imagine amiga in current year

date: 2019-09-06T04:17:16Z

tags: technology tomo walkaway

i think amiga had the right idea with their daughter-board design. the A4000TX[1] was *hand routed* and featured Zorro 3 expansion slots. these have ~20 MB/s tx speed[2], which is slow by modern standards, but still. modern computers share this design, but i like this simpler example of what that architecture can look like. i wonder if we could design something similar. i'm super impressed and excited about the stuff that @niconiconi[3] is doing, and i think it is important for us to revisit and rethink computing in this kind of alternate context.

1: A4000TX

2: ~20 MB/s tx speed

3: @niconiconi

if we go back in time to when computers were nearly always assembled by hand, and then fast forward, what does that look like? i'm not convinced that the complexity of modern computing is needed. we certainly will need to change *how we write software*, but i think we can work around the very large performance penalty that will come from a properly simple hardware stack. cryptography is one place that is this hurts a lot, but with the right combination of daughter-boards maybe this isn't a problem.

slow computing is the retrofuture, which might give us a chance to control our computers in a meaningful way. it won't be as shiny, but that may be the price that we have to pay for autonomy and privacy. in leaving the grid and default society behind, we are trading one set of problems with another. the hope is that by changing the equation and walking in a different direction, we can find a better path to the future.

one example of modern tech that can be used to accelerate important crypto operations is the block eruptor asic[4]. when paired with SPHINCS+[5] you could theoretically verify messages at ~300MH/s. contrast that to the ~500-1000H/s you can expect on a high end x86 CPU. benchmarks are needed to see what the real-world impact is, but i'm expecting USB bandwidth will be a bottleneck.

4: block eruptor asic


the idea here is to use something like the vocore2[6] for main computation and offload specific workloads to other devices over USB or PCIe. when considering mobile applications having a collection of discrete modules is a big win for energy usage. in an extreme case, you could power down everything other than the main cpu and drop down to 74mA draw.

6: vocore2