5 ms·

I have an M1 Max. If anyone wants me to benchmark any workflows, I'd be glad to help!

pvinis 5y ago
Maybe you could try this one? https://github.com/ashfurrow/xcode-hardware-performance https://github.com/ashfurrow/xcode-hardware-performance
KptMarchewa 5y ago
Do you have two 4k displays? If yes, then does running them affects performance significantly? On i9 MBP this was causing extremely severe performance issues, to the point where processor was throttled to 30% performance.
- EricE 5y ago
  Biggest reason to get 64GB of RAM if you intend to do this - the Unified Memory is shared between the CPU and GPU. More pixels in total that the machine has to render, the more RAM will be needed for the GPUs. Don't skimp on RAM or buy into the hype that ARM halves the amount of required RAM - it's total BS.
- ReaLNero 5y ago
  I only have one 4k display, and I don't notice any difference in performance.
- codeisawesome 5y ago
  I run two 5Ks and my 2019 Intel feels quite sluggish. I still pay that price for the real estate.
  - KptMarchewa 5y ago
    That's what I would love to avoid.
- ReaLNero 5y ago
  I only have one external 4k 60hz display. I played DOTA 2 on it with no issues.
bullen 5y ago
I'm having a hard time finding mips/flops for the M1 CPU, can you measure one of them?
- brrrrrm 5y ago
  ~100G fp32 FMA/s per core on my M1 air last I checked.
  - bullen 5y ago
    Is that CPU or GPU? Sounds a bit high when 2Gflops/watt (but I guess at 40W for all cores =2,5Gflops/W it could make sense?!) is the industry standard! Can you turn off the acceleration to get CPU only figures? What is FMA/s? Edit: Ok that is some sort of SIMD/GPU... I need figures for the raw CPU that can also do general purpose calculations and branches! Maybe MIPS is a better "apples to apples" measurement?
    - brrrrrm 5y ago
      Ah, it's actually 100gflops (not 100Gfmas) because the fmla instruction is fused. I just ran it on CPU, single core (no pinning), just using the aarch64 fmla instruction. Using pure BLAS (multithreaded, with the MM ISA extension), I get like 1.2TFlops in fp32 on the whole device. Code for the blas test: https://jott.live/code/blas_test.cc https://jott.live/code/blas_test.cc
      - bullen 5y ago
        So a friend that has a M1 did this test: https://github.com/brianolson/flops/blob/master/flops.c https://github.com/brianolson/flops/blob/master/flops.c And the 5nm M1 has ~2.5Gflops/W which is not a huge increase compared to the 28nm Pi 4 at 2Gflops/W. No-moores law in effect. Game Over!
      - bullen 5y ago
        Can you run it without SIMD (fmla)? Same as these tests: http://web.eece.maine.edu/~vweaver/group/green_machines.html http://web.eece.maine.edu/~vweaver/group/green_machines.html
floatboth 5y ago
Relatively obscure fun thing: CRC32 checksum instructions. https://github.com/srijs/rust-crc32fast/pull/6 https://github.com/srijs/rust-crc32fast/pull/6 (Get rust nightly via https://rustup.rs https://rustup.rs, pull the repo, run: cargo bench --features nightly)
- weatherlight 5y ago
  (16-inch, 2019) - 2.6 GHz 6-Core Intel Core i7 - 16 GB 2667 MHz DDR4 test result: ok. 0 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 0.00s Running unittests (target/release/deps/bench-b9bcbb53a5ef2c28) running 4 tests test bench_kilobyte_baseline ... bench: 310 ns/iter (+/- 70) = 3303 MB/s test bench_kilobyte_specialized ... bench: 48 ns/iter (+/- 5) = 21333 MB/s test bench_megabyte_baseline ... bench: 298,273 ns/iter (+/- 42,157) = 3515 MB/s test bench_megabyte_specialized ... bench: 45,100 ns/iter (+/- 7,249) = 23250 MB/s test result: ok. 0 passed; 0 failed; 0 ignored; 4 measured ----------------------------------------------------------- (14-inch, 2021) - Apple M1 Max - 64gb test result: ok. 0 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 0.00s Running unittests (target/release/deps/bench-3eea1c397faf4328) running 4 tests test bench_kilobyte_baseline ... bench: 223 ns/iter (+/- 8) = 4591 MB/s test bench_kilobyte_specialized ... bench: 100 ns/iter (+/- 0) = 10240 MB/s test bench_megabyte_baseline ... bench: 231,689 ns/iter (+/- 1,885) = 4525 MB/s test bench_megabyte_specialized ... bench: 122,382 ns/iter (+/- 4,651) = 8568 MB/s test result: ok. 0 passed; 0 failed; 0 ignored; 4 measured
  - floatboth 5y ago
    Thanks! Interesting result, baseline performance is impressive (pretty close to my R9 5950X which does 4853/5069 MB/s) but the CRC32 hardware is surprisingly mediocre (Cortex-A72 on Graviton1 does 14840/14483, Neoverse N1 on Graviton2 does 17066/19560)

Maybe you could try this one? https://github.com/ashfurrow/xcode-hardware-performance https://github.com/ashfurrow/xcode-hardware-performance

KptMarchewa 5y ago

Do you have two 4k displays? If yes, then does running them affects performance significantly? On i9 MBP this was causing extremely severe performance issues, to the point where processor was throttled to 30% performance.

EricE 5y ago
Biggest reason to get 64GB of RAM if you intend to do this - the Unified Memory is shared between the CPU and GPU. More pixels in total that the machine has to render, the more RAM will be needed for the GPUs. Don't skimp on RAM or buy into the hype that ARM halves the amount of required RAM - it's total BS.
ReaLNero 5y ago
I only have one 4k display, and I don't notice any difference in performance.
codeisawesome 5y ago
I run two 5Ks and my 2019 Intel feels quite sluggish. I still pay that price for the real estate.
- KptMarchewa 5y ago
  That's what I would love to avoid.
ReaLNero 5y ago
I only have one external 4k 60hz display. I played DOTA 2 on it with no issues.

bullen 5y ago

I'm having a hard time finding mips/flops for the M1 CPU, can you measure one of them?

brrrrrm 5y ago
~100G fp32 FMA/s per core on my M1 air last I checked.
- bullen 5y ago
  Is that CPU or GPU? Sounds a bit high when 2Gflops/watt (but I guess at 40W for all cores =2,5Gflops/W it could make sense?!) is the industry standard! Can you turn off the acceleration to get CPU only figures? What is FMA/s? Edit: Ok that is some sort of SIMD/GPU... I need figures for the raw CPU that can also do general purpose calculations and branches! Maybe MIPS is a better "apples to apples" measurement?
  - brrrrrm 5y ago
    Ah, it's actually 100gflops (not 100Gfmas) because the fmla instruction is fused. I just ran it on CPU, single core (no pinning), just using the aarch64 fmla instruction. Using pure BLAS (multithreaded, with the MM ISA extension), I get like 1.2TFlops in fp32 on the whole device. Code for the blas test: https://jott.live/code/blas_test.cc https://jott.live/code/blas_test.cc
    - bullen 5y ago
      So a friend that has a M1 did this test: https://github.com/brianolson/flops/blob/master/flops.c https://github.com/brianolson/flops/blob/master/flops.c And the 5nm M1 has ~2.5Gflops/W which is not a huge increase compared to the 28nm Pi 4 at 2Gflops/W. No-moores law in effect. Game Over!
    - bullen 5y ago
      Can you run it without SIMD (fmla)? Same as these tests: http://web.eece.maine.edu/~vweaver/group/green_machines.html http://web.eece.maine.edu/~vweaver/group/green_machines.html

floatboth 5y ago

Relatively obscure fun thing: CRC32 checksum instructions. https://github.com/srijs/rust-crc32fast/pull/6 https://github.com/srijs/rust-crc32fast/pull/6 (Get rust nightly via https://rustup.rs https://rustup.rs, pull the repo, run: cargo bench --features nightly)

weatherlight 5y ago
(16-inch, 2019) - 2.6 GHz 6-Core Intel Core i7 - 16 GB 2667 MHz DDR4 test result: ok. 0 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 0.00s Running unittests (target/release/deps/bench-b9bcbb53a5ef2c28) running 4 tests test bench_kilobyte_baseline ... bench: 310 ns/iter (+/- 70) = 3303 MB/s test bench_kilobyte_specialized ... bench: 48 ns/iter (+/- 5) = 21333 MB/s test bench_megabyte_baseline ... bench: 298,273 ns/iter (+/- 42,157) = 3515 MB/s test bench_megabyte_specialized ... bench: 45,100 ns/iter (+/- 7,249) = 23250 MB/s test result: ok. 0 passed; 0 failed; 0 ignored; 4 measured ----------------------------------------------------------- (14-inch, 2021) - Apple M1 Max - 64gb test result: ok. 0 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 0.00s Running unittests (target/release/deps/bench-3eea1c397faf4328) running 4 tests test bench_kilobyte_baseline ... bench: 223 ns/iter (+/- 8) = 4591 MB/s test bench_kilobyte_specialized ... bench: 100 ns/iter (+/- 0) = 10240 MB/s test bench_megabyte_baseline ... bench: 231,689 ns/iter (+/- 1,885) = 4525 MB/s test bench_megabyte_specialized ... bench: 122,382 ns/iter (+/- 4,651) = 8568 MB/s test result: ok. 0 passed; 0 failed; 0 ignored; 4 measured
- floatboth 5y ago
  Thanks! Interesting result, baseline performance is impressive (pretty close to my R9 5950X which does 4853/5069 MB/s) but the CRC32 hardware is surprisingly mediocre (Cortex-A72 on Graviton1 does 14840/14483, Neoverse N1 on Graviton2 does 17066/19560)