• Hello MLAers! We've re-enabled auto-approval for accounts. If you are still waiting on account approval, please check this thread for more information.

Overclock LC475/P475/Q605 without soldering or spicy-o-clock

Very interesting, can anyone confirm the "343S0161" vs "343S1135" thing?
It doesn't seem to matter. Tried yet another board that has a 343S0161 and it works just as expected.
Not sure what's wrong with the very first board I used for testing this and why it wouldn't talk to the clock chip.

Meanwhile I added a few extra values to the code to reach up to 50MHz... 45MHz seems stable, anything faster crashes once you try to do anything after setting the speed. That's with 50ns RAM but only 70ns VRAM as I haven't anything faster at hand. SCSI clock selection resistor is set for bus clock/2 so the SCSI chip isn't overclocked.
 
Last edited:
@Bolle Do you have RAM soldered to the board?
This may limit RAM access to 70ns or 80ns.
I assume there are different ways to disable those (remove chips, remove resistor in RAS line, change code in ROM)
 
Do you have any issues with outputting any resolution aside from 640x480 when changing the clock? Maybe stupid question, gotta ask though.
 
Nevermind, works on a different logicboard. What's interesting, the non-working board has a "343S0161" installed for the clock chip, the working one has a "343S1135" - what's on your board @Mustermann?

All my boards have a MC88916 swapped in place, no problem to go up to 42MHz... we need faster settings now :p
This side is telling:
343S0161GazelleVideo clock generator, LC 475/575, Quadra 605
No information about 343S1135 on that side.

My LC475 has a 343S1135

Is leg 9 of U17 of non cooperative device lifted and pulled to 5V by a resistor to achieve 40Mhz?
 

Attachments

  • Gazelle.jpg
    Gazelle.jpg
    2.1 MB · Views: 22
For reproducibility: how are you validating that you're getting the frequency you think you're getting?
Checked with , Checked with Norton Sysinfo(Post #1), clockometer(Post#11) and Osziloscope at R96(Picture attached, frequency can be found at the bottom line of screen)
 

Attachments

  • Scope25Mhz.jpg
    Scope25Mhz.jpg
    1.3 MB · Views: 21
  • Scope33Mhz.jpg
    Scope33Mhz.jpg
    1.3 MB · Views: 17
Do you have any issues with outputting any resolution aside from 640x480 when changing the clock? Maybe stupid question, gotta ask though.
All my tests until now were operating with 1152*870*8bit @75Hz using 60ns VRAM

Right now I tested 640*480, 800*600, 1024*768, and 1152*870 @38Mhz with 80ns VRAM
and 1152*870 @38Mhz with 60ns VRAM

All of them work but switching resolution @38Mhz stopped with black screen.
When first switched back to 25Mhz resolution change was possible.

I am not sure if ROM is setting VRAM timing in FrameBufferConfiguration

@Bolle If you have a LC475 with 33Mhz resistor mod at hand, may you post the output of my program prior first speed change.

If 4 most significant bits are 0000 when starting with 33Mhz ROM does not set these bits as proposed in MEMCjr datasheet (0010).
 
I haven’t tried this yet @Mustermann but what would happen on a resistor swapped board where it is set to 33mhz on the board?
Starting with the good news: Will work.

But it is not only the "loading at startup" topic.

When starting, Gazelle will read resistors and chose 33Mhz dataset from internal table (if resistors are changed for 33Mhz).

When changing to to any value using the "setlong" function in my code, the 33Mhz entry will be replaced by the variable entry (e.g.38Mhz).
I am not sure if a reboot will reset this or if a power-down cycle is needed.
 
Starting with the good news: Will work.

But it is not only the "loading at startup" topic.

When starting, Gazelle will read resistors and chose 33Mhz dataset from internal table (if resistors are changed for 33Mhz).

When changing to to any value using the "setlong" function in my code, the 33Mhz entry will be replaced by the variable entry (e.g.38Mhz).
I am not sure if a reboot will reset this or if a power-down cycle is needed.

I'm going to try my 475 when I get home this afternoon, unless someone else wants to try. Hopefully my rasPi ethernet bridge still works.
 
Meanwhile I added a few extra values to the code to reach up to 50MHz...
Just to let you know, how to calculate parameters for "setlong" function.

case 29:
setregistermin(0);
setregistermax(3);
setlong(13,7,2);
setregistermin(1);
setregistermax(2);
break;

setlong expect 3 parameters corresponding to Ni, Mi, and Pi in the calculation from the datasheet (attached).

First choose desired Clock-speed (29Mhz)

Than add 0,1 as 29,05Mhz is nicer that 28,95Mhz (29,1Mhz)

Next divide by 2 as MC88920/16 will double the frequency later on(14,55Mhz)

Chose divider P from 1,2,4,8 in that way that 14,55*P is smaller than 100Mhz and bigger than 45Mhz
For 20-25Mhz CPU clock 8 will work, for 25-50Mhz 4 will work, above use 2,

Multiplier 1 is coded as 0 in last parameter
Multiplier 2 is coded as 1 in last parameter
Multiplier 4 is coded as 2 in last parameter
Multiplier 8 is coded as 3 in last parameter

So 14,55Mhz*4=58,2Mhz. Last parameter = 2

Next divide crystal frequency (31,3344Mhz) by the frequency calculated (58,2Mhz)
Results 0.538391753

Ask your spreadsheet calculation to format that as a two-digit fraction:
7/13

Enter last number (counter)(13) as first parameter Ni and first number (denominator)(7) as second parameter Mi.

I am not sure if it is really named counter and denominator.
 

Attachments

  • Limitations.png
    Limitations.png
    32.4 KB · Views: 12
  • Calculations.png
    Calculations.png
    66.2 KB · Views: 16
For Omega, the pixel clock is fully programmable but the CPU clock is set to one of 4 canned settings by resistors being present or not.
 
Omega does not seem to be as similar to the SC11412 as Gazelle.
Omega might have 28 pins (based on LCIII schematics).

Is there a datasheet for Omega?
 
Do you have any issues with outputting any resolution aside from 640x480 when changing the clock? Maybe stupid question, gotta ask though.
Side note:

This well-known page
report about screen resolution issues after changing in resistors for 33Mhz and lifting pin 9 of Gazelle and pull it to 5V using a resistor to achieve 40Mhz.

From this thread we know, that pin 9 is used as DATA line to transfer data from MEMCjr to Gazelle.
Gazelle is also generating Dot_Clk for Antelope Grafik Chip.

If lifted pin 9 and pull to 5V the System is no longer able to change Dot_Clk frequency.
Gazelle will stay at the clock frequency that is internally defined for power-up.

I do not wonder that there are resolution issues with other than minimum resolution when lifting pin 9.

Fortunately this is no longer needed.
 
Sweet find. Here's a quick and dirty extension to set 40mhz based on the code posted by @Mustermann. It loads as early as possible during the boot process to get running fast ASAP. Seems to work correctly: Clockometer reports 40mhz & corresponding speed improvement in Speedometer.

Also includes Symantec C project to compile. It's all hardcoded, if you want to try something else you'll need to pick different parameters and stick them in there.
 

Attachments

Back
Top