wickedpotus wrote: ↑Mon Nov 22, 2021 6:12 pmBut the sad truth is that Apples CPU dont even support hyper-threading so it is more designed for simple one-thread tasks and not for high-perfroamnce computin like chess-apps etc.
This sentence strongly suggests that you don't really know what you're talking about and are just driven by silly Apple hatred.
The M1 is a very interesting cpu because it shows that the ARM architecture may turn out to be superior to the x86 architecture also for high-performance computing. There is a serious possibility that x86 will have disappeared by the end of the decade.
wickedpotus wrote: ↑Mon Nov 22, 2021 6:12 pmBut the sad truth is that Apples CPU dont even support hyper-threading so it is more designed for simple one-thread tasks and not for high-perfroamnce computin like chess-apps etc.
This sentence strongly suggests that you don't really know what you're talking about and are just driven by silly Apple hatred.
The M1 is a very interesting cpu because it shows that the ARM architecture may turn out to be superior to the x86 architecture also for high-performance computing. There is a serious possibility that x86 will have disappeared by the end of the decade.
This sentence strongly suggests you don't really know much about CPU architecture and high performance CPU design beside what you read on Wikipedia and are just driven by silly Apply fanboyism.
The irony is that you think you really understand it just because you went "back and forth" through Intel Architecture Software Developer's Manual.
M1 in all its variants is definitely a bit over-hyped - I got cinebench benchmarks to compare, haven't yet been able to test for chess as I don't own an M1 Mac (the "M" in M1 stands for magical btw... )
However, HT in chess doesn't seem to yield much elo anymore at decent thread counts.
Werewolf wrote: ↑Tue Dec 28, 2021 2:21 pm
M1 in all its variants is definitely a bit over-hyped - I got cinebench benchmarks to compare, haven't yet been able to test for chess as I don't own an M1 Mac (the "M" in M1 stands for magical btw... )
However, HT in chess doesn't seem to yield much elo anymore at decent thread counts.
HT increase the number of independent instructions in the pipeline, when you increase the number of threads.
Werewolf wrote: ↑Tue Dec 28, 2021 2:21 pm
M1 in all its variants is definitely a bit over-hyped - I got cinebench benchmarks to compare, haven't yet been able to test for chess as I don't own an M1 Mac (the "M" in M1 stands for magical btw... )
However, HT in chess doesn't seem to yield much elo anymore at decent thread counts.
HT increase the number of independent instructions in the pipeline, when you increase the number of threads.
Yes.
But for chess 16 cores running 16 threads, on a processor with no HT, yields very close in elo to 16 cores running 32 threads on a processor with HT (assuming same clock speed, efficiency etc).
Werewolf wrote: ↑Tue Dec 28, 2021 2:21 pmHowever, HT in chess doesn't seem to yield much elo anymore at decent thread counts.
But it does allow running more games in parallel.
Allows it, but how does that affect the results ?
Your point is completely moot. .its like arguing 30% speed increase does or does not bring visible elo under specific TCs etc- .... So why speed anything up?... Completely irrelevant and besides the discussion if HT (or clustered multithreading) is useful or not.
Simple question. How many NPS can you get out of :
M1, M1 Max, Apple i7/i9, Windows/Linux notebook i7/i9, Windows/Linux desktop i7/i9?
My testing of the same program compiled for the different platforms gives the following conclusions:
2012 Mac i5: 2,5 * 4
2017 Dell i7: 6 * 8
2012 HP Xeon: 3,5 * 18
2020 HP centreon: 3,8 * 2
The latter is the available threads..... The first is the number of NPS per thread. Wonder to see how well M1 fares. And how this compares to the latest Ryzen.
)