m5@lynx.uucp (Mike McNally) (02/08/89)
Forgive me for what may be an ignorant question, but is there a version of the 386 that doesn't have problems with TLB misses while floating point stuff is going on? We have ``double sigma'' parts, and they exhibit the problem. For a while, I thought that the double sigmas fixed the bug; I guess I was wrong. -- Mike McNally Lynx Real-Time Systems uucp: {voder,athsys}!lynx!m5 phone: 408 370 2233 Where equal mind and contest equal, go.
smcroft@sactoh0.UUCP (Steve M. Croft) (02/09/89)
In article <5219@lynx.UUCP>, m5@lynx.uucp (Mike McNally) writes: > Forgive me for what may be an ignorant question, but is there a version > of the 386 that doesn't have problems with TLB misses while floating > point stuff is going on? I heard of this problem but thought that it was with the 386/287 combo. Could this also be happening with 386/387 equipment? -- ############################################################### # steve "whadda guy" croft # # ...!pacbell!sactoh0!smcroft || ...csusac!athena!crofts # ###############################################################
mslater@cup.portal.com (Michael Z Slater) (02/09/89)
>Forgive me for what may be an ignorant question, but is there a version >of the 386 that doesn't have problems with TLB misses while floating >point stuff is going on? We have ``double sigma'' parts, and they >exhibit the problem. For a while, I thought that the double sigmas >fixed the bug; I guess I was wrong. The double-sigma parts have a 32-bit multiply that works; they have all 21 of the errata, however, including the paging/fpu problem. The "D-step" parts supposedly fix all known problems; they are marked 80386DX. Curiously, the DX is what identifies the chip as a D-step part, but the D is not the step number. It is supposedly just a coincidence that the DX nomenclature was added at the same time that they switched to the D step. SX suffix is for Single-wide (16-bit) data bus, DX is for Double wide (32-bit) I'm told. Perhaps there will be a 486QX ?? (64-bit bus) Michael Slater, Microprocessor Report mslater@cup.portal.com