• Please review our updated Terms and Rules here

32MB RAM in 486, cache problems?

per

Veteran Member
Joined
Jan 21, 2008
Messages
3,006
Location
Western Norway
Some time ago, I maxed my 486-25 systems RAM from 20MB to 32MB RAM. The first thing that happened was that EMM386.EXE hung during memory testing, unless I disabled cache for RAM above 16MB in the CMOS settings.

I've tried to troubleshoot a bit. Everything seems to work fine with the new RAM modules in the old 20MB configuration, and swapping the modules of the two banks in a 32MB config doesn't change much.

Has anyone experienced something like this before?

While just sticking to disabling cache for half the memory might work fine for now, I'd love to know why enabling cache for everything fails with EMM386.EXE.
 

mR_Slug

Veteran Member
Joined
Nov 28, 2006
Messages
941
Location
UK
Is this a very early motherboard? Max 32MB RAM would indicate that it might be. What's the chipset?

I'm not 100% sure if it is related but DMA access above 16MB has some issues. May be unrelated since it worked at 20MB. I'm just thinking along the lines of, early chipset = never tested with a whopping 32MB RAM.
 

per

Veteran Member
Joined
Jan 21, 2008
Messages
3,006
Location
Western Norway
Correction, it's HIMEM.SYS that fails. It's loaded without any parameters or switches. EMM386.EXE is loaded just afterwards with NOEMS.

The motherboard is circa 1993.

I may want to play around with defining custom non cacheable blocks above 16M as well, not just the entire region...

*update* just defining a custom noncacheable block from 16M to 32M did not work... Weird.
 

Attachments

  • DSC_0074.jpg
    DSC_0074.jpg
    94.8 KB · Views: 1
  • 15143069271327846886765316146016.jpg
    15143069271327846886765316146016.jpg
    93.9 KB · Views: 1
Last edited:

krebizfan

Veteran Member
Joined
May 23, 2009
Messages
5,476
Location
Connecticut
Database servers could use 32 MB on a 486.

Might need 32 MB if running Office 95 on Windows NT 3.51. 25 MHz would be slow for that but probably fast enough to prove it works.
 
Top