The Prospects for 128 Bit Processors ( John Mashey SGI 1995)

A 1995 future prediction meets modern commenters who are not impressed

TLDR: In 1995, John Mashey guessed that moving beyond 64-bit computers would be driven by the need for vastly more memory, likely decades later. Commenters now say the future zigged instead of zagged: computing power exploded, memory pressure lagged, and some are still salty we never got the flashy 128-bit upgrades they wanted.

A dusty 1995 post from Silicon Graphics legend John Mashey has resurfaced, and the real spectacle is not just the prediction — it’s the comment section showdown. Back then, Mashey tried to calmly game out when computers might need to jump from 64-bit to 128-bit, basically asking: when do machines need to handle absurdly huge amounts of memory? His rough answer pointed way out into the future, around 2043, based on how fast memory sizes were growing. In plain English: he thought the pressure to go bigger would come from people wanting computers to remember more stuff.

But today’s commenters are side-eyeing that whole premise. One big reaction, from A1kmm, is basically: nice theory, wrong bottleneck. Their hot take is that the world didn’t run into a memory wall nearly as fast as expected because we poured progress into raw computing power instead. Translation for non-nerds: chips got faster and busier, not just bigger in what they could store. That sparked the classic retro-tech drama of “smart forecast for its time” versus “this aged like milk.”

Meanwhile, Aardwolf arrives with a different grievance entirely: not about giant memory, but about missing math muscle. Their complaint? We never got fancy 128-bit number-crunching as a standard feature, and they sound genuinely betrayed. It gives the thread a weirdly emotional twist: one camp says we never needed the giant leap, the other says fine, but we still wanted the cool upgrade. The vibe is half history lesson, half nerd heartbreak, with a side of “wow, the future turned out rude.”

Key Points

  • Mashey argues that forecasting a shift from 64-bit to 128-bit processors should be based on address-space demand driven by memory growth.
  • He states that DRAM density had been increasing by 4x every three years, implying roughly two additional address bits every three years for high-end systems.
  • The article says 36-bit physical addresses support 16 GB memories and notes that some shipped microprocessor systems already reached 16 GB.
  • Mashey explains that multi-memory system designs can consume physical address space faster because address bits may be used to select among memory banks or CPUs.
  • Using continued memory growth and a 4:1 virtual-to-physical memory ratio, he estimates a leading-edge 64-to-128-bit transition around 2043 based on physical memory pressure.

Hottest takes

"the core assumption that didn't pan out" — A1kmm
"we've produced a lot more compute" — A1kmm
"I'm disappointed 128-bit floating point ... never properly made it into CPU's" — Aardwolf
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.