4 ms·

I'd be willing to believe in factory overclocks, one magic speed bin that wasn't common enough to market to the public, but was common enough to yield a few her

by jeffbee 2mo ago

I'd be willing to believe in factory overclocks, one magic speed bin that wasn't common enough to market to the public, but was common enough to yield a few here or there. I worked at a place that was able to secure factory overclocks from Intel, 100MHz faster than what everyone else was buying.

hedgehog 2mo ago
Factory overclock, and in those days Motorola could probably have given them a batch of mid-development DSP chips (sort of proto 56002) that could clock much higher and still sit on the unmodified motherboard, max out the scratchpad memory, etc. Then the add-on cards. Anything that was potentially involved in nuclear response decisions probably could get a lot of budget. It would be interesting to inspect the hardware that still exists to see if there are any signs of what exactly actually happened.
- rasz 2mo ago
  > mid-development DSP chips (sort of proto 56002) "The physical pinout of the DSP56001 is different from the DSP56002" https://digisolutions.de/downloads/Differences56001_56002.pdf https://digisolutions.de/downloads/Differences56001_56002.pd... Buying new unproven workstation just to develop expansion cards to make it fast makes no sense when you can just buy off teh shelf fast SUN or SGI.
  - hedgehog 2mo ago
    I don't know how the contemporaneous IRIS would compare for the target workloads but not necessarily faster than a stock NeXT, let alone one with whatever upgrades the customer wanted to stick in there. The 68000 and by that point the 56000 families both already likely had been proven for similar imaging workloads, with validated code available. At the same time the app dev story with interface builder etc on NeXT was pretty shiny. That compute was a direct component of the latency between something happening on the ground and photos reaching decision makers. In a post-glasnost post-Chernobyl environment with everything shifting around, very valuable to be able to upgrade that workflow rapidly. Someone who was in that group at Motorola in that era would know if there were any special-purpose 56000s that got sold.

by jeffbee 2mo ago

hedgehog 2mo ago

Factory overclock, and in those days Motorola could probably have given them a batch of mid-development DSP chips (sort of proto 56002) that could clock much higher and still sit on the unmodified motherboard, max out the scratchpad memory, etc. Then the add-on cards. Anything that was potentially involved in nuclear response decisions probably could get a lot of budget. It would be interesting to inspect the hardware that still exists to see if there are any signs of what exactly actually happened.

rasz 2mo ago
> mid-development DSP chips (sort of proto 56002) "The physical pinout of the DSP56001 is different from the DSP56002" https://digisolutions.de/downloads/Differences56001_56002.pdf https://digisolutions.de/downloads/Differences56001_56002.pd... Buying new unproven workstation just to develop expansion cards to make it fast makes no sense when you can just buy off teh shelf fast SUN or SGI.
- hedgehog 2mo ago
  I don't know how the contemporaneous IRIS would compare for the target workloads but not necessarily faster than a stock NeXT, let alone one with whatever upgrades the customer wanted to stick in there. The 68000 and by that point the 56000 families both already likely had been proven for similar imaging workloads, with validated code available. At the same time the app dev story with interface builder etc on NeXT was pretty shiny. That compute was a direct component of the latency between something happening on the ground and photos reaching decision makers. In a post-glasnost post-Chernobyl environment with everything shifting around, very valuable to be able to upgrade that workflow rapidly. Someone who was in that group at Motorola in that era would know if there were any special-purpose 56000s that got sold.