Some suggestions, if I may, on your soldering technique. Buy a cheap, heat resistant silicone mat to work on. You may be able to buy a white one if you need it for contrast for video capture. Use kapton tape to affix the pcb on the mat: It will free up your left hand. You wont burn your fingers and you won't need to use the hotair gun's nozzle to push components off the pcb. Other than that, great work! I enjoyed every second of it!!
Thanks for your suggestions! Absolutely appreciate it. You're absolutely right! Will try it out in upcoming projects. Happy to hear that you enjoyed the video!
SIMMs with toggle switches, never would have thought I would see that. Though in retrospect, I am somewhat surprised that one of the SIMM manufactures back in the day didn't think to do make what you created with a tiny EDO/FPM DIP switch. Would have have cut out a sku. Regardless, nice job.
Now that you mentioned it, I wonder that too. Would be interesting to hear a professional opinion about this hack and why or why not it may be a bad idea.
I don't think any of those module manufacturers really knew what they were making; they were probably just following a reference design without an in depth understanding of how system chipsets actually speak to the memory. It took an exceptionally sharp and well versed person with a wide spectrum of applied knowledge to point out that you can make this sort of memory hack.
I was looking forward to this and you did not disappoint. Glad you were able to recycle that dead memory module, it made a nice series that was fun to watch and prevented ewasting all those still good chips.
Awesome work! I was wondering what the point of the switch was, but it never occurred to me that anyone would try using them in a much later board that supports EDO. Now it makes total sense, and even a jumper would be sensible option. Honestly, I should've thought of it. I recycled some of the memory from the family 386 using one of those 30 to 72 pin SIMM adapters when we replaced the machine with a 486, and again when I upgraded to an AMD 586. I bought a much newer motherboard with PCI slots for the upgrade, which I'm sure would've supported EDO memory. I even upgraded it to EDO memory later and noticed a slight speed boost in the process, but at the time I put it down to just having more memory. So yeah, I should've thought of that kind of application. Not sure if there would be many boards that would support 30 pin EDO SIMMs without the 72 pin adapter, but it's still an interesting option.
Hello creator of the awesome SIMM PCB! Your work definitely made my modded PCB a lot easier! Especially the dimensions, chip placements etc. I don't know if you're planning on changing your GitHub, merge my changes, or just leave it. But I forked your repository initially. And yes, the switch is only there to answer my curiosity for future content. Thanks again for your great design! It definitely made this project possible.
@@bitsundbolts You're welcome, and I'm glad you took the project and did something cool and interesting with it! I think it makes sense to keep them as separate projects, since they each have their own different advantages. Kind-of like how different linux distributions might start out as a fork and become a completely new OS.
@@bitsundbolts great work! Now it's time to make switchable 72-pin SIMMs. And... what about parity modules? Honestly, I still don't know why some modules use an extra DRAM chip for parity capability, and some use a sort of larger (?) DRAM chip, that differ from others on module. Of course, we don't need fake parity (logical parity) chips support because it seems cheaper today to add just an extra memory chip than to increase BOM by one position more as its use only for systems that stronlgy run with parity modules. As it was previously said by Artyom (yes, that is how that name actually spells), 386 motherboards with EDO support should exist, at least one chipset has support, ALI M1217
@@nzoomed EDO memory was more of a Pentium-era thing, although I have read threads where people mention using EDO in their 486s back in the day, and when they do mention what kind of SIMMs they were using it was 72 pin, not 30 pin. I found the datasheet for the SIS 85C496/497 chipset, which is a late 486-era PCI chipset that does list EDO compatibility, but I doubt anyone ever made a board that used that chipset and had 30 pin SIMM slots. In other words, I doubt anyone ever made 30 pin SIMMs with EDO memory back in the day, since they would've been more expensive and incompatible with most motherboards that they fit on (unless they used This One Weird Trick covered in this video).
Oh well, I could have gone with a jumper too, but I used those on the Voodoo memory expansion board. So, time for something new and I learned about on/on switches or "SPDT" switches 😃. Happy to hear that you enjoy the videos!
My pleasure! Just if someone is interested, it took me around 3 days in my spare time to make the PCB layout. On 10th May, I placed the order with PCBWay and the box with the green PCBs arrived on 17th May.
I ordered some of these. The price was reasonable for the small board size, but when I priced PCBWay against JLBPCB for another project, the latter won out for the same specs, it was almost 3x cheaper. When I'm done with the project, I'll make a video to share it.
Why did you add a switch instead of just bridging the pad and removing the connection to the original pad? I'm not aware of any EDO 30-pin SIMMs and assume that the standard only supports FPM, so leaving the option for EDO seems peculiar to me. Thanks for your work, and I may acquire some of these, but I'll hard-wire them for FPM mode 😉
You asked about desoldering thoughts, personally for small boards like that, I think you'd be way better off using a hop-plate style heater for pulling IC's off boards that have no components on one side. MHP30 would be the perfect size for pulling IC's one by one and other small solder reflow needs.
Absolut spitze! Klasse Arbeit. Ich habe zwei 486er Boards mit SIMM Sockets, die derzeit mit 8MB bestückt sind. Dort werden jetzt Deine Platinen einziehen, sobald ich auch so gut Chips verlöten kann. Kann Deine Platine auch 1MB Kapazität wenn dort Chips mit geringerer Kapazität aufgesetzt werden? Habe 4 *1MB Module, die als defekt erkannt werden und ich weiß nicht, ob es die Chips oder die alten Platinen sind.
Awesome project! I used your new board design to build a pair of 4mb SIMM modules for my Macintosh Color Classic. I was not able to get them to work and I believe its due to the Mac being picky about memory. I did build working FPM modules with the board design you based this project from. I just wanted to put out a YMMV if you are trying this EDO2FPM on old Macs :)
Sorry to hear that the memory doesn't work with the Macintosh. Maybe you can find someone with a 386/486 to see if the modules work there. I have a follow-up project with those modules and a SIMM adapter to 72-pin slot for a 486. Would be good to hear if others were successful to build their own working modules with those PCBs.
@@bitsundbolts I'll try and hunt down someone with an old PC test rig. Maybe I can send them to you and you can do another test/troubleshooting video 🤓
A followup to this subject: I've tried to do the same with my 4MB EDO modules with 42-pin 16-bit ICs. So, at first attempt I've just shorted /OE with the nearby /lCAS pin and got weird behavior with everything working right on a 486-DX100 machine except for any write operations with FDD. Each tie I've got corrupted data with visible pattern of correct and wrong bytes interleaving each other. Then I've made a test module with two standard 1n4148 diodes draining current from /OE to /uCAS and /lCAS and errors has stopped to appear. A have a theory that this has something to do with DMA, assumed to handle 8-bit operations differently from processor, so generally memory controller just activates /RAS, /lCAS and /uCAS sequently but in some situations the order or a set of signals is different, so with 16-bit modules this additional mod is required for correct operation. By the way it makes absolutely no difference between using a hardware floppy drive or Gotek emulator, the result is always the same: half of the data ends up absolutely random. What surprises me is that the diodes do not seem to affect performance but probably, using a Schottky or another type of fast diodes should be superior. The next goal is doing the same trick with rare 50ns chips.
In DOS 8 MiB is usually enough to run all DOS games for which a 386 is fast enough. However, operating systems such as Windows NT and OS/2 would benefit. They need at least 12 MiB, better 16 MiB to run properly. Windows 3.x would benefit from the more memory in that it could run multiple applications at the same time. The same also applies to Windows 95.
Why switch? Whenever you do switch it over, you always need soldering iron to replace the chips. Small pads that you could short with solder would make a bit more sense. ;)
I may want to try the 4x 30-pin to 72-pin converter and see if I can switch between EDO and FPM in the future. You don't have to use a switch at all as I mentioned in the video. The pads are close to each other to easily bridge them with solder.
Does this mean technically you could get 128MB on one of these boards? My manual says max 64MB, which would be 8x8MB sticks. Would a BIOS patch unlock the rest or is it a memory controller limitation?
It most likely is a limitation of the chipset and memory controller. But it could also be that at the time of release, there were no 16 MB SIMM modules in the market. It could be that boards support more memory than mentioned in the manual. Only testing would verify of it works or not.
@@bitsundbolts OK, would be an interesting test for sure. I also notice that some motherboards have issues with 3 chip SIMMs, so would be interesting to see how these behave. Would be interesting to see if its possible to push a motherboard so 128M, not that there is any point in DOS!
Interesting project. I was thinking of a board adapter so that you don't have to desolder the chips. But I don't know if that would be realistic because does EDO modules have a controller chip of some sort?
Haha, don't say more - you're spoiling my project list :) :) :) No, but seriously, I am thinking of a project to use one 72-pin socket and an adapter to split it to 4 30-pin modules. Drawback is that it may only work with 72-pin FPM modules, not with EDO because /OE pin is not exposed on connector pins. It should work with FPM modules though. It is a reverse 4x30-pin-to-72-pin adapter... Don't know if I want to do that though - routing 72 traces etc...
great video, thanks!! question is this resulting 30 pin module parity or non-parity? i have a tandy 2500 sx-33 that seem to require 4meg x9 modules (i think its parity?) let me know..
The ones I created are non-parity. But it should be possible to use one of the chips mentioned in the original BOM from UpLateGeek to make it a parity module. I may do this in the future - at least the real FPM modules. Yes, x9 means parity. If this is a requirement, then you need the third chip on the PCB which I did not add in this video.
@@bitsundbolts understood. but the chips on ULG BOM are FPM, and the others will be EDO behaving as FPMs... any conflict there? i guess it needs testing?
Yes, it needs testing. But I strongly believe that there shouldn't be a problem because I already mixed EDO and FPM on the 386 in one memory bank (4 slots = 1 bank = 32 bits). So, 8+8+8 bits were EDO-2-FPM modded and another 8 bits were FPM chips - this worked and is in the video at 09:39.
As much heat as you can without melting or disturbing nearby componentry is the answer. In my opinion, these RAM boards would be more heat, less time. If you were working on a chip near a plastic connector, less heat more time.
I'm afraid in this configuration with two memory chips it's either 1MB or 4MB. I guess a 2MB module is possible, but then it would take only half of the address space of the highest not, which is weird. I just have not seen it before.
The switch seems a little silly since it's easy to change a bridge like that. could also just use pinheaders and a jumper, but yea if you have extras of those switches then why not.
It should be possible if you find chips with Address line 11 active. The parity chip, however, is not connected to this line. Therefore, parity check for 16MB modules won't work on those PCBs. If I find chips that have the correct pinout, I may give it a try.
a 386 with 32mb of ram running windows 95. that's such an odd combination. I wanna see that. OSR2 with usb, can you get usb working? I wanna see a 386 with usb support. What about 98?
Absolutely! With 32 MB ram, it would have more space than some hard drives at that time! I haven't looked into this project yet, but it's definitely on my list.
PCBWay gives you a quote right on their website. If you go with 10 PCBs, it will cost 5 USD + shipping. You can use the link in the video description and get a 5 USD welcome bonus. Then you pay for shipping only.
I could try to turn an EDO module into a FPM module, but this would involve cutting one pin for each memory chip and then bridging the remains to the neighboring pin. Wouldn't it be easier to buy a real FPM module for your 486?
@@bitsundbolts I have quite a few modules. Unfortunately, I am having difficulty identifying them and if they work or not. I only have one motherboard and is currently limed to a 5V DX2 66 CPU's. The 3.3 v regulator is not present but the jumper pins, pads and holes for it are.
Maybe you can lookup the data sheets for the chips on the modules - if they're not too many. "Datasheets360 dot com" may be very helpful. At the moment I don't have a better solution than to try one by one. You may have noticed that some of your module have very different looking chips and sometimes only two for single sided and four chips for double sided modules. I could imagine that with a hack, things get more complicated and probably will increase your trouble with memory further.
Manufacturing takes about 2 days for green and 3-5 days in other colors. Shipping takes around 5 days with FedEx or DHL. I cannot find many ICs locally. Mainly checking on AliExpress and Alibaba.
I checked the pinout - looks like this chip should be compatible as well. It even supports FPM, so, no need for the "hack" in older systems that require FPM memory.
The video is already done, but I am almost on my way to the airport now - holidays for 2 weeks. Spoiler: There will be one video with the installation of Windows 95 and a second video repairing a GeForce2 MX - both ready to go, but need to spread them over the next two weeks.
Based on the manual, yes. However, SIMMs should theoretically go up to 16 MB per module. That would be 128 MB with 8 modules. But I doubt the chipsets supports those and the motherboard may not have all the traces in place.
Support of the chipset and traces on the motherboard. The chipset may simply not expose the 12th address pin/internally not connected. Or the motherboard does not route all address lines to the SIMM sockets. I am sure there may be other reasons, but those are the ones I can think of. Now, I don't know if this "Super Low Cost" (SLC) chipset has a physical limitation, or if it could be possible to run this board with 128MB - even if it's not mentioned in the manual. A pinout of the chipset may give some hints.
@@BarsMonster I had a small "Pride Expression" motherboard with an Am386DX processor on an ALi M1429/M1431 chipset with 8 SIMM30 slots and the manual for it contained tables for filling memory slots up to 128 MB. Its specifications also mentioned 128 MB. However, at that time I had only 8 MB of memory.
this is epic, this channel needs more attention. you need to do a write up and post on hack a day or other websites with this sort of content.
I think I'll write up something on vogons and will look for other sites too. Thanks for the suggestion and I hope the channel grows 😃
Some suggestions, if I may, on your soldering technique. Buy a cheap, heat resistant silicone mat to work on. You may be able to buy a white one if you need it for contrast for video capture. Use kapton tape to affix the pcb on the mat: It will free up your left hand. You wont burn your fingers and you won't need to use the hotair gun's nozzle to push components off the pcb. Other than that, great work! I enjoyed every second of it!!
Thanks for your suggestions! Absolutely appreciate it. You're absolutely right! Will try it out in upcoming projects.
Happy to hear that you enjoyed the video!
SIMMs with toggle switches, never would have thought I would see that. Though in retrospect, I am somewhat surprised that one of the SIMM manufactures back in the day didn't think to do make what you created with a tiny EDO/FPM DIP switch. Would have have cut out a sku.
Regardless, nice job.
Now that you mentioned it, I wonder that too. Would be interesting to hear a professional opinion about this hack and why or why not it may be a bad idea.
I don't think any of those module manufacturers really knew what they were making; they were probably just following a reference design without an in depth understanding of how system chipsets actually speak to the memory. It took an exceptionally sharp and well versed person with a wide spectrum of applied knowledge to point out that you can make this sort of memory hack.
One faulty memory boy, started this journey. Awesome project!
With a heat-plate I think you can take the chips faster and safer off.
Great video btw, fun stuff, very nostalgic.
Windows 95 on a 386? That’s exactly the content I am here for 😂
Great content, tweaking this old tech is very rewarding 👍 my first CPU was a Celeron 300A. Ohh the memories 😊
I was looking forward to this and you did not disappoint. Glad you were able to recycle that dead memory module, it made a nice series that was fun to watch and prevented ewasting all those still good chips.
Excellent work! I'm so glad the new PCBs worked perfectly!
I can't wait for this channel (and necroware's channel too) to blow up and get the viewership they deserve 👍 this ram project/mod was superb.
Thanks for mentioning me in the same sentence as Necroware 🙂
Awesome work! I was wondering what the point of the switch was, but it never occurred to me that anyone would try using them in a much later board that supports EDO. Now it makes total sense, and even a jumper would be sensible option.
Honestly, I should've thought of it. I recycled some of the memory from the family 386 using one of those 30 to 72 pin SIMM adapters when we replaced the machine with a 486, and again when I upgraded to an AMD 586. I bought a much newer motherboard with PCI slots for the upgrade, which I'm sure would've supported EDO memory. I even upgraded it to EDO memory later and noticed a slight speed boost in the process, but at the time I put it down to just having more memory.
So yeah, I should've thought of that kind of application. Not sure if there would be many boards that would support 30 pin EDO SIMMs without the 72 pin adapter, but it's still an interesting option.
Hello creator of the awesome SIMM PCB! Your work definitely made my modded PCB a lot easier! Especially the dimensions, chip placements etc. I don't know if you're planning on changing your GitHub, merge my changes, or just leave it. But I forked your repository initially.
And yes, the switch is only there to answer my curiosity for future content.
Thanks again for your great design! It definitely made this project possible.
@@bitsundbolts You're welcome, and I'm glad you took the project and did something cool and interesting with it!
I think it makes sense to keep them as separate projects, since they each have their own different advantages. Kind-of like how different linux distributions might start out as a fork and become a completely new OS.
@@bitsundbolts great work! Now it's time to make switchable 72-pin SIMMs. And... what about parity modules? Honestly, I still don't know why some modules use an extra DRAM chip for parity capability, and some use a sort of larger (?) DRAM chip, that differ from others on module. Of course, we don't need fake parity (logical parity) chips support because it seems cheaper today to add just an extra memory chip than to increase BOM by one position more as its use only for systems that stronlgy run with parity modules.
As it was previously said by Artyom (yes, that is how that name actually spells), 386 motherboards with EDO support should exist, at least one chipset has support, ALI M1217
Didn't know there were 30 pin EDO Simms to begin with!
@@nzoomed EDO memory was more of a Pentium-era thing, although I have read threads where people mention using EDO in their 486s back in the day, and when they do mention what kind of SIMMs they were using it was 72 pin, not 30 pin. I found the datasheet for the SIS 85C496/497 chipset, which is a late 486-era PCI chipset that does list EDO compatibility, but I doubt anyone ever made a board that used that chipset and had 30 pin SIMM slots.
In other words, I doubt anyone ever made 30 pin SIMMs with EDO memory back in the day, since they would've been more expensive and incompatible with most motherboards that they fit on (unless they used This One Weird Trick covered in this video).
you have no idea how much i enjoy your videos. so different and informative. never ever have i imagined i'll see ram with switches :)) cheers
Oh well, I could have gone with a jumper too, but I used those on the Voodoo memory expansion board. So, time for something new and I learned about on/on switches or "SPDT" switches 😃.
Happy to hear that you enjoy the videos!
That was fast. Thank you for modifying the PCB layout and putting it online.
My pleasure! Just if someone is interested, it took me around 3 days in my spare time to make the PCB layout. On 10th May, I placed the order with PCBWay and the box with the green PCBs arrived on 17th May.
I ordered some of these. The price was reasonable for the small board size, but when I priced PCBWay against JLBPCB for another project, the latter won out for the same specs, it was almost 3x cheaper. When I'm done with the project, I'll make a video to share it.
Thank you so much, saved for future reference.
Bestes Projekt bis jetzt.
Dankeschön! Hoffe dass noch viele beste Projekte kommen werden!
Why did you add a switch instead of just bridging the pad and removing the connection to the original pad? I'm not aware of any EDO 30-pin SIMMs and assume that the standard only supports FPM, so leaving the option for EDO seems peculiar to me. Thanks for your work, and I may acquire some of these, but I'll hard-wire them for FPM mode 😉
You asked about desoldering thoughts, personally for small boards like that, I think you'd be way better off using a hop-plate style heater for pulling IC's off boards that have no components on one side. MHP30 would be the perfect size for pulling IC's one by one and other small solder reflow needs.
Absolut spitze! Klasse Arbeit. Ich habe zwei 486er Boards mit SIMM Sockets, die derzeit mit 8MB bestückt sind. Dort werden jetzt Deine Platinen einziehen, sobald ich auch so gut Chips verlöten kann. Kann Deine Platine auch 1MB Kapazität wenn dort Chips mit geringerer Kapazität aufgesetzt werden? Habe 4 *1MB Module, die als defekt erkannt werden und ich weiß nicht, ob es die Chips oder die alten Platinen sind.
Congrats!
Great work. What about using memory chips from EDO 168-pin memory sticks? They look similar.
I suppose that they are rare
@@ChrisSTARS2407and there is another problem, 168-pin sticks are 3.3v, the chips can't handle 5v on 30-pin slots.
My Macintosh iici loves its 8, 8mb 30 pin ram boards. The 16mb boards slow it down, so 64megs seems optimal.
Well done, just as good as necroware videos.
Awesome project!
I used your new board design to build a pair of 4mb SIMM modules for my Macintosh Color Classic.
I was not able to get them to work and I believe its due to the Mac being picky about memory.
I did build working FPM modules with the board design you based this project from.
I just wanted to put out a YMMV if you are trying this EDO2FPM on old Macs :)
Sorry to hear that the memory doesn't work with the Macintosh. Maybe you can find someone with a 386/486 to see if the modules work there.
I have a follow-up project with those modules and a SIMM adapter to 72-pin slot for a 486.
Would be good to hear if others were successful to build their own working modules with those PCBs.
@@bitsundbolts I'll try and hunt down someone with an old PC test rig. Maybe I can send them to you and you can do another test/troubleshooting video 🤓
We could arrange that. We can discuss via email: bitsundbolts at gmail com
I love these videos!
A followup to this subject: I've tried to do the same with my 4MB EDO modules with 42-pin 16-bit ICs. So, at first attempt I've just shorted /OE with the nearby /lCAS pin and got weird behavior with everything working right on a 486-DX100 machine except for any write operations with FDD. Each tie I've got corrupted data with visible pattern of correct and wrong bytes interleaving each other. Then I've made a test module with two standard 1n4148 diodes draining current from /OE to /uCAS and /lCAS and errors has stopped to appear. A have a theory that this has something to do with DMA, assumed to handle 8-bit operations differently from processor, so generally memory controller just activates /RAS, /lCAS and /uCAS sequently but in some situations the order or a set of signals is different, so with 16-bit modules this additional mod is required for correct operation. By the way it makes absolutely no difference between using a hardware floppy drive or Gotek emulator, the result is always the same: half of the data ends up absolutely random.
What surprises me is that the diodes do not seem to affect performance but probably, using a Schottky or another type of fast diodes should be superior. The next goal is doing the same trick with rare 50ns chips.
are 16mb versions of this board available? 16mb 30 pin simms are super $$$ and i'd love to load up my Mac SE/30 with 'em. :D
Let me guess: when you select EDO mode with FPM chips it will still work. It will only (potentially) slow down some timings.
Nice. Great Succes, but what's the pint doing this? Is there any technology from 386 that will utilize that much of memory?
Just creating this crazy configuration system is perfectly good reason IMO.
In DOS 8 MiB is usually enough to run all DOS games for which a 386 is fast enough. However, operating systems such as Windows NT and OS/2 would benefit. They need at least 12 MiB, better 16 MiB to run properly. Windows 3.x would benefit from the more memory in that it could run multiple applications at the same time. The same also applies to Windows 95.
You can never have too much RAM ;-)
Fantastic job Sir! You got one more subscriber to the channel. I personally go for less heat and more time when working with SMD stuff
Did someone research SIMM expansion modules for old HP printers, i.e. HP LaserJet 6L ?
Why switch? Whenever you do switch it over, you always need soldering iron to replace the chips.
Small pads that you could short with solder would make a bit more sense. ;)
I may want to try the 4x 30-pin to 72-pin converter and see if I can switch between EDO and FPM in the future. You don't have to use a switch at all as I mentioned in the video. The pads are close to each other to easily bridge them with solder.
Does this mean technically you could get 128MB on one of these boards? My manual says max 64MB, which would be 8x8MB sticks. Would a BIOS patch unlock the rest or is it a memory controller limitation?
It most likely is a limitation of the chipset and memory controller. But it could also be that at the time of release, there were no 16 MB SIMM modules in the market. It could be that boards support more memory than mentioned in the manual. Only testing would verify of it works or not.
@@bitsundbolts OK, would be an interesting test for sure. I also notice that some motherboards have issues with 3 chip SIMMs, so would be interesting to see how these behave. Would be interesting to see if its possible to push a motherboard so 128M, not that there is any point in DOS!
Interesting project. I was thinking of a board adapter so that you don't have to desolder the chips. But I don't know if that would be realistic because does EDO modules have a controller chip of some sort?
Haha, don't say more - you're spoiling my project list :) :) :) No, but seriously, I am thinking of a project to use one 72-pin socket and an adapter to split it to 4 30-pin modules. Drawback is that it may only work with 72-pin FPM modules, not with EDO because /OE pin is not exposed on connector pins. It should work with FPM modules though. It is a reverse 4x30-pin-to-72-pin adapter... Don't know if I want to do that though - routing 72 traces etc...
@@bitsundbolts Ooh, nice :) you are making me curious already
great video, thanks!! question is this resulting 30 pin module parity or non-parity? i have a tandy 2500 sx-33 that seem to require 4meg x9 modules (i think its parity?) let me know..
The ones I created are non-parity. But it should be possible to use one of the chips mentioned in the original BOM from UpLateGeek to make it a parity module. I may do this in the future - at least the real FPM modules.
Yes, x9 means parity. If this is a requirement, then you need the third chip on the PCB which I did not add in this video.
@@bitsundbolts understood. but the chips on ULG BOM are FPM, and the others will be EDO behaving as FPMs... any conflict there? i guess it needs testing?
Yes, it needs testing. But I strongly believe that there shouldn't be a problem because I already mixed EDO and FPM on the 386 in one memory bank (4 slots = 1 bank = 32 bits). So, 8+8+8 bits were EDO-2-FPM modded and another 8 bits were FPM chips - this worked and is in the video at 09:39.
Do these modules use 2 chips for memory and the third one for parity?
Yes, that is correct.
This 386 system has more memory than my Pentium system had back in the day.
Do you have success with inserting 16MB memory into 386SX M/B? Which chipset? Ali?
As much heat as you can without melting or disturbing nearby componentry is the answer. In my opinion, these RAM boards would be more heat, less time. If you were working on a chip near a plastic connector, less heat more time.
32MB EDO modules are quite valuable. Would it be possible to use lower capacity modules as donors?
Single-sided 16MB modules are the same basically.
The donor had a dead chip. So this wasn't a waste. They are also not particularly rare. Just eBay inflated prices
I'm afraid in this configuration with two memory chips it's either 1MB or 4MB. I guess a 2MB module is possible, but then it would take only half of the address space of the highest not, which is weird. I just have not seen it before.
The switch seems a little silly since it's easy to change a bridge like that. could also just use pinheaders and a jumper, but yea if you have extras of those switches then why not.
great work
Thank you!
If I understand correctly, you bought new memory chips. Do you think it would be possible to create a 16MB stick with new components?
It should be possible if you find chips with Address line 11 active. The parity chip, however, is not connected to this line. Therefore, parity check for 16MB modules won't work on those PCBs. If I find chips that have the correct pinout, I may give it a try.
a 386 with 32mb of ram running windows 95. that's such an odd combination. I wanna see that. OSR2 with usb, can you get usb working? I wanna see a 386 with usb support. What about 98?
I wonder what the newest version of windows you could get on this. Hopefully something newer than Windows Me.
I think it would be windows me and very slowly, I am sure you need a 486 to get nt 5 to run.
32MB on a 386 is completely bonkers! My 386 back in the day had 4MB and it was a dream machine for me.
I absolutely agree. However, I want to create a ram drive one day on this system. It's all for fun and entertainment 😉
@@bitsundbolts a RAM drive is awesome in that era. I never had enough RAM back then to make one that was useful - it was just a novelty.
Absolutely! With 32 MB ram, it would have more space than some hard drives at that time! I haven't looked into this project yet, but it's definitely on my list.
Great! Could you please say the appr. price for PCB manufacturing
PCBWay gives you a quote right on their website. If you go with 10 PCBs, it will cost 5 USD + shipping. You can use the link in the video description and get a 5 USD welcome bonus. Then you pay for shipping only.
Could you show how to mod or make EDO 72 Pin SIMM into a FPM 72 Pin SIMM. i have a old late model 486 but having tribble with memory.
I could try to turn an EDO module into a FPM module, but this would involve cutting one pin for each memory chip and then bridging the remains to the neighboring pin. Wouldn't it be easier to buy a real FPM module for your 486?
@@bitsundbolts I have quite a few modules. Unfortunately, I am having difficulty identifying them and if they work or not. I only have one motherboard and is currently limed to a 5V DX2 66 CPU's. The 3.3 v regulator is not present but the jumper pins, pads and holes for it are.
Maybe you can lookup the data sheets for the chips on the modules - if they're not too many. "Datasheets360 dot com" may be very helpful.
At the moment I don't have a better solution than to try one by one. You may have noticed that some of your module have very different looking chips and sometimes only two for single sided and four chips for double sided modules.
I could imagine that with a hack, things get more complicated and probably will increase your trouble with memory further.
How long does it usually take PCBway to ship to Dubai? Do you source jellybean components and general cheap ICs locally?
Manufacturing takes about 2 days for green and 3-5 days in other colors. Shipping takes around 5 days with FedEx or DHL.
I cannot find many ICs locally. Mainly checking on AliExpress and Alibaba.
Can i use GM71C17400AJ6 from 16/32MB 72 pin as a donor, it seams there are 30 pin 4mb modules with the same chips.
I checked the pinout - looks like this chip should be compatible as well. It even supports FPM, so, no need for the "hack" in older systems that require FPM memory.
What program do you use for pcb design?
I use EasyEDA - I find it really easy to use for beginner and it is web-based.
I wonder how exactly the faulty chip is bad... if only some data lines are bad, maybe it could still be useful as a parity chip
Awesome
very nice :)
I can't wait to see Windows 95 running on this 32MB 386 system!
The video is already done, but I am almost on my way to the airport now - holidays for 2 weeks.
Spoiler: There will be one video with the installation of Windows 95 and a second video repairing a GeForce2 MX - both ready to go, but need to spread them over the next two weeks.
So that was the maximum EDO ram you could run with? 32MB?
Based on the manual, yes. However, SIMMs should theoretically go up to 16 MB per module. That would be 128 MB with 8 modules. But I doubt the chipsets supports those and the motherboard may not have all the traces in place.
If max memory per SIMM is 16Mb - total max memory for 8 slots is 128Mb? What are the obstacles to reach that?
Support of the chipset and traces on the motherboard. The chipset may simply not expose the 12th address pin/internally not connected. Or the motherboard does not route all address lines to the SIMM sockets. I am sure there may be other reasons, but those are the ones I can think of.
Now, I don't know if this "Super Low Cost" (SLC) chipset has a physical limitation, or if it could be possible to run this board with 128MB - even if it's not mentioned in the manual.
A pinout of the chipset may give some hints.
@@bitsundbolts But in principle, is 16Mb SIMM realistic to build?
@@BarsMonster I had a small "Pride Expression" motherboard with an Am386DX processor on an ALi M1429/M1431 chipset with 8 SIMM30 slots and the manual for it contained tables for filling memory slots up to 128 MB. Its specifications also mentioned 128 MB. However, at that time I had only 8 MB of memory.
I still have some of memory sticks still
It was invented at least 5 years ago. A switch is not needed. On all older systems, the EDO memory operates in FPM mode.
I can not give a link, it is almost immediately deleted.
Clearly, you need to make a RAMDISK to use all that extra space.
nice )
32MB EDO Module sind überhaupt nicht "abundant". Das Meiste sind 8MB EDO Module, die man hintergeworfen bekommt.