Athlon problems fixed tweaking BIOS memory settings

Athlon problems fixed tweaking BIOS memory settings

[Anton Altaparmakov]

Hi all,

I had a few random crashes on my Athlon EPOX 8KTA3 mobo Athlon 1.33GHz
(BIOS from 06/14/2001) and Alan Cox suggested to run memtest as the oops
apparently looked like memory corruption. Sure enough memtest showed
problems above 256MiB but swapping the chips around still showed problems
above 256MiB which meant it couldn't be the chips at fault (failing tests
were 5 and 8 in memtest86 supplied with SuSE 7.2). 

Playing with the memory settings I found that using the "DRAM
Timing by SPD" option was unstable but setting it to manual and using the
settings: 133MHz, CL2, Bank interleave disabled as well as Pre charge to
active 3T, active to precharge 6T, active to cmd 3T. Makes it fully
stable even with Page-Mode enabled.

Changing the precharge/act/cmd value to faster caused increasing amounts
of errors in memtest to appear and enabling Timing by SPD did the same.
Memory now running at 23.1 MiB/sec according to memtest86 compared to 26
MiB/sec at previous, faster but unstable settings.

It seems there is some kind of problem with the "Timing by SPD" option. 
Also no matter what the settings are the system is always error free below
256MiB, which means that might explain why some people are having problems
and some are not... Until I upgraded the 256 to 768MiB I didn't have any
problems.

Just a datapoint...

Best regards,

	Anton
-- 
Anton Altaparmakov <aia21 at cam.ac.uk> (replace at with @)
Linux NTFS maintainer / WWW: http://linux-ntfs.sf.net/
ICQ: 8561279 / WWW: http://www-stu.christs.cam.ac.uk/~aia21/

Re: Athlon problems fixed tweaking BIOS memory settings

[Janne Pänkälä]

> above 256MiB which meant it couldn't be the chips at fault (failing tests
> were 5 and 8 in memtest86 supplied with SuSE 7.2). 

I have Tyan S1590S (super 7) motherboard and I had this problem with
memtest too.

It was after I changed 768M of memory to it. (I had AT power supply)

when I ran memtest it went fine until it cme to test 5 after that at
random point errors started to flow in (and didn't stop).

Later it was found out that this was due to the fact that the motherboard
converted the 5V from AT PSU to 3.3v needed by dimm trough a tyristor (or
like) and when it was running with a lot of memory and AT supply it heated
up like hell.

when I changed it to ATX supply and changed one dip on the mobo the
tyristor was as cool as ever and memtest ran trough just fine.

not sure if this is the problem with some of you know. The fact that it is
especially test 5 (32-bit moves?) makes me wonder.

Althou all athlon mobos are ATX to begin with *shrug*.

-- 
Janne  justtrowinghisspooninthesouptomakethisevenmoreconfusing.
echo peufiuhu@tt.lac.nk | tr acefhiklnptu utpnlkihfeca

Re: Athlon problems fixed tweaking BIOS memory settings

[David Balazic]

Anton Altaparmakov (aia21@cus.cam.ac.uk) wrote :

> Hi all, 
> 
> I had a few random crashes on my Athlon EPOX 8KTA3 mobo Athlon 1.33GHz 
> (BIOS from 06/14/2001) and Alan Cox suggested to run memtest as the oops 
> apparently looked like memory corruption. Sure enough memtest showed 
> problems above 256MiB but swapping the chips around still showed problems 
> above 256MiB which meant it couldn't be the chips at fault (failing tests 
> were 5 and 8 in memtest86 supplied with SuSE 7.2). 

To eliminate any doubt about the chips, I suggest testing each of them separately.

> 
> Playing with the memory settings I found that using the "DRAM 
> Timing by SPD" option was unstable but setting it to manual and using the 
> settings: 133MHz, CL2, Bank interleave disabled as well as Pre charge to 
> active 3T, active to precharge 6T, active to cmd 3T. Makes it fully 
> stable even with Page-Mode enabled. 
> 
> Changing the precharge/act/cmd value to faster caused increasing amounts 
> of errors in memtest to appear and enabling Timing by SPD did the same. 
> Memory now running at 23.1 MiB/sec according to memtest86 compared to 26 
> MiB/sec at previous, faster but unstable settings. 
> 
> It seems there is some kind of problem with the "Timing by SPD" option. 

SPD EEPROMs on many ( especially no-name ) RAM modules are programmed incorrectly
( according to C'T magazine - http://www.heise.de/ct/ ) so "Timing by SPD" does
not work at all, or sets a wrong timing.

> Also no matter what the settings are the system is always error free below 
> 256MiB, which means that might explain why some people are having problems 
> and some are not... Until I upgraded the 256 to 768MiB I didn't have any 
> problems. 
> 
> Just a datapoint... 
> 
> Best regards, 
> 
>         Anton 

-- 
David Balazic
--------------
"Be excellent to each other." - Bill S. Preston, Esq., & "Ted" Theodore Logan
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -