RetroFlops
Menu

Personal computer

Apple Mac mini (M1, 2020)

16 000 000 000 transistors on 5 nm, with processor and graphics cores sharing 8 GiB or 16 GiB of unified memory. The computer is rated at 150 W. Its performance and efficiency cores retain separate clock records.

Manufacturer
Apple
Released
November 2020
Region
Worldwide
Record status
approved

About this machine

This record continues a line the catalog has drawn since the Apollo Guidance Computer. The oldest machines here are described almost entirely by time: a memory cycle in microseconds, an addition in milliseconds. Home computers are described by frequency. Consoles add floating-point rates beside the frequency, then stop giving the frequency. The M1 has separate clocks for its performance and efficiency cores, rather than one CPU-wide rate.

Apple gives a transistor count, a process node, a core count split into two kinds of core, and a graphics throughput figure. Only the first two are properly comparable with anything else in the catalog. A process node can be displayed, but node names stopped being physical dimensions long before this one, so the metric forbids ratios between them.

The memory is unified, as the PlayStation 4's is, and fitted at manufacture. Two configurations were sold, 8 GiB and 16 GiB, and neither could be changed afterwards.

Configuration

8 GB unified memory

Component Role Count Clock Capacity
Apple M1 Four performance cores and four efficiency cores with separate clock records. Main CPU 1 3.2 GHz Nominal clock 2064 MHz Nominal clock not applicable Nominal clock —
Apple M1 GPU GPU 1 1278 MHz Nominal clock —
Apple M1 unified memory Unified memory 1 — 8 GiB shared

16 GB unified memory

Component Role Count Clock Capacity
Apple M1 Four performance cores and four efficiency cores with separate clock records. Main CPU 1 3.2 GHz Nominal clock 2064 MHz Nominal clock not applicable Nominal clock —
Apple M1 GPU GPU 1 1278 MHz Nominal clock —
Apple M1 unified memory Unified memory 1 — 16 GiB shared

Figures for the whole machine

Apple Mac mini (M1, 2020): system figures

provisional A figure marked provisional is recorded but not accepted: its evidence is unsettled. It may rely on one source, an unconfirmed announcement, or a digit read from a damaged scan. The reason appears under the figure. The catalog shows the figure but never uses it in a multiplier. How records are judged.

Metric Value Evidence Basis Method Source
Clock frequency cpu · efficiency cores 2064 MHz Confirmed Measured Nominal clock Published independently The four Icestorm efficiency cores, which AnandTech calls Thunder cores. The other half of the same chip, running at the same time. Four of the 8 cores hold this frequency and four hold 3.2 GHz, so neither figure is "the M1's clock" and the catalog does not pretend one of them is. That is also why no multiplier is offered against either: a machine that answers this question twice has no single figure to divide.
Clock frequency cpu · performance cores 3.2 GHz Confirmed Measured Nominal clock Published independently The four Firestorm performance cores, in single-threaded work and, given thermal headroom, in all-core loads too. Apple publishes no frequency for this machine, and this figure is not Apple's: two independent parties measured the running hardware and reported the same number. It is recorded beside the vendor's silence rather than instead of it, and it sits in the same comparability group as every other nominal clock, the Commodore 64's included, because the group is set by method and not by who established the number.
Clock frequency gpu provisional 1278 MHz Reported Measured Nominal clock Published independently Notebookcheck reports this clock in its M1 GPU specification table.
Geekbench score, multi-core whole-system 8915 Confirmed Measured Benchmark chart average Published independently geekbench 6 , multi-core Mac mini (Late 2020), averaged over every Geekbench 6 result uploaded for the model. The same dated snapshot as the single-core figure and a different quantity: the variant is part of the benchmark identity, so this can never be divided by the single-core score to produce a scaling factor.
Geekbench score, single-core whole-system 2198 Confirmed Measured Benchmark chart average Published independently geekbench 6 , single-core Mac mini (Late 2020), averaged over every Geekbench 6 result uploaded for the model. A dated snapshot of a moving number, not a score this machine got. The latest archived Mac chart available to the 26 July curation run is from 25 July; its hash pins this value. It may be divided only by a figure from the same variant and curation run.
Memory capacity whole-system 16 GiB Confirmed Vendor-rated Design capacity Stated by the vendor Unified memory in the larger configuration, shared by the processor cores, the graphics cores and the Neural Engine. Apple writes "16GB", meaning 16 GiB, and offers it only as a build-time option. The same caveat as the base configuration applies: it cannot be changed afterwards.
Memory capacity whole-system 8 GiB Confirmed Vendor-rated Design capacity Stated by the vendor Unified memory in the base configuration, shared by the processor cores, the graphics cores and the Neural Engine. Apple writes "8GB", meaning 8 GiB. Fitted in the package and not expandable, so this is the machine's memory for its whole life rather than a starting point.
Memory bandwidth whole-system provisional 68.25 GB/s Reported Theoretical Theoretical peak Published independently Eight 16 bit channels of LPDDR4X-4266-class memory on a 128 bit bus, shared by the processor cores, the graphics cores and the Neural Engine. Provisional on the sourcing, not on the number. One analysis states it; the other source states the channel count and the memory class it was computed from and stops there. It is also a ceiling the bus allows rather than a rate anything reached: the same page measures a single core copying at 60 to 62 GB/s.
Rated power consumption whole-system 150 W Confirmed Vendor-rated Rated power Stated by the vendor "Maximum continuous power" from the technical specification, on a 100 V–240 V supply. Identical for both memory configurations. A ceiling the machine is permitted to draw, not a figure it reaches: no workload is named, and this is a computer that spends most of its life far below it. Comparable in kind with the PlayStation 2's 79 W rating and with nothing that was actually measured, which is why the metric forbids ratios.
1 documented absence
Apple Mac mini (M1, 2020): system figures: documented absences
Metric Value Evidence Basis Method Source
Clock frequency cpu not applicable Not stated Vendor-rated Nominal clock Stated by the vendor The M1 has separately recorded performance-core and efficiency-core clocks; one CPU clock would erase that distinction.

Figures for its components

Apple M1
Metric Value Evidence Basis Method Source
Transistor count cpu 16 000 000 000 transistors Confirmed Vendor-rated Die specification Stated by the vendor The whole M1 die: processor cores, graphics cores, Neural Engine and memory controller. Apple writes "16 billion", which is two significant digits. The figure covers the entire die and is recorded at cpu scope only because the metric admits no scope that means "whole chip"; it is not a count of the processor cores' transistors, and no vendor breakdown exists.
Process node cpu 5 nm Confirmed Vendor-rated Die specification Stated by the vendor Fabrication process named by Apple for the M1 die. A process name expressed as a length, not a measured dimension: nothing on this die is five nanometers across. The metric forbids ratios for that reason, so this figure can be displayed and never divided.
Apple M1 GPU

provisional A figure marked provisional is recorded but not accepted: its evidence is unsettled. It may rely on one source, an unconfirmed announcement, or a digit read from a damaged scan. The reason appears under the figure. The catalog shows the figure but never uses it in a multiplier. How records are judged.

Metric Value Evidence Basis Method Source
Peak FP32 rate gpu provisional 2.6 TFLOP/s Confirmed Vendor-rated Theoretical peak Stated by the vendor "2.6 teraflops of throughput" for the eight graphics cores, stated on the day the chip was announced. Recorded as single precision, which Apple does not state: the same editorial reading, and the same reason for provisional status, as the Dreamcast's and the PlayStations' floating-point figures. With no clock and no operations-per-cycle published, it cannot be reconciled with the eight graphics cores it belongs to.
Geekbench Compute score, metal gpu 31 387 Confirmed Measured Benchmark chart average Published independently geekbench-compute 6 , metal A dated capture of the benchmark publisher’s rolling average for this graphics processor, read from the 18 July 2026 Metal chart. It may be compared only with another Geekbench 6 Metal figure from this curation round, never with an OpenCL figure, a later version or a processor score. The chart also lists this processor under “Apple M1 GPU”, at 32 442; the row naming the product as its maker does was taken, and the disagreement is recorded in the conflict ledger.
Geekbench Compute score, opencl gpu 17 592 Confirmed Measured Benchmark chart average Published independently geekbench-compute 7 , opencl A dated capture of the benchmark publisher’s rolling average for this graphics processor, read from the 25 July 2026 OpenCL chart. Version 7 calibrates a compute result against a baseline of 100,000, so it may be compared only with another Geekbench 7 OpenCL figure from this curation round, never with a Metal figure, an earlier version or a processor score. The chart also lists this processor under “Apple M1 (G13G B1)”, at 13 136; the row naming the product as its maker does was taken, and the disagreement is recorded in the conflict ledger.

Editorial notes

The release date is the date of the cited announcement. The Neural Engine's 11 trillion operations per second is not recorded: the catalog has no metric for it, the operations are not floating-point, and inventing one for a single record would leave it comparable with nothing. Storage configurations are not recorded either. The catalog had no storage metric when this record was written, and it gained one with the iPhone. Filling this machine's in is a change for whoever next revisits it, not a side effect of another record. No derived claim compares this machine's graphics throughput with the PlayStation 4's because both figures are provisional and they were obtained by different methods. The ratio formula would refuse it twice over.

Sources