Would a smaller & lower power OoO ARM core (e.g. A610/A615) be a valuable addition?

Newguy

Regular
Obviously ARM haven't done so in all these years and have no public plans which is an answer in and of itself, however currently they have:
  • X_ series - highest performance "big" cores which split from the A7x a few years ago, laptop potential. X1, X2, X3
  • A7_ series - now the PPA efficiency "medium" cores, A78, A710, A715
  • A5_ series - very small & low cost, area efficient, low absolute power, In-order "small" cores, A53, A55, A510 (that's 3 gens from ~2014-2023)
For the Snapdragon 8 gen 2 the A715s are 3.3x more efficient in SpecInt 2017 and 4.9x in Specfp2017 vs the A510, the X3's efficiency is similar to the A715. The A510/family will always have a place in entry-mid level devices because of their much lower cost (e.g. A55 is 0.3mm2, A76 is 1.26mm2 w/ 512KB L2, A78 about 1.4mm2 using relative numbers) and they have talked about iterating In-order cores faster than before which will hopefully reduce the gap ("Hayes"being 2023 + 2024 per roadmap). But for example a hypothetical A615 that's about 0.7-0.8mm2, similarly efficient to the larger OoO cores (3-4x vs A510, so 1.5-2x faster & uses about 0.5-0.7x the power for example) would make sense to replace A510s in premium SoCs, be great for laptops and could save money for mid-range SoCs (e.g. 2x A7_, 2x A6_, 4x A5_ config)

I get why they wouldn't want to, another set of cores on top of already strained resources plus you can have 2-4 lower clocked A7_ cores instead which makes all of that pointless anyway. But the tech enthusiast in me wants it just because


Spec scores, power & efficiency calcs:
Int scores & total power, A715/A510 ((3.64 points/1.03)/(5062 Joules/4698))=3.28x efficiency
FP scores & total power, A715/A510 ((5.7 points/1.14)/(4093 Joules/3994))=4.88x efficiency
 
Back
Top