65C02 wrote on 2024-02-04, 18:53:

I would like to see a slot A slotket!

me too. I don't know why we're not funding this. There are native slot A thunderbird Athlons - I presume making an adapter to run Socket A FSB100 athlons in a slot A motherboard is doable. I'd be quite interested - enough to put some money down.

PcBytes wrote on 2024-02-07, 03:27:

Another one from me: 440BX or LX based Socket 8 mobo. Bring AGP to the Pentium Pro.

And I don't mean a slotket, but a whole BX or LX equipped mobo with socket 8 instead of Slot 1.

Another problem that could be easily solved with a slotket - this time Socket 8 to slot 1. It would let you use a pentium pro on most pentium 2 motherboards - so AGP. It would also allow for overclocking the Ppro.

Count me in too, with the addition of a Goldfinger circuit built in 🤣

Socket3 wrote on 2025-03-21, 16:37:

Another problem that could be easily solved with a slotket - this time Socket 8 to slot 1. It would let you use a pentium pro on most pentium 2 motherboards - so AGP. It would also allow for overclocking the Ppro.

Those were made and pop up once in a while. The thing there is that having AGP and by implication 3D graphics, isn't that spectacular without MMX support.

I thought this was a neat way to put old hardware back to work:

Socket 7 NAS aiming for the stars

But it looks like maybe it was only done once, and maybe I'm the only other person interested? As far as a "thing to do," I would be interested in a hardware/software recipe for assembling a NAS using some older hardware like a Socket 7 board, K6-2/3+ CPU, some sort of used SATA/SAS/RAID card, and whatever software to get it working, probably linux of some sort. Used server hardware is very affordable at the moment, and several TB in a NAS would be both fun and useful. Anyway, my 2¢

nickles rust wrote on 2025-03-26, 19:09:

I thought this was a neat way to put old hardware back to work:

Socket 7 NAS aiming for the stars

But it looks like maybe it was only done once, and maybe I'm the only other person interested? As far as a "thing to do," I would be interested in a hardware/software recipe for assembling a NAS using some older hardware like a Socket 7 board, K6-2/3+ CPU, some sort of used SATA/SAS/RAID card, and whatever software to get it working, probably linux of some sort. Used server hardware is very affordable at the moment, and several TB in a NAS would be both fun and useful. Anyway, my 2¢

As a fun project ? Sure.

As an easily accessible retro datastore for stuff one already has archived AND backed up elsewhere ? Possibly.

As something useful for general use ? I would prefer something faster. PCI bus is limited to 133MB/sec minus overhead and software RAID on CPUs that old is meh and older hardware RAID controllers of that era will likely not handle multi-TB drives all that well if at all . Also, hardware that is 30+ years old and of potentially questionable reliable is not a platform that I would use for general purpose NAS purposes.

EDIT: I don't mean to put down your the work of the work of the person who set this up, by the way. There are definitely safe and practical use cases for this approach.

Socket3 wrote on 2025-03-21, 16:37:

65C02 wrote on 2024-02-04, 18:53:

I would like to see a slot A slotket!

me too. I don't know why we're not funding this. There are native slot A thunderbird Athlons - I presume making an adapter to run Socket A FSB100 athlons in a slot A motherboard is doable. I'd be quite interested - enough to put some money down.

Years ago I got a socket A to slot A slotket in a box of basically junk I bought off an auction.

It was only rated for the original Thunderbird up to 1.4ghz 100/200Fsb which I didn’t have, eventually when I downsized it went to ewaste

I never thought much about it until years later when I found out those dinguses never made it out of prototype because they weren’t stable.

My guess is the same applies today

rmay635703 wrote on 2025-03-28, 15:42:

Years ago I got a socket A to slot A slotket in a box of basically junk I bought off an auction. […]
Show full quote

Socket3 wrote on 2025-03-21, 16:37:

65C02 wrote on 2024-02-04, 18:53:

I would like to see a slot A slotket!

me too. I don't know why we're not funding this. There are native slot A thunderbird Athlons - I presume making an adapter to run Socket A FSB100 athlons in a slot A motherboard is doable. I'd be quite interested - enough to put some money down.

Years ago I got a socket A to slot A slotket in a box of basically junk I bought off an auction.

It was only rated for the original Thunderbird up to 1.4ghz 100/200Fsb which I didn’t have, eventually when I downsized it went to ewaste

I never thought much about it until years later when I found out those dinguses never made it out of prototype because they weren’t stable.

My guess is the same applies today

Even if they're unstable with faster CPUs, I think they are still worth it since faster slot A CPUs are expensive. If a socket A slotket is stable with a 900MHz socket A thunderbird I'm happy as a clam. If we can mod mainboard bios to support the Duron I'm even happier.

Socket3 wrote on 2025-03-30, 08:00:

rmay635703 wrote on 2025-03-28, 15:42:

Years ago I got a socket A to slot A slotket in a box of basically junk I bought off an auction. […]
Show full quote

Socket3 wrote on 2025-03-21, 16:37:

me too. I don't know why we're not funding this. There are native slot A thunderbird Athlons - I presume making an adapter to run Socket A FSB100 athlons in a slot A motherboard is doable. I'd be quite interested - enough to put some money down.

Years ago I got a socket A to slot A slotket in a box of basically junk I bought off an auction.

It was only rated for the original Thunderbird up to 1.4ghz 100/200Fsb which I didn’t have, eventually when I downsized it went to ewaste

I never thought much about it until years later when I found out those dinguses never made it out of prototype because they weren’t stable.

My guess is the same applies today

Even if they're unstable with faster CPUs, I think they are still worth it since faster slot A CPUs are expensive. If a socket A slotket is stable with a 900MHz socket A thunderbird I'm happy as a clam. If we can mod mainboard bios to support the Duron I'm even happier.

The instability was architectural so speed didn’t affect it.

rmay635703 wrote on 2025-04-11, 22:10:

Socket3 wrote on 2025-03-30, 08:00:

rmay635703 wrote on 2025-03-28, 15:42:

Years ago I got a socket A to slot A slotket in a box of basically junk I bought off an auction. […]
Show full quote

Years ago I got a socket A to slot A slotket in a box of basically junk I bought off an auction.

It was only rated for the original Thunderbird up to 1.4ghz 100/200Fsb which I didn’t have, eventually when I downsized it went to ewaste

I never thought much about it until years later when I found out those dinguses never made it out of prototype because they weren’t stable.

My guess is the same applies today

Even if they're unstable with faster CPUs, I think they are still worth it since faster slot A CPUs are expensive. If a socket A slotket is stable with a 900MHz socket A thunderbird I'm happy as a clam. If we can mod mainboard bios to support the Duron I'm even happier.

The instability was architectural so speed didn’t affect it.

no, it's not - it doesn't make sense. If you can cram a full thunderbird on a slot A package with no instability then a socket A thunderbird will work in a slotket. The only limiting factor would be speed and power delivery. I can see higher clocks and power draw inducing instability in a slot A to socket A slotket, but don't call it an architectural instability when native slot A thunderbird CPUs exist. The project probably failed because of little interest or cheap socket A motherboards.

so was the instability a bios problem then? Or the socket cpus didn’t like the power supply of the board or what specifically?

1. (reliable and fast) 16bit ISA card providing USB 2.0
2. Very fast 387SX FPU built
4. possibility to add L2 cache to a scamp 386sx (via VL82C325 (SX) Cache controller)
5. Plug In adapter for 386sx CPU upgrades WITHOUT soldering
6. 16 Bit ISA card providing WIFI (yes mentioned several times already)
7. (Auto-)- Re-Assembling of DOS4Gw games to std 16bit real mode
8. PLOP boot manager without need for 46(??) MB RAM
9. DOS ISO mounting software working also from network shares (SUCD won’t)
10. Tool to increase hz frequency for non VESA modes under dos (like text mode, std vga)
11. ultra SCSI for ISA (without need for any mysterious 10MB/s DMA setting 😀
12. composite CGA emulation under dos

Cannot and don’t want to insist that several things dont already exist.

Been thinking about making a fast 386sx board with a replaceable CPU socket, cache, and 16-bit VLB recently, but have a lot of other projects at the moment.

Lots of possibilities.

-I believe the Tseng ET4000-w32i can run in 16-bit VLB, but not 100%.
-RAM can be 72-pin (instead of 30pin), with 50ns modules.
-Asynchronous FPU XTAL set to 40Mhz (Cyrix/Reveal, Ulsi), and READY# AND'd with the local bus instead of BUSY#(?)
-Chipset: Chips 82C836 & 82C835 two-way set associative cache at 50Mhz using 8ns SRAM (W24257AJ-8N), and four-bank interleave RAM for page writes... or
-Chipset: VLSI Scamp VL82C310/113 & VL82C325 direct-map cache with NA pipelining.

Some boards have the pinout for a socket (4x dual rows of 25 pins), and 386sx's have the FLT# pin so they can be disabled, but a real socket would be better.

Would love to see that. Keep us pls informed. If you go for vlsi take care to use the -33 chipset instead of the -25 version.

Marco wrote on 2025-05-24, 20:57:

2. Very fast 387SX FPU built
5. Plug In adapter for 386sx CPU upgrades WITHOUT soldering
7. (Auto-)- Re-Assembling of DOS4Gw games to std 16bit real mode

2. What means "very fast 387SX FPU build"? A 387SX FPU running at 50 MHz synchronously with overclocked Am386SX-40@50 MHz?
5. Actually there are some CPU upgrade PCBs/modules with plug-in square plastic QFP "contact plugs" (I don't know how to call those), but these are very rare and hard to find nowadays.
7. There are 16-bit DOS extenders for 286 CPU (Rational System's DOS/16M and Phar Laps's TNT (286|DOS) DOS extender), but almost nobody used/s them. Recently I've asked about a 16-bit 286 port of FastDoom with some 16-bit DOS extender used in the FD's thread here and immediately started getting comments how this is impossible, how this wouldn't work, or that I have to sit down and write my own code from scratch, 'cause it would be a tremendous effort to adjust the FD code for 16-bit compiler and 16-bit DOS extender.

And where is 3.?

reg. 2.: Indeed good question. Idea: You can upgrade your 386SX socket by e.g. Cyrix, Ti486SXLC2, IBM, ... Maybe there will be a similar option for the FPU - so a high clocked FPU with optimized (486?) code execution etc.
reg. 5.: heard of it as well as rumor but never saw one
reg. 7.: I can fully understand that this "wish" wont be possible at least not in an automated way. Would anyway be interesting to see a performance comparison at least for one example from 32Bit Dos Ext --> 16Bit Ext --> Realmode
reg. 3.: See as for Larry 4 - seems to be lost forever

Marco wrote on 2025-05-25, 13:22:

reg. 7.: I can fully understand that this "wish" wont be possible at least not in an automated way. Would anyway be interesting to see a performance comparison at least for one example from 32Bit Dos Ext --> 16Bit Ext --> Realmode

Skills in x86 assembly and knowledge of DOS memory models, as (Fast)Doom code in some parts is written in 32-bit assembly and most likely needs to be converted for 16-bit registers of 286 CPU. "Real mode" is inherited limiting cr*p of 8086/88 CPUs and the Doom's original code has nothing to do with it. But this is topic for another thread.

An NE2000 compatible network card (NIC) that supports Gigabit ethernet (RJ45, optical) and has both an ISA and PCIe connector?
A flip card, essentially? That works on traditional AT class PCs (286 to 486), as well as the latest DOS PCs with CSM wrapper (CSMwrap)?

I know there are adapters to convert RJ45<>BNC and RJ45<>optical, but in all honesty: Let's make DOS great again™! :)

I hope that some FPGA can be used here, maybe.

Or maybe it's possible to use an Raspberry Pi on an ISA card, at least? Via GPIO pins?
Since Raspberry Pi 4 has Gigabit ethernet already, half the work is done already.

Great idea. Added to my wish list as well 😀

What also comes to mind: Optimizing VGA/VBE BIOSes for a couple of popular DOS cards. Such as ET-4000 or S3 Trio32/64.
Perhaps by modifying the machine code a little bit, so that 32-Bit instructions or String I/O instructions are used. Or so that linear frame buffer is enabled for VBE access.
That will break 80286 or 8086 processor compatiblity but might improve performance in 386/486 PCs and PC/AT emulators. Again, it's just an idea. An experiment.

Optimizing an AT BIOS and DOS (say FreeDOS) for using 32-Bit instructions through-and-through would be another one.
While remaining in pure Real-Mode, I mean. Adding a "cloaking" feature (see Helix software) would be cool, too.
So that large parts of the modified 32-Bit BIOS/DOS would run past the first Megabyte.

Not sure how practical all of this seems to you guys,
but I tink it would be interesting to see what happens if BIOS/DOS ran without bottlenecks (but still with 16-Bit application compatibility left intact).
In addition, if we could get rid of VGA modes in public domain or open source applications and use VBE 2.0, for example, we wouldn't need the VGA framebuffer in A segment so often.
So it could become available to conventional memory. Because the frmebuffer could be located at 15-16 MB (ISA SVGAs) or somewhere between 3,5-4GB (VLB/PCI SVGAs).
Again, just an idea, an Gedankenexperiment. Maybe for a special PC build or a dedicated emulator project. To see how well it goes..

Edit: Another idea that comes to mind: A 64-Bit subsystem for Windows 9x.
If there's Win32s that can run on 16-Bit Windows 3.1x (okay 16/32-Bit hybrid),
then wouldn't it be possible to make a Win64 compatibility layer, too?
Both Win32 API and Win64 API are very closely related.
Problem might be x64 processor mode, though.
Though it's possible to run 64-Bit guest VMs on a 32-Bit Windows XP/7 host using Virtualbox, for example, so there must be some way.
Maybe using hardware-assisted virtualization features of later CPUs, not sure.
If that's too complicated, then our imaginary "Win64s" might just have its own x64 CPU emulator?
Win3mu and OTVDM/WineVDM do use an emulated x86 core in order to run Win16 applications (Windows 3.1x applications) on Windows 10/11 x64 (and ARM64), for example.
And the ARM versions of Windows 11 do emulate x86/x64 in software, too.
Or use that Rosetta 2 on current macOS if available, not sure.
Anyway, being able to run a few 64-Bit utilities or simple 64-Bit games on Windows 98SE would be cool! 😁

I just thought of something that would be interesting...

Voodoo2 on an FPGA on a mini PCI card for late 90s early 00s laptops. Though either internally cobbling to VESA pins or whatever you'd need to do to get it on the LCD might be a pain. Otherwise I guess it's awkward ribbon cable to VGA out for monitor only.

If it did it over PCI completely like a PowerVR or MPEG card that would be easier. Power VR cards aren't super complex, wonder what the chip situation is like, hmmm.

edit: NEC chips might be "around" ... can't read Japanese tho ... https://datasheet.datasheetarchive.com/origin … SASW0064371.pdf