[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference giadev::decstation

Title:DECstation PC Conference
Notice:register note 2, see notes 3 & 4
Moderator:TARKIN::LININD
Created:Tue Jan 10 1989
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:7470
Total number of notes:34994

7360.0. "OK, what is CLKMUL set to on a LPx and MTE?" by SNOFS1::RASMUSSEN () Fri Feb 21 1997 06:49

    I recently purchased a Am5x86-133 chip plus a voltage converter (5v to
    3.45). I tried this combination in a DEC LPx and DEC MTE system and was
    surprised (not really...frustrated was more like it ;-) to find that 
    Norton SI Version 8 and Norton Utilities for Windows 95 both indicated 
    the chip was delivering performance close to
    an Intel 486 DX4-100  (FYI in the case of the LPx I had to set the Write
    Wait States to Zero via DEBUG as the BIOS V1.05(unsupported) did not
    recognise the CPU hence did not set it for me). So I had another challenge
    ...(the last major one, prior to relevant BIOS 1.05.07, was the why does a 
    LPx with an Intel DX4-100 give Norton SI ratings of a DX2-66)...
    Why is it so? Why does it look like the Am5x86 is doing 3xExternal
    Clock rather than 4x? (I even thought some unscrupulous organsisation
    may have shaved a Am486-100 chip and relabled it until I saw a
    diagnostic saw the chip was a genuine Am5x86)
    
    I got a copy of the AMD BIOS Development Quide from AMD's Web site and
    noted: "If CLKMUL is High or Floating, the CPU uses 3x clock mode. If
    CLKMUL is Low, the ... Am5x86 CPU uses 4x clock mode...". 
    
    So let me guess....the LPx and MTE have CLKMUL floating or high? 
    
    Could some kind sole verify this? Also, I guess from looking at the
    LPv+ specs from (DECPC_ENG) it is also likely to be in the same boat!
    
    Thanks
    
    Wayne
    
    (PS yes you guessed it....If my theory is correct I am seriously
    considering some "minor modification" to remedy the situation. Also
    note that the price here in Australia for the Chip and Voltage
    Converter was about half the cost of a Kingston offering - I am quessing 
    that the Kingston offering has the CLKMUL signal forced Low).
T.RTitleUserPersonal
Name
DateLines
7360.1Correction and elaboration to .0SNOFS1::RASMUSSENFri Feb 21 1997 17:1313
    I am going to have to stop putting in notes late at night (the
    second "saw" should be showed, sole should be soul etc :-(. 
        
    CLKMUL is pin R17 on the Am5x86 chip. The "minor modification" I
    may make is to tie CLKMUL (R17) to Vss (P17). My current thoughts are
    to place a small wire bewteen R17 and P17 on the AMD CPU chip and
    remove Pin R17 from the Voltage Converter. I will remove R17 as I am
    concerned about what happens  if the motherboard has CLKMUL effectively tied
    to Vcc and I basically then tie that to Vss...even if the MTE & LPx have
    CLKMUL floating, one day I just might forget this little constraint and
    try it in another board that doesn't have CLKMUL floating.
         
    Wayne
7360.2The answer is...not connectedSNOFS1::RASMUSSENMon Feb 24 1997 16:526
    I have it from reliable sources that CLKMUL is not connected on the MTE
    and LPx. Not sure what the is for other boards but at least my
    questions are answered.
    
    Wayne
    
7360.3NPSS::JOHNSONMike J., Network Products SupportWed Feb 26 1997 08:2217
Wayne,

You didn't indicate who your vendors were for the AMD586 upgrades, but I have 
had a little experience "evaluating" upgrades from both Evergreen and 
Kingston on 466d2-MTE platforms.  It should be noted that the Evergreen 
upgrade has a JUMPER to select the clock multiplier.  Using BOTH vendors'
benchmarks on both upgrades, I got IDENTICAL results, indicating performance
somewhat better than a "native" Intel 60Mhz Pentium, or approximately half
of my Celebris 133.

As far as a comparision of the two upgrades, the real differences were price 
and cooling method.  The Evergreen was $10 cheaper and used a small heat sink. 
The Kingston was $10 more expensive and used a minature muffin fan.  Now I can
debate with myself the relative merits of more aggressive cooling versus the
reliability of moving parts (aka the fan).

/mj
7360.4RE:.3SNOFS1::RASMUSSENWed Feb 26 1997 21:0312
I am not getting a specific upgrade kit. I purchased a AMD 5x86 chip and
a Voltage Converter basically from two different sources. Put them together
(with a Fan) and did not get the performance increase expected. I then went
about finding out why. Putting it into $$$$s, getting an Evergreen or Kingston
upgrade here was going to cost me about $100(US) more than buying the parts.
Now I know about the impact of CLKMUL I will have to add 2 cents worth of
wire & solder to get the performance I expected.

Wayne

PS yes I understand I probably have spent >>$100(US) in time and effort
sorting it all out...but I view that as education/sport.
7360.5BBQ::WOODWARDC...but words can break my heartThu Feb 27 1997 19:126
    Wayne,
    
    I'll keep you in mind when I want to do the same - seems like you will
    need some more 'edjumacation' ;')
    
    Harry