16% higher than the 4MB L3 Cortex X1, the best result is 16.82 points/GHz, and Firestorm is still 26% higher than it. INT IPC, this is still the best result,
This means that if you can’t lead the frequency by a large margin, then the single-core performance of is still suppressed, but Apple has already pulled the frequency to 3.23GHz. The large core frequency of the first generation of Snapdragon processors is almost impossible to exceed it, so there is no way to catch up with the performance of large cores.
The performance of small cores is also an aspect that Qualcomm is almost impossible to catch up.
The small core frequency of the flagship SOC on the mobile phone is basically the same, which means that Icestorm, which has several times the IPC of the A55/A510, is an almost impossible target for the A55/A510, and the new A510 is conventionally released. Small cores will also need to be used for several years, so the performance of small cores is even more difficult to catch up. Apple's small cores can provide good performance at ultra-low power consumption. This is not possible on Android. Performance needs to be done by large cores with much higher power consumption or mid-core
GPU peak performance is the point that Qualcomm is most likely to catch up, and there are already some rumours. The first generation of Qualcomm's flagship SOC will use low-frequency GPUs, and the performance is good, which shows that the Adreno7 series GPUs are still doing considerable improvements in architecture.
Under the same performance, low-frequency and high-scale GPUs will have the advantage of higher energy efficiency under light load compared with high-frequency and low-scale GPUs. So the next-generation Qualcomm flagship SOC should theoretically not be as hot as the Snapdragon 888
Of course, for the current Qualcomm Snapdragon flagship SOC, if you can switch to the TSMC process, it seems more worth looking forward to reducing the power consumption of the processor. Samsung’s process is still a little watery, with density and leakage. The rate is not better than TSMC.
.
