486 Sandwich Socket Blaster: Redesigned Interposer for 3V CPUs on 5V motherboards

[@Vanessaira-Retro]

When that Duke Nukem remix kicks in.

[@GadgetUK164]

Great video! The instabilities will be the length and routing of the traces with the 2 tier approach! Brilliant work! I think your original version is better, even if its hard to assemble!

[@scrap_computing]

Yeah, that's what I am suspecting too. I am also not sure about the connection between the male/female headers, perhaps it's better just to solder them, which could also reduce the height of the adapter.

[@darthtripedacus1]

Always a pleasure to see an upload from you good sir.

[@rasz]

Those pins just sticking in the air are insane :o Surprised it worked this well at all. You could try grounded copper foil between rows of pins to minimize cross pin coupling.

[@g4z-kb7ct]

I left this comment on Epictronics latest video about this adapter but I'll paste it here for you too with some additions. When talking to epictronics I offered to help with the PCB re-design, but since you can do that easily I may as well talk directly to the horse rather than the jockey ;-) There is a very easy way to make the PSU better. Use the same pin header points that the mini buck converter board uses but make your own little power board that is larger and uses all the space inside the PGA socket. The board will sit up on the header pins so you can put parts on both sides of the board. The larger board can have more heatsinking (larger full copper internal power planes) and also have more and larger tantalum capacitors and some inductors which really are necessary for a proper SMPS PSU to work correctly. When mounting the little psu board put a thermal pad under the psu board so it touches the interposer board, push down then solder it down. Those thermal pads can be bought cheaply and cut to any size. A 4mm pad would work well. It really should also be a 4-layer board with vcc and gnd on the inside using large copper planes. The boards are not that big so the cost increase for a 4-layer board won't be too much but it will make a massive difference to power stability, especially with a new larger buck board and more tantalums and inductors.

[@Epictronics1]

Great experiment! Could you redesign the original socket blaster with a more powerful onboard voltage regulator? Unfortunately, even the larger buck converter isn't powerful enough for higher clock speeds. At least that's what I think is the cause of the crashes. Keep up the good work! This is a great project

[@llamallarry]

The late 486 CPUs sometimes required more than 3.3v, like 3.45v. perhaps they also require a smaller ripple than the older CPU models?

[@llamallarry]

After a bit of research, the most current these 486 cpus could draw when extremely overclocked is 3.1 ampere, AFAICT. And a theoretical maximum power draw of 10.8 watt at 3.6v in extreme overclocking. In other words the buck converter should support 3A and 10w to allow for heavy overclocking, like 166 Mhz.

[@llamallarry]

And, for reference, the Intel 486 dx4-100 drew at most 7.5w at 3.3 volts, according to its specs. That means a maximum current draw of 2.3A. The buck converter must support this load as a minimum.

[@n.shiina8798]

@@llamallarry the AP7176 he use wasn't enough apparently. it was spec'd with 1.7W maximum power dissipation. at 2.3A, the voltage drop should be 560mV at most to satisfy the spec

[@Epictronics1]

@@llamallarry I noticed when I ran a 5x86 with the large buck that the voltage dropped. They are spec'd for 3A on eBay and Ali but I'm not sure that can be trusted. I'll see if I can find the time to test one

[@rs-qq3os]

Wow, really good work , thanks for the video.

[@Choralone422]

Great work! I still wish I had my 486 PC from way back when. I had a 3.3V adapter on it in order to use an AMD 5x86 133 MHz CPU as it was a fairly early 5V only board. I think I found the socket adapter via Computer Shopper or PC Magazine. Would be handy to still have it to use it for reverse engineering purposes. I too ended up creating a little tower of sorts in the socket as I also had a heatsink and fan on top of the CPU.

[@waytostoned]

Considering interposers now for 486 cpus are insane priced now , I get why. Me im thankful I got that trinityworks one years ago for under 30 bux.

[@RetroErik]

Wow, really nice. I will try to make one soon. Maybe also a ThrottleBlaster

[@scrap_computing]

Please keep in mind that it's quite experimental. I am not sure why it's not stable on my older motherboard. Perhaps it's better to just solder the two parts together instead of connecting them with male/female headers. This adapter also includes a STPCLK pin which is meant to be used with the Throttle Blaster, though I haven't tested it yet.

[@RetroErik]

@@scrap_computing Do you believe that distance between the two board is the instability cause? How would you solder it directly? Are there other pins/connectors that are shorter? Will you be trying this out?

[@scrap_computing]

@@RetroErik I am not sure what is the cause. It seems to be board-specific. I will try soldering the two parts together in case this fixes the issue.

[@cowrevenge]

Space Quest!!!!!, I know that sound anywhere!

[@laudennn]

A+

[@Megatog615]

In the next version, do you think it would be possible to have a built-in 5v header for a fan? This would make for a very tidy cooling solution with say a 40mm 5v Noctua fan so you don't need to use a molex connector from the PSU.

[@scrap_computing]

That's a great idea. I will try to fit a header for a fan, not sure if there will be enough room for a proper molex connector though.

[@Megatog615]

@scrap_computing no need for a molex connector, just a 3-pin(it could even be a 2-pin) fan connector. make sure you leave room for the off chance that someone might want to use a 4-pin fan.

[@DoomWarriorX]

I wonder why you use 1206 SMD Caps between the pins. If you would use 1608 metric / 0603 the caps should fit without angeling.

[@scrap_computing]

I prefer using 1206 parts because they are easier to handle, but yeah a smaller part would fit better for sure.

[@TheCj71984]

Need one of these for socket 7 cpus

[@scrap_computing]

That would be great, but the pins are more dense on the Socket 7 so we can't use off-the-shelf headers for the pins as we do here.

[@SudosFTW]

This sort of thing needs done with the Socket 423 to 478 adapters as well. When they do show up they go for hundreds of dollars, and there is no open-source project to replace it with. pin-compatible sockets for Socket M/P still exist and can be used instead, and the only difference would be figuring out how to wire up such an adapter so that any socket 423 board can get a socket 478 CPU installed. Ideally, an SL7EY 2.8GHz Northwood.

[@MrGencyExit64]

Don't take this the wrong way... is that your real voice? The whimsical music in the background makes me wonder :)

[@DemianTowers]

tronco has pensado en sintetizar la voz en lugar de forzarla para evitar su captura? Supongo que no supondrá un problema de salud a largo plazo por no tratarse de algo prolongado, pero puede que te resulte incluso satisfactorio terminar la ofuscación con máquinas anteriores a los dos miles.

[@joaoc_PT]

PC tower of power

[@Pickle136]

at 3.29 did you intend to skip that pin?

[@scrap_computing]

Good catch! No, I fixed it later but I just went over all the pins ones more and I did find some dodgy looking ones :)

[@PROSTO4Tabal]

Your voice is so funny lol your nd2 life you should be a singer

[@toseltreps1101]

it sounds like an ugly gay cartoon character

[@g4z-kb7ct]

I think your shorts are a bit too tight, loosen them off a bit hehe! ;-)

[@RetroErik]

Can you explain how you reduced the Multiplier to 2.5X. kzbin.info/www/bejne/mnyzhWZsYsuhrdU

[@scrap_computing]

I meant to say 2x, sorry about that. I just move the CLKMUL jumper from "MOBO" position to the other which reduces the multiplier from 3x to 2x on the DX4.