I have some Intel 10/100 PCI LAN cards with PLCC sockets. Maybe those would work?

Kahenraz wrote on 2022-03-06, 22:30:

I have some Intel 10/100 PCI LAN cards with PLCC sockets. Maybe those would work?

Unsure - PLCC would be an odd socket for a boot ROM. Perhaps its intended for a Flash chip. The question then is where to get the firmware to put on the chip - most manufacturers did not provide this for free (Realtek and AMD did). If the firmware could be found it would probably work fine for network booting DOS.

PCI may still be an issue with Windows 95 though - Microsoft didn't officially support it but possibly the workaround on the LanWorks website would work. I'm not entirely sure what the actual problem is with PCI NICs - when I tested with VirtualBox emulating a PCI AMD PCnet it would get most of the way through startup and then just crashes. The system has to boot using a real-mode network stack and drivers and at some point during statup a protected-mode stack and drivers have to seamlessly take over - possibly this is where it runs into trouble. Maybe the bits needed to configure PCI cards haven't been loaded from the network by that point though that wouldn't explain why the workaround from LanWorks works.

Sadly there isn't very much information on this stuff out there - it took quite a bit of trial and error to get it working as Microsofts limited documentation leaves quite a bit out. There are also bugs that never got fixed - there is a program that runs early on to read a config file on the network which tells it where the machines registry lives. That program in particular is a bit temperamental - it gives all sorts of errors none of which are documented anywhere visible to google. Often just swapping the order of two lines in the ini file its reading makes the problem go away. I'm surprised at how quickly they seemed to give up on Network booted Windows 95 given how much effort they put in to making it work at all - all the extra stuff in Windows 95 setup to deal with this scenario, the special tools for preparing the shared windows directory, etc.

I think it would be good to start a new thread with details of how to connect early PCs (especially pre 95) to a network and how to set up network boots. Would make it far easier for others to find the info.

Around early February I was offered a Sanyo MBC-1250 CPM Computer (told it was identical internally to an MBC-1200) but has dual drives, dual z80's CPUs,
graphic monochrome computer. aside from oldcomputers.org, I cannot seem to be able to find any documentation for it.
After several days of research, I found another source of diskettes for the MBC-1200/1250

I found out that Montezuma Micro CPM v2.32 for TRS-80 model 4, supports Sanyo 80T DSDD 627K format
but the DUP program only had problems copying the IMD/HFE image file to an 80T DSDD diskette
So I backtracked my approach to a previous version of Montezuma's CPM v2.2. and suddenly there, made a Sanyo CP/M v2.2 functional diskette!
I now offer you the result! great system, great graphics! I am super happy with the resolution after the time of research.

davidrg wrote on 2022-03-06, 22:55:

Unsure - PLCC would be an odd socket for a boot ROM. Perhaps its intended for a Flash chip. The question then is where to get th […]
Show full quote

Kahenraz wrote on 2022-03-06, 22:30:

I have some Intel 10/100 PCI LAN cards with PLCC sockets. Maybe those would work?

Unsure - PLCC would be an odd socket for a boot ROM. Perhaps its intended for a Flash chip. The question then is where to get the firmware to put on the chip - most manufacturers did not provide this for free (Realtek and AMD did). If the firmware could be found it would probably work fine for network booting DOS.

PCI may still be an issue with Windows 95 though - Microsoft didn't officially support it but possibly the workaround on the LanWorks website would work. I'm not entirely sure what the actual problem is with PCI NICs - when I tested with VirtualBox emulating a PCI AMD PCnet it would get most of the way through startup and then just crashes. The system has to boot using a real-mode network stack and drivers and at some point during statup a protected-mode stack and drivers have to seamlessly take over - possibly this is where it runs into trouble. Maybe the bits needed to configure PCI cards haven't been loaded from the network by that point though that wouldn't explain why the workaround from LanWorks works.

Sadly there isn't very much information on this stuff out there - it took quite a bit of trial and error to get it working as Microsofts limited documentation leaves quite a bit out. There are also bugs that never got fixed - there is a program that runs early on to read a config file on the network which tells it where the machines registry lives. That program in particular is a bit temperamental - it gives all sorts of errors none of which are documented anywhere visible to google. Often just swapping the order of two lines in the ini file its reading makes the problem go away. I'm surprised at how quickly they seemed to give up on Network booted Windows 95 given how much effort they put in to making it work at all - all the extra stuff in Windows 95 setup to deal with this scenario, the special tools for preparing the shared windows directory, etc.

Are you looking to support Windows 95 specifically? Would you consider a Novell NetWare solution or do you want to only use Microsoft software?

I remember in High School that all of the school's computers (mostly 95/98) lagged to hell for even the most basic of tasks. Years later I would attribute this to most of their software being managed over what was probably an over provisioned network. I remember seeing some of the computers booting with a Novell NetWare login prompt as well. I don't know if the computers actually booted from the network, but it's possible.

TxSnipper wrote on 2022-03-07, 01:03:

Around early February I was offered a Sanyo MBC-1250 CPM Computer (told it was identical internally to an MBC-1200) but has dual […]
Show full quote

Around early February I was offered a Sanyo MBC-1250 CPM Computer (told it was identical internally to an MBC-1200) but has dual drives, dual z80's CPUs,
graphic monochrome computer. aside from oldcomputers.org, I cannot seem to be able to find any documentation for it.
After several days of research, I found another source of diskettes for the MBC-1200/1250

I found out that Montezuma Micro CPM v2.32 for TRS-80 model 4, supports Sanyo 80T DSDD 627K format
but the DUP program only had problems copying the IMD/HFE image file to an 80T DSDD diskette
So I backtracked my approach to a previous version of Montezuma's CPM v2.2. and suddenly there, made a Sanyo CP/M v2.2 functional diskette!
I now offer you the result! great system, great graphics! I am super happy with the resolution after the time of research.

That computer looks amazing! Job well done rescuing and restoring such a relic. Can you explain more about the hardware and what kind of software was used on it?

The demo in your photos is also very impressive and looks wonderful on the green phosphor. Are there any videos of it? Does it animate or just draw things?

Kahenraz wrote on 2022-03-07, 02:10:

davidrg wrote on 2022-03-06, 22:55:

Unsure - PLCC would be an odd socket for a boot ROM. Perhaps its intended for a Flash chip. The question then is where to get th […]
Show full quote

Kahenraz wrote on 2022-03-06, 22:30:

I have some Intel 10/100 PCI LAN cards with PLCC sockets. Maybe those would work?

Unsure - PLCC would be an odd socket for a boot ROM. Perhaps its intended for a Flash chip. The question then is where to get the firmware to put on the chip - most manufacturers did not provide this for free (Realtek and AMD did). If the firmware could be found it would probably work fine for network booting DOS.

PCI may still be an issue with Windows 95 though - Microsoft didn't officially support it but possibly the workaround on the LanWorks website would work. I'm not entirely sure what the actual problem is with PCI NICs - when I tested with VirtualBox emulating a PCI AMD PCnet it would get most of the way through startup and then just crashes. The system has to boot using a real-mode network stack and drivers and at some point during statup a protected-mode stack and drivers have to seamlessly take over - possibly this is where it runs into trouble. Maybe the bits needed to configure PCI cards haven't been loaded from the network by that point though that wouldn't explain why the workaround from LanWorks works.

Sadly there isn't very much information on this stuff out there - it took quite a bit of trial and error to get it working as Microsofts limited documentation leaves quite a bit out. There are also bugs that never got fixed - there is a program that runs early on to read a config file on the network which tells it where the machines registry lives. That program in particular is a bit temperamental - it gives all sorts of errors none of which are documented anywhere visible to google. Often just swapping the order of two lines in the ini file its reading makes the problem go away. I'm surprised at how quickly they seemed to give up on Network booted Windows 95 given how much effort they put in to making it work at all - all the extra stuff in Windows 95 setup to deal with this scenario, the special tools for preparing the shared windows directory, etc.

Are you looking to support Windows 95 specifically? Would you consider a Novell NetWare solution or do you want to only use Microsoft software?

I remember in High School that all of the school's computers (mostly 95/98) lagged to hell for even the most basic of tasks. Years later I would attribute this to most of their software being managed over what was probably an over provisioned network. I remember seeing some of the computers booting with a Novell NetWare login prompt as well. I don't know if the computers actually booted from the network, but it's possible.

At the moment NetWare is exactly what I'm using - the server (virtualised) is NetWare 4.11 with windows 95 loading everything over a 10Mbps network link - even its swap file is on the server. I assume the only reason its preforming well right now is the client has 64MB of RAM and the server is pretty much idle. I bet a whole lab full of machines with 16-32MB of RAM would perform terribly as they're all frantically swapping to some poor network server.

In an hour or two I'll probably make a new thread covering what I've done so far (and what I'm going to attempt next) in a single post in case there might be anyone else out there interested in this unusual corner of vintage networking.

Tested out some motherboards that I got from a scrap lot.

PC-Chips M919 with COAST stick: works

386DX-40 board -removed the leaky battery and used vinegar to neutralize. : works

486 board: doesn't work. Suspect shorted capacitor.

Socket 5 board: doesn't work. Suspect shorted capacitor.

I am using an ATX power supply with an ATX to AT adapter to test with and both boards that aren't working keep the PSU from powering up. Looks like short protection is kicking in on the PSU as the fan just does a little move and then nothing.

I gotta test my K6BV3+/66 and recap my 6VTXE board.

Later on tonight, I'm going to upgrade the floppy drive to a dual floppy drive and install a 1x CD-ROM drive in my Compaq Portable 286.

I wish you all of the luck with the 486 and Socket 5 restorations, cyclone3d.

Took my mt-100 apart and cleaned every last part inside and outside. There was a ton of dust and dirt around all of the buttons, inside wasn't all that bad considering it's age.

I need to do the same for my SD-35 sometime. The FDD was not working so I replaced that with a black Gotek, kinda want to return it to original state and get that black DD FDD working. ALso the stop button is not working on mine, gotta see what's going on there..

Don't have any of the original disks for mine, so I'm not sure if it works or not. Probably doesn't any more, but I really only need the mt-32 part of it.

sirotkaslo wrote on 2022-03-07, 06:43:

Took my mt-100 apart and cleaned every last part inside and outside. There was a ton of dust and dirt around all of the buttons, inside wasn't all that bad considering it's age.

Be sure to also check your SC-55 for corrosion.

PCB destroying adhesive found in my Roland Sound Canvas SC-55mkII (inspect your synthesizer!)

I've already cleaned the SC and replaced battery, all caps look good

davidrg wrote on 2022-03-07, 02:24:

At the moment NetWare is exactly what I'm using - the server (virtualised) is NetWare 4.11 with windows 95 loading everything over a 10Mbps network link - even its swap file is on the server. I assume the only reason its preforming well right now is the client has 64MB of RAM and the server is pretty much idle. I bet a whole lab full of machines with 16-32MB of RAM would perform terribly as they're all frantically swapping to some poor network server.

In an hour or two I'll probably make a new thread covering what I've done so far (and what I'm going to attempt next) in a single post in case there might be anyone else out there interested in this unusual corner of vintage networking.

I wonder how well it would perform with a swap file on a small c:\ partition.

Windows will always swap regardless of how much memory there is in a system, as this is best practice to keep as much fast memory available as possible. You may find there to be more of an improvement by adding a tiny swap partition than more RAM.

Kahenraz wrote on 2022-03-07, 10:27:

Windows will always swap regardless of how much memory there is in a system, as this is best practice to keep as much fast memory available as possible. You may find there to be more of an improvement by adding a tiny swap partition than more RAM.

I second that.
Never fully deactivate the swap file. Reduce it to the minimum and add RAM.

My favorite trick, in anything other than laptops obviously, is to just stick another drive in a windoze box for swap... doesn't matter if it's small and half the speed.. just keeping windoze from swapping on the main drive speeds things up. I mean, it's better if it's a fast drive, but you still get speedup with a slow one... in the early oughts, I was favoring those 20GB Quantum Fireballs which were around, people were dumping them cheap to get 40, 80, 160 etc GB as capacities seemed to explode during that time, but they had fast seek, 7200RPM, decent transfer rate. So I'd give windows a decent size swap file on one, and fill the rest full of media or zipped files for backup.... Though the Fireball name was also getting abused so there were some horrible slow ones around too 5400RPM and slower than the regular "Prodrive" line.

Kahenraz wrote on 2022-03-07, 10:27:

davidrg wrote on 2022-03-07, 02:24:

At the moment NetWare is exactly what I'm using - the server (virtualised) is NetWare 4.11 with windows 95 loading everything over a 10Mbps network link - even its swap file is on the server. I assume the only reason its preforming well right now is the client has 64MB of RAM and the server is pretty much idle. I bet a whole lab full of machines with 16-32MB of RAM would perform terribly as they're all frantically swapping to some poor network server.

In an hour or two I'll probably make a new thread covering what I've done so far (and what I'm going to attempt next) in a single post in case there might be anyone else out there interested in this unusual corner of vintage networking.

I wonder how well it would perform with a swap file on a small c:\ partition.

Windows will always swap regardless of how much memory there is in a system, as this is best practice to keep as much fast memory available as possible. You may find there to be more of an improvement by adding a tiny swap partition than more RAM.

IIRC Microsoft did recommend having a local disk to store the swap file if possible. I guess with a local swap file performance would probably improve a fair bit - the NIC is only 10Mbps or around 1MB/s so its kind of amazing windows runs as fast as it does. Wikipedia says IDE was originally only 8.3MB/s half-duplex so I guess with a 10/100 card you could probably beat that (assuming the server has fast disks).

It's fascinating to see Win95 run without a local hard drive in that manner.

Now we just need to do it over the internet.

"Windows 95 machine in New Zealand, boots fr0m hard disk in USA". That's a headline. (Take the world's attention off this pesky political stuff that's for sure. 😁 )

Shreddoc wrote on 2022-03-07, 20:48:

It's fascinating to see Win95 run without a local hard drive in that manner.

Now we just need to do it over the internet.

"Windows 95 machine in New Zealand, boots fr0m hard disk in USA". That's a headline. (Take the world's attention off this pesky political stuff that's for sure. 😁 )

That would be something to see!

If you want to see what it looks like booting from the LAN I got out my IP-KVM and made a quick recording:
https://www.youtube.com/watch?v=9mOgDrJwWP4

Video performance isn't great - I'd really need some sort of VGA to HDMI converter and some sort of HDMI capture device if I was going to make a habit of this.