486 Socket Blaster, Adding 3.45v support to the supposedly unupgradable IBM PS/1
HTML-код
- Опубликовано: 12 май 2023
- Part 1: Repair of PSU and CPU. here: • You're not going to be...
Part 2: Hacking the MOBO: here: • Motherboard hacking fo...
Part 3, This video!
PCB Prototype the Easy Way. Full feature custom PCB prototype service. www.pcbway.com/
GitHub project page: github.com/scrapcomputing/486...
Vogons post here: www.vogons.org/viewtopic.php?...
--------------------------------------------------------------------------------------
New on patreon.com/Epictronics
--------------------------------------------------------------------------------------
Tools I regularly use
DeoxIT D5 Contact Cleaner
Hanstar 861DW Rework Station
Pro'sKit SS-331 Desoldering Station
UNI-T UT61E Auto Ranging Multimeter
UNI-T UT890D Manual Ranging Multimeter
MESR-100 mk2 ESR meeter
PINECIL Soldering Iron
PinePowerPSU
TS-100 Soldering Iron
AMTECH NC-559-ASM Flux
MaAnt Grinding Pen
Multicore 60/40. 0.38mm and 0.5mm solder
TL866 II Plus Programmer
Tektronix 2246A 100 MHz four-channel analog scope
PCBs from PCBWay.com :)
--------------------------------------------------------------------------------------
patreon.com/Epictronics
Join me on Twitter: / epictronics1
Music by Karl Casey @ White Bat Audio Наука
Part 1: Repair of PSU and CPU. here: ruclips.net/video/49ZOxgc2pXI/видео.html
Part 2: Hacking the MOBO: here: ruclips.net/video/kxbkkpe6wfM/видео.html
Part 3, This video!
Support me on patreon.com/Epictronics
Don't forget to re-pin your comment after editing!
@@UpLateGeek Thanks!
The creator of this has a youtube channel and did a video on its creation. "ScrapComputing"
Scrap Computing does really nice work and is a wonderful member of the retro community
Very happy to see scrap computings socket blaster getting some love.
The socket blaster project deserves it
I think I know how to set the clock multiplier to 4x. On a 5x86 socket 3 cpu, there is a pin called CLKMUL at position R-17. On most of the earlier 486 cpus pin R-17 was marked as INC which means "internal no connect". It's likely that pin isn't connected and since the 5x86 has an internal pull up resistor on that pin, it will default to being "high" which is the "3x" multiplier setting. You need the signal on that pin to go low and that will put it in 4x multiplier mode. Try connecting a 10kohm resistor between R-17 and ground then see if that works. Since the model 2133 motherboard (the one the wiki says came with a 486 cpu) was made in 1992, it's likely there won't be any points that are connected to R-17 besides the pins on the pga168 socket.
I did something like this with a ms-4144 motherboard that supports am486dx4 and intel 486 dx4 cpus but not amd 5x86 cpus. In my case I got away with putting one end of a resistor into a header I found that was connected to R-17 and the other end to a different header that was connected to ground.
Hacking the board to run at a faster bus speed than 25MHz might be hard. Hopefully the isa bus clock is the cpu clock divided by 4 and not 3 in which case you can replace whatever is generating that 25mhz input clock with something else such as a 33MHz or 40MHz osillator. The chipset and motherboard may not be able to handle those speeds though.
That's very helpful, thanks!
Very tight soldering work, great stuff. I enjoyed watching that video, looking forward to the next adventure. Wishing you all the best keep em coming .. Mick.
Thank you :)
The buck converter likely doesn’t have a large enough load on it to stabilise it’s regulation or it’ll change to intermittent cycling to minimise power consumption. Once there is enough load it’ll probably stabilise and achieve the datasheet specifications.
Man I really love this style of retro computing. Old retro gear with modern day tools/solutions to make them do things that are unnatural. 😅
Your soldering skill are absolutely next level. 😱
Thanks!
Great work! I had the same issue with my 486 based PC way back when. It was a Packard Bell machine, the first model they offered with a DX2 66Mhz CPU in it. When I upgraded the machine with an AMD 5x86 133 chip (also using an in socket voltage converter) there was no way of setting the multiplier on the board so the CPU always ran with a 3x multiplier. In the end I used a 40 Mhz board setting which allowed me to run the CPU at 120 Mhz. Even at only 120 Mhz it was a nice step up from the 66 Mhz chip and changing the board speed from 33 to 40 Mhz made the local bus graphics run a little faster as well.
Thanks, that's the next step. I'm sure we can hack the clock speed on this board somehow
I had a friend that was so into the IBM PS/1 and PS/2 that IBM reps would email and ask HIM for advice.
When I saw you struggling with the pin headers, everything inside me screamed: put the pins in the socket first, then fit the PCB on top. Then solder while everything is held in place
Why didn't I think of that... I'll try it the next time, thanks
@@Epictronics1 while watching the video, it dawned on me that you have to flip the board all the time. So it might be a bit cumbersome. But still, having a socket underneath helps with getting the pins all properly aligned
@@TSteffi Ah, yeah, I forgot that the solder joints aren't accessible otherwise
i used to put tiny pieces of wire in the socket to set the multiplier. i once even connected the old turbo switch in front of the case to the pins from the back of the socket to change the multiplier.
Excellent, did you just pull a pin low or high?
@@Epictronics1 it's too long ago. i can't remember.
Amazing video and nice way to speed things up!
Thank you :)
A couple tricks I use for soldering pin headers & sockets: Blue-Tak (the same stuff you’d use to hold a poster to the wall) works great for sticking the headers to the board and keeping them aligned while soldering. The other thing I sometimes do is to use a spare board to keep the headers aligned (it forms a sandwich with the header in between two PCBs).
Nice work on this one!
Thanks!
So I think I saw you skip one of the settings when you installed the P75 (the 5X86). This might be something worth checking, ill explain.
The original Intel DX2 series is of P24 architecture, when the AMD 5X86 is based upon the latter P24D. The P24D would enable L1 WB, while they both are internally clock doubled.
I think one more check would be worth the effort. Set all three jumpers to the DX2 settings, then with the chip installed, check if the FBS will results in 33MHz since the BIOS should set the multiplier accordingly on the identification vector during POST. However, earlier IBM builds did some funny things when it came to clock multipliers. So I can not ensure this will work as stated.
I cant wait to see if this results in a win or loss (if you choose to accept the mission)!! Hope this helps, and enjoying all of your work and dedication. Keep up the great work!! Best...
Thanks. I have made some progress with this project and I'm planning for a new attempt to go even faster soon :)
I love this kind of content, great stuff. It'd be cool to see if you can figure out a way to get the 2x multiplier working.
We'll get there, thanks!
Fun project! That is some fiddly soldering!
It was! Now, let's see if we can double the speed again : )
grattis på födelsedagen pappa
I miss the days when different CPU makes would fit into the same socket.
Those switching regulator boards from Ali might not contain a genuine TI part, so performance could be different.
Basically, it works like this. Chinese manufacturers don't want to rely on a bunch of US-designed ICs in their products, so the domestic market provides (usually inferior) ICs that are pin-compatible with those US-designs. They seldomly get marketed outside of China, some even infringe on patents and couldn't be exported either way.
These parts then sometimes get relabeled as genuine parts from which they were copied, and thus enter the product cycle again as fakes.
yeah, not unlikely that my cheap converter has a fake chip. But it seems to be stable at about 2v below input voltage. 3.45v is a bit challenging to set but at 3.3v it's perfectly stable
@@Epictronics1 Just saying that looking at the original datasheet for the genuine part will not reveal the actual performance characteristics. I generally fault any maker for including questionable Ali parts in their designs.
@@graealex It's dumb to blame the designer, blame the manufacturer. There's a worldwide semiconductor shortage, and often times, knockoff parts are all that's available. So you have the choice of using a knockoff part that probably works, or speccing the real part that isn't available for months/years, or is astronomically expensive because of greedflation.
So good!
Thanks :)
The Socket Blaster has a clock multiplier jumper on it, "CLKMUL". Try changing the jumper to short pins 2 and 3.
Unfortunately, that jumper only reduces the multiplier
You should be able to force a 2x multiplier on the socket blaster which the am5x86 will interpret as 4x unless I'm mistaken.
@@willyarma_uk That's what I'm hoping for. I don't know how though :)
@@Epictronics1 Try removing the jumper entirely.
Yeah!!!
I still think it's funny how the PS/1 came out after the PS/2 lol.
It could still be interesting to benchmark with Am5x86 due to differences in microarchitecture.
So much hacking of that board. Loving it. 😁
We're far from done yet :)
Regarding soldering and camera - have you thought about buying one of digital microscopes?Tomlov DM602 costs about $150 brand new and you can see everything and record. Mark fixes stuff channel had a review recently about the DM601
Yes, that's probably the best solution. I'll have a look at the DM602. Thanks
13:35 know why it's jumpin? The buck needs a load 😁!
yep, it stabilised, but only at around 2v below input voltage
@@Epictronics1 haven't you try 3.45V?
@@paveljelinek772 I did, but it dropped down to 3.3 with load
@@Epictronics1 yea, normal behavior.. to hold the set voltage you'd need to add bigger caps to the module..
It'd be nice to see a modern AM5 to AM4 converter kits.
Not possible, the architecture is completely different.
The only reason adapters like this work is because you're only changing the core voltage, multiplier and clock settings. The more exotic adapters to go from 286 to 386 or 386 to 486 only work because the architecture is backwards compatible, and the bus is close enough to use glue logic to translate between the two. You're still limited to the system board of the older design though. There's a limit to how fast you can make the CPU, because at some point, you're waiting on everything else in the system to catch up.
AM4 and AM5 are two entirely different architectures with different incompatible buses and memory standards.
On the other hand, it is entirely possible to modify LGA115x boards to go up to 8th or 9th gen Intel CPUs with socket and BIOS modification. Really shows how scumbag Intel was forcing socket changes to segment the market and generate artificial profit.
I checked VOGONS, too. There were a couple of threads where people were trying to change from 25 to 33MHz but no dice...;-I
ruclips.net/video/ScpSBcMMuX4/видео.html
I have a related IBM PS/1 board, no idea what model it's from since I got it without a case, but it's a weird one with a 486BLC2 and a CPU socket.
My board uses a similar PLL setup to yours but it has three jumpers on the edge of the board by the SIMM sockets. Those allow for selection of 25 / 33 / 50mhz by moving the jumper to one of the three jumpers (1x jumper, 3x 2-pin jumper locations). The jumper's pins connect to Ground and a pin of a square GAL chip just next to them - which must be doing some work to specify the bus speed.
The board I have has an ICS2694m chip next to the 14.3MHz crystal, which looking at the datasheet has 4 pins dedicated to speed selection, giving 16 possible combinations - which must be what the GAL is for, cutting down on having lots of jumpers and less chance of breaking things.
To find out how to set the bus speed on your board, I recommend looking at the datasheet for that tiny IC next to the 14.3MHz crystal. Hopefully there's a datasheet online for that and you can trace where the CPU selection pins on the PLL go. They might be hardwired to Ground or left floating, so you may have to do some surgery (lifting an IC leg) to change the FSB if there's no provision for it otherwise.
Thanks. I'll start with the chips close to the crystal and see what I can find. Since the seemingly identical board in the 2155 is running at 33MHz there must be a way (EDIT) I'm not aware of a PS/1 with an SLC, could your board be from a PS/2?
@@Epictronics1 No idea what it's from, maybe one of those Ambra PCs or some other off brand.
Check the thread, I spotted something interesting that different between your board and the 33mhz one :)
@@voodler SLC2s weren't that common. The only machine I have with an SLC2 is a PS/2 Model 56. I have posted a new picture in the thread. Hopefully someone with a 2155 will add a high-res image too
there's unwritten rule to solder smallest and least bulky components first.
What clock crystal is installed IBM may have put the lowest possible to meet requirement of 486sx or possibly left circuit unpopulated for those jumpers. Not very experienced but putting ideas to check I overlook small stuff all the time.
This board doesn't have a large crystal, like my Commodore 486 board that we overclocked recently. I think the clock is controlled by a chip on this board
@Epictronics ouch wonder if a clock circuit you be built separately and patched directly to processor.
@@christopherdecorte1599 That's plan B. I'll try to configure the onboard clock chip first
@Epictronics I'm currently working on a duel pentium 3 hp server NAS so hard to find a reasonable gpu for pci ati rage 8m is a joke for anything fun.
J25 it apears to be related to disabling and/or selecting the soldered cpu. But why did you only tested the voltage and not also the clock while testing j23 and j24?
I did, Unfortunately, the clock was at 25MHz in all positions on all three jumpers
I put in my 2133-571 a kingston turbochip and it run flawless (see vogons IBM PS/1 multimedia max restore), otherwise pentium overdrive not, probably bios update problem. I have also an aptiva 2144-921 and PS/1 is much better with the same processor.
Yeah, I remember reading your thread back in 2018 ish. A really nice machine. I think you might be right about it being the firs IBM "multimedia" I have an Aptiva with a Multimedia sticker and the same SB+CDROM kit installed. Btw. I reread the thread now and noticed that you have a GD 5428. That chip can take 2Mb of VRAM. If you can find a zip socket and chip for those unused pads next to the onboard VRAM, you might be able to get really fancy graphics :) PS. you got some crap in the thread for not having VLB slots, but you don't need them anyways. your machine already has local bus graphics onboard. Unfortunately, my board doesn't have a J37 to change the speed, we need to figure out what it actually does and bodge it somehow.
Hey, great project. These adapters... Well I am looking for 775 to 478 adapter for years (to run an old P4 CPU on a more modern s775 motherboard). If anyone has GERBER files, please share. I can do some re-degin, as daily PCB design engineer. I can also help with some of these awesome projects!
it will also require a modified bios for your motherboard, if it even works
Does the 3rd jump[er on the motherboard change the clock multiplier?
I didn't actually check but I don't think it does. The three jumpers are now set the same way as on the 2155 board but my board still runs at 25MHz
Necroware probably has a way?
I wonder if the clock speed isnt controlled in some form of hardcoded firmware or something.
yes, probably an IC controlling the speed
@@Epictronics1 have you found any clock ICs? Or could there be some trace that has to be cut to unlock that feature. I've seen that before inplace of jumpers.
@@xephorce Not yet, but the board is on the bench
@@Epictronics1 idk if its standard in board manufacturing but some will make it easy to find the places that have traces to cut usually by having 2 arrows pointing to each other showing the spot >
@@xephorce Thanks :)
Would be great if you could buy these. Looks really difficult to make.
Well, they do require some patience. But it's really worth it
Could we do a group buy and have PCBway assemble them?
Just never copy that floppy!!!😊
What if you replaced the crystal oscillator
It doesn't have the usual large crystal I have upgraded on other 486 boards. It probably uses an IC to set the speed
Unfortunately that would potentially break everything, the board uses a PLL with a 14.318mhz crystal. Changing that reference clock would change more clocks than just the Bus speed.