Yep. Also, when I do come up with a new solution for something, I pursue it until I prove that it works, then have no interest in productizing it (that part seems so boring).
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
TSMC Paves Way for 5nm A14 Chip in 2020 iPhones
- Thread starter MacRumors
- Start date
- Sort by reaction score
Yep. Also, when I do come up with a new solution for something, I pursue it until I prove that it works, then have no interest in productizing it (that part seems so boring).
They actually are used in the real world, a LOT, in some of the most fundamental ways - that's why Geekbench represents real-world usage, and aren't made up synthetic benchmarks. Geekbench uses real code, such as SQLite, Gumbo, LLVM, etc: https://www.geekbench.com/doc/geekbench4-cpu-workloads.pdf
In the end, if you see a Geekbench result that's 2x faster, that means your app is going to run 2x faster.
Thanks.
[doublepost=1554947443][/doublepost]
Those libraries are the critical path in applications. The GUI and other interfaces aren't.
You don't measure processing speed by how fast you can enter data in a spreadsheet or how fast you can play a game.
You measure processing speed by the core compute algorithms they use.
LoL, Of course it's using real code. Your "logic" doesn't make sense and you haven't contradicted in any way what I wrote.
That's just your opinion.
Er. No? That's the whole point: it does not mean that at all. Not only are there far more variables to consider than the CPU, it's not even true of the CPU, because your app will be more complex than that, and your OS runs dozens to hundreds of processes next to it.
We'll be generous and assume that you mean twice-as-fast single-core results (because your assertion is even sillier for multi-core performance). That will yield nowhere near twice the overall perceived performance of the app.
This implies a separation (e.g. into different threads that never need to sync with each other) that doesn't actually take place in practice.
And no, most GUI apps never have some isolated algorithm as their "critical path". That was kind of true in the 1990s, but these days, you waste far more time either fighting for that single core, synchronizing between threads (if you bother writing it like that), or, most crucially, waiting for I/O.
Yes, which is called a synthetic benchmark.
Last edited:
I've never said these things. Why are you making up lies?
EDIT:
those quotes are by mozumder. not me. not sure why you put my name on those.
Sorry about that. I must've hit the wrong quote button or something. Weird.
[doublepost=1555012088][/doublepost]
There ultimately isn't much to discuss on 5nm chips, is there?
They're a fabricator. They don't design chips. The A~ class chips are designed by Apple engineers and produced by a fabricator.
It would first have to pass regulatory hurdles in several countries, including ones not related to either company. It would also be odd considering TSMC has snapped up parts of Qualcomm in the past. Apple and Qualcomm tried to buy out exclusive production rights from TSMC years ago but they declined the offer because they don't need outside investment and prefer to have their fabs open for everyone to use. AMD is one of TSMC's bigger clients. 7nm and 5nm AMD CPU and GPUs will be produced by TSMC, 12nm and larger will be held by GloFlo and Samsung may be tapped as an additional supplier for 7nm and 5nm production if TSMC can't keep up with demand.
Yeah sure sure.
If you would be interested to know what synthetic benchmarks are and you would search for definitions, descriptions you would immediately realize how perfectly it fits with what Geekbench is.
Synthetic benchmarks are artificial programs that are constructed to try to match the characteristics of a large set of programs. The goal is to create a single benchmark program where the execution frequency of statements in the benchmark matches the statement frequency in a large set of benchmarks.
One major drawback of synthetic benchmarks is that no user would ever run a synthetic benchmark as an application because these programs don’t compute anything a user would find remotely interesting. Furthermore, because synthetic benchmarks are not real programs, they usually do not reflect program behavior, other than the behavior considered when they were created. Finally, compiler and hardware optimizations can inflate performance of these benchmarks, far beyond what the same optimizations would achieve on real programs. Of course, because these benchmarks are not natural programs, they may not reward optimizations of behavior that occur in real programs.
Last edited:
This whole conversation is so bizarre.
You should see the benchmarks the cpu architects use to evaluate design decisions. Snippets of algorithms like SPICE, etc.
Geek bench is practically golden in comparison.
No reason to worry about processor tech at this point especially with Apple.....their UI on mobile devices is not good enough for their processors.......
When run isolated without any purpose other than to measure a number? Yes, absolutely it is.
[doublepost=1555684244][/doublepost]
The conversation isn’t whether Geekbench is good or bad at all. It’s whether it represents real-world use of a machine.