VOGONS


First post, by SSTV2

User metadata
Rank Oldbie
Rank
Oldbie

Always wanted to have one of those interposer boards for motherboards that can only supply 5V Vcc to the CPU. Spent whole weekend assembling it from parts that can be bought basically at any electronics store and it eventually paid off. Though, I would not want to repeat same experience for the second time, nor I would recommend anyone to follow my example (unless you are into masochism). No PCB designing, etching, etc. were needed, just had to widen pin holes with a drill.

1.jpg
Filename
1.jpg
File size
1.49 MiB
Views
1935 views
File comment
Interposer in 4DPS MB.
File license
Fair use/fair dealing exception
2.jpg
Filename
2.jpg
File size
1.83 MiB
Views
1935 views
File comment
Bottom, Vcc wire and 8 bypass caps can be seen.
File license
Fair use/fair dealing exception
3.jpg
Filename
3.jpg
File size
1.02 MiB
Views
1935 views
File license
Fair use/fair dealing exception

Soldering properly that Vcc wire was pain in the ass, probably the hardest part of all.

I'm currently using external power source (5V) from molex, If everything will be OKAY and it'll pass stability test, then i'll add power from the extra row of socket 3. So far, 4DPS posts, no deeper tests were done.

Reply 2 of 25, by mwdmeyer

User metadata
Rank Oldbie
Rank
Oldbie

Very cool. Do you need to to convert the signals down to 3.3volts too or it is just the power the CPU runs off running at 3.3v? I don't know much about it!

Vogons Wiki - http://vogonswiki.com

Reply 3 of 25, by SSTV2

User metadata
Rank Oldbie
Rank
Oldbie

I'm probably going to design one in the future, most likely the same time when I'll run out the patience of using 8 megs of SIMM 30 RAM 😀

Only the Vcc needs to be separated, rest of the pins are 1:1 converted.

Reply 4 of 25, by SSTV2

User metadata
Rank Oldbie
Rank
Oldbie

Tested module in 4DPS and UM486V motherboards with external power source - no problems at all, AMD DX4-100 CPU runs perfectly fine. Removed external power source wires and added 4 extra pins for +5V Vcc from the outer pins row of socket 3, looks pretty decent now.

5.jpg
Filename
5.jpg
File size
1.32 MiB
Views
1788 views
File license
Fair use/fair dealing exception
6.jpg
Filename
6.jpg
File size
1.58 MiB
Views
1788 views
File license
Fair use/fair dealing exception

I could never boot UM486V MB into OS with that CPU running @ 120MHz, no matter the timing/wait states settings, but with the interposer, it can run doom @ 120MHz 😀

Reply 5 of 25, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Interesting. This is the first time I've seen a switching regulator on a socket 3 interposer board. I assume that the blue trimmer is to adjust the voltage? What voltage adjustment range do you have?

Plan your life wisely, you'll be dead before you know it.

Reply 6 of 25, by SSTV2

User metadata
Rank Oldbie
Rank
Oldbie

Yes, trimmer is used to adjust output voltage, with load, output voltage range is from 1.3V to 3.9V @ 5V input. Max output voltage could be increased with a better characteristics catch diode, though.

Reply 8 of 25, by SSTV2

User metadata
Rank Oldbie
Rank
Oldbie

Current SS34 catch diode has a 0.5V forward voltage drop, if it were replaced with a ultralow forward voltage drop diode, lets say 0.2-0.3V, then maybe Vout could be increased up to 4.2V. Still, output voltage heavily depends on load and current that passes LM2596 IC, according to datasheet, its saturation voltage is 1.5V at 3A, which would result in even lower max Vout at heavier loads. I've noticed that @ 120MHz, max output voltage drops to 3.8V.

Reply 10 of 25, by feipoa

User metadata
Rank l33t++
Rank
l33t++
SSTV2 wrote:

Current SS34 catch diode has a 0.5V forward voltage drop, if it were replaced with a ultralow forward voltage drop diode, lets say 0.2-0.3V, then maybe Vout could be increased up to 4.2V. Still, output voltage heavily depends on load and current that passes LM2596 IC, according to datasheet, its saturation voltage is 1.5V at 3A, which would result in even lower max Vout at heavier loads. I've noticed that @ 120MHz, max output voltage drops to 3.8V.

This sounds about right when I did Cyrix and AMD 5x86 overclocking experiments.

Plan your life wisely, you'll be dead before you know it.

Reply 12 of 25, by SSTV2

User metadata
Rank Oldbie
Rank
Oldbie

I think that I'm finally done with this thing. Soldered additionally a micro switch for an easy multiplier selection, no more switches/jumpers needed afaik?

7.jpg
Filename
7.jpg
File size
1.37 MiB
Views
1582 views
File license
Fair use/fair dealing exception

And yes, ability to increase volage helps to stabilize CPU during overclocking, not always neccessary, but most of the time it's needed.

Interesting thing about that motherboard is that it supports undocumented 54.3MHz FSB. Current Am DX4-100 is clocked @ 108.6Mhz and it surely beats 120MHz option in Doom, at least 3FPS avg. increase over 120MHz.

Reply 13 of 25, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Oh, that's very interesting. I've never seen a 54.3 MHz option. What's the model of the clock gen chip?

It would be fun to run the Am5x86-133 at 163 MHz on this board.

Plan your life wisely, you'll be dead before you know it.

Reply 14 of 25, by SSTV2

User metadata
Rank Oldbie
Rank
Oldbie

It uses a ICS2494M clock gen. I had previously ran a 5x86-133 @ ~160Mhz in GA486VM motherboard few years earlier, by setting FSB to 53.2MHz with an oscillator (borrowed it from SEGA Genesis PAL circuit). I remember well that game "Blood" ran on it as smooth as on pentium based PC.

Reply 16 of 25, by SSTV2

User metadata
Rank Oldbie
Rank
Oldbie
treeman wrote:

with stock cache chips? assuming 3.3V since it runs at 160 on 3.3V fine

Yup, stock cache and just 128kB, GA-486VM supplies only 5V to the CPU socket, had no interposer back then 🤣

Reply 17 of 25, by treeman

User metadata
Rank Oldbie
Rank
Oldbie

would be very interesting to see the benchmarks 53.2 vs 40 fsb. was blood in svga by any chance or the stock low res vga? how long were running it on 5v?

Reply 18 of 25, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Normally, reduced wait states above 40 MHz kill any benefit of a 50+ MHz FSB compared to 4x40 MHz.

Plan your life wisely, you'll be dead before you know it.

Reply 19 of 25, by treeman

User metadata
Rank Oldbie
Rank
Oldbie

It would be nice to see how the performance of amd at 50fsb would be vs your ibm cyrix at 66fsb, how do you guys change the multiplier to less then the stock 4x, its a physical pin, but where exactly?