You'd be amazed at how slow even the best dedicated computers were! The early Sargon programs were estimated to run at 15 nps on the 6502 2MHz Apple ][ and Commodore. Level 2 meant a 2 ply search! (perhaps these were only tallying full-width nodes, not capture search nodes)Don wrote:I know some of you have some of the old constellations and other machines - can you tell me how many nodes per second these programs/machines are reporting?
You can see current dedicated state-of-the-art for dedicated Morsch programs, which report the nps in their display. The VIP/Sapphire/Diamond line of Novag units report their nodes-per-second out their serial interface. The TASC R30 figure is reported on the display screen and estimated in their manual (though that seems low to me, given its strength). The Fidelity Excel series (club, Mach II/III/IV) would report the figure on its display.
NPS Year Model
5000 1998 Sapphire II (H8-32)
4500 2006 Citrine (H8S-26)
4000 1997 Atlanta (SH-20)
3000 1994 TASC R30 (ARM-30)
3000 1996 Milano Pro / Master chess (SH-20)
3000 1993 GK-2100 / Cosmos (H8-10)
2000 1988 Fidelity Excel 68000 12 MHz
1400 1992 GK-2000 / Miami (H8-10)
15 1980 Sargon II (6502 2 MHz)
25 MHz embedded processors were sufficient to attain master level play, the maximum strength sustainable in the mass market (apparently). And that was with 1990's level search techniques, mostly without memory for hash tables, and when the null-move heuristic was cutting edge!