'486 compiles like '286

'486 compiles like '286

Post by Eylon Gev » Fri, 10 Sep 1993 23:00:04



OK, I've posted for help before, and received many replies (Thank you), BUT,
unfortunately nothing helped.   Here's the problem:

I have a '486 running at 33 MHz with Linux on a 212 MB HD.
I have a 128K cache.

Now then, I had installed a total of 20 Megs of RAM.  The problem was that
Linux compiled like a '286 (took me 2.5 hours to compile the kernel).
I tried doing what all the letters I received said:  I first tried every
configuration of enabling and disabling external and internal caching,
shadowing, and assorted other CMOS options.  None helped. So, I pulled 4 MBs,
and was thus left with (what seems to be the upper limit) 16 megs of RAM.
Again...no help.  So once again I played with CMOS, but to no avail.

With 128K cache, 16MB RAM, and screwing around with CMOS, my machine STILL
compiles like a slug  (and X windows is incredibly slow as well, when I bother
running it).

  Does anyone have any idea what the problem is?  Any suggestions would be
helpful!

Thank you,
  Eylon Geva

 
 
 

'486 compiles like '286

Post by Christian Se » Sat, 11 Sep 1993 15:13:51



Quote:>I have a '486 running at 33 MHz with Linux on a 212 MB HD.
>I have a 128K cache.
>Now then, I had installed a total of 20 Megs of RAM.  The problem was that
>Linux compiled like a '286 (took me 2.5 hours to compile the kernel).

Hi,

I had the same problem. The only way for me to fix it was to replace the
4MB SIMM Modules with 1MB SIMM Modules. I now have only an 8MB System,
but it works with full speed.

I would be interested in an explanation why this happens - I have no idea.

This is on a 486 mother board from DTK.

regards Christian
--


 
 
 

'486 compiles like '286

Post by Larry Doolitt » Sun, 12 Sep 1993 00:13:07



(Christian Seyb) writes:


> >I have a '486 running at 33 MHz with Linux on a 212 MB HD.
> >I have a 128K cache.

            ^^^^^^^^^^

Quote:> >Now then, I had installed a total of 20 Megs of RAM.  The problem was that
> >Linux compiled like a '286 (took me 2.5 hours to compile the kernel).

> I had the same problem. The only way for me to fix it was to replace the
> 4MB SIMM Modules with 1MB SIMM Modules. I now have only an 8MB System,
> but it works with full speed.
> This is on a 486 mother board from DTK.

This seems to be a common occurrence, and I think it shows up in one
of the FAQ's.  Briefly, the problem is not having all the RAM cache'd.
The fix is either to reduce to 16M or less, or adding cache RAM.
Usually a full configuration of 256K cache will solve the problem.
You need to look carefully at the motherboard and/or its reference
book to make sure
 o The right RAMs are in the right sockets to get 256K cache data RAM
 o The right RAMs are in the right sockets for the cache tag RAM to
support that mode
 o The jumpers (if any) are set properly for this configuration
 o The BIOS settings (if any) are correct for this configuration

The details really do depend on the make *and model number* of
the motherboard (maybe even the revision level of the chipset!).
If you can't fix the problem because of lousy docs, post again
with *all* the details.  Someone else may have access to better
docs.  If you can't fix the problem because the motherboard/chipset
is buggy, post again with *all* the details (make, model, date code)
so that other Linux users can avoid buying it.


 
 
 

'486 compiles like '286

Post by Craig T Mans » Sun, 12 Sep 1993 05:53:09




> (Christian Seyb) writes:


>> >I have a '486 running at 33 MHz with Linux on a 212 MB HD.
>> >I have a 128K cache.
>             ^^^^^^^^^^
>> >Now then, I had installed a total of 20 Megs of RAM.  The problem was that
>> >Linux compiled like a '286 (took me 2.5 hours to compile the kernel).

>> I had the same problem. The only way for me to fix it was to replace the
>> 4MB SIMM Modules with 1MB SIMM Modules. I now have only an 8MB System,
>> but it works with full speed.
>> This is on a 486 mother board from DTK.

A friend of mine had this same problem with his mother board.  It has
nothing to do with the ammount of memory on your board.  It has to do
with a CMOS setting.  Somewhere in you CMOS you should have a casheable
ram option, you should set this to cashe the ram above the first 4 meg.
This solved my friends problem.  I have only seen this on the 486's.
I am running a 386DX/40 and my CMOS doesn't have this option.

Good luck

Albion

 
 
 

'486 compiles like '286

Post by Paul Gortmak » Mon, 13 Sep 1993 17:24:15


[complaints of slow systems deleted]

        This has been discussed many times. If you have heaps of SIMM RAM, and
are *not* running the full 256k cache, your machine may be a complete dog!!!

Spend the few $$$ required, and get the extra 128k cache -- you won't regret
it. If you don't like that, then pull out SIMM until it is cached effectively,
and mail me the extra ones. I'll make sure that they are put to good use.  ;-)

Paul.

--
Paul Gortmaker c/o Microelectronics and Materials Technology Centre.
Royal Melbourne Institute of Technology, GPO Box 2476V, Melbourne 3001,
Victoria, Australia. Ph  (61) 3 660 2601. FAX (61) 3 662 1921.

 
 
 

'486 compiles like '286

Post by M.BE.. » Mon, 13 Sep 1993 18:28:00




> writes:

> >I have a '486 running at 33 MHz with Linux on a 212 MB HD.
> >I have a 128K cache.

> >Now then, I had installed a total of 20 Megs of RAM.  The problem was that
> >Linux compiled like a '286 (took me 2.5 hours to compile the kernel).

> Hi,

> I had the same problem. The only way for me to fix it was to replace the
> 4MB SIMM Modules with 1MB SIMM Modules. I now have only an 8MB System,
> but it works with full speed.

> I would be interested in an explanation why this happens - I have no idea.

> This is on a 486 mother board from DTK.

I guess that could be caused by the cache...
Did you try expanding the cache to 256 kB?
Does the motherboard use write back caching?
If it does, what about the dirty tag ram (also called altered tag ram)?
Is it there?

If the cache causes the problem, you should see it all the time when running
programs using lots of data.

If that is the case, I could get you a program checking out the cache.
The program is called 'cachemess', it was published by a german computer
magazine, magazin fuer computer technik, and is running on DOS.
Perhaps you'll need TurboPascal, since the source code is a strange mixture
of Turbo Pascal and inline assembler statements and I don't know whether I'll
find a compiled version somewhere.

-Michael

 - Be a loyal plastic robot for a world that doesn't care - (Frank Zappa)

## CrossPoint v2.1 ##

 
 
 

'486 compiles like '286

Post by Christian Se » Tue, 14 Sep 1993 14:45:34



Quote:>[complaints of slow systems deleted]
>    This has been discussed many times. If you have heaps of SIMM RAM, and
>are *not* running the full 256k cache, your machine may be a complete dog!!!

Hi,

the board is a DTK PKM-0031Y 486 Cache Mainboard and does have 256k cache
installed (including the tag Rams). The cache is implemented as a write
through cache. I use a 486DX2-66 Cpu.

regards Christian
--

 
 
 

'486 compiles like '286

Post by Paul Gortmak » Wed, 15 Sep 1993 16:42:33



Quote:>A friend of mine had this same problem with his mother board.  It has
>nothing to do with the ammount of memory on your board.  It has to do
>with a CMOS setting.  Somewhere in you CMOS you should have a casheable
>ram option, you should set this to cashe the ram above the first 4 meg.
>This solved my friends problem.  I have only seen this on the 486's.
>I am running a 386DX/40 and my CMOS doesn't have this option.

...yes, this is true, *but* the above users did state that they did not
have the full 256k cache, and yet were trying to run large (ie >=16Mb)
RAM configurations. If you want to effectively cache this much SIMM, then
you really should have the 256k cache. Besides, it is not very expensive.

Paul.

--
Paul Gortmaker c/o Microelectronics and Materials Technology Centre.
Royal Melbourne Institute of Technology, GPO Box 2476V, Melbourne 3001,
Victoria, Australia. Ph  (61) 3 660 2601. FAX (61) 3 662 1921.

>Good luck
>Albion


 
 
 

'486 compiles like '286

Post by James Mo » Thu, 16 Sep 1993 11:39:38



>OK, I've posted for help before, and received many replies (Thank you), BUT,
>unfortunately nothing helped.   Here's the problem:
>I have a '486 running at 33 MHz with Linux on a 212 MB HD.
>I have a 128K cache.
>Now then, I had installed a total of 20 Megs of RAM.  The problem was that
>Linux compiled like a '286 (took me 2.5 hours to compile the kernel).
>I tried doing what all the letters I received said:  I first tried every
>configuration of enabling and disabling external and internal caching,
>shadowing, and assorted other CMOS options.  None helped. So, I pulled 4 MBs,
>and was thus left with (what seems to be the upper limit) 16 megs of RAM.
>Again...no help.  So once again I played with CMOS, but to no avail.
>With 128K cache, 16MB RAM, and screwing around with CMOS, my machine STILL
>compiles like a slug  (and X windows is incredibly slow as well, when I bother
>running it).
>  Does anyone have any idea what the problem is?  Any suggestions would be
>helpful!
>Thank you,
>  Eylon Geva


Have you messed with your disk system?  If you have got a REALLY slow
disk/disk controller it could do what you are talking about... if you
can try a cached controller, or if you've got VESA slots try a cached
VESA controller, otherwise this stumps me.
jim

--
James Moss        University of South Carolina |  "Linux is cool."

-----------------------------------------------|  "Yea!  He-he-he"
Blessed Be:  Pagan not by birth but by thought |        <Beavis>

 
 
 

'486 compiles like '286

Post by Paul Gortmak » Thu, 16 Sep 1993 13:51:54




>>[complaints of slow systems deleted]
>>        This has been discussed many times. If you have heaps of SIMM RAM, and
>>are *not* running the full 256k cache, your machine may be a complete dog!!!
>Hi,
>the board is a DTK PKM-0031Y 486 Cache Mainboard and does have 256k cache
>installed (including the tag Rams). The cache is implemented as a write
>through cache. I use a 486DX2-66 Cpu.

...well then, as was discussed in another article, make sure your BIOS is
set to allow the cache to work on *all* of your memory, and not just the lower
4, 8, or 16Mb.

Paul.

--
Paul Gortmaker c/o Microelectronics and Materials Technology Centre.
Royal Melbourne Institute of Technology, GPO Box 2476V, Melbourne 3001,
Victoria, Australia. Ph  (61) 3 660 2601. FAX (61) 3 662 1921.

>regards Christian
>--


 
 
 

1. 486 Doesn't see >16MB...no it's not compiled that way!

I posted a problem I've been having with a Compaq 486/25 with augmented
memory.  I went from 12MB (4 on main board, and 8MB SIMM on mem
expansion
board) to 36 (3 additional 8MB SIMMs).  The problem is, only 16MB are
seen by the system (kernel 2.0.30, then upgraded to 2.0.33).

People here suggested I fix my append line in lilo.conf which I
successfully
did, forcing recognition of 36MB.  However, as soon as free reports
about
16MB used, the system will hang.  Obviously, it still thinks there is no
more than 16MB.

Someone also suggested I check that I didn't have the 16MB_MAX setting
in my
kernel config file, which would indeed limit memory recognition.  That
wasn't
the case, although I tried both config settings.

At bootup, the Compaq does a memory check and reports 36864KB OK.  I did
have
to obtain a system config disk from Compaq's web site to tell BIOS what
the
memory amount is.  The memory config utility is a bit funky, asking how
much
memory at memory address 0, 1M and 16M (I report 15MB at 1M and 20MB at
16MB,
and this is what the system will set automatically).

However Linux just doesn't see it!  Anyone know what else I can do?  I
tested
the SIMM by replacing the original, Compaq 8MB SIMM with one of the new
ones
and it seemed fine.  Basically, my 12-->36MB has turned into a 12-->16MB
upgrade.  The Compaq system config utility does not say I need to reset
any
board switches, and Compaq says I should be able to put the SIMMs in any
of
the slots.

Might there be some incompatibility between Compaq BIOS response and
what
Linux expects?  Any pointers on where in the kernel code I should look
for
Linux memory mapping functions (mm, memory.c, etc.)?  I really doubt
this is
a Linux error.  Any ideas are welcome.

Thanks

SMH

2. Executibles for Solaris-x86 from Proworks-C, 3.0.1 (Sparc)

3. Newbie questions re: X windows & AfterStep

4. REQ: best conf for 486 + 286 + slip + dynamic ip

5. USB floppy

6. Linux on a 286-to-486 machine: how good?

7. Apache log files reporting tools

8. Trouble compiling g++ 2.8.1 under SCO 3.2v4.2 on '486

9. Question: Greater than 500MB Hd's in older 486's

10. Free UNIX for 286's?????

11. Can I network 286's to Linux server?

12. C++ WANTED FOR AIX/RT and SCO XENIX '286