|
|
|
|
|
|
by StudyAnimal
2460 days ago
|
|
|
Mainframes are general purpose machines optimized for high (IO) throughput of bulk business operations rather than number crunching. The answer to your question is probably marketing and avoiding confusion. You wouldn’t call a supercomputer a mainframe even though it could probably do the work. |
|
|
S/370-based supercomputers like the Hitachi S-3800 are both IBM-compatible mainframes AND supercomputers at the same time. The Fujitsu FACOM VP and VP 2000 supercomputer series were also based on IBM-compatible mainframe architecture. In a similar category is the NEC SX-1 (1983) and SX-2 (1985) supercomputers (but not the SX-3 and later), but rather than being S/370-compatible, they are based on an ACOS-4 mainframe architecture, which is a Japanese derivative of GE/Honeywell/Bull GECOS mainframes.
Arguably, 3090+VF belongs to the same category, the mainframe-supercomputer overlap. Technically it is quite similar to Hitachi/Fujitsu/NEC's approaches, in being a mainframe-based supercomputing solution. IBM introduced VF to try to sell mainframes to the supercomputer market. It wasn't very successful, whereas their POWER-based supercomputers were much more successful. On the other hand, Japanese manufacturers like Hitachi that tried the same strategy had significantly more success with it than IBM did.
(I understand from a marketing/PR perspective how vendors want to forget their old unsuccessful products when introducing new ones. Nobody wants to say "this is our 2nd attempt to address this market, the first one was not very successful, hopefully this new one will succeed where the prior one failed" at a new product launch. You just pretend the prior attempt never happened.)
(Going back to the early 1960s, arguably IBM 7030 Stretch and IBM 7950 Harvest are both IBM 7000 series mainframes and supercomputers.)