Computers All The Way Down

Here is an example of the ill-defined nature of the question: on your desk or in your pocket, how many computers do you currently have? How many computers are in your “computer”? Did you think just one? Let’s take a closer look—it’s computers all the way down. You might think you have just the one large CPU occupying pride of place on your motherboard, and perhaps the GPU too, but the computational power available goes far beyond just the CPU/​GPU, for a variety of reasons: mass-produced transistors and processor cores are so cheap now that it often makes sense to use a separate core for realtime or higher performance, for security guarantees, to avoid having to burden the main OS with a task, for compatibility with an older architecture or existing software package, because a DSP or core can be programmed faster than a more specialized ASIC can be created, or because it was the quickest possible solution to slap down a small CPU and they couldn’t be bothered to shave some pennies⁠⁠1… (Halvar Flake dubs this phenomenon “cheap complexity”⁠.) Whenever a peripheral or device is made, the Wheel of Reincarnation begins to turn. Further, many of these components can be used as computational elements even if they were not intended to be or try to hide that functionality. (For example, the Commodore 64’s floppy drive’s CPU running Commodore DOS was used as a source of spare compute power & for defeating copy-protection schemes, because it was as powerful; one can offload to the ‘co-processor’ tasks like computing fractals, 3D math routines for animating demos, computer chess…)

Down

Thus:

• A common AMD/​Intel x86 CPU has billions of transistors, devoted to many discrete components:

  • Each of the >2 CPU cores can run independently, shutting on or off as necessary, and has its own private caches L1–L3 (often bigger than desktop computers’ entire RAM a few decades ago⁠⁠2⁠, and likely physically bigger than their CPUs were to boot), and must be regarded as individuals. Further, the CPU as a whole is reprogrammable through microcode, such as to work around errors in the chip design, so what those CPUs compute changes.

  • CPUs sport increasingly opaque features, such as the management engines/​secure enclaves, like the Intel Management Engine (with a JVM for programmability⁠; Ruan2014 & SGX), or AMD’s Platform Security Processor (PSP) or ⁠Android’s TEEs or Titan chips⁠; these hardware modules typically are full computers in their own right, running independently of the host and able to tamper with it (and have contributed to terminological confusion, as the formerly simple protection ring levels of 0–3 have had to be augmented with “ring −1”, “ring −2”⁠, and “ring −3”). Intel’s ME runs a proprietary unauditable fork of MINIX (an OS better known for its role in the creation of Linux), whose security implications concerned Google enough to launch a project to remove MINIX from its CPUs & its Titan chip to cryptographically verify firmware on boot.

    • any floating-point unit may be Turing-complete through encoding into floating-point operations in the spirit of FRACTRAN⁠⁠3
    • the MMU can be programmed into a page-fault weird machine driven by a CPU stub (see above)

• DSP units, custom silicon: ASICs for video formats like h.264 probably are not Turing-complete (despite their support for complicated deltas and compression techniques which might allow something like Wang tiles), but for example Apple’s A9 mobile system-on-a-chip goes far beyond simply a dual-core ARM CPU and GPU as like Intel/​AMD desktop CPUs, it includes the secure enclave (a ⁠physically separate dedicated CPU core), but it also includes an image co-processor, a motion/​voice-recognition coprocessor (partially to support Siri), and apparently a few other cores. These ASICs are sometimes there to support AI tasks, and presumably specialize in matrix multiplications for neural networks; as recurrent neural networks are Turing-complete… Other companies have rushed to expand their system-on-chips as well, like Motorola or Qualcomm

  • Mark Ermolov notes that a given system on a chip (SoC) may have not just multiple CPUs, but easily 5 different CPU architectures all represented:

    It’s amazing how many heterogeneous CPU cores were integrated in Intel Silvermont’s Moorefield SoC (ANN): x86, ARC⁠, LMT, 8051⁠, Audio DSP, each running own firmware and supporting JTAG interface.

  • even weirder special-purpose chips like the Macbook Touch Bar’s chips which run little WatchOS apps & fingerprint authentication on a Apple T1⁠/​T2 SoC (using an ARM CPU) independent of the laptop OS, which can, of course, be hacked to run Linux⁠.

    Apple’s “Always-on Processor” in its smartphones doesn’t just enable location tracking of the phones even when ‘turned off’, but implements wakeup for Siri as well.

• motherboard BIOS & management engines (on top of the CPU equivalents), typically with network access

  These management or debugging chips may be ‘accidentally’ left enabled on shipping devices, like the Via C3 CPUs’s embedded ARM CPUs.

• GPUs have >100s of simple GPU cores, which can run neural networks well (which are highly expressive or Turing-complete), or do general-purpose computation (albeit slower than the CPU)⁠⁠4

• smartphones: in addition to all the other units mentioned, there is an independent baseband processor running a proprietary realtime OS for handling radio communications with the cellular towers/​GPS/​other things, or possibly more than one virtualized using something like L4⁠. Baseband processors have been found with backdoors, in addition to all their vulnerabilities.

  Given ARM CPUs are used in most of these embedded applications, it’s no surprise ARM likes to boast that “a modern smartphone will contain somewhere between 8 and 14 ARM processors, one of which will be the application processor (running Android or iOS or whatever), while another will be the processor for the baseband stack.”.

• SIM cards for smartphones are much more than simple memory cards recording your subscription information, as they are smart cards which can independently run Java Card applications (apparently NFC chips may also be like this as well), somewhat like the JVM in the Intel Management Engine. Naturally, SIM cards can be hacked too and used for surveillance etc.

• the IO controllers for tape drives, hard drives, flash drives, or SSD drives etc typically all have ARM processors to run the on-disk firmware for tasks like hiding bad sectors from the operating system or accelerating disk encryption or doing wear leveling⁠; these can be hacked⁠.

• network chips do independent processing for DMA⁠. (This sort of independence, in conjunction with the motherboard & CPU management engines, is why features like Wake-on-LAN for netboot work.)

• USB/​Thunderbolt cables/​devices, or motherboard-attached devices: an embedded processor on device is needed for negotiation of data/​power protocols at the least for cables/​batteries/​chargers⁠⁠5⁠, and may be even more heavy duty with multiple additional specialized processors themselves like WiFi adapters or keyboards or mice or SD cards⁠.

  In theory, most of these are separate and are at least prevented from directly subverting the host via DMA by in-between IOMMU units, but the devil is in the details…

• monitor-embedded CPU for control of displaying whatever the GPU sends it (part of a tradition going back to smart teletypes)