Ran ChkCpu on my portable 95 machine (Libretto 50CT) since I couldn't get CPU-Z (tried every old version possible) to work. It reported that my system has a non-MMX, 75 MHz Pentium.

That's cool, but this thing has a Pentium MMX sticker on it! Perhaps ChkCpu is misidentifying the processor, or maybe the only stickers available when this thing was made were the MMX ones? Definitely strange. Still love this little thing though!

How much cache does it come up with? I've got one of the early Libretto 50CTs that unfortunately predates the free MMX upgrade (which is excellent to have if you want to overclock it).

There's no L2 cache in the libretto, but an MMX should have 16KB of cache in cachechk, while my non-MMX has 8KB. Also speedsys does some extra tests if there are MMX instructions available.

It looks like CPU-Z (1.78) doesn't work on my Windows 95 either 😀

Thermalwrong wrote:

How much cache does it come up with? I've got one of the early Libretto 50CTs that unfortunately predates the free MMX upgrade (which is excellent to have if you want to overclock it).

There's no L2 cache in the libretto, but an MMX should have 16KB of cache in cachechk, while my non-MMX has 8KB. Also speedsys does some extra tests if there are MMX instructions available.

It looks like CPU-Z (1.78) doesn't work on my Windows 95 either 😀

It tells me I have 16 KB of L1 cache. So it is an MMX! Strange how Chkcpu doesn't notice the MMX instructions. Perhaps the program doesn't behave well with these CPUs. Also, what kind of error do you get? I'm trying to run 1.20-something and it gives me an OLE32.DLL error. I'm probably missing some kind of framework it requires.

Tested a bunch of 386 & 486 stuff out, trying to piece together The One That I'll Keep. Some motherboards I've had in my closet for ages, CPUs I had lying around and from the bag I found in a thrift shop a couple weeks ago, and RAM from who knows where.

Results: somewhat disappointing. All the CPUs work, after spending most of an hour meticulously un-bending pins. (Protip: a 72-pin SIMM works great as a Socket 3 pin straightener!) The DataExpertHiNT EISA+VLB combo board seems to work fine, but unfortunately I had no luck with the Gigabyte GA-486VS. Tried most of the CPUs, a bunch of different RAM including FPM & EDO, re-doing all the jumpers, etc. but the result was the same: no POST, not even any beeps.

The DataExpert is a pretty neat board, but I'd really prefer to use the Gigabyte for a few reasons - ZIF socket, 3.3V CPU support, and a more modern BIOS presumably supporting larger hard drives being the main ones. I haven't given up on it, but I ran out of stuff I could try for the time being. I got it with the BIOS chip missing so I had to burn my own; it's possible I did it wrong or I used an incompatible EPROM, although it verified OK.

Then I got my stray 386 stuff out and tried to put together a functional system. Same deal - I got both these boards with parts missing and never quite had enough spares to make them both run. I only have one CPU and one main clock oscillator between the two of them, and I still don't have a BIOS for the MSI/Contaq board (left). But I finally managed to get keyboard controllers for both and burnt a BIOS for the Symphony board (right.) With everything installed, I wasn't able to get it to POST either. Maybe my BIOS, maybe something else wrong... I just don't know.

Finally, tired and frustrated from hours of swapping things around & not having any luck, I accidentally installed the oscillator backwards & cooked it. So that's done.

So out of two 486 & three 386 boards (one in my existing system), I still only have one of each that's confirmed good, but I can't say any of the others are definitely bad either. I hate having projects in limbo.

Endlessly trying things & having them just not work isn't fun. This is really reminding me of why I'm just kinda done with this generation of hardware... I can have a Win98 rig on a Pentium 2 board built from scratch and running 3DMarks in an hour, but this took most of the day and left my parts closet & work table a disaster. My original plan was to upgrade my existing 386 with one of the 486 boards, hang onto that (& probably keep a few choice CPUs since I have a nice spread of speed grades), and Ebay as much of the rest as I can in "tested & working condition." But now I'm thinking I should just flog the lot and build my K5/75 on a nice ATX board with integrated IDE controllers and large-ish HDD support instead. Bleah.

Ultrax wrote:

It tells me I have 16 KB of L1 cache. So it is an MMX!

No, it's not. Thermalwrong is wrong here. MMX P55C has 32 KB L1 cache, non-MMX has 16 KB. There never was any P5 CPU with only 8 KB of L1 cache.

wiretap wrote:

Just organized some of my towers..

Do I see a tower there with a built-in monitor?

xjas wrote:

Results: somewhat disappointing. All the CPUs work, after spending most of an hour meticulously un-bending pins. (Protip: a 72-pin SIMM works great as a Socket 3 pin straightener!) The DataExpertHiNT EISA+VLB combo board seems to work fine, but unfortunately I had no luck with the Gigabyte GA-486VS. Tried most of the CPUs, a bunch of different RAM including FPM & EDO, re-doing all the jumpers, etc. but the result was the same: no POST, not even any beeps.

The DataExpert is a pretty neat board, but I'd really prefer to use the Gigabyte for a few reasons - ZIF socket, 3.3V CPU support, and a more modern BIOS presumably supporting larger hard drives being the main ones. I haven't given up on it, but I ran out of stuff I could try for the time being. I got it with the BIOS chip missing so I had to burn my own; it's possible I did it wrong or I used an incompatible EPROM, although it verified OK.

I think I found a dump at this link, have you tried it?

http://th2chips.freeservers.com/ga486vs/index.html

derSammler wrote:

Do I see a tower there with a built-in monitor?

Yes -- It is a Dash OPS-1000 with amber CRT in the 5.25" bays.

That monitor looks really cool ~

dionb wrote:

Good luck there... the number of people into vintage networking is tiny and just as with the video stuff, some things are more sought after than others, and ISA/PCI modems, RTL8139 (typo I assume there) and NE2000 are bottom of the pile. I only manage to shift them by making bigger lots and adding that stuff for free. Apart from perhaps 3c509 cards the chance of flogging that stuff is equally tiny (unless that's an ISA WiFi card, in which case PM me 😉 )

I figured nobody will want these things, especially the two bottom barrel things (and yes, RTL8139), or I wouldn't have so many myself hahahaha. All my machines use Intel or 3com or other fancier NICs... Networking makes file transfer much nicer, the main reason I have NICs in all my computers anyway. One day I should try to set up a DOS LAN party 🤣.
No ISA WiFi card sadly, but I have an ISA PCMCIA adaptor and I will try to get it work with a WiFi card at some point 🤣.

derSammler wrote:

Ultrax wrote:

It tells me I have 16 KB of L1 cache. So it is an MMX!

No, it's not. Thermalwrong is wrong here. MMX P55C has 32 KB L1 cache, non-MMX has 16 KB. There never was any P5 CPU with only 8 KB of L1 cache.

That was the bit I didn't want to elaborate on 😜
The non-MMX has 8KB of instruction cache, that I don't think CacheChk can see, and there's 8KB of data cache, which it can. The MMX has a 16KB+16KB configuration instead.

I've been trying to get my Abit AN4 working - a very old VLB board which I bought for parts a couple of months ago, with 2 other corroded boards. This one is now back to operating properly after replacing the corroded bits and patching some wires 😁

This was my second go, the first time it gave a keyboard controller error. Last time, those wires that are under the hot glue were just long wires going straight to the KBC, it apparently did not like that. I've replaced them with short wires, they're held safely in place with glue and it works again. It got glued because the traces were coming away from the PCB, since the vias are long gone and putting those wires on tugged them a bit too much 😒
It boots and I need to put it into the case with the VLB cards I have. Now I'm up to 2 out of 3 corroded boards working.
Does anyone know if there are any BIOS files available for the AN4?

I also got a Shuttle HOT 317 motherboard, with an AM386DX40 and cache. The only way to identify it was it says so on the BIOS EEPROM.
It wasn't in such bad condition and may have booted / worked without cleaning up the corrosion. Now I've added in a coin cell instead of the NICAD battery. I disconnected the resistor R1, which connects to a diode feeding into the battery, to stop the computer from trying to charge lithium cell.

wiretap wrote:

Yes -- It is a Dash OPS-1000 with amber CRT in the 5.25" bays. […]
Show full quote

derSammler wrote:

Do I see a tower there with a built-in monitor?

Yes -- It is a Dash OPS-1000 with amber CRT in the 5.25" bays.

as i understand this is self-made project?

Tiido wrote:

That monitor looks really cool ~

Indeed. Nice stuff!

[...]

No ISA WiFi card sadly, but I have an ISA PCMCIA adaptor and I will try to get it work with a WiFi card at some point 🤣.

The holy grail would be a PCMCIA (not Cardbus) adapter with WiFi-g and so also WPA support. All the ISA and PCMCIA stuff I've seen is WiFi-b only, which limits security to WEP, which in turn means you have to leave the network wide open to allow them to connect.

Talking about PC-Card stuff, Installed my PCI -> Cardbus adapter in my P3. Good news: it's detected perfectly, drivers install nicely. Bad news: whenever I stick anything in there, it hangs the system when looking for drivers :'(

Not my day in other ways either. Spent half the afternoon getting all the documentation and .cfg files for my EISA cards and transferring all of that plus the AMI cfg.exe to the Gotek. Then hooked up the 486 EISA motherboard and... 0b on POST. Previously messing around with CPU in LIF-socket solved that. Not this time. Also happened when I removed the DS1387 RealtimeRAMified RTC+CMOS module for battery mod. So either some iffy traces in the board have finally died, or that DS1387 is dead. Hope that it's the DS1387, otherwise that board will need major surgery. Either way, my EISA stuff goes back into the box again.

Fortunately a re-check of other old stuff went better. Two 486 boards I want to sell are working fine, as is my old Hedaka 286 (even if it does like scaring me my taking a VERY long time to timeout on bad HDD settings).

wiretap wrote:

Just organized some of my towers..

They are beautiful!

PcBytes wrote:

I think I found a dump at this link, have you tried it?

http://th2chips.freeservers.com/ga486vs/index.html

Thanks, I think that's the one I used. It's also mirrored on here (in the 486 BIOS thread), but I THINK I compared them a while ago & they were the same.

Is there some kind of compatibility guide to what EPROMs to use (or EEPROMs) for PC BIOSes? I literally used an old BIOS chip for this so I thought it'd work, and it verified fine when I burned it. Could this board just be expecting a different type of chip?

@Wiretap: I've said this before, but words can't express how cool I think that thing is. The amber display with a smoked cover in a dark grey case really works. It looks like it came straight off the Nostromo. I also quite like the silver monolith on the middle row - I'd expect to see that cooling liquid nitrogen in a lab somewhere instead of housing PC internals. What is is?

Tried to clean up the hobby room a bit. Also built a shelf on top of my old table there to hopefully clear the clutter on the work area a bit. I need to find a way to mount the power strip on the bottom of the shelf to get some more space.

retropol wrote:

as i understand this is self-made project?

No, it came like that from the integrator.

Thermalwrong wrote:

I also got a Shuttle HOT 317 motherboard, with an AM386DX40 and cache. The only way to identify it was it says so on the BIOS EEPROM.
It wasn't in such bad condition and may have booted / worked without cleaning up the corrosion. Now I've added in a coin cell instead of the NICAD battery. I disconnected the resistor R1, which connects to a diode feeding into the battery, to stop the computer from trying to charge lithium cell.

ShuttleHOT317-CorrosionRepair&LithiumBatteryMod.jpg

I have this exact same board. It's my favourite 386DX board, very small foot print. I have 8 megs of RAM installed.

appiah4 wrote:

Also, the PCI version of the M64 is much more abundant than the PCI version of a normal TNT2, and they work very well in Socket 7 systems.

Does a proper 128-bit bus TNT2 even exist in PCI? I can't remember ever having seen one.

I was able to sort of fix a really sick IBM Model2 buckling spring Keyboard.

It had a few broken plastic tabs, being the most important ones those that pressed the pcb to the membrane. Not only that but it also had two bad caps (being that the reason why it was dead).
Luckily i was able to add these screws you see here to hold the pcb in place and help it make contact.

Also the foam that was there to make tension between the two was rotten. I had to make a new one from a piece of rubber.

I had to disassemble it two time in row until i got it right. Did i mention it is really infuriating to disassemble this keyboard?..

Had to make a RJ11 to PS/2 converter since this came from a IBM 3153 terminal.

It was sort of worth it. Sometimes A, B and K register twice, but for now i'm done for with this. I'm typing this with it and it kind of feels good. Also i had to retrobright the bottom cover since it was a bit yellowed. I still have another Model2 keyboard waiting for it's turn.

Ultrax wrote:

Ran ChkCpu on my portable 95 machine (Libretto 50CT) since I couldn't get CPU-Z (tried every old version possible) to work. It reported that my system has a non-MMX, 75 MHz Pentium.

That's cool, but this thing has a Pentium MMX sticker on it! Perhaps ChkCpu is misidentifying the processor, or maybe the only stickers available when this thing was made were the MMX ones? Definitely strange. Still love this little thing though!

No, the 50CT was not an MMX machine. I believe the 70CT, 100CT, and 110CT had MMX CPUs, but not the 50CT, which was a normal Pentium.

On the bright side, it's a hugely underclocked part and can do like 200MHz safely

This week I got my hands on an Intel Advanced/ATX motherboard (codename Thor):

It's pretty much the first ATX motherboard ever, and I'm pretty excited because I like Intel boards, I like ATX stuff and I like mid-90s and this board has it all 😀. It also comes with onboard S3 Trio64V+, Crystal 4232 codec and Yamaha OPL3 chip, so it's a foundation for a fairly high-end '95 computer. I've always wanted to have a dedicated Windows 3.11 machine, but was stuck, because I don't really enjoy 486 hardware, but I would also feel weird putting WFG on a very fast machine. This board actually came out just before Windows 95, so it's kind of perfect. Also it has "turbo" capabilities (disabling L2 cache I guess) that slows down the CPU to mid-386 level.

So I've spent the weekend making a pretty "period-correct" ATX build. I got an early ATX case with another Intel board (a TC430HX) and spent almost all Saturday cleaning it, removing rust stains and also oiling all the fans.

I must say, after spending some time with this case, early generic ATX cases were kind of horrible — you can't even put it back without using some bruteforce and I hate that I can't access the motherboard without removing the front bezel. I kind of just want to get an InWin A500 now (which is also one of the original ATX cases, but has a very nice design).

Another observation: "Advanced/ATX" is a horrible name when you're trying to Google an updated BIOS. I've been browsing intel.com through archive.org for hours, but there was no mention of the board in 1996 or 1997. Finally, I've reach 1999 just to find a dead download link. However, it gave me a filename: 10006cn0.exe. This one is definitely google-able.

Windows 3.11 is a fun experience, but I can see how it could be a nightmare back in the day. Compared to Win95 it's pretty much a DIY OS. There are just loads of extensions: TCP/IP, 32-bit code support, 32 bit HDD access (don't think I've ever needed a driver for a hard drive), video, fonts, CD-ROM support, etc. All very fragile, ready to go wrong at any moment. I wonder if there is any kind of an "unofficial service pack" that combines all the popular add-ons. I also want to try some novelty stuff like After Dark.

However, it's also interesting to see WFG as a much more "useful" OS than I remember it to be. I've never really used it daily and always thought it to be just a DOS shell. However, there is really a ton of 3rd party software. I might even try to go online with the help of WebOne proxy and something like Netscape.

At the end of the day, however, I've tried downloading a bunch of games from my home FTP server. When I came back in an hour I've got a "disk full" message and the system started giving me some really obscure error messages. After a reboot I've found that my whole file system was corrupted. I figured it was just one of those examples of WFG being really fragile, but then I ran Scandisk:

So, lesson learned: never install a bunch of software on a hard disk you've never used without running surface scan first.

Going to replace the Samsung disk with a 1.6GB Western Digital I know to be free of bad blocks and start over. Next week I'll receive a CT3930 SB32 (Vibra-based AWE32) to play around. I wonder if I can make soundfonts work with Win3.11 somehow because I know the Creative drivers support it.