Rugxulo, Ralph, Thank you both for your inputs! Here's the endgame :)
I already tried to test the RAM on this machine using both "AleGr MEMTEST 2.00" and "MEMTEST86+ 4.20" (from a boot floppy), but both fail miserably - AleGr MEMTEST imemdiately freezes, while MEMTEST86+ immediately reboots the PC. The BIOS test passes though, so even if RAM chips are bad, they are at least seemingly behaving correctly. Now, I tested two things: MS HIMEM v2.06 (from the XMS developer disk, with sources and all) - works fine, mem detects all XMS memory, Wolf3D is happy, no wild reboots, no freezes, everything works (yaay!). This is not a 'free' solution, so it can't be considered viable, but still - WORKS (meaning my 386 is not totally wasted). Also, MS HIMEM consumes 30K of conventional memory, which is not cool. HIMEMX v3.34 (as per Rugxulo's suggestion) - miracle, this one works perfectly - mem sees all the RAM, Wolf3D is happy (as well as all other applications I tested), no crash, no reboots, no freezes, no nothing. Solution found! I didn't test XMGR, since I don't really see the point, XMGR being a dead end anyway. About HIMEMX: 1. The 'TESTMEM:ON' switch is not working - with my limited ASM understanding, I think it checks for "TESTMEM:OFF" only, while the default is OFF anyway. 2. What makes the v3.34 "unofficial"? Maybe might it be a good time to advertise it as "official", so other people would avoid going through hoops like I had to? cheers, Mateusz On 25/08/2015 07:28, Rugxulo wrote: > Hi again, > > On Mon, Aug 24, 2015 at 11:19 PM, Mateusz Viste <mate...@viste.fr> wrote: >> >> Yes, I have considered the possibility that it might be a hardware >> problem. Wouldn't be that surprising given that's a 25 years old >> machine. I also resoldered lots of stuff fixing PCB traces and replacing >> a few passive components because the mainboard was in a very poor state >> when I got it, due to an evil battery leak. >> But the problem I have is happening only with XMS, and it does work >> almost perfectly with FDXMS (with the exception of Wolf3D not detecting >> XMS), so I'd rather think there is something non-standard about this >> 386SX, and that it does something differently than expected by most XMS >> managers. That's one of the first 32bit machines after all - plenty of >> space for non-standard solutions. Plus, the BIOS memory test passes all >> right, and other XMS-relying programs work fine, too. > > It could actually be a cpu bug. Didn't early 386s have bugs with some > instructions? > > "Early in production, Intel discovered a marginal circuit that could > cause a system to return incorrect results from 32-bit multiply > operations." (Wikipedia) > > Also see here ("80386 Bugs"): > > http://www.informit.com/articles/article.aspx?p=130978&seqNum=27 > >> I will follow your suggestion and test newer versions of HIMEMX > > IIRC, HIMEMX had a quirk/bug where a 386 [sic] needed to do "jmp $+2" > after enabling pmode. I don't think this was officially ever fixed > (only in unofficial later releases), so if you're still using 3.32, > it's probably still there. > > http://www.ibiblio.org/pub/micro/pc-stuff/freedos/files/dos/himem/himemx/old/himemx-3.33-unofficial-old-386-patch.zip > > Ah, here's the old discussion we had on BTTR (circa 2009): > > http://www.bttr-software.de/forum/board_entry.php?id=6445#p6445 > > (Rod P.'s unofficial 3.34 is mostly for workaround "modern" memory > holes, probably irrelevant for your 386, but IIRC it does include the > above fix too.) > >> and XMSGR, even if unofficial/closedsource, > > No, I meant the one still on iBiblio, circa this year (2015), not old > from 2012 (not sure why you're using that one). Though honestly I > don't know if anything significant changed since then anyways. > > http://freedos.sourceforge.net/software/?prog=xmgr > http://www.ibiblio.org/pub/micro/pc-stuff/freedos/files/dos/xms/xmgr/drivers-2015-03-05.zip > >> just for the sake of knowing whether it changes anything. Then indeed, >> my next plan was to install MSDOS and see if it works fine, but I >> thought it might be a good idea to ask around here first for some >> first-hand experience. > > I'm surprised I even remembered any of this. Maybe I'm not completely > useless after all (just close enough!). :-P > ------------------------------------------------------------------------------ _______________________________________________ Freedos-user mailing list Freedos-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freedos-user