New Snapdragon 8 Gen 3 benchmarks indicate massive GPU upgrades and a solid CPU boost

[Redaktion]

The Snapdragon 8 Gen 3 will power the majority of next-gen Android flagships and is scheduled to debut later this month. The premium chipset appears to have made an early trip over to AnTuTu, however, with the benchmark revealing significant GPU performance upgrades.

https://www.notebookcheck.net/New-Snapdragon-8-Gen-3-benchmarks-indicate-massive-GPU-upgrades-and-a-solid-CPU-boost.758120.0.html

[NikoB]

What about a boost in the battery life of a smartphone from 3 days with 8 hours of work with the screen and 20-30% of the load cores in the background without a screen with 12-14 hours  every day? Why 3 days? Because then a regular battery has a lifespan of 5 years - 600-700 cycles before the capacity drops by 20-25%. And modern smartphones, carefully configured by a smart owner, make no sense to change more than once every 5 years...

[ArsLoginName]

I don't agree with NikoB often, but I agree with what I think he wants - which is not so much of a performance increase at the same power consumption in the SD G3. But he wants a much longer usable time before the battery needs to be recharged so the usable lifetime of the phone increases to 5-6 years from 2-3 years assuming the phone lifetime is primarily determined by how long it takes before you need a battery replacement (neglecting drops and broken screens).

This makes a lot of sense since the leading flagship phones typically receive 5 years of software & security upgrades. If battery life could be extended to 5 years, at least it would be in sync with software support and at that point, the phone could either be sent for recycling or it could get refurbished with a battery replacement and sold for cheap since it is out of software support.

[NikoB]

could either be sent for recycling

If the iron is functional, there is no point in sending it for recycling - the iron must exhaust its resource completely. Anything else is selfish excess.

The flatter the energy efficiency growth curve becomes, and this is exactly the case and is becoming increasingly clear for the majority, the less desire people will have to change gadgets. And as I wrote earlier, new classes of tasks at a qualitative level, feasible only in the presence of powerful local neural networks (you don't want to depend on clouds and the decisions of guys there), will require a jump in performance by several orders of magnitude, and not several times ( evolutionary path), as well as the size of RAM and non-volatile memory - just estimate the scale of modern data centers designed for neural networks - there the bill has already gone to exabytes just to make at least some kind of neural network that is really useful in practice.

All modern neuro-coprocessors in the SoC of laptops and smartphones are essentially meaningless and useless toys without hundreds of terabytes of RAM and petabyte disk systems.

For this reason, humanity has been frozen for at least 50 years at this level of productivity and there will definitely not be any significant leap in the near future - the world has already been rapidly degrading since 2019, it had no time for fundamental science before, and now it has no time for it at all .

Radical changes are needed in the fundamental scientific achievements of humanity, embodied technologically, so that a new qualitative leap can be made in the "power supply" (in all senses) of the individual. Which will naturally change the entire face of civilization.

We are now approximately frozen at the level of "steam locomotives" and "airplanes" still require decades of accumulation of qualitative technological changes. While there were steam locomotives, humanity could not travel around the planet as easily as it does now and so quickly, but this required only 50 years of progress. New tasks are much more difficult than airplanes and even a flight to Mars with people, but not with autonomous robots...

This is something like the "hyperdrives" and jumps through subspace invented long ago by science fiction writers. It's easy to dream up, just like about real AI, but to implement it will require centuries and millennia of fundamental research in a favorable environment for the creativity of scientists.

At least, in contrast to "hyperdrive", we now see the living embodiment of AI - a human brain weighing 1.3-2 kg, albeit with poor long-term memory and a lot of stability problems, in reality, but capable of creating and moving into the unknown, albeit not so quickly as we would like now, i.e. here, at least, humanity already knows in advance that the problem is 100% solvable in principle, in contrast to the hypothetical "hyperdrive" and "hyper-jumps" in subspace.