Yeah, I was working on a Windows NT image processing program in 01998. My desktop had two processors, and the machines in the dev lab and in the field had four. Mine were Pentium Pro. Some friends of mine had accounts on pioneer.unm.edu, a Sequent Symmetry with I think 8 80386 processors on a single memory bus, up until it was lost or stolen in about 01994. Then I got an account on the MHPCC SP2 UNM had the contract to administer, with I think 384 RS/6000 POWER CPUs, but without shared memory (not even NUMA).
My first Unix account was on a Sequent. I tried to put together a dual processor box in 1995 but they were having supply issues and it didn't work out.
I think it is important to differentiate multi core from multiprocessor, but not so much for the context of that article.
Oh, maybe he was talking about having multiple CPUs on the same piece of silicon? That seems like a more plausible explanation for the dates (which are still somewhat wrong, as you point out), but its connection to software optimization strategies seems tenuous at best. The software can't really find out if the other CPU is on a separate die unless the hardware chooses to tell it.
He’s careful to say cores but multiprocessor also means multiple cores, just not on the same silicon.
And Sun had 32 thread machines not long after his timeline. IIRC they were so goddamned expensive as to be academic, but my customers only ran Sun hardware in their data centers (cellular carriers). Mid tier hardware mostly. Our target was 4 cores per server around 2008.
