| Bofferbrauer2 said: ARM is getting quite speedy in synthetic benchmarks: https://www.xda-developers.com/snapdragon-x-elite-vs-intel-core-ultra-7-155h/#:~:text=Key%20Takeaways,Life%20Extreme%20and%20Aztec%20Ruins. Not only does the Qualcomm Snapdragon Elite X defeat the Core Ultra 155H, it practically pulverizes the Apple M2 in the MacBook Air. I wonder how that translates into real-world performance though, especially on the GPU, as almost all tests were synthetic benches. |
The issue with ARM SoC's is that they bench well, as they build the SoC's for specific work loads due to the limited use-case scenario that the devices housing these chips usually have.
Usually power consumption is the main aim of the game... So there is a heavy reliance on fixed function blocks.
Case in point: Often if you do a HEVC encode on a PC, it's often done on the CPU unless you explicitly ask for the GPU to do it... On an ARM SoC it's done on fixed function hardware specific for that task.
Step outside of that and performance tends to tank.
It's a different design philosophy that Intel and AMD tend to take which is to be good at everything, but master of none... As those core designs need to scale to high-end server, down to workstations and then into laptops, tablets, IoT and more, they are meant to be good at everything.
But there is a reason why most comparisons just use Cinebench when comparing ARM vs x86, it's been known for years that certain behind the scenes "tricks" happen...
I.E. When a certain benchmark is detected, things like Thermal limits get raised... In Samsungs case they even went as far as boosting voltage and frequency to maximum levels, battery life be damned to get the best benchmark result... But when you don't run that benchmark, you would never hit this thermal limits in the real world.
--::{PC Gaming Master Race}::--
