It would certainly be nice to have at least one set (10 pieces) of good 12 ns 32kx8 for reference.

I thought of making a SOJ to DIP adaptor so that I can use modern real 10ns parts and I went through with it. Only complication right now are the pins, normal 2.54mm pin headers have waaaaaay too thick pins and will damage normal chip sockets. 1.27mm pin headers seem ok but having to remove every other pin is kind of a PITA... but disregarding work it is possible to get 2€ per module cost.

Tiido wrote:

I thought of making a SOJ to DIP adaptor so that I can use modern real 10ns parts and I went through with it. Only complication […]
Show full quote

I thought of making a SOJ to DIP adaptor so that I can use modern real 10ns parts and I went through with it. Only complication right now are the pins, normal 2.54mm pin headers have waaaaaay too thick pins and will damage normal chip sockets. 1.27mm pin headers seem ok but having to remove every other pin is kind of a PITA... but disregarding work it is possible to get 2€ per module cost.

This is the way to go for longevity - very nice!

I bought some ready made converters but the company went down the toilette and I couldn't acquire more. De-/Soldering was a pain and it did not correctly work as I was too unskilled.
VikiWAT had SO28 To SDIP28 boards, so I tried buying all remaing ones but again - "sorry, this must be an error on the site, we don't have any".

Your self made option is the best. I even talked yesterday to a person about making such little converter boards. He said he etches all stuff by himself.

Is the process expensive?

Chinese will make the boards and when making hundreds the cost per PCB is under 10 cents even with shipping included, and this is 4 layer stuff too for best performance. SRAMs approach near 1usd when bought in bulk and so will all other bits. Only thing left is actual work and I cannot probably apply reflow methods to the entire process. Things also needing to stay in SDIP footprint also puts limits to what I could do. The main hurdle right now is the pins side, I have seen flat chips pins like electrodes in old chinese adaptor boards and in more recent stuff named as flip-pins but they're stupidly expensive for what they are and not an option. I will have to see if I can solder machine-pins via reflow methods without needing any extra manual soldering steps, if that is possible I will be capable of making nice amounts in not much time and thus at good costs too. (I am flat broke right now and won't get any income until next month so things cannot happen sooner on that front)

@Tiido: You need a Patreon account so people can give you money for such projects. 😀

I am a very lazy person and don't want the pressure 🤣. I do have donations link on my sound card page, maybe one day someone will use it. Next month I will not be broke anymore, I'll have substantial income and can then advance with various things in an effort to sustain myself on them.

In the mean time I tried out something and results are encouraging :
http://www.tmeeco.eu/BitShit/PCschit/HeaderExperiment0.jpg
http://www.tmeeco.eu/BitShit/PCschit/HeaderExperiment1.jpg
http://www.tmeeco.eu/BitShit/PCschit/HeaderExperiment2.jpg

With a little jig to hold the headers in place I can have them soldered to main board via reflow methods, so only thing that's needed is just right alignments. I will have to see if there are any headers made of glass fiber reinforced plastic so the melting would be lot less of a problem. But as is, I'm pretty happy, no shorts or anything even with excessive amount of solder paste and completely wrong pads 🤣. And test with surface mounted capacitors was successful too, they didn't suffer any damage in my cooking setup.

Looking forward to the adapters - just say how much money you want per piece & take your time.

I would also be interested in such adapters. I wonder if you can get 2-1-1-1 timings with 66 MHz FSB that way.

I would be happy to get 2-2-2-2. 3-1-1-1 would be amazing. Isn't there 8 ns SRAM we could use?

Even authentic 10 ns DIP SRAM would be a good start.

Real chips tend to be expensive and/or a gamble. All the faster SRAMs that are still made are 3.3V only at faster than 10ns speeds (and none are DIP anyway).

3.3 V is fine if you are using the PCB proposed. interrupt the 5 V pin with a VRM or 2 diodes in series.

It depends on how the cache is wired, it most designs it shares lines with VLB and other parts and any 5V signalling on the bus will be harmful for the 3.3V chips. Almost none of the 3.3V SRAMs have 5V tolerance which is required in such a case.

Make a 32 pin version too! Some of us have the ability to use 128k x 8 chips!

Grrr. I'm thinking more and more often of converting my 3D printer to a PCB-milling machine.

luckybob wrote:

Make a 32 pin version too! Some of us have the ability to use 128k x 8 chips!

All 128KByte chips I can buy from reliable sources at reasonable cost have too wide footprint to fit within normal DIP area unfortunately.

Tiido wrote:

luckybob wrote:

Make a 32 pin version too! Some of us have the ability to use 128k x 8 chips!

All 128KByte chips I can buy from reliable sources at reasonable cost have too wide footprint to fit within normal DIP area unfortunately.

Vertical PCB with pins on one side. Sorta like SIPP were.

That's not manufacturable (with means I have) and I have doubts about the speed stability with one side having very long paths. I'm only aiming at the fastest possible chips and performance tiers only.