MacBook Neo

 ## Basic information - Board URL (official): https://www.apple.com/macbook-neo/ - Board purchased from: Apple - Board purchase date: March 11, 2026 - Board specs (as tested): A18 Pro / 8GB / 256GB / No Touch ID - Board price (as tested): $599 ## Linux/system information ``` # output of `screenfetch` -/+:. jgeerling@macbook-neo :++++. OS: 64bit macOS /+++/. Kernel: arm64 Darwin 25.3.0 .:-::- .+/:-``.::- Uptime: 2h 9m .:/++++++/::::/++++++/:` Packages: 146 .:///////////////////////:` Shell: zsh 5.9 ////////////////////////` Resolution: 3840x2160 -+++++++++++++++++++++++` DE: Aqua /++++++++++++++++++++++/ WM: Quartz Compositor /sssssssssssssssssssssss. WM Theme: Blue (Dark) :ssssssssssssssssssssssss- Font: FMonoMedium osssssssssssssssssssssssso/` Disk: 108G / 245G (49%) `syyyyyyyyyyyyyyyyyyyyyyyy+` CPU: Apple A18 Pro `ossssssssssssssssssssss/ GPU: Apple A18 Pro :ooooooooooooooooooo+. RAM: 892MiB / 8192MiB `:+oo+/:-..-:/+o+/- # output of `uname -a` Darwin macbook-neo.local 25.3.0 Darwin Kernel Version 25.3.0: Wed Jan 28 20:56:11 PST 2026; root:xnu-12377.91.3~2/RELEASE_ARM64_T8140 arm64 ``` ### System topology  ## Benchmark results ### CPU - Geekbench 6: (3011 single / 6823 multi - https://browser.geekbench.com/v6/cpu/17009998) (TODO: Re-run after file sync complete) - TODO Gflops ([geerlingguy/top500-benchmark](https://github.com/geerlingguy/top500-benchmark) HPL result) ### Power - Idle power draw (at wall): 2.8 W (possibly lower, measurement a few hours after setup) - Maximum simulated power draw (`stress-ng --matrix 0`): 9.3 W (throttles to 7 W after about 10s) - During Geekbench multicore benchmark: 12.8 W - During `top500` HPL benchmark: TODO W ### Disk #### APPLE SSD AP0256Z | Benchmark | Result | | -------------------------- | ------ | | iozone 4K random read | TODO MB/s | | iozone 4K random write | TODO MB/s | | iozone 1M random read | TODO MB/s | | iozone 1M random write | TODO MB/s | | iozone 1M sequential read | TODO MB/s | | iozone 1M sequential write | TODO MB/s | ### Network `iperf3` results: - `iperf3 -c $SERVER_IP`: TODO Mbps - `iperf3 -c $SERVER_IP --reverse`: TODO Mbps - `iperf3 -c $SERVER_IP --bidir`: TODO Mbps up, TODO Mbps down (Be sure to test all interfaces, noting any that are non-functional.) ## GPU ### GravityMark GravityMark results: ``` 1. Download the latest version of GravityMark: https://gravitymark.tellusim.com 2. Run `chmod +x [downloaded_filename].run` 3. Run `sudo ./[downloaded_filename].run` and press `y` to accept the terms. 4. Open the link it prints, and run the Benchmark defaults, changing to 720p resolution and 50,000 asteroids. ``` Note: These benchmarks require an active display on the device. Not all devices may be able to run `glmark2-es2`, so in that case, make a note and move on! ### AI / LLM Inference Basic AI inference results with llama.cpp build: f2ab047f2 (8300) #### tinyllama-1.1b-1t-openorca.Q4_K_M.gguf | model | size | params | backend | threads | test | t/s | | ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: | | llama 1B Q4_K - Medium | 636.18 MiB | 1.10 B | MTL,BLAS | 2 | pp512 | 600.04 ± 18.50 | | llama 1B Q4_K - Medium | 636.18 MiB | 1.10 B | MTL,BLAS | 2 | pp4096 | 415.56 ± 4.62 | | llama 1B Q4_K - Medium | 636.18 MiB | 1.10 B | MTL,BLAS | 2 | tg128 | 61.27 ± 1.42 | | llama 1B Q4_K - Medium | 636.18 MiB | 1.10 B | MTL,BLAS | 2 | pp4096+tg128 | 279.91 ± 2.25 | #### Llama-3.2-3B-Instruct-Q4_K_M.gguf | model | size | params | backend | threads | test | t/s | | ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: | | llama 3B Q4_K - Medium | 1.87 GiB | 3.21 B | MTL,BLAS | 2 | pp512 | 202.86 ± 6.81 | | llama 3B Q4_K - Medium | 1.87 GiB | 3.21 B | MTL,BLAS | 2 | pp4096 | 157.81 ± 3.71 | | llama 3B Q4_K - Medium | 1.87 GiB | 3.21 B | MTL,BLAS | 2 | tg128 | 19.98 ± 0.15 | | llama 3B Q4_K - Medium | 1.87 GiB | 3.21 B | MTL,BLAS | 2 | pp4096+tg128 | 117.09 ± 0.81 | ## Memory `tinymembench` results: <details> <summary>Click to expand memory benchmark result</summary> ``` tinymembench v0.4.10 (simple benchmark for memory throughput and latency) ========================================================================== == Memory bandwidth tests == == == == Note 1: 1MB = 1000000 bytes == == Note 2: Results for 'copy' tests show how many bytes can be == == copied per second (adding together read and writen == == bytes would have provided twice higher numbers) == == Note 3: 2-pass copy means that we are using a small temporary buffer == == to first fetch data into it, and only then write it to the == == destination (source -> L1 cache, L1 cache -> destination) == == Note 4: If sample standard deviation exceeds 0.1%, it is shown in == == brackets == ========================================================================== C copy backwards : 19026.2 MB/s (3.3%) C copy backwards (32 byte blocks) : 18087.6 MB/s (1.8%) C copy backwards (64 byte blocks) : 18563.5 MB/s (1.1%) C copy : 19013.1 MB/s (1.3%) C copy prefetched (32 bytes step) : 19298.5 MB/s (2.1%) C copy prefetched (64 bytes step) : 18302.2 MB/s (1.0%) C 2-pass copy : 16694.4 MB/s (2.0%) C 2-pass copy prefetched (32 bytes step) : 16622.7 MB/s (2.2%) C 2-pass copy prefetched (64 bytes step) : 16862.7 MB/s (2.3%) C fill : 31609.9 MB/s (3.1%) C fill (shuffle within 16 byte blocks) : 31731.3 MB/s (2.4%) C fill (shuffle within 32 byte blocks) : 32944.7 MB/s (3.7%) C fill (shuffle within 64 byte blocks) : 32223.6 MB/s (2.7%) NEON 64x2 COPY : 30157.7 MB/s (2.2%) NEON 64x2x4 COPY : 29959.1 MB/s (2.1%) NEON 64x1x4_x2 COPY : 29691.1 MB/s (1.5%) NEON 64x2 COPY prefetch x2 : 24306.7 MB/s (3.1%) NEON 64x2x4 COPY prefetch x1 : 26457.9 MB/s (3.5%) NEON 64x2 COPY prefetch x1 : 25596.7 MB/s (3.5%) NEON 64x2x4 COPY prefetch x1 : 26708.1 MB/s (4.3%) --- standard memcpy : 30569.4 MB/s (7.2%) standard memset : 36902.3 MB/s (4.3%) --- NEON LDP/STP copy : 30937.6 MB/s (2.2%) NEON LDP/STP copy pldl2strm (32 bytes step) : 29902.0 MB/s (3.0%) NEON LDP/STP copy pldl2strm (64 bytes step) : 27884.2 MB/s (8.6%) NEON LDP/STP copy pldl1keep (32 bytes step) : 30740.1 MB/s (15.5%) NEON LDP/STP copy pldl1keep (64 bytes step) : 30469.7 MB/s (2.5%) NEON LD1/ST1 copy : 29726.6 MB/s (2.4%) NEON STP fill : 40044.9 MB/s (8.4%) NEON STNP fill : 51650.4 MB/s (12.7%) ARM LDP/STP copy : 29774.8 MB/s (4.7%) ARM STP fill : 38568.0 MB/s (6.5%) ARM STNP fill : 34132.8 MB/s (2.6%) ========================================================================== == Memory latency test == == == == Average time is measured for random memory accesses in the buffers == == of different sizes. The larger is the buffer, the more significant == == are relative contributions of TLB, L1/L2 cache misses and SDRAM == == accesses. For extremely large buffer sizes we are expecting to see == == page table walk with several requests to SDRAM for almost every == == memory access (though 64MiB is not nearly large enough to experience == == this effect to its fullest). == == == == Note 1: All the numbers are representing extra time, which needs to == == be added to L1 cache latency. The cycle timings for L1 cache == == latency can be usually found in the processor documentation. == == Note 2: Dual random read means that we are simultaneously performing == == two independent memory accesses at a time. In the case if == == the memory subsystem can't handle multiple outstanding == == requests, dual random read has the same timings as two == == single reads performed one after another. == ========================================================================== block size : single random read / dual random read 1024 : 0.0 ns / 0.0 ns 2048 : 0.0 ns / 0.0 ns 4096 : 0.0 ns / 0.0 ns 8192 : 0.0 ns / 0.0 ns 16384 : 0.0 ns / 0.0 ns 32768 : 0.0 ns / 0.0 ns 65536 : 0.0 ns / 0.0 ns 131072 : 0.0 ns / 0.0 ns 262144 : 2.3 ns / 3.5 ns 524288 : 3.4 ns / 4.5 ns 1048576 : 3.9 ns / 4.7 ns 2097152 : 4.2 ns / 4.9 ns 4194304 : 5.7 ns / 6.6 ns 8388608 : 6.6 ns / 7.4 ns 16777216 : 9.4 ns / 11.9 ns 33554432 : 51.9 ns / 102.0 ns 67108864 : 110.4 ns / 137.3 ns ``` </details> ### Core to Core Memory Latency TODO: If this is a new CPU/SoC, run `c2clat` to generate a core to core memory access latency graph, and paste it here. For instructions, see: https://gist.github.com/geerlingguy/842974c0e49c201c28f4be54a05cc89c ## `sbc-bench` results Run sbc-bench and paste a link to the results here: ``` wget https://raw.githubusercontent.com/ThomasKaiser/sbc-bench/master/sbc-bench.sh sudo /bin/bash ./sbc-bench.sh -r ``` ## Phoronix Test Suite Results from [pi-general-benchmark.sh](https://gist.github.com/geerlingguy/570e13f4f81a40a5395688667b1f79af): - pts/encode-mp3: TODO sec - pts/x264 1080p: TODO fps - pts/x264 4K: TODO fps - pts/phpbench: TODO - pts/build-linux-kernel (defconfig): TODO sec